a combined approach for characterisation of fresh and brined vine leaves by x-ray powder...

9
This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/authorsrights

Upload: independent

Post on 10-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/authorsrights

Author's personal copy

A combined approach for characterisation of fresh and brined vine leavesby X-ray powder diffraction, NMR spectroscopy and direct infusion highresolution mass spectrometry

Antonino Rizzuti a, Rocco Caliandro b, Vito Gallo a,c,⇑, Piero Mastrorilli a,c, Giuseppe Chita b,Mario Latronico a,c

a Dipartimento di Ingegneria Civile Ambientale del Territorio Edile e di Chimica, Politecnico di Bari, Via Orabona 4, I-70125 Bari, Italyb Istituto di Cristallografia, Consiglio Nazionale delle Ricerche, Via Amendola 122/O, I-70126 Bari, Italyc Innovative Solutions S.r.l., Spin off of Politecnico di Bari, Zona H 150/B, 70015 Noci (BA), Italy

a r t i c l e i n f o

Article history:Available online 24 May 2013

Keywords:Vine leavesX-ray powder diffractionDirect infusion mass spectrometryNuclear magnetic resonancePrincipal component analysisCovariance analysis

a b s t r a c t

X-ray powder diffraction was combined, for the first time, with Nuclear Magnetic Resonance spectros-copy and direct infusion mass spectrometry to characterise fresh and brined grape leaves. Covarianceanalysis of data generated by the three techniques was performed with the aim to correlate informationderiving from the solid part with those obtained for soluble metabolites. The results obtained indicatethat crystalline components can be correlated to the metabolites contained in the grape leaves, pavingthe way to the use of X-ray diffraction analysis for food fingerprinting purposes. Moreover it was ascer-tained that, differently from most of the metabolites present in the fresh vine leaves, linolenic acid (anomega-3-fatty acid) and quercetin-3-O-glucuronide (a polyphenol metabolite) do not undergo sensibledegradation during the brining process, which is used as preservative method for the grape leaves.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Grapevines represent one of the most important plant speciesdue to the wide use of their related food products, wine beingthe most popular in terms of historical, cultural and economic va-lue. Besides the grapes cultivars used for wine production, manyvarieties are used for production of table grapes which are con-sumed as fresh or dried fruit. Moreover, vine leaves are used asfood products, in fresh and brined forms, in several countriesincluding Saudi Arabia, Turkey, Greece, Bulgaria, Romania andVietnam. Vine leaves are considered a healthy food due to theirastringent and hemostatic properties and phenolic composition(Kos�ar et al., 2007). The growing demand of vine leaves for the foodmarket, mainly supported by Arabic regions, has been attractingstrong economic interest also in those countries, like Italy, not par-ticularly devoted to vine leaves distribution.

Concerning vine leaves production and commercialisation,quality assessment and authentication tools are not so developedas for wine and table grapes. Efficient discrimination tools can sup-port the producers in optimising the production process and theconsumers in recognising the leaves quality, but require robust

analytical methods to give a complete overview of foodstuff(Sumner, Duran, Huhman, & Smith, 2002). In this framework, evergrowing attention is paid to metabolomics and to unbiased analyt-ical techniques which are able to guarantee reproducible results(Kopka, Fernie, Weckwerth, Gibon, & Stitt, 2004). Metabolomicapproach, which involve extensive use on Nuclear MagneticResonance (NMR) spectroscopy and mass spectrometry (MS)(Colquhoun, 2007; Kim, Choi, & Verpoorte, 2010; Nicholson &Lindon, 2008; Robinette, Brüschweiler, Schroeder, & Edison,2012; Schripsema, 2010; Sumner, Mendes, & Dixon, 2003; Wishart,2008), provides information on a wide range of detectable chemi-cal compounds contained in food products. The large amount ofdata deriving from metabolome analyses using NMR and MStechniques permitted a widespread use of powerful multivariatestatistical methods, for grapevine based products (Ali et al., 2011;Bevilacqua et al., 2012; Ferrara et al., 2013; Forveille, Vercauteren,& Rutledge, 1996; Gallo et al., 2012; Lamikanara & Kassa, 1999;Mulas et al., 2011; Nikolaou, Koukourikou, & Koragiannidis,2000; Pereira et al., 2005; Pereira, Gaudillere, Pieri, et al., 2006;Pereira, Gaudillere, van Leeuwan, et al., 2006; Picone et al., 2011;Satisha, Pooja Doshi, & Adsule, 2008; Son, Hwang, Ahn, et al.,2009; Son, Hwang, Kim, et al., 2009), with the aim to select theinformation peculiar to the samples under investigation. X-raypowder diffraction (XRPD) studies were applied on plants to deter-mine the crystalline order of their epicuticular waxes, for example

0308-8146/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.foodchem.2013.05.044

⇑ Corresponding author. Address: Dipartimento di Ingegneria Civile, Ambientale,del Territorio, Edile e di Chimica, Politecnico di Bari, Via Orabona, 4, I-70125, Bari,Italy. Tel.: +39 080 5963607; fax: +39 080 5963611.

E-mail address: [email protected] (V. Gallo).

Food Chemistry 141 (2013) 1908–1915

Contents lists available at SciVerse ScienceDirect

Food Chemistry

journal homepage: www.elsevier .com/locate / foodchem

Author's personal copy

on the anti-adhesive property of Lotus leaves (Ensikat, Boese, Ma-der, Barthlott, & Koch, 2006). New methods based on multivariatestatistical analysis of XRPD data were developed for the determina-tion of the crystallinity in cellulose samples (Bansal, Hall, Realff,Lee, & Bommarius, 2010), but, to the best of our knowledge, thistechnique has not been yet applied for food fingerprinting.

It is generally accepted that a single analytical technique doesnot provide sufficient information on the metabolome. For in-stance, among the techniques widely used in metabolomic studies,NMR spectroscopy provides good specificity for compounds con-taining non-zero magnetic moment nuclei such as 1H and 13C, i.e.biological species, but allows for detection of a limited number ofmetabolites due to its low sensitivity (Sumner et al., 2003). Onthe contrary, mass spectrometry produces information about avery high number of metabolites thanks to its high sensitivity,but suffers from insufficient coverage of MS profiling and difficul-ties in comparing analytical data from different studies and labora-tories (Scalbert et al., 2009). Therefore, a combined approach isdesirable to gain a comprehensive view of the sample under inves-tigation. Concerning food characterisation, several studies onhyphenation of diverse analytical instruments (Seger et al., 2005)and the correlation and/or the combination of information pro-vided by different techniques (Aliferis & Jabaji, 2010; Casale,Armanino, Casolino, & Forina, 2007; Casale, Casolino, Olivieri, &Forina, 2010) are reported.

In the present paper we report on the characterisation of freshand brined vine leaves by using, for the first time, XRPD combinedwith 1H NMR and direct infusion MS analyses as sources of com-plementary information. The choice of these techniques derivesfrom the fact that (i) XRPD characterises the crystalline parts ofthe samples which are not detectable neither by solution NMRnor by HRMS; (ii) NMR gives information on primary and second-ary metabolites which are present in not negligible concentration;(iii) direct infusion MS adds information on metabolites which arepresent in traces. The application on XRPD patterns of multivariatestatistics such as the covariance tool, permitted the correlation ofthe crystalline components with the metabolites detectable inthe extracts by NMR and the direct infusion MS, introducing XRPDas a technique for metabolomic analysis.

2. Materials and methods

2.1. Plant materials and sample preparation

Fresh and brined vine leaves were received by producers andstored at �20 �C before freeze drying at �50 �C and 0.045 atm for16 h in a lyophilizer (Martin-Christ GmbH, Model Alpha 1-4 LSC).Dry leaves were ground by a blender equipped with stainless steelblades and the resulting powder was stored at room temperaturein the desiccator using calcium chloride as drying agent.

2.2. X-ray powder diffraction

Vine leaf powders were used without any further treatment. X-ray powder diffraction data were collected on a Debye–Scherrerdiffractometer equipped with a rotating anode generator (Rigaku)and an INEL linear detector. The sample was mounted in a1.0 mm diameter capillary which is continuously rotated duringdata collection occurring with a sampling step of 0.02�. 4 h werenecessary, on average, to collect a single diffraction pattern fromsamples. Data collection in transmission geometry, by means ofthe capillary, was found to be more reliable than that made inreflection geometry, where the sample is placed on a flat sampleholder. In fact, in the latter case the preparation of the sample ismore complex and the reproducibility of the measurements in

Debye–Scherrer geometry is limited by the difficulty of reproduc-ing the angle of incidence of the X-ray beam on the sample.

2.3. Direct infusion mass spectrometry

In a test tube, 7.5 ll of 29% aqueous ammonia and 50 mg ofground lyophilized grape leaf sample were added to 1.5 ml of asolution containing methanol and deionized water (resistivity:18 MX cm) in a 70:30 volume ratio. The resulting mixture wasshaken for 2 min in a VORTEX at 2000 rpm, sonicated for 20 minat room temperature and centrifuged for 30 min at 4000 rpm.

Each grape leaf extract was analysed in three replicates byautomatic direct infusion ESI-MS by means of a syringe pump(100 L KD Scientific) at a flow rate of 3 ll/min and a high perfor-mance LC autosampler (G1377A, Agilent) at an eluent flow rateof 0.150 ml/min with an injection volume of 15 ll. Eluent was pre-pared in stocks consisting of 700 ml of methanol, 300 ml of deion-ized water (resistivity: 18 MX cm) and 7.5 ml of 29% aqueousammonia. Mass spectra were acquired in negative ion mode usingMicrOTOF-Q II mass spectrometer (Bruker Daltonics, Macerata,Italy) in the range m/z 50–1000. Capillary and cone voltages wereset to �2300 V and �50 V, respectively. De-solvation temperaturewas set at 200 �C, nitrogen was used as nebulizer gas at a pressureof 1.5 bar and as dry gas at a flow rate of 8 l/min while argon wasthe collision gas. Calibration was performed analysing a referencesolution made up of 10 ll of 98% formic acid, 10 ll of sodiumhydroxide (1.0 M), 490 ll of i-propanol and 490 ll of deionizedwater. The raw data were collected as continuum mass spectra ata regular time interval (spectra rate of 1 spectrum/s with a rollingaverages of 3), and the corresponding chromatogram represents ca.0.8 minutes wide one peak. Mass spectra were processed usingData Analysis 4.0 (Bruker Daltonik GmbH, Bremen, Germany).

Bucketing of mass data was performed using AMIX 3.9.13 in theadvanced bucketing mode with a m/z displacement of 0.05 in therange of 50.50 and 600.50 Da, scaling the intensities of individual[M�H]� ions recorded between 0.2 and 0.3 min to total intensity.The as-prepared mass data buckets were submitted to the PrincipalComponent Analysis (PCA).

2.4. Nuclear magnetic resonance

520 ll of oxalate buffer at pH 4.1 containing NaN3

(2.5 � 10�3 M), 750 ll of CD3OD and 230 ll of 0.16% trimethylsilylpropanoic acid-d4 (TSP) in D2O was added to 50 mg of lyophilizedleaf sample in an 1.5 ml Eppendorf� vial. The mixture was shakedfor 2 min by VORTEX at 2000 rpm, sonicated for 20 min at roomtemperature at a power of 100% and centrifuged for 30 min at4000 rpm. 800 ll of the extract were used for the NMR analysis.

One-dimensional 1H NOESY spectra were recorded on a BrukerAvance I 400 MHz spectrometer equipped with a 5 mm inverseprobe and with an autosampler. 1H NOESY spectra were acquiredwith 128 scans of 64 K data points with a spectral width of8013 Hz, a pulse angle of 90�, an acquisition time of 4.09 s, a mix-ing time of 10 ms and a recycle delay of 3.0 s. Each spectrum wasacquired using TOPSPIN 3.0 software (Bruker BioSpin GmbH,Rheinstetten, Germany) under an automatic procedure lastingapproximately 22 min and consisting of sample loading, tempera-ture stabilisation for 5 min, tuning, matching, shimming and 90�pulse calibration. Free induction decays (FIDs) were Fourier trans-formed, the phase was manually corrected, the baseline was auto-matically corrected and the spectra were aligned by setting the TSPsinglet to 0 ppm.

NMR raw data were processed using TOPSPIN 3.0 and convertedin regular rectangular buckets by Amix 3.9.13 (Bruker BioSpinGmbH, Rheinstetten, Germany). Bucketing was performed with0.04 ppm bucket width in the range 10.00-0.50 ppm excluding

A. Rizzuti et al. / Food Chemistry 141 (2013) 1908–1915 1909

Author's personal copy

the spectral region between 3.32 and 3.28 ppm, which contains thesignal of deuterated methanol, and the region between 5.00 and4.28 ppm, which contains the residual signals of water and the sig-nal of the tartaric acid. This latter signal was excluded since its highvariability in terms of chemical shift may prevent a correct inter-pretation of statistical analysis. Buckets were scaled to total inten-sity and submitted to the PCA.

2.5. Data analysis

Each sample was represented by a data vector in the N-dimen-sional space of XRPD, NMR or ESI-MS variables. Data vectors wereanalysed by PCA and covariance analysis. Both of them wereperformed by means of the program MultiSpectra (Caliandro, Diprofio, & Nicolotti, 2013) which is available on request atusers.ba.cnr.it/ic/crisrc25, and has been developed within the Rootframework (Brun & Rademakers, 1998). For XRPD patterns, full-profile fitting was performed by QUANTO (Altomare et al., 2001),a stand-alone software available at www.ic.cnr.it.

2.5.1. XRPD data pre-processingPCA was applied to spectral XRPD data yðiÞ obtained by two

subsequent rescaling steps applied to the original spectra y(i):

(1) Normalisation (NORM). Each spectrum was normalised withrespect to the area under the intensity-2h curve. This areacan vary from one sample to the other due to the poor con-trol over the amount of sample put on the capillary. The nor-malisation has been applied according to (Bansal et al.,2010):

y0ðiÞ ¼ yðiÞ

hXN

i¼1

Ið2hiÞ; ð1Þ

where h is the scanning step size of the 2h angle used during mea-surement (0.02�), and Ið2hiÞ is the intensity measured at the diffrac-tion angle 2hi. The above mentioned area coincides with thequantity at denominator of (1).

(1) Background subtraction (BS) was applied, according to:

yðiÞ ¼ y0ðiÞ � bðiÞ ð2Þ

where b(i) represents the background of the spectrum. This hasbeen estimated by means of the Sensitive Nonlinear Iterative Peak(SNIP) clipping algorithm (Burgess & Tervo, 1983; Morhác, Kliman,Matoušek, Veselsky, & Turzo, 1997; Ryan, Clayton, Griffin, Sie, &Cousens, 1988). Given the original pattern y(i), where i = 1,. . .n indi-cate the channel of the spectrum, a new value in the ith channel iscalculated at the pth iteration as:

ypðiÞ ¼ Min yp�1ðiÞ;yp�1ðiþ pÞ þ yp�1ði� pÞ

2

� �: ð3Þ

At the end of the process, the estimated background isbðiÞ ¼ yNclipðiÞ, where Nclip is a free parameter of the algorithm, rep-resenting both the number of iterations and the width of the clip-ping window. It’s value should be chosen by taking into accountthe widths of the peaks one wants to preserve in the spectrum(ideally Nclip channels should cover half-width of the peaks).

2.5.2. Rietveld analysisA full-pattern fitting of XRPD patterns was performed by means

of the Rietveld method (Rietveld, 1969) which aims at deriving theweight fraction of each crystalline phase present in the polycrystal-

line mixtures. This method requires for each phase, the priorknowledge of the structural model.

2.5.3. Covariance analysisThe correlation among XRPD, NMR and ESI-MS data vectors

were studied by calculating the covariance matrix

COVABði; jÞ ¼X

k

ðykðiÞ � hyðiÞiÞAðykðjÞ � hyðjÞiÞB; ð4Þ

where A and B stand for XRPD, NMR or ESI-MS, ykðiÞ is the pre-pro-cessed spectrum of sample k and hyðiÞi is the average value calcu-lated over the samples for the i-th variable. XRPD spectra werepreprocessed according to Eqs. (1) and (2), while NMR and ESI-MSspectra were mean centred.

3. Results and discussion

Grape leave samples considered in the present study are listedin Table 1 and consist of fresh and brined samples deriving fromdifferent cultivars: Victoria, Sangiovese and Italia for fresh leavesand Superior Seedless, Thompson Seedless and Italia for brinedsamples. Among the fresh samples, those of cv. Italia includedleaves from vineyards managed with different agronomical prac-tises (conventional and organic).

3.1. Analysis of XRPD patterns

XRPD patterns show a huge background with a common shapeand a number of sharp peaks, whose position and height vary fromsample to sample. A common 2h range [10�, 60�] was chosen forthe analysis, so to avoid pronounced variations in the backgroundoccurring at 2h < 10� and negligible features of the patterns at2h > 60�.

3.1.1. Contribution of crystalline celluloseSamples treated by manual and automatic grinding are shown

in Fig. S1: particle size and form of the fragments are very differentin the two cases. The presence of white flakes emerging from somefragments was observed in both cases. It was found that suchflakes, that could be separated from leaf fragments by further man-ual grinding, are composed of cellulose in its crystalline Ib form(Nishiyama, Langan, & Chanzy, 2002), as ascertained by fittingthe measured diffraction pattern from this sample with that calcu-lated from the Ib phase through the Rietveld approach (Fig. S2).

The agreement with the experimental pattern was obtained bystrongly enlarging the Ib peaks (i.e. by assigning large widthparameters to the profile functions associated to the Ib reflectionsin the framework of the Rietveld approach). This indicates thatthe coherent crystalline domains contained in the sample havelimited size in the scale of nanometers. Some sharp peaks of theexperimental pattern are not reproduced by the calculated pattern(in fact they are still present in the difference profile in Fig. S2).These peaks can be attributed to mineral components and the fact

Table 1Vine leaves samples used for the analyses.

Sample No. Cultivar Description

0, 1, 2 Victoria Fresh3 Sangiovese Fresh4, 5 Italia Fresh6, 7, 8, 9, 10 Italia Fresh (from organic management)11, 12 Superior Seedless Brined13 Thompson Seedless Brined14 Italia Brined

1910 A. Rizzuti et al. / Food Chemistry 141 (2013) 1908–1915

Author's personal copy

that they are narrower could be explained by the fact that theircrystalline domains are larger than those of cellulose.

The diffraction pattern from the separated crystalline cellulosewas compared with two representative diffraction patterns fromgrape leaf samples in Fig. S3. It emerges that the two representa-tive samples have narrow peaks not attributable to crystalline cel-lulose, some of them occurring in the same position and havingdifferent heights depending on the sample. Some of these peaksare also present in the spectrum of the fibre component. Moreover,the enlarged peaks of crystalline cellulose are not evident in thespectra of the representatives. This can be interpreted as follows:

– the peaks emerging from the background are due to the pres-ence of crystalline compounds of different nature and in differ-ent concentrations. Samples can be discriminated both by theirspecific positions and height;

– the small size of crystalline domains of cellulose and the smallfraction of crystalline cellulose in the sample does not allow theIb peaks to emerge from the background.

In order to verify this latter hypothesis, we performed a Rietveldanalysis aiming to assess the presence of the crystalline cellulosecomponent in the sample. It resulted that the Ib peaks almost com-pletely disappeared, indicating that the crystalline component isalmost entirely enclosed in the background of the sample. Thisindicated that it is not possible to use a Rietveld procedure to char-acterise the crystalline component of the samples, and that thepeaks of the samples emerging from the background are not dueto crystalline cellulose.

3.1.2. Contribution of backgroundIn view of the above considerations, two features are present in

the grape leaf diffraction patterns: the amount and shape of back-ground and the height and position of the peaks emerging from it.The two components can be separated by using background-sub-traction (BS) filtering. The results of the application of the BSalgorithm to sample no. 1 are shown in Fig. S3: the estimated back-ground (red line) is subtracted to the experimental pattern (blackline), to produce the BS pattern (blue curve), which represents afingerprinting of the original pattern. The SNIP algorithm was ap-plied by using an optimum value of 20 for the Nclip parameter(see Section 2.5.1).

3.1.3. Contribution of NaClPatterns of brined samples (No. 11–14) showed intense peaks at

the same characteristic 2h values (27.0�, 32.0�, 45.5�, 53.9� and56.6�, Fig. S4, green pattern). Such peaks were attributed to a NaClcrystal phase present in those samples as an effect of the brinedtreatment.

PCA analysis applied to XRPD patterns of fresh (samples 0–10)and brined leaves (samples 11–14) indicates that the presence ofNaCl in the samples causes a dramatic effect on diffraction pattern,certainly due to the perfection and size of the crystals. PCA analysiswas also used for XRPD data acquired from brined leaf samples(11–14) which were washed with large volumes of water. Theseadditional data were acquired with the aim to distinguish the con-tribution of NaCl contained in the leaf tissues from that of NaClpossibly remained on leaf surface. PC1/PC2 scores plot (Fig. S5)showed that the NaCl peaks dominate the data variability: the firstprincipal component explains 96.4% of the total variance and dif-ferentiates samples only for the presence or absence of NaCl, asshown in PC1 loading plot in Fig. S5. The second component ex-plains 3.3% of the total variance, and accounts for the variation inNaCl signals intensity. This finding suggests that the contributionof NaCl residues on leaf surface does not significantly affect thePCA output.

The effectiveness in grouping brined and fresh samples alongPC1 hampers PCA to appreciate the differences between fresh sam-ples. In order to improve the discriminating ability of XRPD datatowards fresh samples, and then to increase the fingerprinting per-formance of the technique, brined samples should be removedfrom the data matrix, or NaCl contribution should be removedfrom the pattern. We decided to follow this latter strategy, byassigning to all the channels lying under the NaCl peaks an inter-mediate value between those immediately before and after thepeak.

3.2. PCA applied to XRPD data

The NaCl-cleared diffractograms were submitted to PCA analy-sis. We found that the addition of the BS pre-processing performsbetter than the NORM pre-processing alone, or, in other words,using information from height and position of the peaks emergingfrom the background is preferable than using information aboutthe intensity and shape of the background. This can be explainedby the fact that the presence of NaCl also affects the shape of thebackground, therefore by subtracting it we completely eliminateany dependence on the NaCl content. Moreover, the backgroundmay be influenced by factors related to the experimental set-up(i.e., absorption of the sample, scattering with air, granulometryof the sample) and not characteristic of the sample.

The score plot of the BS patterns in the first two principal com-ponents, explaining, respectively, 49.4% and 16.4% of the total var-iance, is shown in Fig. 1. From the analysis of the loading plots ofthe first two principal components, it can be seen that the originof the variability is endowed in few, well defined diffraction peaks.In particular, peaks at 2h = 23�, 24�, 28�, and 36� are responsible forthe PC1 variation whereas peaks at 2h = 12�, 13� 15�, 24� and 35�are responsible for the differentiation along PC2. Principal compo-nents higher than 2 do not introduce relevant discriminationamong data points, as their loading vectors are noisier than thoseof the first two principal components. It is important to note thatexclusion of the NaCl contribution to XRPD patterns allows PCAto give information not related to brining. In fact, although the lim-ited number of the samples prevents a reliable discrimination onthe basis of agronomical practises, PCA performed on NaCl-cleareddiffractograms indicates a tendency of the samples to groupaccording to agronomical management. In particular, data pointscorresponding to samples of the same cultivar (Italia, sample 4–10 and 14) are differentiated in two groups according to agronom-ical management, and not to post-harvest treatment: brined andfresh conventional samples 4, 5 and 14 are differentiated from or-ganic samples 6–10. This results encourage further investigationon the discriminating potential of XRPD analysis.

3.3. Analysis of ESI-Q-TOF-MS spectra

High resolutions mass spectra, recorded in negative ion mode,show characteristic profiles consisting of peaks attributable to or-ganic acids, sugars and polyphenols contained in the grape leaves.

A typical MS spectrum of a fresh grape leaf extract is reported inFig. S6. It shows peaks deriving from ions having m/z of 72.99,87.01, 115.00, 133.01, 149.01, 179.04, 191.02, 195.05, 277.21,311.04, 341.11 and 477.06. The assignment of these ions was madeanalysing their mass, isotopic pattern and fragmentation patternby comparison with literature data and data available in on-linepublic databases (PubChem <http://pubchem.ncbi.nlm.nih.gov/>,Metlin <http://metlin.scripps.edu/> and Chemspider <http://www.chemspider.com>). The list of the identified compounds isreported in Table 2.

A typical MS(�) spectrum of brined grape leaves is reported inFig. S7. All ions detected in fresh grape leaves, except those of

A. Rizzuti et al. / Food Chemistry 141 (2013) 1908–1915 1911

Author's personal copy

linolenic acid (m/z 277.21) and of quercetin-3-glucuronide (m/z477.06), were absent in brined leaf samples which were character-ised by new peaks at m/z 92.92, 150.89, 210.85, 268.81, 326.77,384.73, 444.68, 502.64 and 560.60. These new signals are ascribedto the chloride adduct of sodium chloride clusters [(NaCl)nCl]�

(1 6 n 6 9).In the light of these results, it is evident that organic acids such

as pyruvic acid (87.01 m/z), fumaric acid (115.00 m/z), malic acid(133.01 m/z) and tartaric acid (149.01 m/z), along with sucrose(341.11 m/z), are consumed during brining whereas linolenic acid(277.21 m/z) and quercetin-3-glucuronide (477.06 m/z) resist fer-mentation conditions. This is interesting if one considers the ben-eficial effects of the last two metabolites. In fact, linolenic acid is anomega-3-fatty acid with beneficial effect on cardiovascular health(Burillo et al., 2012) and quercetin-3-glucuronide is a polyphenolicmetabolite identified as a novel intervention for Alzheimer’s dis-ease (Ho et al., 2013), thus rendering brined grape leaves an impor-tant food for a healthy diet.

3.4. PCA applied to ESI-TOF-MS data

PCA applied to MS data gave a clear distinction between brinedand fresh leaves with PC1 explaining 78.35% of the total variance.The ions mainly responsible for the separation of the samples arethose of the deprotonated form of fumaric acid (m/z 115.00), malicacid (m/z 133.01) and tartaric acid (m/z 149.01), characterising thefresh leaves, and of [NaCl+Cl]� (m/z 92.92), the cluster more abun-dant in brined leaves. Considering MS spectra devoid of peaksbelonging to all sodium chloride clusters, the discrimination ofthe brined samples is preserved in PC1/PC2 scores plot, but the

PC1 loadings plot indicates linolenic acid as the main compoundresponsible for the differentiation of brined samples (Fig. 2). With-out contribution of NaCl, PC1 explains 64.6% of the total variance.

Along PC2, explaining 20.9% of the total variance, samples 4and 5 are differentiated from the rest of the samples, includingthose of the same variety. In fact, samples 4, 5 and 6–10 representfresh leaves of cv. Italia, but differ for the agronomical manage-ment, being conventional in the first case (4 and 5) and organicfor the latter samples (6–10). According to PC2 loadings plot, con-ventional cv. Italia samples are characterised by a higher amountof sucrose (m/z 341.11) and a lower content of tartaric acid (m/z149.01) with respect to organic leaves of the same variety. Thisfinding referred to cv. Italia, parallels the trend of glucose andfructose quantities observed when passing from conventional toorganic grape berries of cv. Superior Seedless (Gallo et al., 2012).In grapevines, sucrose is produced as a result of photosynthesisin the leaf and, there is its transportation via phloem to the berry,where is cleaved to glucose and fructose by acid and neutralinvertases (Conde, Agasse, Glissant, Tavares, & Gerós Delrot,2006; Davies & Robinson, 1996; Wu, Liu, Guan, Fan, & Li, 2011).In the light of such a relationship between sucrose and its constit-uent monosaccharides, and since our results indicate that the or-ganic management results in a reduced content of sugars in leavesas well as in the berries, it can be concluded that discrimination oforganic productions may occur by fingerprinting of both leavesand grape berries (after collection of data from a proper numberof samples).

Along PC3, which explain 6.8% of the total variance, samples ofcv. Victoria differentiate from other samples for the higher amountof fumaric (115.00 m/z) and malic (133.01 m/z) acids.

Fig. 1. Score plots and loading vectors of the first three principal components of the XRPD patterns, after removal of NaCl contribution. The percentage of variance explainedby the principal components is reported on the axes. For the sake of clarity, scores are grouped according to the cultivar they belong to.

Table 2HRMS(�) peaks assignment for vine leaves.

ESI (�) TOF

Experimental accurate mass [M�H]� Formula for the detected [M�H]� Fragmentation ion mass (MS/MS) Assignment

87.0063 C3H3O3 – Pyruvic acid115.0028 C4H3O4 71.0124 [M�H�CO2]� Fumaric acid133.0132 C4H5O5 89.0243 [M�H�CO2]� 115.0054 [M�H�H2O]� Malic acid149.0077 C4H5O6 87.0084 [M�H�CO2�H2O]� 105.0201 [M�H�CO2]� Tartaric acid179.0413 C9H7O4 135.0459 [M�H�CO2]� Caffeic acid191.0189 C6H7O7 111.0106 [M�H�CO2�2H2O]� Citric acid195.0501 C6H11O7 153.0329 [M�H�2H2O]� Gluconic/galactonic acid277.2149 C18H29O2 – Linolenic acid311.0394 C13H11O9 149.0116 [Tart-H]� 179.0379 [caff-H]� Caftaric acid341.1055 C12H21O11 – Sucrose477.0643 C21H17O13 301.0391 [M�H�C6H9O6] Quercetin-3-glucuronide

Tart: tartaric acid; caff: caffeic acid.

1912 A. Rizzuti et al. / Food Chemistry 141 (2013) 1908–1915

Author's personal copy

3.5. Analysis of NMR spectra

NMR spectra of vine leaves (Fig. S8) show basically signals ofphenolic compounds (spectral region d 9.0–5.5), carbohydratesand organic acids (d 5.5–2.5) and amino and fatty acids (d 2.5–0.5). Attributions were made by comparison with spectra ofauthentic samples or with literature data (Ali et al., 2009, 2011;Fan, 1996; Figueiredo et al., 2008; Forveille et al., 1996; Mulaset al., 2011; Pereira et al., 2005; Pereira, Gaudillere, Pieri, et al.,2006; Pereira, Gaudillere, van Leeuwan, et al., 2006; Son, Hwang,Ahn, et al., 2009; Son, Hwang, Kim, et al., 2009).

Signal assignments in fresh vine leaves are reported in Table 3.In the aromatic region, the signals at d 7.74 (d), 7.57 (dd), 6.97 (d),6.50 (d), 6.30 (d) and 5.21 (d) are ascribable to the aromatic pro-tons of quercetin-3-O-glucuronide, and the signals at d 7.62 (d),7.15 (d), 7.06 (dd), 6.86 (d) and 6.41 (d) to the aromatic protonsof caffeic acid. Olefinic protons of fumaric acid are observed at d6.47 as a singlet. In the region for carbohydrate and organic acidprotons, the two doublets at 5.39 and d 4.16 are assigned to su-crose, while the doublet at d 5.17 and 4.57 are assigned to a andb anomeric protons of glucose. The singlet at d 4.40 is attributableto tartaric acid, the doublets of doublets at d 4.35, 2.81 and 2.59 areascribable to the protons of malic acid, and the two doublets at d2.83 and 2.74 are assigned to citric acid. In the high field region,the triplet of doublets at d 2.49 is due to glutamic acid, the singletat d 2.03 to acetic acid, the doublet at d 1.47 to alanine methyl

protons and the triplet at d 0.95 along with the overlapped multi-plets centred at d 1.22, to a-linolenic acid.

Brined samples showed simpler 1H NMR spectra in which thesignals of quercetin-3-O-glucuronide (d 7.74, 7.57, 6.97, 6.50,6.30 and 5.21) and of a-linolenic acid (d 0.95 and 1.22) could beclearly identified.

3.6. PCA applied to NMR data

As opposed to XRPD and MS data, 1H NMR spectra do not con-tain either information on NaCl or information about tartaric acid,since the signal was excluded due to the high variability in chem-ical shift. Nevertheless, PCA applied to NMR data (Fig. 3) gave re-sults very similar to those described above for MS analyses. Infact, as shown in the PC1/PC2 scores plot, along PC1 (explaining88.5% of the total variance) brined samples (11–14) are well sepa-rated from fresh samples, while leaves of cv. Italia belonging toconventional productions (4 and 5) are clearly distinguished alongPC2 (explaining 5.0% of the total variance). Leaf samples of cv. Vic-toria (0–2) are separated from the other samples along PC3. Load-ings plot indicates that brined samples are distinguishable alongPC1 due to the presence of linolenic acid (signals in the range d1.24–1.32) and the absence of sugars, which are completely lostduring brining. The samples of conventional cv. Italia leaves arewell separated from the other samples along PC2 due to the higheramount of sucrose. Finally, the contribution to discrimination of cv.Victoria samples along PC3 is given by glucose signals.

3.7. Covariance analyses

PCA results obtained with the three different analytical tech-niques indicate that NMR spectroscopy and MS spectrometry gavecomparable information and allow the identification of metabo-lites responsible for the differentiation of the samples.NaCl-cleared XRPD data gave different results in terms of samplesdiscrimination, allowing to better appreciating grouping based oncultivar and agronomical management. Assignment of the XRPDpeaks to specific molecules in a food matrix is not trivial. In orderto correlate XRPD data to metabolites contained in the vine leavesMS–XRPD and NMR–XRPD covariance analyses was performed. Infact, a covariance matrix results in a normal 2D spectrum contain-ing intra-molecular and inter-molecular cross-peaks relating thedata of the two different considered techniques. Such cross-peaksindicate correlated fluctuations of intensity throughout the ensem-ble of underlying spectra. Signals arising from the same molecule

Fig. 2. Scores plots and loading vectors of the first three principal components of the MS spectra, after removal of NaCl contribution. The percentage of variance explained bythe principal components is reported on the axes.

Table 31H NMR signal assignment of a typical fresh grape leaf extract.

1H NMR

Metabolite Chemical Shift (d)

a-Linolenic acid 0.95 (t), 1.22 (m)Alanine 1.48 (d)Acetic acid 2.03 (s)Glutamic acid 2.49 (td)Malic acid 4.35 (dd), 2.81 (dd), 2.59 (dd)Citric acid 2.83 (d), 2.74 (d)Tartaric acid 4.40 (s)Fumaric acid 6.47 (s)Sucrose 5.39 (d), 4.16 (d)a-Glucose 5.17 (d)b-Glucose 4.57 (d)Caffeic acid 7.62 (d), 7.15 (d), 7.06 (dd), 6.86 (d), 6.41 (d)Quercetin-3-O-

glucuronide7.74 (d), 7.57 (dd), 6.97 (d), 6.50 (d), 6.30 (d), 5.21(d)

A. Rizzuti et al. / Food Chemistry 141 (2013) 1908–1915 1913

Author's personal copy

are positively correlated because their intensities vary coherentlyto molecule concentration (or crystalline phase weight fraction,in case of XRPD). The inter-molecular cross-peaks indicate corre-lated fluctuations of different molecules and are positive whenfluctuations have the same sign and negative when fluctuationshave opposite sign.

Cross peaks obtained in MS–XRPD and NMR–XRPD covariancespectra are listed in Table S1. The set of XRPD patterns at 2h of13.3�, 24.3� and 36.4� (2hset1) is positively correlated to MS sig-nals of tartaric acid and negatively correlated to linolenic acidand sucrose. Moreover, the set at 2h of 12.4�, 21.4� and 43.8�(2hset2) is positively correlated to MS signals of linolenic acid.By NMR–XRPD covariance analysis, 2hset1 and 2hset2 are posi-tively correlated to a-glucose and linolenic acid, respectively.Again, 2hset1 is negatively correlated to linolenic acid and su-crose, thus suggesting the possibility to ascertain a decrease inthe amount of these two metabolites by the increase of theXRPD pattern at 13.3�, 24.3� and 36.4� revealed in leaf powderXRPD spectrum. On the other hand, the positive correlation of2hset2 with MS and NMR signals of linolenic acid might be usedas a tool for monitoring the increment of such an acid in thegrape leaves.

4. Conclusion

The characterisation of grape leaves for fingerprinting purposeswas carried out by using three different analytical techniques:XRPD, NMR and direct infusion MS. The three techniques gavecomplementary information, converging in the grouping of thesamples based on cultivars and agronomical practises, and in thediscrimination of brined and fresh leaves.

XRPD is very sensitive to the presence of crystalline mineralsin the sample. On the contrary, diffraction signal from the crys-talline fraction of cellulose does not contribute in discriminatingthe samples, as it is completely masked by the background. Thus,good performances in the XRPD characterisation are achievedprovided that cellulose and NaCl contributions to the spectrumare removed. The correlation of XRPD patterns to MS and NMRspectral data (covariance analysis) paves the way to the com-bined use of the three analytical techniques for fingerprintingpurposes, which is desirable to gain a deeper knowledge of foodproducts in terms of quality control. Moreover, MS–XRPD andNMR–XRPD covariance analyses can be considered a promisingtool for correlating the solid and the soluble fraction of theleaves.

Acknowledgements

We thank Italian Ministry of Education, University and Researchfor financial support in the framework of the announcement‘‘MIUR prot. No. 713/Ric. – 29/10/2010’’ – Project No.PON02_00186_2866121, Regione Puglia for financial support inthe framework of the announcement ‘‘Reti di Laboratori pubblicidi Ricerca’’ – Project No. 68 ‘‘Apulian Food Fingerprint’’, and Giov-anni Filograsso for valuable help.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.foodchem.2013.05.044.

References

Ali, K., Maltese, F., Fortes, A. M., Pais, M. S., Choi, Y. H., & Verpoorte, R. (2011).Monitoring biochemical changes during grape berry development inPortuguese cultivars by NMR spectroscopy. Food Chemistry, 124, 1760–1769.

Ali, K., Maltese, F., Zyprian, E., Rex, M., Choi, Y. H., & Verpoorte, R. (2009). NMRmetabolic fingerprinting based identification of grapevine metabolitesassociated with Downy mildew resistance. Journal of Agricultural and FoodChemistry, 57, 9599–9606.

Aliferis, K. A., & Jabaji, S. (2010). 1H NMR and GC–MS metabolic fingerprinting ofdevelopmental stages of Rhizoctonia solani sclerotia. Metabolomics, 6, 96–108.

Altomare, A., Burla, M. C., Giacovazzo, C., Guagliardi, A., Moliterni, A. G. G., Polidori,G., et al. (2001). Quanto: A Rietveld program for quantitative phase analysis ofpolycrystalline mixture. Journal of Applied Crystallography, 34, 392–397.

Bansal, P., Hall, M., Realff, M. J., Lee, J. H., & Bommarius, A. S. (2010). Multivariatestatistical analysis of X-ray data from cellulose: A new method to determinedegree of crystallinity and predict hydrolysis rate. Bioresource Technology, 101,4461–4471.

Bevilacqua, V., Triggiani, M., Gallo, V., Cafagna, I., Mastrorilli, P., & Ferrara, G. (2012).An expert system for an innovative discrimination tool of commercial tablegrapes. Lecture Notes in Computer Science – LNAI, 7390, 95–102.

Brun, R., & Rademakers, F. (1998). ROOT – An object. Oriented data analysisframework. Linux Journal, 51.

Burgess, D. D., & Tervo, R. J. (1983). Background estimation for gamma-rayspectroscopy. NIM, 214, 431–434.

Burillo, E., Mateo-Gallego, R., Cenarro, A., Fiddyment, S., Bea, A. M., Jorge, I., et al.(2012). Beneficial effects of omega-3 fatty acids in the proteome of high-densitylipoprotein proteome. Lipids in Health and Disease, 11, 116.

Caliandro, R., Di Profio, G., & Nicolotti, O. (2013). Journal of Pharmaceutical andBiomedical Analysis, 78–79, 269–279.

Casale, M., Armanino, C., Casolino, C., & Forina, M. (2007). Combining informationfrom headspace mass spectrometry and visible spectroscopy in theclassification of the Ligurian olive oils. Analytica Chimica Acta, 589, 89–95.

Casale, M., Casolino, C., Olivieri, P., & Forina, M. (2010). The potential of couplinginformation using three analytical techniques for identifying the geographicalorigin of Liguria extra virgin olive oil. Food Chemistry, 118, 163–170.

Colquhoun, I. J. (2007). Use of NMR for metabolic profiling in plant systems. Journalof Pesticide Science, 32, 200–212.

Fig. 3. Scores plots and loadings vectors of the first three principal components of the NMR spectra. The percentage of variance explained by the principal components isreported on the axes.

1914 A. Rizzuti et al. / Food Chemistry 141 (2013) 1908–1915

Author's personal copy

Conde, C., Agasse, A., Glissant, D., Tavares, R., & Gerós Delrot, S. (2006). Pathways ofglucose regulation of monosaccharide transport in grape cells. Plant Physiology,141, 1563–1577.

Davies, C., & Robinson, S. P. (1996). Sugar accumulation in grape berries. PlantPhysiology, 111, 275–283.

Ensikat, H. J., Boese, M., Mader, W., Barthlott, W., & Koch, K. (2006). Crystallinity ofplant epicuticular waxes: Electron and X-ray diffraction studies. Chemistry andPhysics of lipids, 144, 45–59.

Fan, T. W. M. (1996). Metabolite profiling by one- and two-dimensional NMRanalysis of complex mixture. Progress in Nuclear Magnetic ResonanceSpectroscopy, 28, 161–219.

Ferrara, G., Mazzeo, A., Matarrese, A. M. S., Pacucci, C., Pacifico, A., Gambacorta, G.,et al. (2013). Application of abscisic acid (S-ABA) to ‘Crimson Seedless’ grapeberries in a Mediterranean climate: Effects on color, chemical characteristics,metabolic profile and S-ABA concentration. Journal of Plant Growth Regulation.http://dx.doi.org/10.1007/s00344-012-9316-2 (in press).

Figueiredo, A., Fortes, A. M., Ferreira, S., Sebastiana, M., Choi, Y. H., Souza, L., et al.(2008). Transcriptional and metabolic profiling of grape (Vitis vinifera L.) leavesunravel possible innate resistance against pathogenic fungi. Journal ofExperimental Botany, 59(12), 3371–3381.

Forveille, L., Vercauteren, J., & Rutledge, D. N. (1996). Multivariate statisticalanalysis of two-dimensional NMR data to differentiate grapevine cultivars andclones. Food Chemistry, 57, 441–450.

Gallo, V., Mastrorilli, P., Cafagna, I., Nitti, G. I., Latronico, M., Romito, V. A., et al.(2012). Multivariate statistical analysis of 1H NMR data for evaluation ofmetabolic profile in commercial table grapes (Vitis vinifera): inter- vs intra-vineyard variability. Journal of Food Composition and Analysis (submitted forpublication).

Ho, L., Ferruzzi, M. G., Janle, E. M., Wang, J., Gong, B., Chen, T.-Y., et al. (2013).Identification of brain-targeted bioactive dietary quercetin-3-O-glucuronide asa novel intervention for Alzheimer’s disease. The Journal of the Federation ofAmerican Societies for Experimental Biology, 27, 769–781.

Kim, K. H., Choi, Y. H., & Verpoorte, R. (2010). NMR-based metabolomic analysis ofplants. Nature Protocols, 5, 536–549.

Kopka, J., Fernie, A., Weckwerth, W., Gibon, Y., & Stitt, M. (2004). Metaboliteprofiling in plant biology: Platforms and destinations. Genome Biology, 5,109.1–109.9.

Kos�ar, M., Küpeli, E., Malyer, H., Uylas�er, V., Türkben, C., & Hüsnü Can Bas�er, K.(2007). Effect of brining on biological activity of leaves of Vitis Vinifera L. (Cv.Sultani Çekirdeksiz) from Turkey. Journal of Agricultural and Food Chemistry, 55,4596–4603.

Lamikanara, O., & Kassa, A. K. (1999). Changes in the free amino acid compositionwith maturity of noble cultivar of Vitis rotundifolia Micho grapes. Journal ofAgricultural and Food Chemistry, 47, 4837–4841.

Morhác, M., Kliman, J., Matoušek, V., Veselsky, M., & Turzo, I. (1997). Backgroundelimination methods for multidimensional gamma-ray spectra. NIM, A401,113–132.

Mulas, G., Galaffu, M. G., Pretti, L., Nieddu, G., Mercenaro, L., Tonelli, R., et al. (2011).NMR Analysis of seven selections of Vermentino grape berry: Metabolitescomposition and development. Journal of Agricultural and Food Chemistry, 59,793–802.

Nicholson, J. K., & Lindon, J. C. (2008). Metabonomics. Nature, 455, 1054–1056.Nikolaou, N., Koukourikou, M. A., & Koragiannidis, N. (2000). Effect of various

rootstocks on xylem exudates cytokinins, nutrient uptake, and growth patternof grape vines Vitis Vinifera L. cv. Thompson seedless. Agronomie, 20,363–373.

Nishiyama, Y., Langan, P., & Chanzy, H. (2002). Crystal structure and hydrogen-bonding system in cellulose 1b from synchrotron X-ray and neutron fiberdiffraction. Journal of the American Chemical Society, 124, 9074–9082.

Pereira, G. E., Gaudillere, J. P., Pieri, P., Hilbert, G., Maucourt, M., Deborde, C., et al.(2006). Microclimate influence on mineral and metabolic profiles of grapeberries. Journal of Agricultural and Food Chemistry, 54, 6765–6775.

Pereira, G. E., Gaudillere, J. P., van Leeuwen, C., Hilbert, G., Lavialle, O., Maucourt, M.,et al. (2005). 1H NMR and chemometrics to characterise mature grape berries infour wine-growing areas in Bordeaux, France. Journal of Agricultural and FoodChemistry, 53, 6382–6389.

Pereira, G. E., Gaudillere, J. P., van Leeuwen, C., Hilbert, G., Maucourt, M., Deborde, C.,et al. (2006b). 1H NMR metabolite fingerprints of grape berry: Comparison ofvintage and soil effects in Bordeaux grapevine growing areas. Analytica ChimicaActa, 563, 346–352.

Picone, G., Mezzetti, B., Babini, E., Capocasa, F., Placucci, G., & Capozzi, F. (2011).Journal of Agricultural and Food Chemistry, 59, 9271–9279.

Rietveld, H. M. (1969). A profile refinement method for nuclear and magneticstructures. Journal of Applied Crystallography, 2, 65–71.

Robinette, S. L., Brüschweiler, R., Schroeder, F. C., & Edison, A. S. (2012). NMR inmetabolomics and natural products research: Two sides of the same coin.Accounts of Chemical Research, 45, 288–297.

Ryan, C. G., Clayton, E., Griffin, W. L., Sie, S. H., & Cousens, D. R. (1988). SNIP, astatistics-sensitive background treatment for the quantitative analysis of PIXEspectra in geoscience applications. NIM, B34, 396–402.

Satisha, J., Pooja Doshi, D., & Adsule, P. G. (2008). Influence of rootstocks onchanging pattern of phenolic compounds in Thompson Seedless grapes and itsrelation to the incidence of powdery mildew. Turkish Journal of Agriculture andForestry, 32, 1–9.

Scalbert, A. S., Brennan, L., Fiehn, O., Hankemeier, T., Kristal, B. S., van Ommen, B.,et al. (2009). Mass-spectrometry-based metabolomics: Limitations andrecommendations for future progress with particular focus on nutritionresearch. Metabolomics, 5, 435–458.

Schripsema, J. (2010). Application of NMR in plant metabolomics: Techniques,problems and prospects. Phytochemical Analysis, 21, 14–21.

Seger, C., Godejohann, M., Tseng, L.-H., Spraul, M., Girtler, A., Sturm, S., et al. (2005).LC–DAD–MS/SPE–NMR hyphenation. A tool for the analysis of pharmaceuticallyused plant extracts: Identification of isobaric iridoid glycoside regioisomersfrom Harpagophytum procumbens. Analytical Chemistry, 77(3), 878–885.

Son, H. S., Hwang, G. S., Ahn, H. J., Park, W. M., Lee, C. H., & Hong, Y. S. (2009).Characterisation of wines from grape varieties through multivariate statisticalanalysis of 1H NMR spectroscopic data. Food Research International, 42,1483–1491.

Son, H. S., Hwang, G. S., Kim, K. M., Ahn, H. J., Park, W. M., Van Den Berg, F., et al.(2009). Metabolomic studies on geographical grapes and their wines using 1HNMR analysis coupled with multivariate statistics. Journal of Agricultural andFood Chemistry, 57, 1481–1490.

Sumner, L. W., Duran, A. L., Huhman, D. H., & Smith, J. T. (2002). Metabolomics: Adeveloping and integral component in functional genomic studies of MedicagoTruncatula. In J. T. Romeo & R. A. Dixon (Eds.), Recent advances in phytochemistry(pp. 31–61). Oxford: Pergamon.

Sumner, L. W., Mendes, P., & Dixon, R. A. (2003). Plant metabolomics: Large-scalephytochemistry in the functional genomics era. Phytochemistry, 62, 818–836.

Wishart, D. S. (2008). Quantitative metabolomics using NMR. Trends in AnalyticalChemistry, 27, 228–237.

Wu, B. H., Liu, H. F., Guan, L., Fan, P. G., & Li, S. H. (2011). Carbohydrate metabolismin grape cultivars that differ in sucrose accumulation. Vitis, 50, 51–57.

A. Rizzuti et al. / Food Chemistry 141 (2013) 1908–1915 1915