- Email: [email protected]
Discrimination of extra-virgin-olive oils from different cultivars and geographical origins by untargeted metabolomics
Accepted Manuscript Discrimination of extra-virgin-olive oils from different cultivars and geographical origins by untargeted metabolomics
Silvia Ghisoni, Luigi Lucini, Federica Angilletta, Gabriele Rocchetti, Daniela Farinelli, Sergio Tombesi, Marco Trevisan PII: DOI: Reference:
S0963-9969(18)30999-2 https://doi.org/10.1016/j.foodres.2018.12.052 FRIN 8177
To appear in:
Food Research International
Received date: Revised date: Accepted date:
17 September 2018 4 December 2018 23 December 2018
Please cite this article as: Silvia Ghisoni, Luigi Lucini, Federica Angilletta, Gabriele Rocchetti, Daniela Farinelli, Sergio Tombesi, Marco Trevisan , Discrimination of extravirgin-olive oils from different cultivars and geographical origins by untargeted metabolomics. Frin (2018), https://doi.org/10.1016/j.foodres.2018.12.052
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Discrimination of extra-virgin-olive oils from different cultivars and geographical origins by untargeted metabolomics
Silvia Ghisonia, Luigi Lucinia* , Federica Angillettaa, Gabriele Rocchettia,b, Daniela Farinellic,
Department for Sustainable Food Process, Università Cattolica del Sacro Cuore, Via Emilia
RI
a
PT
Sergio Tombesid, Marco Trevisana
b
SC
Parmense 84, 29122 Piacenza, Italia
Department of Animal Science, Food and Nutrition, Università Cattolica del Sacro Cuore, Via
c
Department of Agricultural, Food and Environmental Sciences, Università degli Studi di Perugia,
MA
Via Borgo 20 giugno 74, 06154, Perugia, Italy d
NU
Emilia Parmense 84, 29122 Piacenza, Italia
Department for Sustainable Crop Production, Università Cattolica del Sacro Cuore, Via Emilia
AC C
EP T
ED
Parmense 84, 29122 Piacenza, Italia
*
Corresponding author: via Emilia Parmense 84, 29122 Piacenza, Italy. Telephone: +39 0523 599156. E-mail: [email protected] 1
ACCEPTED MANUSCRIPT Abstract In
this preliminary study,
quadrupole-time-of-flight
ultra-high-pressure liquid
chromatography (UHPLC)
mass spectrometry (QTOF) metabolomics followed
coupled
to
by multivariate
statistics was applied to discriminate nine extra-virgin olive oil (EVOO) cultivars according to their phenolic and sterolic fingerprints. The same approach was then used to discriminate EVOO samples
PT
from different geographical origins, namely six blends representative of the main growing regions
RI
in Italy (i.e., Sicily, Puglia, Umbria, Liguria, Lombardy and Tuscany).
SC
This approach allowed to putatively identify more than 1000 compounds, considering both polyphenols and sterols. The unsupervised hierarchical cluster analysis (HCA) and the orthogonal
NU
projections to latent structures discriminant analysis (OPLS-DA) discriminated EVOO samples according to both cultivar and geographical origin. In particular, flavonoids (i.e., anthocyanins and
MA
flavonols), hydroxycinnamic acids and cholesterol derivatives were found to be the most representative classes of compounds discriminating EVOO samples according to the two
ED
parameters (cultivar or geographical origin) selected. However, the following Venn analysis allowed to point out the discriminant markers being exclusive for cultivar or origin discrimination.
EP T
In this regard, only the 9.6% of phenolics and 13.6% of sterols were in common, thus indicating that several of these discriminant compounds were exclusive of a single condition. Indeed,
AC C
considering that most of the commercial EVOOs are blended, the contribution of both the cultivars used and the geographical origin must be taken into account.
Keywords: Extra-Virgin Olive Oil; Food metabolomics; Polyphenols; Sterols; Traceability.
2
ACCEPTED MANUSCRIPT 1. Introduction Extra-Virgin Olive Oil (EVOO) is a main component of the Mediterranean diet, displaying health-promoting effects thanks to both its unsaturated fatty acids profile and the presence of bioactive compounds such as polyphenols and tocopherols (Blasi et al., 2018; Suarez, Caimani, del Bas, & Arola, 2017). EVOO is produced in all the countries in the Mediterranean basin, with Spain,
PT
Italy, Tunisia and Greece being the most important producers worldwide (Kalogiouri, Aalizadeh, &
RI
Thomaidis, 2018).
SC
The quality and organoleptic properties of EVOOs depend on many factors such as cultivar, geographical origin, climatic conditions, agronomic techniques and processing technologies, with
NU
consequences also in the final price of the product in the market (Cubero-Leon, Penalver, & Maquet, 2014). Therefore, the protection of olive oils according to their origin, such as protected
MA
designation of origin (PDO), protected geographical indication (PGI) and traditional specialties guaranteed (TSG), has become an essential information in the label, thus entailing the adoption of
ED
methods for authentication (Merchak, et al., 2017).
EVOO is considered a common target for economically motivated adulteration, because of its
EP T
high demand and profitability potential (Esteki et al., 2018). As recently reviewed, chromatographic methods coupled to chemometrics, including untargeted metabolite profiling have been applied for
AC C
authentication purposes in foods, including olive oils (Esteki et al., 2018). Efforts have been done in the area of PDOs and PGIs to avoid fraudulent labelling, but a definitive and robust solution to this problem is still not present. Both PDO and PGI designate products that are linked to the area of which they bear the name and include niche products from small-scale processes in limited areas where growth conditions confer a particular quality to the product (Cubero-Leon et al., 2014). In the last years, several research attempts tried to use the chemical fingerprint of EVOO samples for authentication purposes. For example, according to Aparicio, Morales, Aparicio-Ruiz, Tena and Garcìa-Gonzàlez (2013), sterols represent a very useful tool for the authentication of olive oils. These authors used sterols as index of fraudulent oil mixtures, underlying the importance of 3
ACCEPTED MANUSCRIPT chromatographic techniques to quantify the free and esterified forms in short time. Different studies have been carried out also to authenticate olive oils from Italy (Alonso-Salces et al., 2010), Greece (Longobardi et al., 2012; Alonso-Salces et al., 2010) and Spain (Becerra-Herrera et al., 2018). Furthermore, Ben Mohamed and co-authors (2018) have recently used both sterol and phenolic fingerprints to discriminate Tunisian and Italian EVOOs, outlining the potential of untargeted
PT
metabolomics for authenticity evaluations. Indeed, metabolomics, defined as “the unbiased, global
RI
screening approach to classify samples based on metabolite patterns or fingerprints that change in response to disease, environmental or genetic perturbations with the ultimate goal to identify
SC
discriminating metabolites” (Sales, et al., 2017) might represents a powerful approach to ensure the
NU
authenticity of olive oils in relation with the geographical origin (Esteki et al., 2018) or sensorial properties (Sales, Portolés, Johnsen, Danielsen & Beltran, 2019).
MA
However, to the best of our knowledge, the application of untargeted metabolomics for authenticity markers discovering in different EVOO cultivars from different origins is still scarce.
ED
Therefore, the aim of this work was to apply a metabolomic approach followed by multivariate statistics to identify sterol and phenolic biomarkers for authenticity purposes. Most of the
EP T
commercial oils are sold as blends, where a contribution of both the cultivars used and the geographical origin is expected. On this basis, the possibility to discriminate different EVOOs as a
AC C
function of the cultivar and according to the actual geographical origin has been investigated.
2. Materials and methods 2.1. Samples
Fifteen EVOO samples were included in this study. Six of them are blends representative of the main growing regions in Italy (i.e., Sicily, Puglia, Umbria, Liguria, Lombardy, Tuscany) were kindly provided by COOP Italia SpA (Casalecchio di Reno, Bologna, Italy) a commercial supplier who owns the supply chain and could ensure geographical origins. The oil from each origin was
4
ACCEPTED MANUSCRIPT provided in triplicate; two additional samples (from Sicily and Tuscany, respectively) have been provided to be used as quality control for validation purposes. The remaining nine samples Mediterranean
basin
(i.e.,
represented
‘Leccino’,
some widely diffused
‘Picual’,
‘Picholine marocaine’,
olive cultivars of the ‘Frantoio’,
‘Sourani’,
‘Kalamon’, ‘Manzanilla de Sevilla’, ‘Arbequina’ and ‘Koroneiki’), and were obtained by olives
PT
harvested in the germplasm repository of the University of Perugia (Italy). EVOO was obtained
RI
from 2 kg of fruit samples harvested on November 15th , 2016 from three trees per each cultivar. Oil
SC
was extracted within 24 hours by a custom-built plant, allowing to process small amount of fruits. Oil samples were immediately placed in cold storage (10 ± 2 °C) until further analyses. Three
NU
replicate samples were obtained per accession.
MA
2.2. Untargeted profiling of phenolic and sterol compounds
Each oil sample was analysed in duplicate, with the exception of quality controls that were
ED
analysed in triplicate. For the extraction step, the method reported by Ben Mohamed and co-authors (2018) was used. Briefly, three individual sample replicates from each sample (3 g each) were
EP T
extracted in 3 mL of 80% methanol solution (v/v) (LCMS grade, VWR, Milan, Italy) using a vortex mixer. The suspension was kept at room temperature for 30 minutes and then centrifuged at 6,000 x
analysis.
AC C
g for 10 min at 4 °C. The extracts were collected in 2 mL amber glass vials and stored at 4 °C until
Phenolic compounds and
sterols were then screened
through ultra-high-pressure liquid
chromatography (UHPLC) coupled to an electrospray quadrupole-time-of-flight hybrid mass spectrometer (UHPLC-ESI/QTOF-MS), as previously described (Rocchetti et al., 2017). With this purpose, a 1290 series liquid chromatograph and a G6550 iFunnel mass spectrometer detector equipped with a JetStream dual electrospray (all from Agilent Technologies, Santa Clara, CA, USA), were used. Chromatographic separation was achieved on an Agilent Zorbax Extend-C18 column (75 × 2.1 mm i.d., 1.8 μm) in reverse phase mode (gradient elution from 5% to 90% LC-MS 5
ACCEPTED MANUSCRIPT grade methanol). The mass spectrometer data acquisition was MS-only mode in positive polarity (range 100–1000 m/z) and injection volume was 3 μL. Source conditions were as follows: nitrogen was used as both sheath gas (10 L min-1 and 350 °C) and drying gas (8 L min-1 and 330 °C); nebulizer pressure was 60 psi, nozzle voltage was 300 V, and capillary voltage was 3.5 kV. Raw data were processed using Profinder B.07 (from Agilent Technologies), for mass and
PT
retention time alignment and for the following annotations based on the ‘find-by-formula’ algorithm
RI
(Pretali, Bernardo, Butterfield, Trevisan & Lucini, 2016). According this approach, the entire
SC
isotopic pattern (monoisotopic accurate mass, isotope spacing, and isotope ratio - mass accuracy < 5 ppm) was considered for compounds identification by linking the isotopic profile to a custom
NU
database obtained by combining Phenol-Explorer 3.6 (http://phenol-explorer.eu/) and LIPID MAPS
2.3. Multivariate statistical analysis
MA
(Lipidomics Gateway, http://www.lipidmaps.org/).
ED
Metabolomics data on phenolic and sterol profiling were interpreted in Agilent Mass Profiler Professional B.12.06 as previously reported (Lucini et al., 2016). Compounds were filtered by
EP T
abundance (only those compounds with an area > 10000 counts were considered), normalized at the 75th percentile and baselined to their median in the dataset. Post-acquisition processing included
AC C
also filtering by frequency: only those compounds identified within 100% of replications in at least one treatment were retained. Statistics were finally carried out on this latter dataset; unsupervised hierarchical cluster analysis (HCA) was carried out setting the similarity measure as ‘Euclidean’ and ‘Wards’ as the linkage rule. Thereafter, the raw dataset was exported into SIMCA 13 (Umetrics, Malmo, Sweden), Pareto scaled, and elaborated for orthogonal partial least squares discriminant analysis (OPLS-DA) supervised modelling. Herein, the variation between the groups was separated into predictive and orthogonal (i.e., ascribable to technical and biological variation) components. The presence of outliers was investigated according to Hotelling’s T2 using 95% and 99% confidence limits for 6
ACCEPTED MANUSCRIPT suspect and strong outliers respectively. Cross-validation of the OPLS-DA model was carried out using CV-ANOVA (p < 0.01) and permutation testing carried out to exclude overfitting (N=300), after inspecting model parameters (goodness-of-fit R2 Y and
goodness-of-prediction Q 2 Y).
Regarding Q 2 Y prediction ability, a value > 0.5 was adopted as a threshold to identify acceptable models, according to software recommendation and as set out in the literature (Rombouts et al.,
PT
2017). Finally, the variables importance in projection (VIP analysis) selection was then used to
RI
evaluate the importance of metabolites and to select those having the highest discrimination
SC
potential (VIP score > 1.2).
Finally, to discern among the markers being specifically responsible for origin discrimination
NU
from those related to cultivars, a Venn analysis has been carried out using an online tool (Venny 2.1
MA
- http://bioinfogp.cnb.csic.es/tools/venny/index.html).
3. Results and discussion
ED
3.1. Discrimination of different EVOO cultivars according to phenolic and sterols profiles
EP T
In the first part of this work, the untargeted profiling of EVOO samples was carried out in order to shed light onto the differences in secondary metabolites as a sole function of the cultivar, since
conditions.
AC C
the olive trees were grown in the same field and were therefore affected by the same pedo-climatic
Overall, the metabolomics-based approach allowed us to putatively annotate 1354 compounds (444 polyphenols and 910 sterols; supplementary material). The metabolomic profiling allowed to discriminate the nine different EVOO cultivars according to their sterols and phenolic fingerprints. In particular, both unsupervised and supervised multivariate statistics (i.e., HCA and OPLS-DA) let us to discriminate the different cultivars analyzed. In this regard, the HCA provided good clusters, considering that biological replicates belonging to the same variety grouped together and no mixed clusters could be observed (supplementary material). In more detail, two main clusters could be identified, suggesting a very similar metabolites profile between ‘Manzanilla’ and ‘Kalamon’ 7
ACCEPTED MANUSCRIPT cultivars when comparing to the others. These results were confirmed through the supervised OPLS discriminant analysis, using the cultivar as class membership criterion. As shown in the OPLS score plot (Figure 1), a clear discrimination of the cultivars was obtained, thus suggesting a strong relationship with the secondary metabolites’ profile. Looking in detail the score plot outlined by OPLS-DA, three main groups were obtained: ‘Manzanilla’ and ‘Kalamon’ cultivars clustered
PT
together, while a second major cluster comprising ‘Frantoio’, ‘Arbequina’, ‘Picual’ and ‘Koroneiki’
RI
oils was noticed. However, ‘Leccino’, ‘Sourani’ and ‘Picholine marocaine’ showed a more distinct metabolites profile when compared to the others. The model cross-validation parameters were
SC
found to be very robust, being the goodness-of-fit (R2 Y) and the goodness-of-prediction (Q2 Y) of
NU
0.97 and 0.91, respectively. Besides, no outliers could be observed by Hotelling’s T2, whereas both CV-ANOVA cross-validation (p < 0.01 for regression) and permutation test (N = 100) for
MA
overfitting were more than adequate (supplementary material). The most representative compounds for discrimination purposes were selected by using the VIP
ED
variable selection method, following OPLS-DA supervised modeling, and then grouped into chemical classes and subclasses (Table 1). The most represented class of polyphenols was
EP T
flavonoids (28 compounds), with anthocyanins such as delphinidin 3-O-sambubioside, peonidin 3O-(6''-acetyl-glucoside) and petunidin 3-O-galactoside, possessing a VIP score > 1.3. When
AC C
considering the other flavonoid-subclasses namely dihydrochalcones, flavonols and flavones, phloretin, quercetin 3-O-glucosyl-xyloside and tetramethylscutellarein possessed the highest VIP score (1.44, 1.33 and 1.28, respectively). Interestingly, among the phenolic acids outlined by VIP analysis, hydroxycinnamic acids were more discriminating than hydroxybenzoic acids, with the isomeric forms of feruloylquinic acid possessing the highest VIP score (1.35). The most discriminating sterols outlined by VIP were exported and classified in subclasses, by exploiting the information provided by the LIPID MAPS database (Lipidomics Gateway, http://www.lipidmaps.org/)
(Table
1).
Several sterol-derivatives characterized
the differences
among the cultivars considered, with an abundance of cholesterols, cardanolides, furostanols, 8
ACCEPTED MANUSCRIPT spirostanols, stigmasterols and other compounds. The highest discrimination potential was found for a stigmasterol (ajugalactone; VIP score: 1.40) followed by a cholesterol derivative (3α,7α,12αTrihydroxy-5β-cholestanoyl-CoA; VIP score: 1.37) and two furostanols (i.e., isomeric forms of pseudoprotodioscin; VIP scores: 1.36). The cultivars analyzed in this work are generally suitable for the production of table olive,
PT
further than for olive oil extraction. Interestingly, ‘Manzanilla’ and ‘Kalamon’, two of the most
RI
important cultivars for oil production, clustered together into the OPLS score plot, whereas
SC
‘Frantoio’, ‘Arbequina’ and ‘Koroneiki’, i.e., those cultivars characterized by a smaller fruit size, where found to be very close into the OPLS score plot according to their sterols and phenolic
NU
fingerprints. A recent work used QTOF-based metabolomics followed by multivariate statistics to discriminate EVOO cultivars from Greece (Kalogiouri et al., 2018). Notably, despite these authors
MA
used a wider database for annotations (i.e., the food database FooDB), phenolic compounds and flavonoids in particular were selected as the compounds displaying the highest discrimination
ED
potential. Conversely, Piccinonna et al. (2016) reported oleic acid and fatty chains as discriminants of Apulian olive oils. However, the metabolomic approach they applied was based on NMR
EP T
spectroscopy, and likely had a compounds coverage complementary to our QTOF screening.
AC C
3.2. UHPLC/QTOF-MS discrimination of EVOOs from different geographical origin The previous approach was used also to discriminate EVOO samples from different regions of Italy. The hierarchical cluster analysis (supplementary material) showed a clear separation between each oil sample, with all the biological replicates clustering together. In particular, it is possible to notice that “Regionale del Garda” was clearly separated from all the other samples, thus suggesting a different secondary metabolites profile. These findings are quite interesting considering that the main cultivars used for producing this EVOO are ‘Casaliva’, ‘Leccino’ and ‘Frantoio’ and that ‘Casaliva’ is considered as a possible synonym of ‘Taggiasca’ and ‘Frantoio’ (Muzzalupo, Vendramin, & Chiappetta, 2014). However, considering that ‘Taggiasca’ is the main cultivar 9
ACCEPTED MANUSCRIPT characterizing ‘Regionale ligure” EVOO, it was surprising to observe a so clear segregation of these two samples. However, it must be considered that ‘Regionale ligure” is produced from olives grown on the seaside, whereas “Regionale del Garda” is produced in the Northern area among those considered, hundred kilometers far from the coast, in a very particular climatic niche. Given the HCA outcomes from unsupervised analysis, the following OPLS-DA supervised
PT
modeling was carried out to identify markers for geographical origin. The score scatter plot (Figure
RI
2) confirmed the results obtained from the HCA, thus proving that the metabolomic profiling proposed was able to discriminate EVOO samples according to the geographical origin. In this
SC
regard, the model parameters were found to be excellent, being R2 Y = 0.97 and Q 2 Y = 0.93. No
NU
outlier samples could be observed by Hotelling’s T2, whereas both CV-ANOVA cross-validations (p < 0.01 for regression) and test for overfitting (permutation test, N = 100) were adequate
MA
(supplementary material). In order to perform an external validation of the chemometric approach applied, two quality controls of known origin (namely Tuscany and Sicily) were also analyzed
ED
using the same method, and their metabolomic profiles included into the OPLS-DA model. The resulting misclassification table (a data elaboration displaying class prediction accuracy), allowed
EP T
including both quality controls within the correct class, with a Fisher’s P of 3.6 E-26 (supplementary material). Therefore, this further investigation confirmed the prediction ability of
AC C
the multivariate approach we developed. The variable selection methods applied to OPLS-DA model (VIP analysis) allowed identifying the most discriminant phenolic and sterols among the compounds annotated. These compounds are reported in Table 2. Overall, 47 phenolic compounds possessed a VIP score > 1.2. In particular, flavonoids were found to be the most common class (21 compounds), with anthocyanins (7 compounds) and flavonols (10 compounds) being the most frequent compounds. Notably, glycosidic forms of kaempferol and cyanidin possessed very represented among these two latter phenolic subclasses. Phenolic acids were the second most frequent class (17 compounds), with a clear abundance of hydroxycinnamic acids, such as isomeric forms of caffeoylquinic and 10
ACCEPTED MANUSCRIPT sinapoylquinic acids. In addition, other phenolics such as lignans and stilbenes were reported. However, the compounds possessing the highest VIP score were found to be 6’’-O-Acetylglycitin (belonging to isoflavonoids) and p-Coumaroyl glucose (hydroxycinnamic acid) with VIP scores of 1.43 and 1.40, respectively. Moving to the sterols identified by VIP analysis, cholesterol (5 compounds), spirostanols (4 compounds), ergosterol (4 compounds), steryl esters (3 compounds)
PT
together with stigmasterol, furostanol and cycloartanol (2 compounds) derivatives were the most
RI
represented compounds. In this case, iokundjoside (i.e., cardanolides) and solaspigenin (i.e.,
SC
spirostanols) were characterized by the highest VIP scores (i.e., 1.41 and 1.35, respectively). It is possible that sterols profile in EVOO is affected also by the oil microbiota, typically present in solid
NU
particles and micro drops of vegetation water, where yeasts are predominant (Ciafardini & Zullo, 2018).
MA
The importance of a metabolomics-based approach to discriminate and authenticate EVOOs from different geographical origin was recently reported. For example, Gil-Solsona and co-authors
ED
(2016) used an UHPLC-QTOF MS approach to identify possible biomarkers able to discriminate different EVOO samples, while Ben Mohamed et al. (2018) using the same analytical conditions
EP T
were able to discriminate Tunisian and Italian EVOOs according to their sterols and phenolic fingerprints. The need to authenticate and differentiate EVOOs olive oils from different origins is
AC C
an essential procedure for both the industry and the final consumer. In our work, a metabolomic profiling approach has been proposed, focusing on two classes of secondary metabolites that are affected by the environment and are widely represented in EVOO. A previous work successfully using metabolomics to authenticate geographical origin of EVOOs from Spain (Gil-Solsona et al., 2016), selected twelve markers, seven of which could be identified; all these markers were lipids, even though no sterols were selected. These differences can be likely ascribed to the differences in acquisition and annotation between the present and the previous work. Conversely, Peršurić, Saftić, Mašek & Kraljević Pavelić (2018), used a combination of MALDI/TOF, GC/MS and NIRS to
11
ACCEPTED MANUSCRIPT authenticate geographical origin of Croatian EVOOs. According to the approach proposed, triacylglycerols were reported to be the most discriminating compounds. Therefore, several markers have been proposed for geographical origin of EVOOs, each of them having its peculiarities. Considering that the very most of commercial EVOOs are blend of different cultivars, it is important to discern whether the proposed markers are specifically related to the
PT
origin rather than being the result of the cultivar(s) considered. With this purpose, we carried out a
RI
Venn analysis to point out to which extent the markers for cultivars discrimination are shared with
SC
those for origin discrimination (Figure 3). As it can be noted, 9.6% of phenolics and 13.6% of sterols were in common, thus indicating that the very most of discriminant compounds are exclusive
NU
of a single condition. Likely, this is the result of using OPLS-DA supervised modeling approaches, where the predictive covariance related to a given class membership (either cultivar or origin) is
MA
separated from orthogonal non-predictive covariance.
Despite the novelty and encouraging findings of the approach proposed, based on untargeted
ED
phenolic and sterolic profiling associated to supervised multivariate OPLS-DA, further research is still needed. In fact, it is known that several factors might affect secondary metabolite profile in
EP T
olive fruit and consequently also in oil; therefore, broader investigations based on a higher number
AC C
of samples and geographical origins might be advisable.
4. Conclusions
Adulteration and misleading label information are a serious problem in the food sector; with this regard, particular attention is being paid to olive oils, thus demanding robust and effective techniques for authentication purposes. In this work, an untargeted metabolomics-based approach was used to identify phenolic and sterol compounds able to discriminate nine samples of EVOOs from different cultivars (grown in the same conditions) and blend samples of EVOOs having different geographical origin. The results showed that the profile of these secondary metabolites can be a powerful tool for discrimination purposes, with regards to cultivars and geographical origin, 12
ACCEPTED MANUSCRIPT confirming that UHPLC-MS metabolomic profiling represents an effective approach in food traceability. In particular, although a clear effect of the cultivar can be identified (thus indicating a strong genetic basis), the interaction with the environment results in a distinctive profile in phenolics and sterols. Holistic approaches such as untargeted profiling, associated to supervised multivariate analyses, allowed identifying the markers specifically related to geographical origin.
PT
Notwithstanding the preliminary nature of our research, the two classes of secondary metabolites
RI
considered in this work are entitled to represent a complementary and promising solution for plant
SC
foods traceability, thus deserving further investigations.
From a practical point of view, it is advisable that the markers proposed will be included in a
NU
validated targeted quantification method, e.g. based on less sophisticated and cheaper mass spectrometers such as triple quadrupole MS/MS. This would result in offering a simple, robust and
MA
cost-effective way to improve authentication of EVOOs on routine basis. Furthermore, given the promising outcomes of metabolomics, research aimed to investigate whether metabolomics could
ED
help to identify fraudulent activities (e.g., mixing lower quality oils in EVOOs) is worthwhile in the
EP T
next future.
AC C
Funding
This research did not receive any specific grant from funding agencies in the public, commercial and not-for-profit sectors.
Acknowledgements SG and GR were recipient of a Ph.D. fellowship from the Doctoral School AgriSystem, Piacenza, Italy. The authors thank the “Romeo ed Enrica Invernizzi” foundation for kindly supporting the metabolomic platform.
13
ACCEPTED MANUSCRIPT
References Alonso-Salces, R. M., Moreno-Rojas, J. M., Holland, M. V., Reniero, F., Guillou, C., & Héberger, K. (2010). Virgin olive oil authentication by multivariate analyses of 1 H NMR fingerprints and
PT
13 C and 2 H data. Journal of Agricultural and Food Chemistry, 58(9), 5586-5596.
RI
Aparicio, R., Morales, M., Aparicio-Ruiz, R., Tena, N., & Garcìa-Gonzàlez, D. (2013). Authenticity of olive oil: Mapping and comparing official methods and promising alternative. Food Research
SC
International, (54), 2025-2038.
NU
Becerra-Herrera, M., Vélez-Martín, A., Ramos-Merchante, A., Richter, P., Beltrán, R., & Sayago, A. (2018). Characterization and evaluation of phenolic profiles and color as potential
ED
origin. Food Chemistry, 241, 328-337.
MA
discriminating features among Spanish extra virgin olive oils with protected designation of
Ben Mohamed, M., Rocchetti, G., Montesano, D., Ben Ali, S., Guasmi, F., Grati-Kamoun, N., &
EP T
Lucini, L. (2018). Discrimination of Tunisian and Italian extra-virgin olive oils according to their phenolic and sterolic fingerprints. Food Research International, 106, 920-927.
AC C
Blasi, F., Rocchetti, G., Montesano, D., Lucini, L., Chiodelli, G., Ghisoni, S., Baccolo, G., Simonetti, M. S., & Cossignani, L. (2018). Changes in extra-virgin olive oil added with Lycium barbarum L. carotenoids during frying: Chemical analyses and metabolomic approach. Food Research International, 105, 507-516. Ciafardini, G., & Zullo, B. A. (2018). Virgin olive oil yeasts: A review. Food Microbiology, 70, 245-253. Cubero-Leon, E., Penalver, R., & Maquet, A. (2014). Review on metabolomics for food authentication. Food Research International, 60, 95-107.
14
ACCEPTED MANUSCRIPT Esteki, M., Simal-Gandara, J., Shahsavari, Z., Zandbaaf, S., Dashtaki, E., & Vander Heyden, Y. (2018). A review on the application of chromatographic methods, coupled to chemometrics, for food authentication. Food Control, 93, 165-182. Gil-Solsona, R., Raro, M., Sales, C., Lacalle, L., Díaz, R., Ibánez, M., Beltran, J., Sancho, J. V., & Hernández, F. J. (2016). Metabolomic approach for Extra virgin olive oil origin discrimination
PT
making use of ultra-high performance liquid chromatography - Quadrupole time-of-flight mass
RI
spectrometry. Food Control, (70), 350-359.
SC
Kalogiouri, N. O., Aalizadeh, R., & Thomaidis, N. S. (2018). Application of an advanced and wide scope non-target screening workflow with LC-ESI-QTOF-MS and chemometrics for the
NU
classification of the Greek olive oil varieties. Food Chemistry, 256, 53-61.
MA
Longobardi, F., Ventrella, A., Casiello, G., Sacco, D., Tasioula-Margari, M., Kiritsakis, A., & Kontominas, M. (2012). Characterisation of the geographicasl origin of Western Greek virgin
ED
olive oils based on instrumental and multivariate statistical analysis. Food Chemistry, 133, 169-
EP T
175.
Lucini, L., Borgognone, D., Rouphael, Y., Cardarelli, M., Bernardi, J. & Colla, G. (2016). Mild Potassium Chloride Stress Alters the Mineral Composition, Hormone Network, and Phenolic
AC C
Profile in Artichoke Leaves. Frontiers in Plant Scence, 7, 948. Merchak, N., El Bacha, E., Khouzam, R. B., Rizk, T., Akoka, S., & Bejjan, J. (2017). Geoclimatic, morphological, and temporal effects on Lebanese olive oils composition and classification: A 1 H NMR metabolomic study. Food Chemistry, 217, 379-388. Muzzalupo, I., Vendramin, G. G., & Chiappetta, A. (2014). Gemetic biodiversity of Italian olives (Olea europaea) germplasm analyzed by SSR markers. The Scientific World Journal, vol. 2014, Article ID 296590, 12 pages. https://doi.org/10.1155/2014/296590.
15
ACCEPTED MANUSCRIPT Peršueić, Z., Saftić, L., Mašek, T., & Pavelić, S. K. (2018). Comparison of triacylglycerol analysis by MALDI-TOF/MS, fatty acid analysis by GC-MS and non-selective analyis by NIRS in combination with chemometrics for determination of extra virgin olive oil geographical origin. A case study. LWT, 95, 326-332. Piccinonna, S., Ragone, R., Stocchero, M., Del Coco, L., De Pascali, S. A., Schena, F. P., &
PT
Fanizzi, F. P. (2016). Robustness of NMR-based metabolomics to generate comparable data sets
RI
for olive oil cultivar classification. An inter-laboratory study on Apulian olive oils. Food
SC
Chemistry, 199, 675-683.
Pretali, L., Bernardo, L., Butterfield, T. S., Trevisan, M., & Lucini, L. (2016). Botanical and elicit
a
similar
Induced
Systemic Response in tomato
NU
biological pesticides
(Solanum
MA
lycopersicum) secondary metabolism. Phytochemistry, 130, 56-63. Rocchetti, G., Chiodelli, G., Giuberti, G., Masoero, F., Trevisan, M., & Lucini, L. (2017).
ED
Evaluation of phenolic profile and antioxidant capacity in gluten-free flours. Food Chemistry, 228, 367-373.
EP T
Rombouts, C., Hemeryck, L. Y., Van Hecke, T., De Smet, S., De Vos, W. H., & Vanhaecke, L. (2017). Untargeted metabolomics of colonic digests reveals kynurenine pathway metabolites,
AC C
dityrosine and 3-dehydroxycarnitine as red versus white meat discriminating metabolites. Scientific Reports, 7, 42514. Sales, C., Cervera, M., Gil, R., Portolés, T., Pitarch, E., & Beltran, J. (2017). Quality classification of Spanish olive oils by untargeted gas chromatography coupled to hybrid quadrupole-time of flight mass spectrometry with atmospheric pressure chemical ionization and metabolomics-based statistical approach. Food Chemistry, (216), 365-373. Sales, C., Portolés, T., Johnsen, L. G., Danielsen, M., & Beltran, J. (2019). Olive oil quality classification and measurement of its organoleptic attributes by untargeted GC-MS and multivariate statistical-based approach. Food Chemistry, 271, 488-496. 16
ACCEPTED MANUSCRIPT Suarez, M., Caimani, A., del Bas, J., & Arola, L. (2017). Metabolomics: An emerging tool to evaluate the impact of nutritional and physiological challenges. Tends in Analiytical Chemistry, (96), 79-88.
PT
FIGURE CAPTIONS Figure 1. Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) on
RI
EVOO samples phenolic and sterolic profile, using the ‘cultivar-type’ as class membership
SC
criterion. Individual replications are given in the class prediction model score plot.
NU
Figure 2. Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) on EVOO samples phenolic and sterolic profile according to their geographical origin. Individual
MA
replications are given in the class prediction model score plot. Figure 3. Venn diagrams considering phenolic and sterolic markers outlined by VIP selection
ED
method according to both geographical origin and cultivar-type. The upper diagram [A] represents
AC C
TABLE CAPTIONS
EP T
the output for phenolic compounds, while the lower one [B] is related to sterols.
Table 1. Discriminant phenolic and sterolic compounds identified by VIP (Variable Importance in Projection) selection method, following supervised OPLS-DA of different EVOO samples and using the ‘cultivar-type’ as class membership criterion. Compounds are provided together with VIP scores (measure of variables importance in the OPLS-DA model). Table 2. Discriminant phenolic and sterolic compounds identified by VIP (Variable Importance in Projection) selection method, following supervised OPLS-DA of EVOO samples from different
17
ACCEPTED MANUSCRIPT geographical origin. Compounds are provided together with VIP scores (measure of variables importance in the OPLS-DA model).
SUPPLEMENTARY MATERIAL
PT
Supplementary Table 1. Comprehensive dataset of identified phenolic and sterolic compounds,
RI
together with abundances, annotations (identification scores and raw formulas) and composite
SC
spectrum (masses and their abundances).
Supplementary Table 2. External validation of the OPLS-DA model for prediction of EVOO
NU
origin. Quality Controls QC oils from Tuscany and Sicily were included in the analysis and a misclassification table was then produced. Such table is presented together with the corresponding
MA
OPLS-DA score plot.
Supplementary Figure 1. Non-averaged unsupervised cluster analysis on the phenolic and sterolic
ED
profile of different EVOO samples representative of nine olive cultivars (i.e., Leccino, Picual,
EP T
Picholine marocaine, Frantoio, Sourani, Kalamon, Manzanilla de Sevilla, Arbequina and Koroneiki) (similarity: Euclidean; linkage rule: Ward). Compounds intensity was used to build up heat map, on
AC C
the basis of which the clusters were generated. Supplementary Figure 2. Validation of OPLS-DA discriminant model on phenolic and sterolic profile as a function of ‘cultivar-type’; Hotelling's T2 using 95% and 99% confidence limits is given in upper pane [A], whereas outcome of permutation test (N = 100) is given in the lower one [B]. Supplementary Figure 3. Non-averaged unsupervised cluster analysis on the phenolic and sterolic profile of different EVOO samples from Sicily, Puglia, Umbria, Liguria, Lombardy and Tuscany (similarity: Euclidean; linkage rule: Ward). Compounds intensity was used to build up heat map, on the basis of which the clusters were generated. 18
ACCEPTED MANUSCRIPT Supplementary Figure 4. Validation of OPLS-DA discriminant model on phenolic and sterolic profile as a function of geographical origin; Hotelling's T2 using 95% and 99% confidence limits is given in upper pane [A], whereas outcome of permutation test (N = 100) is given in the lower one
Class
Subclass
Marker
Flavonoids
Anthocyanins
Delphinidin 3-O-sambubioside
PT
[B].
SC
Petunidin 3-O-galactoside
RI
Peonidin 3-O-(6''-acetyl-glucoside)
Delphinidin 3-O-(6''-acetyl-glucoside)
NU
Delphinidin 3-O-(6''-acetyl-galactoside) Cyanidin 3-O-xyloside/arabinoside
MA
Pelargonidin 3-O-arabinoside Peonidin 3-O-(6''-acetyl-galactoside)
ED
Delphinidin 3-O-arabinoside/xyloside
EP T
Petunidin 3-O-glucoside
Dihydrochalcones
Phloridzin Phloretin
AC C
Flavanols
Flavones
(+)-Catechin 3-O-gallate (-)-Epicatechin 3-O-gallate Tetramethylscutellarein Chrysoeriol 7-O-glucoside Rhoifolin/Isorhoifolin
Flavonols
Quercetin 3-O-glucosyl-xyloside Quercetin 3-O-arabinoside/xyloside Isorhamnetin 3-O-rutinoside Kaempferol 3-O-(6''-acetyl-galactoside) 7-O-rhamnoside
VIP score 1.34 0.09 1.33 0.16 1.30 0.09 1.27 0.41 1.27 0.41 1.27 0.10 1.27 0.08 1.27 0.13 1.26 0.13 1.25 0.09 1.25 0.11 1.44 0.20 1.25 0.10 1.25 0.10 1.28 0.40 1.24 0.17 1.23 0.13 1.33 0.07 1.27 0.13 1.26 0.18 1.24 19
ACCEPTED MANUSCRIPT
Spinacetin 3-O-glucosyl-(1-6)-[apiosyl(1-2)]-glucoside Spinacetin 3-O-glucosyl-(1-6)-glucoside Phenolic acids
Hydroxybenzoics
Galloyl glucose
Hydroxycinnamics
3/4/5-Feruloylquinic acid
PT
Cinnamic acid
Stigmastanol ferulate
SC
Sinapic acid
RI
Avenanthramide 2f
1,2-Disinapoylgentiobiose
-
Isolariciresinol
MA
Lignans
NU
1,2'-Disinapoyl-2-feruloylgentiobiose
Pinoresinol
Matairesinol
ED
Secoisolariciresinol
Sterols
Phenolic terpenes
Rosmadial
Curcuminoids
Curcumin
Calysterols and cyclopropyl sidechain derivatives Cardanolides and derivatives
(24S)-isocalysterol
AC C
Other polyphenols
EP T
Anhydro-secoisolariciresinol
3beta-(6-deoxy-4-O-beta-D-glucopyranosyl-3-O-methylbeta-D-galactopyranosyloxy)-14,16beta-dihydroxy-5betacard-20(22)-enolide Adonitoxin Convallatoxin
Cholesterol and derivatives
3alpha,7alpha,12alpha-Trihydroxy-5beta-cholestanoylCoA Ecdysone 25-O-D-glucopyranoside Cholesterol-5-hydroperoxide
0.24 1.30 0.26 1.28 0.29 1.29 0.19 1.34 0.19 1.34 0.17 1.26 0.08 1.26 0.13 1.26 0.18 1.24 0.09 1.22 0.20 1.34 0.19 1.29 0.10 1.29 0.10 1.26 0.24 1.23 0.24 1.23 0.24 1.38 0.22 1.30 0.14 1.30 0.14 1.26 0.26 1.26 0.26 1.37 0.14 1.33 0.13 1.30 0.20 20
ACCEPTED MANUSCRIPT 24S,25-dihydroxycholesterol Ianostane 20-hydroxyecdysone 22-phosphate
RI
PT
3beta-hydroxy-4beta-methyl-5alpha-cholest-7-ene4alpha-carboxylic acid 4alpha-carboxy-4beta-methyl-5alpha-cholesta-8-en-3betaol 3beta-hydroxy-4alpha-methyl-5alpha-cholest-7-ene4beta-carboxylic acid Pinnasterol 7alpha-hydroxy-4-cholesten-3-one-d7
Nebrosteroid L
NU
Ergosterols and C24-methyl derivatives
SC
7-oxocholesterol (d7)
Pseudoprotodioscin
EP T
Furostanols and derivatives
ED
MA
campest-22E-en3beta,4beta,5alpha,6alpha,8beta,14alpha,15alpha,25R,26nonol campest-22E-en3beta,4beta,5alpha,6alpha,8beta,14alpha,15alpha,25,28nonol Conicasterol D
AC C
3-O-(Rhaa1-4(Rhaa1-2)Glcb)-26-O-(Glcb)-(25R)-furost-5en-3beta,22,26-triol Pseudoprotogracillin
Solanidines and alkaloid derivatives
Spirostanols and derivatives
26-O-[beta-D-glucopyranosyl]-25R-furostan3beta,22alpha,26-triol Solasodine Etioline Diospolysaponin A Trillin 3-O-(Galb)-(25R)-12-oxo-5alpha-spirostan-3beta-ol Alliogenin 6-O-(Glcb)-(25R)-5alpha-spirostan-3beta,6alpha-diol Karatavegenin C
1.30 0.20 1.25 0.51 1.24 0.14 1.24 0.31 1.24 0.31 1.24 0.31 1.23 0.17 1.22 0.79 1.22 0.79 1.26 0.18 1.25 0.26 1.25 0.26 1.24 0.31 1.36 0.15 1.36 0.14 1.30 0.23 1.29 0.21 1.23 0.35 1.23 0.35 1.29 0.44 1.29 0.12 1.27 0.15 1.25 0.29 1.24 0.16 1.23 21
ACCEPTED MANUSCRIPT
Protometagenin Steroid lactones
4beta-Hydroxywithanolide E
Steryl esters
15:0 Cholesteryl ester 17:1 Cholesteryl ester
PT
16:1 Campesteryl ester 14:1 Cholesteryl ester Ajugalactone
RI
Stigmasterols and C24-ethyl derivatives
SC
24Z-ethylidene-cholesta-5,7-dien-3beta-ol
Withanolide
1.30 0.14 1.22 0.18 1.22 0.09
AC C
EP T
ED
MA
Withanolides and derivatives
NU
Stigmastanyl glucoside
0.17 1.23 0.17 1.22 0.12 1.31 0.28 1.29 0.13 1.29 0.13 1.24 0.62 1.40 0.14
22
ACCEPTED MANUSCRIPT Subclass Anthocyanins
Marker Cyanidin 3-O-xylosylrutinoside Cyanidin 3-Osambubioside 5-Oglucoside Delphidin 3-Oarabinoside/xyloside Cyanidin 3-O-(6’’malonyl-glucoside) Peonidin 3-O-(6’’-acetylglucoside/galactoside) Phloretin 2’-O-xylosylglucoside Apigenin 7-O-glucoside Kaempferol 3-O-glucoside Kampferol 3-O-(2’’rhamnosyl-galactoside) 7O-rhamnoside Kaempferol 3-Orhamnosyl-rhamnosylglucoside Kaempferol 3-O-xylosylrutinoside Kaempferol 3-O-(2’’rhamnosyl-6’’acetylgalactoside) 7-Orhamnoside Quercetin 3-O-glucoside Patuletin 3-O-glucosyl-(16)-[apiosyl(1-2)]glucoside Spinacetin 3-O-glucosyl(1-6)-glucoside Quercetin 3-O-xylosylrutinoside Quercetin 3-O-galactoside 6’’-O-Acetylglycitin 3,5-Dihydroxybenzoic acids Protocatechuic acid p-Coumaroyl glucose p-Coumaric acid 4-Oglucoside 24-Methylcholesterol ferulate 24-Methyllathosterol ferulate 24-Methylenecholestanol ferulate p-Coumaroyl glycolic acid 3/4/5-Caffeoylquinic acid 3/4/5-Sinapoylquinic acid 1,2’-Disinapoyl-2feruloylgentiobiose
PT
Class Flavonoids
Dihydrochalcones
EP T
ED
MA
NU
SC
RI
Flavones Flavonols
Isoflavonoids Hydroxybenzoics
AC C
Phenolic acids
Hyddroxycinnamics
VIP score 1.36 0.38 1.35 0.11 1.33 0.13 1.31 0.31 1.28 0.08 1.35 0.13 1.33 0.17 1.35 0.21 1.35 0.25 1.34 0.24 1.34 0.24 1.32 0.15
1.32 0.17 1.35 0.19 1.34 0.16 1.30 0.14 1.27 0.16 1.43 0.15 1.31 0.38 1.31 0.38 1.40 0.32 1.39 0.32 1.36 0.19 1.36 0.19 1.36 0.19 1.36 0.29 1.32 0.17 1.28 0.20 1.28 0.12
23
ACCEPTED MANUSCRIPT
Ergosterols and C24-methyl derivatives
Furanostanols
Solanidines and alkaloid derivatives Spirostanols
PT
RI
1.31 0.13 1.34 0.14 1.28 0.13 1.33 0.19 1.29 0.18 1.31 0.50
Iokundjoside 3alpha,7alpha,12alphaTrihydroxy-5betacholestanoyl-CoA 3beta-hydroxy-4betamethyl-5alpha-cholest-7ene-4alpha-carboxylic acid 4alpha-carboxy-4betamethyl-5alpha-cholesta-8en-3beta-ol 3beta-hydroxy-4alphamethyl-5alpha-cholest-7ene-4beta-carboxylic acid Cholestanyl glucoside Trihydroxycoprostanoic acid Ganosporeric acid A Cimicifoetiside A Cycloartenol Campesteryl glucoside
1.41 0.27 1.31 0.48
Conicasterol D Mutasterol Castasterone 26-O-[beta-Dglucopyranosyl]-25Rfurostan-3beta,22alpha,26triol Pseudoprotogracillin Tomatine
1.29 0.37 1.29 0.18 1.28 0.13 1.29 0.21
Epimetagenin Solaspigenin Trillenogenin
1.35 0.22 1.35 0.22 1.31 0.20
NU
Cycloartanols and derivatives
1.36 0.17 1.36 0.17 1.36 0.17 1.34 0.17 1.31 0.12 1.28 0.13 1.29 0.27 1.31 0.12 1.28 0.25 1.34 0.17
Aragusteroketal
MA ED EP T
Stilbenes Sterols
Brassinolides and derivatives Bufanolides and derivatives C24-propyl sterols and derivatives Calysterols and cyclopropyl sidechain derivatives Cardanolides and derivatives Cholesterol and derivatives
AC C
Lignans
Homovanillic acid Dihydrocaffeic acid Syringaldehyde Rosmadial 3-Methoxyacetophenone Esculetin Estragole 4-Vinylguaiacol 4-Ethylphenol Anhydrosecoisolariciresinol Syringaresinol Piceatannol 3-O-glucoside 7-oxotyphasterol Scillaren A 24-Isopropenylcholesterol
SC
Other polyphenols
Hydroxyphenylacetics Hydroxyphenylpropanoics Hydroxybenzaldehydes Phenolic terpenes Hydroxybenzoketones Hydroxycoumarins Hydroxyphenylpropenes Alkylmehtoxyphenols Alkylphenols -
1.29 0.37
1.29 0.37 1.29 0.37 1.29 0.13 1.28 0.13 1.27 0.20 1.30 0.18 1.29 0.18 1.30 0.17
1.28 0.43 1.35 0.20
24
ACCEPTED MANUSCRIPT
Stigmasterols and C24-ethyl derivatives
18:0 Cholesteryl ester 16:0 Sitosteryl ester 15:0 Cholesteryl ester Cyasterone
1.31 0.08 1.30 0.26 1.31 0.08 1.31 0.23
(20S,24S)-24ethylthornasterol Minabeolide-7
1.30 0.24
1.34 0.26
1.28 0.22
SC
RI
Withanolides and derivatives
1.28 0.13 1.27 0.11
PT
Steroids and steroid derivatives Steryl esters
Hongguanggenin 3-O-(Galb)-(25R)-12-oxo5alpha-spirostan-3beta-ol Withaperuvin B
NU
Discrimination of extra-virgin-olive oils from different cultivars and geographical origins by untargeted
MA
metabolomics
ED
Silvia Ghisoni, Luigi Lucini, Federica Angilletta, Gabriele Rocchetti, Daniela Farinelli, Sergio Tombesi, Marco Trevisan
EP T
HIGHLIGHTS
Sterolic and phenolic profiles in EVOO accounted for more than 1000 compounds.
The chemical fingerprints of each EVOO cultivar analyzed showed differences.
Multivariate statistics discriminated EVOO according to geographical origin.
Flavonoids, hydroxycinnamics and cholesterol derivatives were the best markers.
AC C
25
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Silvia Ghisoni, Luigi Lucini, Federica Angilletta, Gabriele Rocchetti, Daniela Farinelli, Sergio Tombesi, Marco Trevisan PII: DOI: Reference:
S0963-9969(18)30999-2 https://doi.org/10.1016/j.foodres.2018.12.052 FRIN 8177
To appear in:
Food Research International
Received date: Revised date: Accepted date:
17 September 2018 4 December 2018 23 December 2018
Please cite this article as: Silvia Ghisoni, Luigi Lucini, Federica Angilletta, Gabriele Rocchetti, Daniela Farinelli, Sergio Tombesi, Marco Trevisan , Discrimination of extravirgin-olive oils from different cultivars and geographical origins by untargeted metabolomics. Frin (2018), https://doi.org/10.1016/j.foodres.2018.12.052
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Discrimination of extra-virgin-olive oils from different cultivars and geographical origins by untargeted metabolomics
Silvia Ghisonia, Luigi Lucinia* , Federica Angillettaa, Gabriele Rocchettia,b, Daniela Farinellic,
Department for Sustainable Food Process, Università Cattolica del Sacro Cuore, Via Emilia
RI
a
PT
Sergio Tombesid, Marco Trevisana
b
SC
Parmense 84, 29122 Piacenza, Italia
Department of Animal Science, Food and Nutrition, Università Cattolica del Sacro Cuore, Via
c
Department of Agricultural, Food and Environmental Sciences, Università degli Studi di Perugia,
MA
Via Borgo 20 giugno 74, 06154, Perugia, Italy d
NU
Emilia Parmense 84, 29122 Piacenza, Italia
Department for Sustainable Crop Production, Università Cattolica del Sacro Cuore, Via Emilia
AC C
EP T
ED
Parmense 84, 29122 Piacenza, Italia
*
Corresponding author: via Emilia Parmense 84, 29122 Piacenza, Italy. Telephone: +39 0523 599156. E-mail: [email protected] 1
ACCEPTED MANUSCRIPT Abstract In
this preliminary study,
quadrupole-time-of-flight
ultra-high-pressure liquid
chromatography (UHPLC)
mass spectrometry (QTOF) metabolomics followed
coupled
to
by multivariate
statistics was applied to discriminate nine extra-virgin olive oil (EVOO) cultivars according to their phenolic and sterolic fingerprints. The same approach was then used to discriminate EVOO samples
PT
from different geographical origins, namely six blends representative of the main growing regions
RI
in Italy (i.e., Sicily, Puglia, Umbria, Liguria, Lombardy and Tuscany).
SC
This approach allowed to putatively identify more than 1000 compounds, considering both polyphenols and sterols. The unsupervised hierarchical cluster analysis (HCA) and the orthogonal
NU
projections to latent structures discriminant analysis (OPLS-DA) discriminated EVOO samples according to both cultivar and geographical origin. In particular, flavonoids (i.e., anthocyanins and
MA
flavonols), hydroxycinnamic acids and cholesterol derivatives were found to be the most representative classes of compounds discriminating EVOO samples according to the two
ED
parameters (cultivar or geographical origin) selected. However, the following Venn analysis allowed to point out the discriminant markers being exclusive for cultivar or origin discrimination.
EP T
In this regard, only the 9.6% of phenolics and 13.6% of sterols were in common, thus indicating that several of these discriminant compounds were exclusive of a single condition. Indeed,
AC C
considering that most of the commercial EVOOs are blended, the contribution of both the cultivars used and the geographical origin must be taken into account.
Keywords: Extra-Virgin Olive Oil; Food metabolomics; Polyphenols; Sterols; Traceability.
2
ACCEPTED MANUSCRIPT 1. Introduction Extra-Virgin Olive Oil (EVOO) is a main component of the Mediterranean diet, displaying health-promoting effects thanks to both its unsaturated fatty acids profile and the presence of bioactive compounds such as polyphenols and tocopherols (Blasi et al., 2018; Suarez, Caimani, del Bas, & Arola, 2017). EVOO is produced in all the countries in the Mediterranean basin, with Spain,
PT
Italy, Tunisia and Greece being the most important producers worldwide (Kalogiouri, Aalizadeh, &
RI
Thomaidis, 2018).
SC
The quality and organoleptic properties of EVOOs depend on many factors such as cultivar, geographical origin, climatic conditions, agronomic techniques and processing technologies, with
NU
consequences also in the final price of the product in the market (Cubero-Leon, Penalver, & Maquet, 2014). Therefore, the protection of olive oils according to their origin, such as protected
MA
designation of origin (PDO), protected geographical indication (PGI) and traditional specialties guaranteed (TSG), has become an essential information in the label, thus entailing the adoption of
ED
methods for authentication (Merchak, et al., 2017).
EVOO is considered a common target for economically motivated adulteration, because of its
EP T
high demand and profitability potential (Esteki et al., 2018). As recently reviewed, chromatographic methods coupled to chemometrics, including untargeted metabolite profiling have been applied for
AC C
authentication purposes in foods, including olive oils (Esteki et al., 2018). Efforts have been done in the area of PDOs and PGIs to avoid fraudulent labelling, but a definitive and robust solution to this problem is still not present. Both PDO and PGI designate products that are linked to the area of which they bear the name and include niche products from small-scale processes in limited areas where growth conditions confer a particular quality to the product (Cubero-Leon et al., 2014). In the last years, several research attempts tried to use the chemical fingerprint of EVOO samples for authentication purposes. For example, according to Aparicio, Morales, Aparicio-Ruiz, Tena and Garcìa-Gonzàlez (2013), sterols represent a very useful tool for the authentication of olive oils. These authors used sterols as index of fraudulent oil mixtures, underlying the importance of 3
ACCEPTED MANUSCRIPT chromatographic techniques to quantify the free and esterified forms in short time. Different studies have been carried out also to authenticate olive oils from Italy (Alonso-Salces et al., 2010), Greece (Longobardi et al., 2012; Alonso-Salces et al., 2010) and Spain (Becerra-Herrera et al., 2018). Furthermore, Ben Mohamed and co-authors (2018) have recently used both sterol and phenolic fingerprints to discriminate Tunisian and Italian EVOOs, outlining the potential of untargeted
PT
metabolomics for authenticity evaluations. Indeed, metabolomics, defined as “the unbiased, global
RI
screening approach to classify samples based on metabolite patterns or fingerprints that change in response to disease, environmental or genetic perturbations with the ultimate goal to identify
SC
discriminating metabolites” (Sales, et al., 2017) might represents a powerful approach to ensure the
NU
authenticity of olive oils in relation with the geographical origin (Esteki et al., 2018) or sensorial properties (Sales, Portolés, Johnsen, Danielsen & Beltran, 2019).
MA
However, to the best of our knowledge, the application of untargeted metabolomics for authenticity markers discovering in different EVOO cultivars from different origins is still scarce.
ED
Therefore, the aim of this work was to apply a metabolomic approach followed by multivariate statistics to identify sterol and phenolic biomarkers for authenticity purposes. Most of the
EP T
commercial oils are sold as blends, where a contribution of both the cultivars used and the geographical origin is expected. On this basis, the possibility to discriminate different EVOOs as a
AC C
function of the cultivar and according to the actual geographical origin has been investigated.
2. Materials and methods 2.1. Samples
Fifteen EVOO samples were included in this study. Six of them are blends representative of the main growing regions in Italy (i.e., Sicily, Puglia, Umbria, Liguria, Lombardy, Tuscany) were kindly provided by COOP Italia SpA (Casalecchio di Reno, Bologna, Italy) a commercial supplier who owns the supply chain and could ensure geographical origins. The oil from each origin was
4
ACCEPTED MANUSCRIPT provided in triplicate; two additional samples (from Sicily and Tuscany, respectively) have been provided to be used as quality control for validation purposes. The remaining nine samples Mediterranean
basin
(i.e.,
represented
‘Leccino’,
some widely diffused
‘Picual’,
‘Picholine marocaine’,
olive cultivars of the ‘Frantoio’,
‘Sourani’,
‘Kalamon’, ‘Manzanilla de Sevilla’, ‘Arbequina’ and ‘Koroneiki’), and were obtained by olives
PT
harvested in the germplasm repository of the University of Perugia (Italy). EVOO was obtained
RI
from 2 kg of fruit samples harvested on November 15th , 2016 from three trees per each cultivar. Oil
SC
was extracted within 24 hours by a custom-built plant, allowing to process small amount of fruits. Oil samples were immediately placed in cold storage (10 ± 2 °C) until further analyses. Three
NU
replicate samples were obtained per accession.
MA
2.2. Untargeted profiling of phenolic and sterol compounds
Each oil sample was analysed in duplicate, with the exception of quality controls that were
ED
analysed in triplicate. For the extraction step, the method reported by Ben Mohamed and co-authors (2018) was used. Briefly, three individual sample replicates from each sample (3 g each) were
EP T
extracted in 3 mL of 80% methanol solution (v/v) (LCMS grade, VWR, Milan, Italy) using a vortex mixer. The suspension was kept at room temperature for 30 minutes and then centrifuged at 6,000 x
analysis.
AC C
g for 10 min at 4 °C. The extracts were collected in 2 mL amber glass vials and stored at 4 °C until
Phenolic compounds and
sterols were then screened
through ultra-high-pressure liquid
chromatography (UHPLC) coupled to an electrospray quadrupole-time-of-flight hybrid mass spectrometer (UHPLC-ESI/QTOF-MS), as previously described (Rocchetti et al., 2017). With this purpose, a 1290 series liquid chromatograph and a G6550 iFunnel mass spectrometer detector equipped with a JetStream dual electrospray (all from Agilent Technologies, Santa Clara, CA, USA), were used. Chromatographic separation was achieved on an Agilent Zorbax Extend-C18 column (75 × 2.1 mm i.d., 1.8 μm) in reverse phase mode (gradient elution from 5% to 90% LC-MS 5
ACCEPTED MANUSCRIPT grade methanol). The mass spectrometer data acquisition was MS-only mode in positive polarity (range 100–1000 m/z) and injection volume was 3 μL. Source conditions were as follows: nitrogen was used as both sheath gas (10 L min-1 and 350 °C) and drying gas (8 L min-1 and 330 °C); nebulizer pressure was 60 psi, nozzle voltage was 300 V, and capillary voltage was 3.5 kV. Raw data were processed using Profinder B.07 (from Agilent Technologies), for mass and
PT
retention time alignment and for the following annotations based on the ‘find-by-formula’ algorithm
RI
(Pretali, Bernardo, Butterfield, Trevisan & Lucini, 2016). According this approach, the entire
SC
isotopic pattern (monoisotopic accurate mass, isotope spacing, and isotope ratio - mass accuracy < 5 ppm) was considered for compounds identification by linking the isotopic profile to a custom
NU
database obtained by combining Phenol-Explorer 3.6 (http://phenol-explorer.eu/) and LIPID MAPS
2.3. Multivariate statistical analysis
MA
(Lipidomics Gateway, http://www.lipidmaps.org/).
ED
Metabolomics data on phenolic and sterol profiling were interpreted in Agilent Mass Profiler Professional B.12.06 as previously reported (Lucini et al., 2016). Compounds were filtered by
EP T
abundance (only those compounds with an area > 10000 counts were considered), normalized at the 75th percentile and baselined to their median in the dataset. Post-acquisition processing included
AC C
also filtering by frequency: only those compounds identified within 100% of replications in at least one treatment were retained. Statistics were finally carried out on this latter dataset; unsupervised hierarchical cluster analysis (HCA) was carried out setting the similarity measure as ‘Euclidean’ and ‘Wards’ as the linkage rule. Thereafter, the raw dataset was exported into SIMCA 13 (Umetrics, Malmo, Sweden), Pareto scaled, and elaborated for orthogonal partial least squares discriminant analysis (OPLS-DA) supervised modelling. Herein, the variation between the groups was separated into predictive and orthogonal (i.e., ascribable to technical and biological variation) components. The presence of outliers was investigated according to Hotelling’s T2 using 95% and 99% confidence limits for 6
ACCEPTED MANUSCRIPT suspect and strong outliers respectively. Cross-validation of the OPLS-DA model was carried out using CV-ANOVA (p < 0.01) and permutation testing carried out to exclude overfitting (N=300), after inspecting model parameters (goodness-of-fit R2 Y and
goodness-of-prediction Q 2 Y).
Regarding Q 2 Y prediction ability, a value > 0.5 was adopted as a threshold to identify acceptable models, according to software recommendation and as set out in the literature (Rombouts et al.,
PT
2017). Finally, the variables importance in projection (VIP analysis) selection was then used to
RI
evaluate the importance of metabolites and to select those having the highest discrimination
SC
potential (VIP score > 1.2).
Finally, to discern among the markers being specifically responsible for origin discrimination
NU
from those related to cultivars, a Venn analysis has been carried out using an online tool (Venny 2.1
MA
- http://bioinfogp.cnb.csic.es/tools/venny/index.html).
3. Results and discussion
ED
3.1. Discrimination of different EVOO cultivars according to phenolic and sterols profiles
EP T
In the first part of this work, the untargeted profiling of EVOO samples was carried out in order to shed light onto the differences in secondary metabolites as a sole function of the cultivar, since
conditions.
AC C
the olive trees were grown in the same field and were therefore affected by the same pedo-climatic
Overall, the metabolomics-based approach allowed us to putatively annotate 1354 compounds (444 polyphenols and 910 sterols; supplementary material). The metabolomic profiling allowed to discriminate the nine different EVOO cultivars according to their sterols and phenolic fingerprints. In particular, both unsupervised and supervised multivariate statistics (i.e., HCA and OPLS-DA) let us to discriminate the different cultivars analyzed. In this regard, the HCA provided good clusters, considering that biological replicates belonging to the same variety grouped together and no mixed clusters could be observed (supplementary material). In more detail, two main clusters could be identified, suggesting a very similar metabolites profile between ‘Manzanilla’ and ‘Kalamon’ 7
ACCEPTED MANUSCRIPT cultivars when comparing to the others. These results were confirmed through the supervised OPLS discriminant analysis, using the cultivar as class membership criterion. As shown in the OPLS score plot (Figure 1), a clear discrimination of the cultivars was obtained, thus suggesting a strong relationship with the secondary metabolites’ profile. Looking in detail the score plot outlined by OPLS-DA, three main groups were obtained: ‘Manzanilla’ and ‘Kalamon’ cultivars clustered
PT
together, while a second major cluster comprising ‘Frantoio’, ‘Arbequina’, ‘Picual’ and ‘Koroneiki’
RI
oils was noticed. However, ‘Leccino’, ‘Sourani’ and ‘Picholine marocaine’ showed a more distinct metabolites profile when compared to the others. The model cross-validation parameters were
SC
found to be very robust, being the goodness-of-fit (R2 Y) and the goodness-of-prediction (Q2 Y) of
NU
0.97 and 0.91, respectively. Besides, no outliers could be observed by Hotelling’s T2, whereas both CV-ANOVA cross-validation (p < 0.01 for regression) and permutation test (N = 100) for
MA
overfitting were more than adequate (supplementary material). The most representative compounds for discrimination purposes were selected by using the VIP
ED
variable selection method, following OPLS-DA supervised modeling, and then grouped into chemical classes and subclasses (Table 1). The most represented class of polyphenols was
EP T
flavonoids (28 compounds), with anthocyanins such as delphinidin 3-O-sambubioside, peonidin 3O-(6''-acetyl-glucoside) and petunidin 3-O-galactoside, possessing a VIP score > 1.3. When
AC C
considering the other flavonoid-subclasses namely dihydrochalcones, flavonols and flavones, phloretin, quercetin 3-O-glucosyl-xyloside and tetramethylscutellarein possessed the highest VIP score (1.44, 1.33 and 1.28, respectively). Interestingly, among the phenolic acids outlined by VIP analysis, hydroxycinnamic acids were more discriminating than hydroxybenzoic acids, with the isomeric forms of feruloylquinic acid possessing the highest VIP score (1.35). The most discriminating sterols outlined by VIP were exported and classified in subclasses, by exploiting the information provided by the LIPID MAPS database (Lipidomics Gateway, http://www.lipidmaps.org/)
(Table
1).
Several sterol-derivatives characterized
the differences
among the cultivars considered, with an abundance of cholesterols, cardanolides, furostanols, 8
ACCEPTED MANUSCRIPT spirostanols, stigmasterols and other compounds. The highest discrimination potential was found for a stigmasterol (ajugalactone; VIP score: 1.40) followed by a cholesterol derivative (3α,7α,12αTrihydroxy-5β-cholestanoyl-CoA; VIP score: 1.37) and two furostanols (i.e., isomeric forms of pseudoprotodioscin; VIP scores: 1.36). The cultivars analyzed in this work are generally suitable for the production of table olive,
PT
further than for olive oil extraction. Interestingly, ‘Manzanilla’ and ‘Kalamon’, two of the most
RI
important cultivars for oil production, clustered together into the OPLS score plot, whereas
SC
‘Frantoio’, ‘Arbequina’ and ‘Koroneiki’, i.e., those cultivars characterized by a smaller fruit size, where found to be very close into the OPLS score plot according to their sterols and phenolic
NU
fingerprints. A recent work used QTOF-based metabolomics followed by multivariate statistics to discriminate EVOO cultivars from Greece (Kalogiouri et al., 2018). Notably, despite these authors
MA
used a wider database for annotations (i.e., the food database FooDB), phenolic compounds and flavonoids in particular were selected as the compounds displaying the highest discrimination
ED
potential. Conversely, Piccinonna et al. (2016) reported oleic acid and fatty chains as discriminants of Apulian olive oils. However, the metabolomic approach they applied was based on NMR
EP T
spectroscopy, and likely had a compounds coverage complementary to our QTOF screening.
AC C
3.2. UHPLC/QTOF-MS discrimination of EVOOs from different geographical origin The previous approach was used also to discriminate EVOO samples from different regions of Italy. The hierarchical cluster analysis (supplementary material) showed a clear separation between each oil sample, with all the biological replicates clustering together. In particular, it is possible to notice that “Regionale del Garda” was clearly separated from all the other samples, thus suggesting a different secondary metabolites profile. These findings are quite interesting considering that the main cultivars used for producing this EVOO are ‘Casaliva’, ‘Leccino’ and ‘Frantoio’ and that ‘Casaliva’ is considered as a possible synonym of ‘Taggiasca’ and ‘Frantoio’ (Muzzalupo, Vendramin, & Chiappetta, 2014). However, considering that ‘Taggiasca’ is the main cultivar 9
ACCEPTED MANUSCRIPT characterizing ‘Regionale ligure” EVOO, it was surprising to observe a so clear segregation of these two samples. However, it must be considered that ‘Regionale ligure” is produced from olives grown on the seaside, whereas “Regionale del Garda” is produced in the Northern area among those considered, hundred kilometers far from the coast, in a very particular climatic niche. Given the HCA outcomes from unsupervised analysis, the following OPLS-DA supervised
PT
modeling was carried out to identify markers for geographical origin. The score scatter plot (Figure
RI
2) confirmed the results obtained from the HCA, thus proving that the metabolomic profiling proposed was able to discriminate EVOO samples according to the geographical origin. In this
SC
regard, the model parameters were found to be excellent, being R2 Y = 0.97 and Q 2 Y = 0.93. No
NU
outlier samples could be observed by Hotelling’s T2, whereas both CV-ANOVA cross-validations (p < 0.01 for regression) and test for overfitting (permutation test, N = 100) were adequate
MA
(supplementary material). In order to perform an external validation of the chemometric approach applied, two quality controls of known origin (namely Tuscany and Sicily) were also analyzed
ED
using the same method, and their metabolomic profiles included into the OPLS-DA model. The resulting misclassification table (a data elaboration displaying class prediction accuracy), allowed
EP T
including both quality controls within the correct class, with a Fisher’s P of 3.6 E-26 (supplementary material). Therefore, this further investigation confirmed the prediction ability of
AC C
the multivariate approach we developed. The variable selection methods applied to OPLS-DA model (VIP analysis) allowed identifying the most discriminant phenolic and sterols among the compounds annotated. These compounds are reported in Table 2. Overall, 47 phenolic compounds possessed a VIP score > 1.2. In particular, flavonoids were found to be the most common class (21 compounds), with anthocyanins (7 compounds) and flavonols (10 compounds) being the most frequent compounds. Notably, glycosidic forms of kaempferol and cyanidin possessed very represented among these two latter phenolic subclasses. Phenolic acids were the second most frequent class (17 compounds), with a clear abundance of hydroxycinnamic acids, such as isomeric forms of caffeoylquinic and 10
ACCEPTED MANUSCRIPT sinapoylquinic acids. In addition, other phenolics such as lignans and stilbenes were reported. However, the compounds possessing the highest VIP score were found to be 6’’-O-Acetylglycitin (belonging to isoflavonoids) and p-Coumaroyl glucose (hydroxycinnamic acid) with VIP scores of 1.43 and 1.40, respectively. Moving to the sterols identified by VIP analysis, cholesterol (5 compounds), spirostanols (4 compounds), ergosterol (4 compounds), steryl esters (3 compounds)
PT
together with stigmasterol, furostanol and cycloartanol (2 compounds) derivatives were the most
RI
represented compounds. In this case, iokundjoside (i.e., cardanolides) and solaspigenin (i.e.,
SC
spirostanols) were characterized by the highest VIP scores (i.e., 1.41 and 1.35, respectively). It is possible that sterols profile in EVOO is affected also by the oil microbiota, typically present in solid
NU
particles and micro drops of vegetation water, where yeasts are predominant (Ciafardini & Zullo, 2018).
MA
The importance of a metabolomics-based approach to discriminate and authenticate EVOOs from different geographical origin was recently reported. For example, Gil-Solsona and co-authors
ED
(2016) used an UHPLC-QTOF MS approach to identify possible biomarkers able to discriminate different EVOO samples, while Ben Mohamed et al. (2018) using the same analytical conditions
EP T
were able to discriminate Tunisian and Italian EVOOs according to their sterols and phenolic fingerprints. The need to authenticate and differentiate EVOOs olive oils from different origins is
AC C
an essential procedure for both the industry and the final consumer. In our work, a metabolomic profiling approach has been proposed, focusing on two classes of secondary metabolites that are affected by the environment and are widely represented in EVOO. A previous work successfully using metabolomics to authenticate geographical origin of EVOOs from Spain (Gil-Solsona et al., 2016), selected twelve markers, seven of which could be identified; all these markers were lipids, even though no sterols were selected. These differences can be likely ascribed to the differences in acquisition and annotation between the present and the previous work. Conversely, Peršurić, Saftić, Mašek & Kraljević Pavelić (2018), used a combination of MALDI/TOF, GC/MS and NIRS to
11
ACCEPTED MANUSCRIPT authenticate geographical origin of Croatian EVOOs. According to the approach proposed, triacylglycerols were reported to be the most discriminating compounds. Therefore, several markers have been proposed for geographical origin of EVOOs, each of them having its peculiarities. Considering that the very most of commercial EVOOs are blend of different cultivars, it is important to discern whether the proposed markers are specifically related to the
PT
origin rather than being the result of the cultivar(s) considered. With this purpose, we carried out a
RI
Venn analysis to point out to which extent the markers for cultivars discrimination are shared with
SC
those for origin discrimination (Figure 3). As it can be noted, 9.6% of phenolics and 13.6% of sterols were in common, thus indicating that the very most of discriminant compounds are exclusive
NU
of a single condition. Likely, this is the result of using OPLS-DA supervised modeling approaches, where the predictive covariance related to a given class membership (either cultivar or origin) is
MA
separated from orthogonal non-predictive covariance.
Despite the novelty and encouraging findings of the approach proposed, based on untargeted
ED
phenolic and sterolic profiling associated to supervised multivariate OPLS-DA, further research is still needed. In fact, it is known that several factors might affect secondary metabolite profile in
EP T
olive fruit and consequently also in oil; therefore, broader investigations based on a higher number
AC C
of samples and geographical origins might be advisable.
4. Conclusions
Adulteration and misleading label information are a serious problem in the food sector; with this regard, particular attention is being paid to olive oils, thus demanding robust and effective techniques for authentication purposes. In this work, an untargeted metabolomics-based approach was used to identify phenolic and sterol compounds able to discriminate nine samples of EVOOs from different cultivars (grown in the same conditions) and blend samples of EVOOs having different geographical origin. The results showed that the profile of these secondary metabolites can be a powerful tool for discrimination purposes, with regards to cultivars and geographical origin, 12
ACCEPTED MANUSCRIPT confirming that UHPLC-MS metabolomic profiling represents an effective approach in food traceability. In particular, although a clear effect of the cultivar can be identified (thus indicating a strong genetic basis), the interaction with the environment results in a distinctive profile in phenolics and sterols. Holistic approaches such as untargeted profiling, associated to supervised multivariate analyses, allowed identifying the markers specifically related to geographical origin.
PT
Notwithstanding the preliminary nature of our research, the two classes of secondary metabolites
RI
considered in this work are entitled to represent a complementary and promising solution for plant
SC
foods traceability, thus deserving further investigations.
From a practical point of view, it is advisable that the markers proposed will be included in a
NU
validated targeted quantification method, e.g. based on less sophisticated and cheaper mass spectrometers such as triple quadrupole MS/MS. This would result in offering a simple, robust and
MA
cost-effective way to improve authentication of EVOOs on routine basis. Furthermore, given the promising outcomes of metabolomics, research aimed to investigate whether metabolomics could
ED
help to identify fraudulent activities (e.g., mixing lower quality oils in EVOOs) is worthwhile in the
EP T
next future.
AC C
Funding
This research did not receive any specific grant from funding agencies in the public, commercial and not-for-profit sectors.
Acknowledgements SG and GR were recipient of a Ph.D. fellowship from the Doctoral School AgriSystem, Piacenza, Italy. The authors thank the “Romeo ed Enrica Invernizzi” foundation for kindly supporting the metabolomic platform.
13
ACCEPTED MANUSCRIPT
References Alonso-Salces, R. M., Moreno-Rojas, J. M., Holland, M. V., Reniero, F., Guillou, C., & Héberger, K. (2010). Virgin olive oil authentication by multivariate analyses of 1 H NMR fingerprints and
PT
13 C and 2 H data. Journal of Agricultural and Food Chemistry, 58(9), 5586-5596.
RI
Aparicio, R., Morales, M., Aparicio-Ruiz, R., Tena, N., & Garcìa-Gonzàlez, D. (2013). Authenticity of olive oil: Mapping and comparing official methods and promising alternative. Food Research
SC
International, (54), 2025-2038.
NU
Becerra-Herrera, M., Vélez-Martín, A., Ramos-Merchante, A., Richter, P., Beltrán, R., & Sayago, A. (2018). Characterization and evaluation of phenolic profiles and color as potential
ED
origin. Food Chemistry, 241, 328-337.
MA
discriminating features among Spanish extra virgin olive oils with protected designation of
Ben Mohamed, M., Rocchetti, G., Montesano, D., Ben Ali, S., Guasmi, F., Grati-Kamoun, N., &
EP T
Lucini, L. (2018). Discrimination of Tunisian and Italian extra-virgin olive oils according to their phenolic and sterolic fingerprints. Food Research International, 106, 920-927.
AC C
Blasi, F., Rocchetti, G., Montesano, D., Lucini, L., Chiodelli, G., Ghisoni, S., Baccolo, G., Simonetti, M. S., & Cossignani, L. (2018). Changes in extra-virgin olive oil added with Lycium barbarum L. carotenoids during frying: Chemical analyses and metabolomic approach. Food Research International, 105, 507-516. Ciafardini, G., & Zullo, B. A. (2018). Virgin olive oil yeasts: A review. Food Microbiology, 70, 245-253. Cubero-Leon, E., Penalver, R., & Maquet, A. (2014). Review on metabolomics for food authentication. Food Research International, 60, 95-107.
14
ACCEPTED MANUSCRIPT Esteki, M., Simal-Gandara, J., Shahsavari, Z., Zandbaaf, S., Dashtaki, E., & Vander Heyden, Y. (2018). A review on the application of chromatographic methods, coupled to chemometrics, for food authentication. Food Control, 93, 165-182. Gil-Solsona, R., Raro, M., Sales, C., Lacalle, L., Díaz, R., Ibánez, M., Beltran, J., Sancho, J. V., & Hernández, F. J. (2016). Metabolomic approach for Extra virgin olive oil origin discrimination
PT
making use of ultra-high performance liquid chromatography - Quadrupole time-of-flight mass
RI
spectrometry. Food Control, (70), 350-359.
SC
Kalogiouri, N. O., Aalizadeh, R., & Thomaidis, N. S. (2018). Application of an advanced and wide scope non-target screening workflow with LC-ESI-QTOF-MS and chemometrics for the
NU
classification of the Greek olive oil varieties. Food Chemistry, 256, 53-61.
MA
Longobardi, F., Ventrella, A., Casiello, G., Sacco, D., Tasioula-Margari, M., Kiritsakis, A., & Kontominas, M. (2012). Characterisation of the geographicasl origin of Western Greek virgin
ED
olive oils based on instrumental and multivariate statistical analysis. Food Chemistry, 133, 169-
EP T
175.
Lucini, L., Borgognone, D., Rouphael, Y., Cardarelli, M., Bernardi, J. & Colla, G. (2016). Mild Potassium Chloride Stress Alters the Mineral Composition, Hormone Network, and Phenolic
AC C
Profile in Artichoke Leaves. Frontiers in Plant Scence, 7, 948. Merchak, N., El Bacha, E., Khouzam, R. B., Rizk, T., Akoka, S., & Bejjan, J. (2017). Geoclimatic, morphological, and temporal effects on Lebanese olive oils composition and classification: A 1 H NMR metabolomic study. Food Chemistry, 217, 379-388. Muzzalupo, I., Vendramin, G. G., & Chiappetta, A. (2014). Gemetic biodiversity of Italian olives (Olea europaea) germplasm analyzed by SSR markers. The Scientific World Journal, vol. 2014, Article ID 296590, 12 pages. https://doi.org/10.1155/2014/296590.
15
ACCEPTED MANUSCRIPT Peršueić, Z., Saftić, L., Mašek, T., & Pavelić, S. K. (2018). Comparison of triacylglycerol analysis by MALDI-TOF/MS, fatty acid analysis by GC-MS and non-selective analyis by NIRS in combination with chemometrics for determination of extra virgin olive oil geographical origin. A case study. LWT, 95, 326-332. Piccinonna, S., Ragone, R., Stocchero, M., Del Coco, L., De Pascali, S. A., Schena, F. P., &
PT
Fanizzi, F. P. (2016). Robustness of NMR-based metabolomics to generate comparable data sets
RI
for olive oil cultivar classification. An inter-laboratory study on Apulian olive oils. Food
SC
Chemistry, 199, 675-683.
Pretali, L., Bernardo, L., Butterfield, T. S., Trevisan, M., & Lucini, L. (2016). Botanical and elicit
a
similar
Induced
Systemic Response in tomato
NU
biological pesticides
(Solanum
MA
lycopersicum) secondary metabolism. Phytochemistry, 130, 56-63. Rocchetti, G., Chiodelli, G., Giuberti, G., Masoero, F., Trevisan, M., & Lucini, L. (2017).
ED
Evaluation of phenolic profile and antioxidant capacity in gluten-free flours. Food Chemistry, 228, 367-373.
EP T
Rombouts, C., Hemeryck, L. Y., Van Hecke, T., De Smet, S., De Vos, W. H., & Vanhaecke, L. (2017). Untargeted metabolomics of colonic digests reveals kynurenine pathway metabolites,
AC C
dityrosine and 3-dehydroxycarnitine as red versus white meat discriminating metabolites. Scientific Reports, 7, 42514. Sales, C., Cervera, M., Gil, R., Portolés, T., Pitarch, E., & Beltran, J. (2017). Quality classification of Spanish olive oils by untargeted gas chromatography coupled to hybrid quadrupole-time of flight mass spectrometry with atmospheric pressure chemical ionization and metabolomics-based statistical approach. Food Chemistry, (216), 365-373. Sales, C., Portolés, T., Johnsen, L. G., Danielsen, M., & Beltran, J. (2019). Olive oil quality classification and measurement of its organoleptic attributes by untargeted GC-MS and multivariate statistical-based approach. Food Chemistry, 271, 488-496. 16
ACCEPTED MANUSCRIPT Suarez, M., Caimani, A., del Bas, J., & Arola, L. (2017). Metabolomics: An emerging tool to evaluate the impact of nutritional and physiological challenges. Tends in Analiytical Chemistry, (96), 79-88.
PT
FIGURE CAPTIONS Figure 1. Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) on
RI
EVOO samples phenolic and sterolic profile, using the ‘cultivar-type’ as class membership
SC
criterion. Individual replications are given in the class prediction model score plot.
NU
Figure 2. Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) on EVOO samples phenolic and sterolic profile according to their geographical origin. Individual
MA
replications are given in the class prediction model score plot. Figure 3. Venn diagrams considering phenolic and sterolic markers outlined by VIP selection
ED
method according to both geographical origin and cultivar-type. The upper diagram [A] represents
AC C
TABLE CAPTIONS
EP T
the output for phenolic compounds, while the lower one [B] is related to sterols.
Table 1. Discriminant phenolic and sterolic compounds identified by VIP (Variable Importance in Projection) selection method, following supervised OPLS-DA of different EVOO samples and using the ‘cultivar-type’ as class membership criterion. Compounds are provided together with VIP scores (measure of variables importance in the OPLS-DA model). Table 2. Discriminant phenolic and sterolic compounds identified by VIP (Variable Importance in Projection) selection method, following supervised OPLS-DA of EVOO samples from different
17
ACCEPTED MANUSCRIPT geographical origin. Compounds are provided together with VIP scores (measure of variables importance in the OPLS-DA model).
SUPPLEMENTARY MATERIAL
PT
Supplementary Table 1. Comprehensive dataset of identified phenolic and sterolic compounds,
RI
together with abundances, annotations (identification scores and raw formulas) and composite
SC
spectrum (masses and their abundances).
Supplementary Table 2. External validation of the OPLS-DA model for prediction of EVOO
NU
origin. Quality Controls QC oils from Tuscany and Sicily were included in the analysis and a misclassification table was then produced. Such table is presented together with the corresponding
MA
OPLS-DA score plot.
Supplementary Figure 1. Non-averaged unsupervised cluster analysis on the phenolic and sterolic
ED
profile of different EVOO samples representative of nine olive cultivars (i.e., Leccino, Picual,
EP T
Picholine marocaine, Frantoio, Sourani, Kalamon, Manzanilla de Sevilla, Arbequina and Koroneiki) (similarity: Euclidean; linkage rule: Ward). Compounds intensity was used to build up heat map, on
AC C
the basis of which the clusters were generated. Supplementary Figure 2. Validation of OPLS-DA discriminant model on phenolic and sterolic profile as a function of ‘cultivar-type’; Hotelling's T2 using 95% and 99% confidence limits is given in upper pane [A], whereas outcome of permutation test (N = 100) is given in the lower one [B]. Supplementary Figure 3. Non-averaged unsupervised cluster analysis on the phenolic and sterolic profile of different EVOO samples from Sicily, Puglia, Umbria, Liguria, Lombardy and Tuscany (similarity: Euclidean; linkage rule: Ward). Compounds intensity was used to build up heat map, on the basis of which the clusters were generated. 18
ACCEPTED MANUSCRIPT Supplementary Figure 4. Validation of OPLS-DA discriminant model on phenolic and sterolic profile as a function of geographical origin; Hotelling's T2 using 95% and 99% confidence limits is given in upper pane [A], whereas outcome of permutation test (N = 100) is given in the lower one
Class
Subclass
Marker
Flavonoids
Anthocyanins
Delphinidin 3-O-sambubioside
PT
[B].
SC
Petunidin 3-O-galactoside
RI
Peonidin 3-O-(6''-acetyl-glucoside)
Delphinidin 3-O-(6''-acetyl-glucoside)
NU
Delphinidin 3-O-(6''-acetyl-galactoside) Cyanidin 3-O-xyloside/arabinoside
MA
Pelargonidin 3-O-arabinoside Peonidin 3-O-(6''-acetyl-galactoside)
ED
Delphinidin 3-O-arabinoside/xyloside
EP T
Petunidin 3-O-glucoside
Dihydrochalcones
Phloridzin Phloretin
AC C
Flavanols
Flavones
(+)-Catechin 3-O-gallate (-)-Epicatechin 3-O-gallate Tetramethylscutellarein Chrysoeriol 7-O-glucoside Rhoifolin/Isorhoifolin
Flavonols
Quercetin 3-O-glucosyl-xyloside Quercetin 3-O-arabinoside/xyloside Isorhamnetin 3-O-rutinoside Kaempferol 3-O-(6''-acetyl-galactoside) 7-O-rhamnoside
VIP score 1.34 0.09 1.33 0.16 1.30 0.09 1.27 0.41 1.27 0.41 1.27 0.10 1.27 0.08 1.27 0.13 1.26 0.13 1.25 0.09 1.25 0.11 1.44 0.20 1.25 0.10 1.25 0.10 1.28 0.40 1.24 0.17 1.23 0.13 1.33 0.07 1.27 0.13 1.26 0.18 1.24 19
ACCEPTED MANUSCRIPT
Spinacetin 3-O-glucosyl-(1-6)-[apiosyl(1-2)]-glucoside Spinacetin 3-O-glucosyl-(1-6)-glucoside Phenolic acids
Hydroxybenzoics
Galloyl glucose
Hydroxycinnamics
3/4/5-Feruloylquinic acid
PT
Cinnamic acid
Stigmastanol ferulate
SC
Sinapic acid
RI
Avenanthramide 2f
1,2-Disinapoylgentiobiose
-
Isolariciresinol
MA
Lignans
NU
1,2'-Disinapoyl-2-feruloylgentiobiose
Pinoresinol
Matairesinol
ED
Secoisolariciresinol
Sterols
Phenolic terpenes
Rosmadial
Curcuminoids
Curcumin
Calysterols and cyclopropyl sidechain derivatives Cardanolides and derivatives
(24S)-isocalysterol
AC C
Other polyphenols
EP T
Anhydro-secoisolariciresinol
3beta-(6-deoxy-4-O-beta-D-glucopyranosyl-3-O-methylbeta-D-galactopyranosyloxy)-14,16beta-dihydroxy-5betacard-20(22)-enolide Adonitoxin Convallatoxin
Cholesterol and derivatives
3alpha,7alpha,12alpha-Trihydroxy-5beta-cholestanoylCoA Ecdysone 25-O-D-glucopyranoside Cholesterol-5-hydroperoxide
0.24 1.30 0.26 1.28 0.29 1.29 0.19 1.34 0.19 1.34 0.17 1.26 0.08 1.26 0.13 1.26 0.18 1.24 0.09 1.22 0.20 1.34 0.19 1.29 0.10 1.29 0.10 1.26 0.24 1.23 0.24 1.23 0.24 1.38 0.22 1.30 0.14 1.30 0.14 1.26 0.26 1.26 0.26 1.37 0.14 1.33 0.13 1.30 0.20 20
ACCEPTED MANUSCRIPT 24S,25-dihydroxycholesterol Ianostane 20-hydroxyecdysone 22-phosphate
RI
PT
3beta-hydroxy-4beta-methyl-5alpha-cholest-7-ene4alpha-carboxylic acid 4alpha-carboxy-4beta-methyl-5alpha-cholesta-8-en-3betaol 3beta-hydroxy-4alpha-methyl-5alpha-cholest-7-ene4beta-carboxylic acid Pinnasterol 7alpha-hydroxy-4-cholesten-3-one-d7
Nebrosteroid L
NU
Ergosterols and C24-methyl derivatives
SC
7-oxocholesterol (d7)
Pseudoprotodioscin
EP T
Furostanols and derivatives
ED
MA
campest-22E-en3beta,4beta,5alpha,6alpha,8beta,14alpha,15alpha,25R,26nonol campest-22E-en3beta,4beta,5alpha,6alpha,8beta,14alpha,15alpha,25,28nonol Conicasterol D
AC C
3-O-(Rhaa1-4(Rhaa1-2)Glcb)-26-O-(Glcb)-(25R)-furost-5en-3beta,22,26-triol Pseudoprotogracillin
Solanidines and alkaloid derivatives
Spirostanols and derivatives
26-O-[beta-D-glucopyranosyl]-25R-furostan3beta,22alpha,26-triol Solasodine Etioline Diospolysaponin A Trillin 3-O-(Galb)-(25R)-12-oxo-5alpha-spirostan-3beta-ol Alliogenin 6-O-(Glcb)-(25R)-5alpha-spirostan-3beta,6alpha-diol Karatavegenin C
1.30 0.20 1.25 0.51 1.24 0.14 1.24 0.31 1.24 0.31 1.24 0.31 1.23 0.17 1.22 0.79 1.22 0.79 1.26 0.18 1.25 0.26 1.25 0.26 1.24 0.31 1.36 0.15 1.36 0.14 1.30 0.23 1.29 0.21 1.23 0.35 1.23 0.35 1.29 0.44 1.29 0.12 1.27 0.15 1.25 0.29 1.24 0.16 1.23 21
ACCEPTED MANUSCRIPT
Protometagenin Steroid lactones
4beta-Hydroxywithanolide E
Steryl esters
15:0 Cholesteryl ester 17:1 Cholesteryl ester
PT
16:1 Campesteryl ester 14:1 Cholesteryl ester Ajugalactone
RI
Stigmasterols and C24-ethyl derivatives
SC
24Z-ethylidene-cholesta-5,7-dien-3beta-ol
Withanolide
1.30 0.14 1.22 0.18 1.22 0.09
AC C
EP T
ED
MA
Withanolides and derivatives
NU
Stigmastanyl glucoside
0.17 1.23 0.17 1.22 0.12 1.31 0.28 1.29 0.13 1.29 0.13 1.24 0.62 1.40 0.14
22
ACCEPTED MANUSCRIPT Subclass Anthocyanins
Marker Cyanidin 3-O-xylosylrutinoside Cyanidin 3-Osambubioside 5-Oglucoside Delphidin 3-Oarabinoside/xyloside Cyanidin 3-O-(6’’malonyl-glucoside) Peonidin 3-O-(6’’-acetylglucoside/galactoside) Phloretin 2’-O-xylosylglucoside Apigenin 7-O-glucoside Kaempferol 3-O-glucoside Kampferol 3-O-(2’’rhamnosyl-galactoside) 7O-rhamnoside Kaempferol 3-Orhamnosyl-rhamnosylglucoside Kaempferol 3-O-xylosylrutinoside Kaempferol 3-O-(2’’rhamnosyl-6’’acetylgalactoside) 7-Orhamnoside Quercetin 3-O-glucoside Patuletin 3-O-glucosyl-(16)-[apiosyl(1-2)]glucoside Spinacetin 3-O-glucosyl(1-6)-glucoside Quercetin 3-O-xylosylrutinoside Quercetin 3-O-galactoside 6’’-O-Acetylglycitin 3,5-Dihydroxybenzoic acids Protocatechuic acid p-Coumaroyl glucose p-Coumaric acid 4-Oglucoside 24-Methylcholesterol ferulate 24-Methyllathosterol ferulate 24-Methylenecholestanol ferulate p-Coumaroyl glycolic acid 3/4/5-Caffeoylquinic acid 3/4/5-Sinapoylquinic acid 1,2’-Disinapoyl-2feruloylgentiobiose
PT
Class Flavonoids
Dihydrochalcones
EP T
ED
MA
NU
SC
RI
Flavones Flavonols
Isoflavonoids Hydroxybenzoics
AC C
Phenolic acids
Hyddroxycinnamics
VIP score 1.36 0.38 1.35 0.11 1.33 0.13 1.31 0.31 1.28 0.08 1.35 0.13 1.33 0.17 1.35 0.21 1.35 0.25 1.34 0.24 1.34 0.24 1.32 0.15
1.32 0.17 1.35 0.19 1.34 0.16 1.30 0.14 1.27 0.16 1.43 0.15 1.31 0.38 1.31 0.38 1.40 0.32 1.39 0.32 1.36 0.19 1.36 0.19 1.36 0.19 1.36 0.29 1.32 0.17 1.28 0.20 1.28 0.12
23
ACCEPTED MANUSCRIPT
Ergosterols and C24-methyl derivatives
Furanostanols
Solanidines and alkaloid derivatives Spirostanols
PT
RI
1.31 0.13 1.34 0.14 1.28 0.13 1.33 0.19 1.29 0.18 1.31 0.50
Iokundjoside 3alpha,7alpha,12alphaTrihydroxy-5betacholestanoyl-CoA 3beta-hydroxy-4betamethyl-5alpha-cholest-7ene-4alpha-carboxylic acid 4alpha-carboxy-4betamethyl-5alpha-cholesta-8en-3beta-ol 3beta-hydroxy-4alphamethyl-5alpha-cholest-7ene-4beta-carboxylic acid Cholestanyl glucoside Trihydroxycoprostanoic acid Ganosporeric acid A Cimicifoetiside A Cycloartenol Campesteryl glucoside
1.41 0.27 1.31 0.48
Conicasterol D Mutasterol Castasterone 26-O-[beta-Dglucopyranosyl]-25Rfurostan-3beta,22alpha,26triol Pseudoprotogracillin Tomatine
1.29 0.37 1.29 0.18 1.28 0.13 1.29 0.21
Epimetagenin Solaspigenin Trillenogenin
1.35 0.22 1.35 0.22 1.31 0.20
NU
Cycloartanols and derivatives
1.36 0.17 1.36 0.17 1.36 0.17 1.34 0.17 1.31 0.12 1.28 0.13 1.29 0.27 1.31 0.12 1.28 0.25 1.34 0.17
Aragusteroketal
MA ED EP T
Stilbenes Sterols
Brassinolides and derivatives Bufanolides and derivatives C24-propyl sterols and derivatives Calysterols and cyclopropyl sidechain derivatives Cardanolides and derivatives Cholesterol and derivatives
AC C
Lignans
Homovanillic acid Dihydrocaffeic acid Syringaldehyde Rosmadial 3-Methoxyacetophenone Esculetin Estragole 4-Vinylguaiacol 4-Ethylphenol Anhydrosecoisolariciresinol Syringaresinol Piceatannol 3-O-glucoside 7-oxotyphasterol Scillaren A 24-Isopropenylcholesterol
SC
Other polyphenols
Hydroxyphenylacetics Hydroxyphenylpropanoics Hydroxybenzaldehydes Phenolic terpenes Hydroxybenzoketones Hydroxycoumarins Hydroxyphenylpropenes Alkylmehtoxyphenols Alkylphenols -
1.29 0.37
1.29 0.37 1.29 0.37 1.29 0.13 1.28 0.13 1.27 0.20 1.30 0.18 1.29 0.18 1.30 0.17
1.28 0.43 1.35 0.20
24
ACCEPTED MANUSCRIPT
Stigmasterols and C24-ethyl derivatives
18:0 Cholesteryl ester 16:0 Sitosteryl ester 15:0 Cholesteryl ester Cyasterone
1.31 0.08 1.30 0.26 1.31 0.08 1.31 0.23
(20S,24S)-24ethylthornasterol Minabeolide-7
1.30 0.24
1.34 0.26
1.28 0.22
SC
RI
Withanolides and derivatives
1.28 0.13 1.27 0.11
PT
Steroids and steroid derivatives Steryl esters
Hongguanggenin 3-O-(Galb)-(25R)-12-oxo5alpha-spirostan-3beta-ol Withaperuvin B
NU
Discrimination of extra-virgin-olive oils from different cultivars and geographical origins by untargeted
MA
metabolomics
ED
Silvia Ghisoni, Luigi Lucini, Federica Angilletta, Gabriele Rocchetti, Daniela Farinelli, Sergio Tombesi, Marco Trevisan
EP T
HIGHLIGHTS
Sterolic and phenolic profiles in EVOO accounted for more than 1000 compounds.
The chemical fingerprints of each EVOO cultivar analyzed showed differences.
Multivariate statistics discriminated EVOO according to geographical origin.
Flavonoids, hydroxycinnamics and cholesterol derivatives were the best markers.
AC C
25
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7