of
se
J.M
ctio
17,
Oli e
grap
fication results for all classes. Moreover, 1H NMR spectra of the bulk oil, and its corresponding
linked to the concept of terroir – a sense of place discernible in the
flavour of the food. PDO products must be produced, processed and
Spectroscopic, isotopic, chromatographic, and trace element
analysis prove to be very interesting in the study of large natural
variability in chemical composition of foodstuffs due to differences,
such as species, maturity, environment, growing region, as well as
processing and storage techniques. Analysis of the large amount of
data generated using each of these techniques and the often subtle
differences between authentic and adulterated samples require the
use of statistical programs. Chemometric analysis is therefore a
very important feature of modern analytical approaches for the
characterisation of complex matrices. A considerable number of
approaches [Commission Regulation (EEC) No. 2568/91], both
Abbreviations: VOO, virgin olive oil; PDO, protected designation of origin; PGI,
protected geographical indication; TSG, traditional specialty guaranteed; HF,
hydrocarbon fraction; TF, tocopherol fraction; AF, alcohol fraction; SF, sterol
fraction; PCA, principal component analysis; LDA, linear discriminant analysis; PLS-
DA, partial least squares discriminant analysis; CART, classification and regression
trees; SIMCA, soft independent modelling of class analogy; LOO, leave-one-out
cross-validation; SSOG, Stazione Sperimentale per le Industrie degli Oli e dei Grassi.
* Corresponding author. Tel.: +39 0332785986; fax: +39 0332789303.
Food Chemistry 118 (2010) 956–965
Contents lists availab
Food Che
journal homepage: www.else
E-mail address: rosa-maria.alonso-salces@jrc.it (R.M. Alonso-Salces).
The consumption of olive oil dates back to biblical times, and
the cultivation of the olive tree as well as the production of olive
oil from the mature drupe (fruit) remain an essential part of farm-
ing practices in the Mediterranean basin. At present, 79% of the
global production of olive oil takes place in that area, namely Spain,
Italy, and Greece. The characterisation of the geographical origin of
virgin olive oil (VOO), which is permitted to be marketed under a
protected designation of origin (PDO), protected geographical indi-
cation (PGI) or traditional specialty guaranteed (TSG) label, on the
basis of their area and methods of production [Council Regulations
(EEC) Nos. 2081/92 and 2082/92], is becoming increasingly impor-
tant. According to the EU definition, PDO products are most closely
the product bears. The quality or characteristics of the product
must be due essentially or exclusively to its place of origin, i.e., cli-
mate, the nature of the soil and local know-how. Food products
with a PGI status must have a geographical link in at least one of
the stages of production, processing or preparation. As can be ex-
pected, given the financial benefits associated with these presti-
gious labels, it is very likely that economic fraud occurs (e.g.
labelling a non-PDO product as a PDO one or adulteration with ol-
ive oils that do not fulfil the PDO requirements). Therefore, analyt-
ical methods are necessary to guarantee the authenticity and
traceability of PDO and PGI olive oils to prevent illicit practices
in this sector.
Keywords:
Olive oil
Unsaponifiable
NMR fingerprinting
Pattern recognition
Multivariate data analysis
1. Introduction
0308-8146/$ - see front matter � 2008 Elsevier Ltd. A
doi:10.1016/j.foodchem.2008.09.061
unsaponifiable fraction, as well as the subfractions of the unsaponifiable fraction (alcohol, sterol, hydro-
carbon, and tocopherol fractions) were studied in the search for the markers that multivariate techniques
revealed to be related to the geographical origin of olive oils. Additionally, a preliminary study regarding
1H NMR data of the bulk oil and the corresponding unsaponifiable fraction of VOOs suggested that these
spectral data contained complementary information for the geographical characterisation of VOOs.
� 2008 Elsevier Ltd. All rights reserved.
prepared in a specific region using traditional production methods.
The raw materials must also be from the defined area whose name
Received 28 November 2007
Received in revised form 14 July 2008
Accepted 22 September 2008
gerprint of the unsaponifiable matter is presented. The 1H NMR spectra of the unsaponifiable fraction of
virgin olive oils from Spain, Italy, Greece, Tunisia, Turkey, and Syria were analysed by several pattern rec-
ognition techniques (LDA, PLS-DA, SIMCA, and CART). PLS-DA (PLS-1 approach) obtained the best classi-
Multivariate analysis of NMR fingerprint
virgin olive oils for authentication purpo
R.M. Alonso-Salces a,*, K. Héberger b, M.V. Holland a,
G. Bellan c, F. Reniero a, C. Guillou a
a European Commission - Joint Research Centre, Institute for Health and Consumer Prote
bChemical Research Centre, Hungarian Academy of Sciences, H-1525 Budapest, P.O. Box
c Laboratorio oli e grassi vegetali e animali, Stazione Sperimentale per le Industrie degli
a r t i c l e i n f o
Article history:
a b s t r a c t
A new approach to the geo
ll rights reserved.
the unsaponifiable fraction of
s
. Moreno-Rojas a, C. Mariani c,
n, Physical and Chemical Exposure Unit, Via Fermi 2, I-21020 Ispra, Italy
Hungary
dei Grassi (SSOG), Via Giuseppe Colombo 79, I-20133 Milan, Italy
hical characterisation of virgin olive oils (VOOs) based on the 1H NMR fin-
le at ScienceDirect
mistry
vier .com/locate / foodchem
d Ch
chemical and sensorial, combined with statistical analysis, has
greatly facilitated the fight against fraud in this sector. Moreover,
the contribution of databases and mathematical algorithms is nec-
essary in order to provide the classification required to fully guar-
antee both the geographical origin and authenticity of VOO
(Aparicio & Aparicio-Ruiz, 2000).
Olive oil is mainly made up of triglycerides (more than 98%),
whereas the unsaponifiable fraction of virgin olive oil (VOO) repre-
sents 1–2% of the oil. This fraction is made up of minor constitu-
ents, which may vary both qualitatively and quantitatively
depending on vegetal species, climatic conditions, extraction and
refining procedures, and storage conditions (Cañabate-Diaz et al.,
2007; Harwood & Aparicio, 2000). These also greatly influence
the organoleptic quality and stability of the oil. VOO may contain
up to 0.7% hydrocarbons, which is mainly constituted by squalene,
a precursor of other sterols and triterpenic alcohols (Bor-
tolomeazzi, Berno, Pizzale, & Conte, 2001; Lercker & Rodriguez-
Estrada, 2000). Besides squalene, the hydrocarbon fraction of olive
oil comprises of low quantities of epoxy-squalene isomers and al-
kanes (C16–C35) (Grob, Lanfranchi, & Mariani, 1990; Mariani &
Fedeli, 1986). Plant sterols or phytosterols make up the main part
of the unsaponifiable fraction of olive oil. The most abundant olive
oil sterol is b-sitosterol, followed by D5-avenasterol. Campesterol
and stigmasterol are present in lower concentrations (Harwood &
Aparicio, 2000). Regarding the tracking of commercial fraud, the
sterol fraction has many applications, especially where the con-
tamination of some vegetable oils with other cheaper ones is con-
cerned. Positional isomers of the double bond in the sterol ring
have been detected and can be used as fraud tracers in VOO (Apari-
cio & Aparicio-Ruiz, 2000; Biedermann, Grob, Mariani, & Schmidt,
1996; Lercker & Rodriguez-Estrada, 2000; Mariani, Venturin, &
Grob, 1995). Of the tocopherols, a-tocopherol comprises about
90% of the total tocopherol fraction. The stability of VOO is mainly
due to their relatively low fatty acid unsaturation and to the anti-
oxidant activity of some of the unsaponifiable components: the
activity of a-tocopherol and the effect it has on the oxidative sus-
ceptibility of an olive oil was researched by Deiana et al. (2002).
Moreover, a-tocopherol showed a synergistic effect in association
with some phenolic compounds with significant activity. The ma-
jor phenolic compounds identified and quantified in olive oil be-
long to three different classes: simple phenols (hydroxytyrosol,
tyrosol), secoiridoids, and the lignans. The cultivar, the system of
extraction, and the conditions of processing and storage are critical
factors for the polyphenol content of an olive oil (Aparicio & Luna,
2002). Other constituents of the unsaponifiable matter are the pig-
ments which impart the characteristic colour to an olive oil. They
are made up of carotenoids, most importantly b-carotene, which
gives a yellow colour; chlorophylls, responsible for the green
shades; and pheophytins (Minguez-Mosquera, Gandul-Rojas, Garr-
ido-Fernandez, & Gallardo-Guerrero, 1990). Other components in-
clude aliphatic alcohols made up of docosanol, tetracosanol,
hexacosanol and octacosanol; and at trace levels, tricosanol, penta-
cosanol, and heptacosanol. In smaller quantities the following
triterpenic alcohols are present: cycloartenol, 24-methylen-cyclo-
artenol, and a- and b-amirines; diterpenic alcohols: fitol and gera-
nilgeraniol; and triterpenic dialcohols: erythrodiol and uvaol.
These numerous compounds, present in tiny percentages in olive
oil, hold great biological importance and are characteristic of each
type of oil (Harwood & Aparicio, 2000). In fact, the composition of
the unsaponifiable fraction of VOO is affected by several factors
(Harwood & Aparicio, 2000) such as olive cultivar (Aparicio, Mor-
ales, & Alonso, 1997; Esti, Cinquante, & La Notte, 1998; Pinelli
et al., 2003), altitude (Mousa, Gerasopoulos, Metzidakis, & Kiritsa-
R.M. Alonso-Salces et al. / Foo
kis, 1996), climatology (Aparicio, Ferreiro, & Alonso, 1994), agro-
nomic factors (Gutierrez, Albi, Palma, Rios, & Olias, 1989), time of
harvest (Aparicio & Morales, 1998), olive storage after harvest
(Mariani, Fedeli, Grob, & Artho, 1991), and oil extraction system
(Angerosa & Di Giovacchino, 1996; Ranalli & Angerosa, 1996).
The diversity and interrelation amongst all these factors is re-
flected in the chemical composition of VOO, and it is highly unli-
kely that this influence would be the same in different regions.
So, the geographical characterisation of VOO regards all these agro-
nomic, pedoclimatic and botanical aspects that characterise the oil
of each origin (Aparicio, Alonso, & Morales, 1994; Esti et al., 1996).
Therefore, it can be expected that the unsaponifiable fraction of
VOOs may contain information for the geographical characterisa-
tion of olive oils.
The more commonly used methods for minor constituent deter-
mination in olive oil usually require isolation and several proce-
dures of separation, identification and quantification. The
conventional method for quantifying sterols and triterpenic
alcohols involves capillary GC–FID of the fraction isolated by TLC
[Commission Regulation (EEC) No. 2568/91]. Regarding the quanti-
fication of minor components in the unsaponifiable matter of olive
oil, chromatographic methods have proven to be particularly suit-
able (Lercker & Rodriguez-Estrada, 2000). Thus, alkanols, squalene,
a-tocopherol, and sterols were determined by GC (Giacometti,
2001); sterols were quantified by GC–FID for the classification
and authentication of monovarietal and PDO olive oils (Alves, Cun-
ha, Amaral, Pereira, & Oliveira, 2005); phytosterols were analysed
by GC–MS in order to compare their content with that of sunflower
oil and butter (Zhang et al., 2005); SPE and GC–MS were used to
characterise free and esterified sterols (Cunha, Fernandes, & Oli-
veira, 2006); tocopherols, and phytosterols, quantified by reversed
phase HPLC, discriminate between very similar oils (Lopez Ortiz,
Prats Moya, & Berenguer Navarro, 2006); and seven phytosterols
analysed by LC–MS distinguished different qualities of olive oils
(Cañabate-Diaz et al., 2007). FT-Raman and FT-NIR spectroscopy
has also been applied to characterise olive oil. Moreover, the spec-
tra of the unsaponifiable matter of olive oil obtained by these tech-
niques together with univariate and multivariate statistical models
enabled the detection of hazelnut oil in olive oil at a level as low as
8% (Baeten et al., 2005). FT-Raman bands due to the major unsa-
ponifiable series of compounds, i.e. squalene, sterolic, and terpenic
fractions, analysed by unsupervised multivariate techniques allow
us to differentiate between olive oils and other seed oils as well as
and amongst varietal VOO (Baeten, Dardenne, & Aparicio, 2001).
1H, 13C and/or 31P NMR analysis of the bulk oil, in combination with
multivariate techniques, have been used to distinguish VOOs
according to their geographical origin (Rezzi et al., 2005), as well
as to detect adulteration of the oil (Fragaki, Spyros, Siragakis, Saliv-
aras, & Dais, 2005; Garcia-Gonzalez, Mannina, D’Imperio, Segre &
Aparicio, 2004). IRMS methods have also been used for the authen-
tication of olive oil by analysing the bulk oil (Angerosa, Camera,
Cumitini, Gleixner, & Reniero, 1997; Bianchi, Angerosa, Camera,
Reniero, & Anglani, 1993; Breas, Guillou, Reniero, Sada, & Angerosa,
1998). Isotopic measurements of alcohol and sterol fractions of ol-
ive oil also proved to be useful for its geographical characterisation
(Angerosa et al., 1999). 1H NMR and the more recently developed
hyphenated LC–SPE–NMR technique have been applied to study
phenolic compounds in the polar fraction of olive oil for authenti-
cation purposes (Christophoridou, Dais, Tseng, & Spraul, 2005; Sac-
co et al., 2000).
In the present work, a new approach based on 1H NMR finger-
printing of the unsaponifiable fraction of VOO is presented. VOO
from six countries, namely Spain, Italy, Greece, Tunisia, Turkey,
and Syria, were analysed. The spectral data was subjected to pat-
tern recognition techniques, which can be used to classify a VOO
according to its geographical origin. Moreover, 1H NMR spectra of
emistry 118 (2010) 956–965 957
the bulk oil, and its corresponding unsaponifiable fraction, as well
as the subfractions of the unsaponifiable fraction (alcohol, sterol,
hydrocarbon, and tocopherol fractions) were studied in the search
visible under ultraviolet light: the hydrocarbon fraction (HF);
tocopherol fraction (TF); alcohol fraction (AF), containing cycloar-
Ch
tenol; and the sterol fraction (SF) and erythrodiol. Then, each single
band is scraped off and extracted with diethyl ether (10 ml), fil-
for the markers that multivariate techniques revealed to be related
to the geographical origin of olive oils.
2. Materials and methods
2.1. Samples and chemicals
Commercial virgin olive oils (99 samples) from six countries of
the Mediterranean basin, namely Italy (19 VOOs), Spain (36 VOOs),
Greece (12 VOOs), Tunisia (15 VOOs), Turkey (8 VOOs), and Syria (9
VOOs), were provided by the SSOG (Stazione Sperimentale per le
Industrie degli Oli e dei Grassi, Milan, Italy), which was able to as-
sure the true type (virgin) and origin of the olive oils at least at the
national level. Virgin olive oils (23 samples) from different Italian
(5 VOOs), Spanish (4 VOOs), and Greek (5 VOOs) PDOs as well as
samples from Tunisia (5 VOOs) and Turkey (4 VOOs) were collected
directly in the country of origin by the SSOG.
Deuterated chloroform for NMR analysis (99.8 at % D), chloro-
form (p.a.), 1-eicosanol, a-tocopherol, b-sitosterol, stigmasterol,
campesterol, and silica gel on TLC plates diethyl ether were pro-
vided by Sigma–Aldrich Chemie (Steinheim, Germany); potassium
hydroxide (p.a.), anhydrous sodium sulphate (p.a.), hexane (p.a.),
and 20,70-dichlorofluorescein (TLC grade) by Merck (Darmstadt,
Germany); diethyl ether (HPLC grade) by Fluka Chemie (Buchs,
Switzerland); methanol (HPLC grade) by Carlo Erba (Rodano, Italy);
and erythrodiol by Extrasynthèse (Genay France). Cycloartenol
standard was prepared by extracting the unsaponifiable fraction
from flax oil and performing a further purification of the extract
by thin layer chromatography (TLC) (see Sections 2.2 and 2.3).
2.2. Preparation of the unsaponifiable fraction of olive oil
The unsaponifiable fraction was prepared using a modification
of the method described in the regulation EEC-2568/91 (annex V,
Section 5.1). The sample of olive oil is dried under a nitrogen flow
and filtered. Then, 50 ml of methanolic potassium hydroxide 2 N is
added to an aliquot of 5 g of the dried and filtered olive oil, and
heated to a gentle boil in a water bath with continuous vigorous
stirring under reflux for 1 h. Then the content is transferred quan-
titatively into a 500 ml funnel using several rinses of distilled
water (about 100 ml), and three successive extractions with ethyl
ether (80 ml) are performed. The ether phase is washed with dis-
tilled water until the wash water reaches a neutral pH. Once the
water is removed, the extract is dried with anhydrous sodium sul-
phate for 30 min, filtered and the solvent removed using a rotava-
por at 40 �C to dryness. The repeatability of the method was
evaluated by extracting separately four aliquots of eight different
samples of VOOs of four different geographical origins.
2.3. Preparation of the alcohol, sterol, hydrocarbon, and tocopherol
fractions
The unsaponifiable fraction of the oil is dissolved in chloroform
(5% w/v) and fractionated by TLC using hexane-diethyl ether
(50:50, v/v) as the eluent. Once the plate is developed, it is dried
by leaving it for a short time under a fume hood. Then, it is sprayed
with an ethanolic solution of 2,7-dichlorofluorescein (0.2%, w/v), in
order to make the bands of the different fractions and components
958 R.M. Alonso-Salces et al. / Food
tered under vacuum, washed several times with diethyl ether,
and evaporated to dryness by mild heating in a gentle flow of nitro-
gen. An aliquot of each fraction was analysed by GC–FID (regula-
tion EEC-2568/91, annex V, Sections 5.3 and 5.4.2.1.) in order to
control the performance of the fractionation procedure.
2.4. NMR analysis
Each unsaponifiable fraction of VOO or 40 ll of the bulk oil was
dissolved in 200 ll of deuterated chloroform, shaken in a vortex,
and placed in a 2 mm NMR capillary. The 1H NMR experiments
were performed at 300 K on a Bruker (Rheinstetten, Germany)
Avance 500 (nominal frequency 500.13 MHz) equipped with a
2.5 mm broadband inverse probe. The spectra were recorded using
a 7.5 ls pulse (90�), an acquisition time of 3.0 s (32k data points)
and a total recycling time of 4.0 s, a spectral width of 5500 Hz
(11 ppm), 64 scans (+4 dummy scans), with no sample rotation.
Prior to Fourier transformation, the free induction decays (FIDs)
were zero-filled to 64k and a 0.3 Hz line-broadening factor was ap-
plied. The chemical shifts are expressed in d scale (ppm), refer-
enced to the residual signal of chloroform (7.26 ppm) (Hoffman,
2006). The interesting regions of the NMR spectra for the analysis
of the unsaponifiable fraction are from 0 to 5.44 ppm, and for the
bulk oil, from 0 to 7 ppm. The spectra were phase- and baseline-
corrected manually, and normalised to total intensity over the
region 0–5.44 ppm for the unsaponifiable fraction and over the re-
gion 4.08–4.28 ppm for the bulk oil. The region of interest of the
NMR spectra was binned with 0.04 ppm-wide buckets. TopSpin
1.3 (2005) and Amix-Viewer 3.7.7 (2006) from Bruker BioSpin
GMBH (Rheinstetten, Germany) were used to perform the process-
ing of the spectra. The data table generated with the spectra of all
samples was then used for pattern recognition. One bucket in the
region 3.96–4.00 ppm (artifact) in the unsaponifiable fraction anal-
ysis and five buckets in the region 4.08–4.28 ppm (reference re-
gion) in the bulk oil analysis were excluded in the multivariate
data analysis.
2.5. Multivariate data analysis
The data set consisted of a 99 � 135 matrix, in which rows rep-
resented the samples (99 unsaponifiable fractions of VOO ana-
lysed), and columns the 135 buckets of the 1H NMR spectrum.
Each VOO was represented in the 135-dimension