lable at ScienceDirect

Plant Physiology and Biochemistry 86 (2015) 24e33

Contents lists avai

Plant Physiology and Biochemistry

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

Research article

Metabolite profiling elucidates communalities and differences in thepolyphenol biosynthetic pathways of red and white Muscat genotypes

Asfaw Degu a, b, Caterina Morcia c, Giorgio Tumino c, Uri Hochberg a, b, David Toubiana a, b,Fulvio Mattivi e, Anna Schneider f, Polina Bosca d, Luigi Cattivelli c, Valeria Terzi c,Aaron Fait b, *

a The Albert Katz International School, Beer-Sheva, Israelb The French Associates Institute for Biotechnology and Agriculture of Dryland, The Jacob Blaustein Institutes for Desert Research, Ben-Gurion University ofthe Negev, Sede Boqer Campus, 84990, Israelc Consiglio per la Ricerca e la Sperimentazione in Agricoltura, Genomics Research Centre, Via S. Protaso 302, 29017 Fiorenzuola d' Arda, PC, Italyd BOSCA S.p.A, Canelli, AT 14053, Italye IASMA Research and Innovation Center, Fondazione Edmund Mach, S. Michele a/Adige, I-38010, Italyf Consiglio Nazionale delle Ricerche, Istituto di Virologia Vegetale, Sezione Grugliasco, Torino 10095, Italy

a r t i c l e i n f o

Article history:Received 20 June 2014Accepted 13 November 2014Available online 15 November 2014

Keywords:MuscatFlavonoidsGrapevineMetabolite-profiling

* Corresponding author.E-mail address: fait@bgu.ac.il (A. Fait).

http://dx.doi.org/10.1016/j.plaphy.2014.11.0060981-9428/© 2014 Elsevier Masson SAS. All rights re

a b s t r a c t

The chemical composition of grape berries is varietal dependent and influenced by the environment andviticulture practices. In Muscat grapes, phenolic compounds play a significant role in the organolepticproperty of the wine. In the present study, we investigated the chemical diversity of berries in a Muscatcollection. Metabolite profiling was performed on 18 Moscato bianco clones and 43 different red andwhite grape varieties of Muscat using ultra-performance liquid chromatographyequadrupole time offlightemass spectrometry (UPLC-QTOF-MS/MS) coupled with SNP genotyping. Principle componentanalysis and hierarchical clustering showed a separation of the genotypes into six main groups, three redand three white. Anthocyanins mainly explained the variance between the different groups. Additionally,within the white varieties mainly flavonols and flavanols contributed to the chemical diversity identified.A genotype-specific rootstock effect was identified when separately analyzing the skin of the clones, andit was attributed mainly to resveratrol, quercetin 3-O-galactoside, citrate and malate.

The metabolite profile of the varieties investigated reveals the chemical diversity existing amongdifferent groups of Muscat genotypes. The distribution pattern of metabolites among the groups dictatesthe abundance of precursors and intermediate metabolite classes, which contribute to the organolepticproperties of Muscat berries.

© 2014 Elsevier Masson SAS. All rights reserved.

1. Introduction

Natural variety in fruit metabolism is a fundamental aspect ofcrop breeding used to develop novel strategies for fruit qualityenhancement. In tomato, exotic germplasm resources are beingexploited for the identification of agriculturally valuable traits orfor their potential use for metabolic engineering (Schauer et al.,2005). The screening of wild or natural varieties can be later usedto generate mapping populations to study the genetic regulation ofcomplex traits (Schauer et al., 2006). This approach was used to

served.

elucidate the genetic basis of fruit metabolism in several species(Schauer et al., 2006; Harel-Beja et al., 2010). In grape, the genomicresources available have increased significantly in recent years(Grimplet et al., 2009, 2011; Cipriani et al., 2008; Vezzulli et al.,2008; Denoeud et al., 2008; Cipriani et al., 2010; Laucou et al.,2011), and research efforts are dedicated to determine the geneticvariation among grape varieties and its impact on fruit quality andmetabolism (This et al., 2006). Yet, while extensive research hasbeen conducted on the metabolism of developing berries of singlevarieties (Deluc et al., 2007), only a few works focused on themetabolic diversity between berries of different varieties (Mattiviet al., 2006; Cantos et al., 2002; Dimitrovska et al., 2011). More-over, the distinguishing between the environmental and the

A. Degu et al. / Plant Physiology and Biochemistry 86 (2015) 24e33 25

genetic components and their interactions remains challenging(Vincent et al., 2007).

Muscat grapes are typically associated with a sweet floral flavor(Rib�ereau-Gayon et al., 1975), which is, in large part, defined by itslevels of geraniol, nerol and linalool (Gunata et al., 1985; Mateo andJim�enez, 2000). A general classification of Muscat-flavored varietiesbased on monoterpene concentration was proposed by Mateo andJimenez (Mateo and Jim�enez, 2000). Others characterized acollection of over 20 Muscat varieties for ampelographic-codifieddescriptors and for DNA polymorphism (Botta et al., 1999). Gener-ally, the two characterization methods consistently categorized thevarieties; nevertheless, in the white Muscat group, accessions witha different berry color showed the same DNA profile. Recently, there-sequencing of 1-deoxy-D-xylulose 5-phosphate synthase(VvDXS), in a study on 148 grape varieties, which included Muscat-flavored, aromatic and neutral accessions, mutants and non-aromatic offsprings, revealed the key role of this gene on Muscatdiversification (Emanuelli et al., 2010). The aforementioned studiesadd to the understanding of diversification of Muscat metabolismbetween varieties. Nevertheless a global analysis of the metabolicdiversity of Muscat grapes is lacking.

A collection of 18Moscato bianco clones and 43Muscat varietieswas established in Canelli, Piedmont (260 masl), grafted onto‘Kober 5 BB’ and ‘420 A’ rootstocks (Botta et al., 1999). The collectionwas intensively studied, and traits, including yield, percentage ofbunches affected by gray rot and quality, i.e., the sugar content, totalacidity, pH and terpene content, showed significant differencesamong the clones. Associations were eventually established be-tween sugars and terpene compounds, while no significant rela-tionship was noticed between total acidity, terpene content of themust and the percentage of Botrytis infection. Here, a UPLC-QTOF-MS-basedmetabolite profiling was used to characterize the varietaldiversity in flavonoids and nonflavonoids of grape origins. The re-sults are described on the background of the current knowledge ofgrape metabolism. Moreover, the metabolite analysis was coupledwith SNP genotyping, using the 20K Infinium array from Illuminaon a subset of the genotypes subjected to metabolic analysis.

2. Materials and ethods

2.1. Plant material and sampling

Samples were taken from a collection of Muscat grape varietiesand clones located in Canelli, Italy. A set of varieties used both forwine and table consumption were included (Appendix A e

Table A1). Sixty-one accessions were, examined, among which 18were clones of Moscato bianco collected in Piedmont (Italy) andgrafted either on Kober 5BB and 420A, 17 were synonyms or colormutants of Moscato bianco from different countries and theremaining 26 were different wine grape Muscat-flavored varietiesand their synonyms. Berry maturity was monitored using arefractometer at weekly intervals during maturation. Samples werecollected and measured for potential alcohol, titratable acidity, pH,berry weight and dimension over three years. All the samples werecollected when the berries reached about 12% potential alcohol andwere referred to by the date of designatedmaturity according to thewinery (BOSCA) regulation. Each year berry weight was averagedfrom eight randomly selected berries of each genotype. Total sol-uble solids (TSS) in oBrix and potential alcohol (% Vol) were assayedyearly from the juice of the pooled crushed berries using a digitalMaselli refractometer. Yearly titratable acidity measurements (g/Ltartaric acid equivalents) were performed according to the stan-dard procedures used in Guymon and Ough (Guymon and Ough,1962). All the above berry quality measurements were repeatedduring three growth seasons (2010, 2011 and 2012). The list of the

accessions included in the analyses, along with their country oforigin, rootstock, berry color, parentage and their use, is presentedin Appendix A (Table A1).

2.2. Metabolite extraction

Metabolic profiling was performed on berries from four inde-pendent replications (from different bunches) for each genotypeduring 2011 growing season. Each replicate included a pool ofberry-skin tissue from five berries separated from the pulp. Sam-ples were snap-freezed immediately with liquid nitrogen and keptat �80 �c until further analysis as described in Weckwerth et al.(2004). Skin tissue was dried in a lyophilizer, ground under pre-chilled holders and grinding beads using a RETCH-mill (RetschGmbh, 42787 Haan, Germany). A sample of 70 mg frozen tissuepowder was transferred to 2 mL Eppendorf tube, and metaboliteswere extracted in a pre-chilled methanol/chloroform/waterextraction solution (2.5/1/1 v/v/v). The mixture was then brieflyvortexed and centrifuged for 2 min at 14,000 RPM (microcentrifuge5417R), and the supernatant was decanted into the new tubes. Thesupernatant was mixed with 300 ml of chloroform and 300 ml ofUPLC grade water and then centrifuged at 14,000 RPM for 2 min.The upper water/methanol phase was transferred to UPLC vials forLC-MS analysis.

2.3. UPLCeMS analysis

For LC-MS analysis, 4 ml of extracted sample was injected onto aUPLC-QTOF-MS system equipped with an ESI interface (Waters Q-TOF Xevo™:Waters MS Technologies, Manchester, UK) operating innegative and positive ion mode. Chromatographic separation wascarried out on an Acquity UPLC BEH C18 column (100mm� 2.1 mm,1.7 mm). The column and autosampler were maintained at 40 �Cand 10 �C, respectively. During each sample running, the mobilephase comprised 95%water, 5% acetonitrile, 0.1% formic acid (phaseA), and 0.1% formic acid in acetonitrile (phase B). The solventgradient program was conditioned exactly as described previously(Hochberg et al., 2013). All analyses were acquired using leucineenkephalin for lock mass calibration to ensure accuracy andreproducibility, at a concentration of 0.4 ng L�1, in 50/50 of ACN/H2O with 0.1% v/v formic acid. The MS conditions were set essen-tially as described previously (Hochberg et al., 2013).

2.4. UPLC data processing

MassLynx software (Waters) version 4.1 was used as the systemcontrolling the UPLC and for data acquisition as described previ-ously (Hochberg et al., 2013). The raw data acquiredwere processedusing MarkerLynx application manager (Waters) essentially asdescribed previously (Hochberg et al., 2013). To verify metaboliteidentification, representative samples from each group were runusing the same instruments and under the same operating condi-tions at the metabolomics facility of the Edmund Mach Foundationin San Michele all'Adigee Italy, where an in-house standard librarydescribed in details in Arapitsas et al. (2012), Degu et al. (2014) wasused to validate metabolite annotation based on retention timeorder of commercial standards (Appendix A e Table A2), and me-tabolites were also identified based on a fragmentation patternsearched against the Chemspider metabolite database (http://www.chemspider.com/) and further confirmed (Appendix A e

Table A3) with previous works (S�anchez-Rabaneda et al., 2004;Liang et al., 2008; Iijima et al., 2008; Hanhineva et al., 2008;Monagas et al., 2003, 2005; Moco et al., 2006). Each metabolitemarkers identified with the waters markerlynx software were thennormalized to the total peak area in each run, on the internal

A. Degu et al. / Plant Physiology and Biochemistry 86 (2015) 24e3326

standards and initial tissue weight so that all samples could becompared in the same condition as described previously(Katajamaa and Ore�si�c, 2007).

2.5. Statistical analysis

The processed datawere transformed to log10 and subjected to aHierarchical Clustering Module (HCL) based on a Pearson correla-tion and principal component analysis (PCA) using MEV software(http://www. tm4.org/mev/). Based on PCA and HCL clustering, thedata were analyzed for metabolic variability between the identifieddifferent groups. The data were subjected to an ANOVA, and toascertain possible significant differences between the groups ofstudied Muscat grape varieties, a TukeyeKramer HSD of multiplecomparisons of mean values was applied at p < 0.05. Pairwisecorrelations to all annotated metabolites on red and white varietiesseparately were performed using the Pearson correlation algo-rithm. The corresponding p-values were computed using thecor.test function of R as previously described (Toubiana et al.,2012a).

2.6. Genotyping

For genotyping, 25 Muscat wine varieties from all the six groupswere considered. DNA samples obtained from the leaf tissues ofaccessions grown in Bosca field collection were hybridized to theGrapeReSeq_Illumina_20K_SNP_chip containing 15022 SNPs fromVitis vinifera genotypes and 4978 SNPs from Vitis species. The se-quences for the chip development were computed by IGA (http://www.appliedgenomics.org/) and URGI (http://urgi.versailles.inra.fr/). The filtered SNPs were submitted to Illumina to be scored bythe Assay Design Tool (https://icom.illumina.com). The hybridiza-tion step was performed with TraitGenetics (Gatersleben, Ger-many). The data related to 18,000 SNPs were processed with Rpackages, and a neighbor-joining (NJ) dendrogram, based onEuclidean distance, was obtained.

Fig. 1. Principal component analyses of 61 Muscat grape collections using 46 identified mrepresent average of four biological replicate. Picture of the bunch are representative pictu

3. Results and discussion

TheMuscat collection, previously analyzed for its morphologicaldiversity and which showed high variability (Botta et al., 1999), washere profiled for its chemical and genetic diversity using a UPLC-Q-TOF-MS-based protocol and SNP markers.

3.1. Juice and berry quality-related parameters

During sampling, berry quality traits, including berry diameter,berry weight, total acidity, pH, �Brix and probable alcohol level,were measured. The total acidity, pH and �Brix at harvest are in therange of values previously reported (Deluc et al., 2007; Bindonet al., 2008). During the period under investigation, no significantdifferences were observed in berry diameter, berry weight, pH,�Brix and probable alcohol across the collection (Appendix A e

Table A4). Significantly higher levels of titratable acidity wereseen in the red grape group R2 as compared to the white grapegroup W3 (Appendix A e Table A4).

3.2. UPLC-qTOF-MS-based clustering analysis and grouping ofaccessions

UPLC-qTOF-MS-based metabolic profiling detected 46 special-ized metabolites and few central metabolites, including sugars,malate, tartrate, citrate, and the shikimate derived amino acids Pheand Trp.

Principle component analysis of the metabolite dataset sepa-rated the Muscat collection into defined clusters (Fig. 1). The PCAand HCL clustering analysis, used to elucidate the metabolic dif-ferentiation in the Muscat collection, yielded generally comparableresults. The first three principal components explained 46.29, 13.67and 7.3% of the variance within the dataset, respectively (AppendixA e Table A5). The variance explained by the first principalcomponent (PC1) was mainly contributed by the anthocyaninmalvidin-3-O-glucoside, peonidin 3-O-(600-p-coumaroyl-gluco-side), malvidin 3-O-(600-p-coumaroyl-glucoside), petunidin 3-O-

etabolites. Each number represent vine code number in the field and each data pointre from each group. Triangular symbols represent clones in the groups.

A. Degu et al. / Plant Physiology and Biochemistry 86 (2015) 24e33 27

glucoside, peonidin 3-O-glucoside, delphinidin 3-O-glucoside,malvidin 3-O-(600-acetyl-glucoside), cyanidin 3-O-(600-p-cou-maroyl-glucoside), petunidin 3-O-(600-p-coumaroyl-glucoside),peonidin 3-O-(600-acetyl-glucoside), cyanidin-3-O-glucoside petu-nidin 3-O-(600-caffeoyl-glucoside). Quercetin 3-O-galactosidecontributed the most to the separation along the second principalcomponent (PC2) (Appendix A e Table A6). Projection of themetabolic dataset of the accessions on the two-dimensional planedefined by the first two principal components revealed five distinctgroups (Fig. 1). Further projection of the data set on the second andthird principal components clearly separated the white varieties intwo groups (W2 and W3 in Appendix B e Fig. B1). Cyanidin-3-O-glucoside, procyanidin dimer B1, peonidin 3-O-glucoside, p-

Fig. 2. Hierarchical clustering result for the 61 Muscat varieties/clones based on berry skinmetrics.

coumaroyl-tartrate, quercetin 3-O-glucoside, myricetin and vitisinlargely contributed to the separation of these two groups along PC3(Appendix A e Table A6). In summary, based on the LC-MSmetabolite profiling, the Muscat collection consisted of six groups(white 1, white 2, white 3, red 1, red 2 and red 3) with characteristicmetabolic features.

HCL (Fig. 2) shows a clustering similar to the one suggested bythe PCA, except for the affiliation of Tamiioasa de Bohotin andRouge Beaumes Venise to the R2 and the different clustering ofMuscat de la miaut SCA, which was part of W3 in the PCA. Thismight arise from the fact that the methods used different algo-rithms, yet both are based on similarity measures (Sneath andSokal, 1973). The HCL analysis produced six distinct clusters, each

metabolite profile. The clustering was performed using Pearson Correlation distance

A. Degu et al. / Plant Physiology and Biochemistry 86 (2015) 24e3328

characterized by qualitatively and quantitatively different meta-bolic features (Fig. 2). Genotypes in the R3 group were mainlycomposed of the glucoside forms of anthocyanin and lacked acet-ylated and coumarylated anthocyanins. The relative pattern ofmetabolite abundance in R3 is comparable to that of the W1 groupfor most of the metabolites. This similarity is shown in the PCA,where R3 is located in between the two red groups and the whiteclusters, and in the HCL analysis, where R3 clusters with W1. Thismight suggest that R3 are intermediate genotypes between thewhite and red grape clusters in respect to polyphenol metabolism(Clarke and Bakker, 2004). The R3 group includes Muscat rose INRAand Muscat de Frontignan rouge SCA, two berry color mutations ofMoscato bianco with pink and red berries used for producingMuscat-flavored pale ros�e/white wines and ros�e wines,respectively.

It should be noted that the evaluated genotypes originated fromdifferent countries of Europe and the USA (Appendix A e Table A1).Furthermore two different rootstocks (Kober and 420 A) were usedfor a number of Moscato bianco clones (Appendix A e Table A1).Apparently, neither the PCA grouping nor the HCL clusteringanalysis showed appreciable effects of either factor (rootstock andgeography). Nevertheless, when separately analyzing the clones inthe collection, an effect largely attributed to the rootstock wasidentified, in particular for clones 171, 4 and 57 (PCA in Appendix Be Fig. B2). In these plants, the metabolites contributing the most tothe separation were quercetin 3-O-glucoside, resveratrol, andprocyanidin dimer B1 (Appendix A e Table A7). Other clones didnot show a rootstock component in their metabolite profiles, e.g.,132 and 16. Taken together, these results suggest that the effect of

Fig. 3. Relative amount of hydroxycinnamate and flavonols in the berry skin of Muscat collecfollowed by same latter are not significantly different (p < 0.05).

rootstock on Muscat berries’ specialized metabolism was limitedand genotype specific.

3.3. Organic acids and the shikimate pathway metabolism

The progressive decline in fruit acidity during the post-veraisonstage of berry development derives from the degradation of car-boxylates (Etienne et al., 2013; Sweetman et al., 2009). Thecomposition of organic acids has a substantial effect on theorganoleptic characteristics of the fruit (Esteves et al., 2004). In thepresent study, no significant differences among the differentMuscat groups were observed in identified organic acids (AppendixA e Table A8). However, on average, the white genotypes (W1, W2and W3 together) contained relatively higher organic acids thanthe red genotypes (Appendix A e Table A8). For example, amongthe whites, 190K had the highest relative content of all threeidentified organic acids, and compared to the red varieties, it hadfold changes up to 2.68, 2.27 and from 1.9 to 5.09 for malate,tartrate and citrate, respectively (Appendix A e Table A8). Thedifferences between white and red varieties in the accumulation ofcarboxylates suggest that anthocyanin metabolism, in the latervarieties, can lead to the enhanced utilization of carbon pools at theexpenses of organic acids.

The shikimate pathway links carbohydrate metabolism to thedownstream flavonoid pathway via the aromatic amino acids, e.g.Phe (Bar-Akiva et al., 2010; Tzin and Galili, 2010). Among theMuscat groups identified, W1 had the highest amount of Phe, fol-lowed by W3 and R3, while in the W2 and R2 clusters Phe wassignificantly (p < 0.05) lower in abundance (Appendix B e Fig. B3).

tions in the red and white groups. Values are mean of each group þ SE and mean values

A. Degu et al. / Plant Physiology and Biochemistry 86 (2015) 24e33 29

A similar trend was also observed for the level of Trp (Appendix Be

Fig. B3). Generally, the white had higher Phe than the red varieties.This might suggest a reduced partitioning of Phe to the down-stream flavonoid pathways in the white varieties, likely due to thelack of anthocyanin. Chassy et al. (Chassy et al., 2014) also reportedrecently the higher incorporation of labeled 13C phenylalanine intoanthocyanins than other classes of flavonoids in grape berries. Thathaving said, in the present study only peonidin showed a significantnegative correlation with Phe in the red group while no significantcorrelation was found between Phe and resveratrol in either red orwhite groups.

3.4. Polyphenol metabolism

Hydroxycinnamic acids are precursors of non-flavonoid poly-phenol stilbenes, lignins and of flavonoid intermediates (Strack,1997; Hichri et al., 2011). Four hydroxycinnamic acids were iden-tified in all varieties/clones, including p-coumarate, which is thecommitting metabolite produced in the pathway (Fraser and

Fig. 4. Relative amount of querecitine-3-O-galactoside in the skin of different Muscat grape cgroups in red (R) and white (W) Muscat.

Chapple, 2011) (Fig. 3), coumarate hexoside, p-coumaroyl-tartrateand ferulate, p-coumaroyl-tartrate being the dominant metabolite(Appendix Ae Table A8). All groups exhibited comparable amountsof p-coumarate and coumarate hexoside, while all three of thewhite groups contained higher ferulate contents than the red va-rieties. The W2 group contained significantly lower (p < 0.05) p-coumaroyl-tartrate levels than those of the other five groups(Appendix A e Table A8). Three stilbene compounds were identi-fied, namely resveratrol, piceid and d-viniferin. On average, andsurprisingly when compared to the current knowledge on stilbenedistribution in grape varieties (Bavaresco et al., 2009), the whitevarieties had higher resveratrol amounts than the red varieties.Varieties in the R1 group had the highest resveratrol and d-vin-iferin levels (Appendix A e Table A8).

Extensive variability across the collection was shown for otherpolyphenols. Several varieties, notably Moscato di Scanzo, werequite distinct from the rest of the varieties, exhibiting higherabundance in most of the identified polyphenols. Most grapephenolics are localized in the berry skin and released into the wine

ollections. Values presented are mean þ SE (n ¼ 4) and colors with letters represent the

A. Degu et al. / Plant Physiology and Biochemistry 86 (2015) 24e3330

during the process of maceration. These metabolite classescontribute to the flavor, color, antioxidant properties and astrin-gency of the final wine. A higher accumulation of berry skin fla-vonoids implies an increased concentration of these metabolites inthe wine (Kennedy et al., 2002). In this regard, the selection ofberry-skin-phenolic-rich Muscat varieties with acceptable senso-rial flavors may have a profound impact in diversifying the existingMuscat wine. For example, large variability was observed in thelevels of the derived naringenin glucoside (Fig. 3), a precursormolecule for the production of flavonoid metabolic classes (Heet al., 2010). The R1 group had a fifth the amount of naringeninglucoside than the R2, and ~6 times lower than the W2 groups(Fig. 3). Generally, the R1 group, however, accumulated relativelyhigher levels of flavanols (compared to the other two red groups),anthocyanins and other flavonoids, compared to the other groups.When taken together, these results suggest an enhanced rate ofconversion of naringenin chalcone towards the downstream fla-vanol and anthocyanin pathway. Roubelakis-Angeakis and Kliewer(Roubelakis-Angelakis and Kliewer, 1986) suggested an indepen-dent regulation of these two flavonoid classes, while sharing thesame precursor, dihydroflavonols, in the flavonoid pathway (Koeset al., 1994).

3.4.1. Quercetin 3-O-galactoside differentiates between whiteMuscat varieties

Most striking among the changes in the flavonoid-derived me-tabolites was a high varietal fluctuation in the relative amount ofquercetin 3-O-galactoside (Fig. 4), which contributed the most tothe separation of W1 and W2 from the W3 clusters (Fig. 1) and R2and R3 from the R1 groups. Quercetin 3-O-galactoside belongs tothe family of flavonoid O-glycosides that occurs widely in plants,e.g., in apple peel and Hypericum perforatum (St John's wort). It is amedicinally active compound that can be isolated from severalplant species, including Drosera rotundifolia, Prunella vulgaris,Rumex acetosella and from St John's wort (Li et al., 2005). It has beensuggested as a lipid and protein protectant in renal and neuralmodel systems against supra-lethal (0.25 kGy) gamma radiation(Chawla et al., 2005). The relative abundance of quercetin 3-O-galactoside varied from 10 to 400 in theW1 andW2, and from 10 to50 in the R2 and R3; in R1 and W3, this compound was below thedetection level. The level of quercetin, on the other hand, wassignificantly higher in the W3 group compared to the other groups(Fig. 3), suggesting a reduced conversion rate of quercetin in theW3

Fig. 5. Relative amount of different anthocyanins in the skin of red Muscat grape groups. Vsignificantly different (p < 0.05).

Muscat cluster. In our analysis, we could distinguish between theglucoside form, predominantly stored in plants (Hollman et al.,1997), and the galactoside form of quercetin (Appendix B e

Fig. B4). The former did not show significant differences in abun-dance between varieties.

Other flavonol classes displayed significant varietal differences.The R1 and R2 groups had significantly higher levels of myricetinand its glucuronide form compared to the W groups and to R3.Ivanova et al. (Ivanova et al., 2011) reported a lack of myricetin inthe skins of non-Muscat white grape varieties, Smederevka andChardonnay, compared to the red grape varieties Vranec andMerlot. Similarly, Montealegre et al. (Rodriguez Montealegre et al.,2006) also reported an absence of myricetin glucuronide andglucoside in the skin of six non-Muscat white grape varieties. Takentogether the evidences suggest that the metabolic pathway mightbe highly active in red-skinned grape varieties and be under aunique regulation in Muscat white grape varieties. Further work isneeded to dissect the regulation of the involved enzymes indifferent Muscat and non Muscat varieties.

The degree of variability between W1 and W2 groups wasestimated as marginal, showing a similar pattern of metaboliteaccumulation. However, the levels of kaempferol-3-glucuronideand rutin distinguished between the two groups, being higher inW2 than in W1 (Fig. 3). On average, the white skin genotypes hadhigher flavanols, including catechin, epicatechin and procyanidin-O-gallate, but not myricetin (Appendix Ae Table A8). Flavonols andanthocyanins share a commonprecursor, i.e. dihydroflavonols, withtheir respective enzymes, flavonol synthase and dihydroflavonol 4-reductase, in the general flavonoid pathways (Gou et al., 2011).Hence, a competition might exist between these two classes ofmetabolites in red grapes. Flavonols impart a yellow pigment andcontribute directly to the color of white wine (Boulton, 2001) andare also involved in copigmentation in red wines (Castillo-Mu~nozet al., 2007). Flavonols and flavanols are groups of flavonoids andshare the same precursor molecules with anthocyanin. The higherflavonols and flavanols in the white varieties, as compared to thered varieties, suggest an enhanced metabolic partitioning in thewhite varieties at the expense of anthocyanin accumulation.

3.4.2. AnthocyaninsAnthocyanins are the main flavonoids in red skin varieties, and

their production is developmentally dependent (Boss et al., 1996).The mono-glucosides, acetyl glucosides and coumaroyl glucosides

alues are mean of each group þ SE and mean values followed by same letter are not

Fig. 6. Dendrogram showing leaf genetic markers relationship of Muscat grape col-lections based on 20K SNP.

A. Degu et al. / Plant Physiology and Biochemistry 86 (2015) 24e33 31

forms of the five anthocyanidins (cyanidin, delphinidin, petunidin,peonidin, andmalvidin) were, as expected, identified in the skins ofthe red grape varieties. In agreement with the previous report(Ivanova et al., 2011; Wulf and Nagel, 1978), the mono-glucosideforms of anthocyanins were present in a higher abundance thanthe other forms, regardless of the groups. The malvidin and peo-nidin anthocyanin forms were found in a higher abundance thanthe cyaniding, delphinidin, and petunidin forms (Fig. 5).

Morphologically, the three red groups vary in the extent of berrycolor, R1 being dark red, R3 a light red and R2 displaying an in-termediate color intensity (Fig. 1). Luo and He (Luo and He, 2004)stated the direct relationship of berry skin color to anthocyanincontent. Accordingly, the R1 group possessed a higher amount(p < 0.05) of most of the anthocyanins than the other two groups.On the other hand, R3 possessed only the mono-glucoside antho-cyanin forms and the lowest amount of all anthocyanins, except forcyanidin glucoside (Fig. 5), which was significantly higher than thatin R2 and R1 (Fig. 5) These results, together with the increasedcontent of W-specific dimers (procyanidin B1, B2 and B3, AppendixA e Table A8) in R3, suggest the occurrence of an intermediatemetabolic regulation between the white and the red groups. Inaddition, the higher accumulation of procyanidin dimers in R3, atthe expense of acetylated and acylated anthocyanins, suggests acompetition between these two groups for the same precursors(Leucocyanidin) in the flavonoid pathway.

3.5. Integrated analysis to investigate metabolite accumulation

Correlation-based analysis has been used to study co-regulationbetween metabolic pathways and underlying regulatory mecha-nisms (Urbanczyk-Wochniak et al., 2007). The existence of corre-lations between groups of metabolites was suggested to reflect theoccurrence of transcriptional or post-translational events, meta-bolic regulation (feedback, feed-forward), and allosteric regulation(Toubiana et al., 2012a; Steinhauser et al., 2008).

To evaluate the interactions between metabolites and their co-ordinated changes across varieties, a correlation analysis was per-formed for all pairs of metabolites on red and white varietiesseparately, using the approach described in Toubiana et al.(Toubiana et al., 2012b). There were highly significant negativecorrelations of tartrate (r ranging from�0.74 to�0.92), procyanidindimer B1 & B2 (r ¼ �0.72 to �0.82) and catechin (r ¼ �0.69to �0.84) with anthocyanins in the red genotype group (AppendixA e Table A9). Myricetin and D-viniferin exhibited highly signifi-cant positive correlations with anthocyanins (r ¼ 0.74 to 0.92 andr ¼ 0.73 to 0.82, respectively) in the red genotype group (AppendixA e Table A9). The white group, on the other hand, showed weakerand non-significant correlations between most of the metabolites(Appendix A e Table A10) suggesting a tighter coordination at theinterplay between the central and specialized metabolisms in thered varieties than in the white varieties. Previously, it was reportedthat sugars are initial precursors for the synthesis of both tartrate(Saito and Kasai, 1978) and anthocyanins (Hrazdina et al., 1984;Teusch et al., 1987). The negative correlation of tartrate with theanthocyanins in the red genotypes might suggest the mobilizationof carbon towards the anthocyanin pathway, and the possiblecompetition between the parallel biosynthesis pathways ofanthocyanin against procyanidin dimer B1, procyanidin dimer B2and catechin.

To investigate a possible genetic basis for the variability insecondary metabolite abundance across the collection, the geneticvariability was assessed by subjecting 25 varieties to theGrapeReSeq_Illumina_20K_SNP_chip-based analysis for singlenucleotide polymorphism. A significant genetic variability wasobserved between varieties (Fig. 6). Nevertheless, a few varieties

clustered together e for example, Moscato bianco cn4, Muscat vi-olet cyperus, Muscat de Frontognan rouge, Muscat rouge deBaumes de Venice and Muscat rose-that are all strictly related eachother. A subset of SNPs, located on gene sequences related to ter-penes, flavonoids, flavones and flavonols, starting from thoseincluded in the 20K SNP array, was further investigated. A DAPCanalysis was performed, integrating the groups found via metab-olite profiling. No clear overlap between genetic and metabolicclustering was found. Comparing SNP genetic mapping to metab-olite profiling, the latter better resolved the variability existing inthe Muscat collection. For example, the white genotype (MuscatoBianco) was closely related to the red genotypes in the SNP arraygenetic mapping, while grouping based on metabolite profilingmade a clear separation between the red and white genotypes.Taken together, these results suggest that the basis for the meta-bolic variability identified in the Muscat collection studied may befound at the transcriptional or post-transcriptional level. More

A. Degu et al. / Plant Physiology and Biochemistry 86 (2015) 24e3332

detailed genetic mapping is required for a better resolution of thecollection genetics.

In conclusion, the LC-MS based metabolite profiling was shownto separate the genotypes of Muscat collections into six distinctgroups, three reds and three whites. The level of quercetin 3-O-galactoside in the whites and the form of anthocyanin in the reddefined the varietal affiliation. The higher overall averages of fla-vonols and flavanols of the white varieties, as compared with thered genotypes, likely originated from the lack of anthocyanin in thewhite genotypes. The extensive metabolite variability not repre-sented at the genetic level, according to the genotyping methodused, corroborate the use of metabolite profiling in protocols forvarietal identification and tracing.

Authors' contribution

AD conceived and conducted the experiment, analyzed LC-MSdata and wrote the body of the paper with AF; CM and GT per-formed sampling and berry quality related measurements, UHhelped running the experiment, processed the samples formetabolite profiling, DT helpedwith the data analysis, FM helped invalidating metabolite annotation at FEM, AF conceived and coor-dinated the project together with AS, PB, LC and VT. All authorsreviewed, edited and approved the final version of the manuscript.

Notes

The authors declare that they have no competing financialinterest.

Acknowledgments

We thank BOSCA winery for allowing us sampling their Muscatcollection and for their financial support. The study was supportedby the Eureka funded ItalyeIsrael cooperation «MetagenecodingBosca», the Italian Ministry of Foreign Affairs and the Israeli Min-istry of Industry, Trade and Labor (grant number 87174721).

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.plaphy.2014.11.006.

References

Arapitsas, P., Scholz, M., Vrhovsek, U., Di Blasi, S., Bartolini, A.B., Masuero, D.,Perenzoni, D., Rigo, A., Mattivi, F., 2012. A metabolomic approach to the study ofwine micro-oxygenation. PloS one 7 (5), e37783.

Bar-Akiva, A., Ovadia, R., Rogachev, I., Bar-Or, C., Bar, E., Freiman, Z., Nissim-Levi, A.,Gollop, N., Lewinsohn, E., Aharoni, A., Weiss, D., Koltai, H., Oren-Shamir, M.,2010. Metabolic networking in Brunfelsia calycina petals after flower opening.J. Exp. Bot. 61 (5), 1393e1403.

Bavaresco, L., Fregoni, C., de Macedo Basto, MI van Zeller, Vezzulli, S., 2009. Phys-iology & molecular biology of grapevine stilbenes: an update. In: Anonymous(Ed.), Grapevine molecular Physiology & Biotechnology. Springer, pp. 341e364.

Bindon, K., Dry, P., Loveys, B., 2008. Influence of partial rootzone drying on thecomposition and accumulation of anthocyanins in grape berries (Vitis viniferacv. Cabernet Sauvignon). Aust. J. Grape Wine Res. 14 (2), 91e103.

Boss, P.K., Davies, C., Robinson, S.P., 1996. Expression of anthocyanin biosynthesispathway genes in red and white grapes. Plant Mol. Biol. 32 (3), 565e569.

Botta, R., Akkak, A., Schneider, A., 1999. Caratterizzazione di Moscati in collezione aCanelli mediante marcatori molecolari: analisi di sinonimie, omonimie e rela-zioni di parentela. Quad. Vitic. Enol. Univ. Torino 23, 33e44.

Boulton, R., 2001. The copigmentation of anthocyanins and its role in the color ofred wine: a critical review. Am. J. Enol. Vitic. 52 (2), 67e87.

Cantos, E., Espin, J.C., Tom�as-Barber�an, F.A., 2002. Varietal differences among thepolyphenol profiles of seven table grape cultivars studied by LC-DAD-MS-MS.J. Agric. Food Chem. 50 (20), 5691e5696.

Castillo-Mu~noz, N., G�omez-Alonso, S., García-Romero, E., Hermosín-Guti�errez, I.,2007. Flavonol profiles of Vitis vinifera red grapes and their single-cultivarwines. J. Agric. Food Chem. 55 (3), 992e1002.

Chassy, A.W., Adams, D.O., Waterhouse, A.L., 2014. Tracing phenolic metabolism inVitis vinifera berries with 13C6-phenylalanine: implication of an unidentifiedintermediate reservoir. J. Agric. Food Chem. 62, 2321e2326.

Chawla, R., Arora, R., Sagar, R.K., Singh, S., Puri, S.C., Kumar, R., Singh, S.,Sharma, A.K., Prasad, J., Khan, H.A., 2005. 3-O-b-D-galactopyranoside of quer-cetin as an active principle from high altitude Podophyllum hexandrum andevaluation of its radioprotective properties. Z. Nat. C Biosci. 60 (9e10),728e738.

Cipriani, G., Marrazzo, M.T., Di Gaspero, G., Pfeiffer, A., Morgante, M., Testolin, R.,2008. A set of microsatellite markers with long core repeat optimized for grape(Vitis spp.) genotyping. BMC Plant Biol. 8 (1), 127.

Cipriani, G., Spadotto, A., Jurman, I., Di Gaspero, G., Crespan, M., Meneghetti, S.,Frare, E., Vignani, R., Cresti, M., Morgante, M., 2010. The SSR-based molecularprofile of 1005 grapevine (Vitis vinifera L.) accessions uncovers new synonymyand parentages, and reveals a large admixture amongst varieties of differentgeographic origin. Theor. Appl. Genet. 121 (8), 1569e1585.

Clarke, R.J., Bakker, J., 2004. Wine Flavour Chemistry. Wiley Online Library.Degu, A., Hochberg, U., Sikron, N., Venturini, L., Buson, G., Ghan, R., Plaschkes, I.,

Batushansky, A., Chalifa-Caspi, V., Mattivi, F., 2014. Metabolite and transcriptprofiling of berry skin during fruit development elucidates differential regula-tion between Cabernet Sauvignon and Shiraz cultivars at branching points inthe polyphenol pathway. BMC Plant Biol. 14 (1), 188.

Deluc, L.G., Grimplet, J., Wheatley, M.D., Tillett, R.L., Quilici, D.R., Osborne, C.,Schooley, D.A., Schlauch, K.A., Cushman, J.C., Cramer, G.R., 2007. Transcriptomicand metabolite analyses of Cabernet Sauvignon grape berry development. BMCGenomics 8 (1), 429.

Denoeud, F., Aury, J., Da Silva, C., Noel, B., Rogier, O., Delledonne, M., Morgante, M.,Valle, G., Wincker, P., Scarpelli, C., 2008. Annotating genomes with massive-scale RNA sequencing. Genome Biol. 9 (12), R175.

Dimitrovska, M., Bocevska, M., Dimitrovski, D., Murkovic, M., 2011. Anthocyanincomposition of Vranec, Cabernet Sauvignon, Merlot and Pinot Noir grapes asindicator of their varietal differentiation. Eur. Food Res. Technol. 232 (4),591e600.

Emanuelli, F., Battilana, J., Costantini, L., Le Cunff, L., Boursiquot, J., This, P.,Grando, M., 2010. A candidate gene association study on muscat flavor ingrapevine (Vitis vinifera L.). BMC Plant Biol. 10 (1), 241.

Esteves, V.I., Lima, S.S., Lima, D.L., Duarte, A.C., 2004. Using capillary electrophoresisfor the determination of organic acids in Port wine. Anal. Chim. Acta 513 (1),163e167.

Etienne, A., G�enard, M., Lobit, P., Mbegui�e-A-Mb�egui�e, D., Bugaud, C., 2013. Whatcontrols fleshy fruit acidity? A review of malate and citrate accumulation infruit cells. J. Exp. Bot. 64 (6), 1451e1469.

Fraser, C.M., Chapple, C., 2011. The Phenylpropanoid Pathway in Arabidopsis, vol. 9.The Arabidopsis Book/American Society of Plant Biologists.

Gou, J., Felippes, F.F., Liu, C., Weigel, D., Wang, J., 2011. Negative regulation ofanthocyanin biosynthesis in Arabidopsis by a miR156-targeted SPL transcrip-tion factor. Plant Cell Online 23 (4), 1512e1522.

Grimplet, J., Cramer, G.R., Dickerson, J.A., Mathiason, K., Van Hemert, J., Fennell, A.Y.,2009. VitisNet:“Omics” integration through grapevine molecular networks.PloS one 4 (12), e8365.

Grimplet, J., Dickerson, J., Adam-Blondon, A., Cramer, G., 2011. Bioinformatics Toolsin Grapevine Genomics.

Gunata, Y., Bayonove, C., Baumes, R., Cordonnier, R., 1985. The aroma of grapes I.Extraction and determination of free and glycosidically bound fractions of somegrape aroma components. J. Chromatogr. A 331, 83e90.

Guymon, J., Ough, C., 1962. A uniform method for total acid determination in wines.Am. J. Enol. Vitic. 13 (1), 40e45.

Hanhineva, K., Rogachev, I., Kokko, H., Mintz-Oron, S., Venger, I., K€arenlampi, S.,Aharoni, A., 2008. Non-targeted analysis of spatial metabolite composition instrawberry (Fragaria � ananassa) flowers. Phytochemistry 69 (13), 2463e2481.

Harel-Beja, R., Tzuri, G., Portnoy, V., Lotan-Pompan, M., Lev, S., Cohen, S., Dai, N.,Yeselson, L., Meir, A., Libhaber, S., 2010. A genetic map of melon highly enrichedwith fruit quality QTLs and EST markers, including sugar and carotenoidmetabolism genes. Theor. Appl. Genet. 121 (3), 511e533.

He, F., Mu, L., Yan, G., Liang, N., Pan, Q., Wang, J., Reeves, M.J., Duan, C., 2010.Biosynthesis of anthocyanins and their regulation in colored grapes. Molecules15 (12), 9057e9091.

Hichri, I., Barrieu, F., Bogs, J., Kappel, C., Delrot, S., Lauvergeat, V., 2011. Recent ad-vances in the transcriptional regulation of the flavonoid biosynthetic pathway.J. Exp. Bot. 62 (8), 2465e2483.

Hochberg, U., Degu, A., Toubiana, D., Gendler, T., Nikoloski, Z., Rachmilevitch, S.,Fait, A., 2013. Metabolite profiling and network analysis reveal coordinatedchanges in grapevine water stress response. BMC Plant Biol. 13 (1), 184.

Hollman, P.C., van Trijp, J.M., Buysman, M.N., Mengelers, M.J., de Vries, J.H.,Katan, M.B., 1997. Relative bioavailability of the antioxidant flavonoid quercetinfrom various foods in man. FEBS Lett. 418 (1), 152e156.

Hrazdina, G., Parsons, G.F., Mattick, L.R., 1984. Physiological and biochemical eventsduring development and maturation of grape berries. Am. J. Enol. Vitic. 35 (4),220e227.

Iijima, Y., Nakamura, Y., Ogata, Y., Tanaka, K., Sakurai, N., Suda, K., Suzuki, T.,Suzuki, H., Okazaki, K., Kitayama, M., 2008. Metabolite annotations based on theintegration of mass spectral information. Plant J. 54 (5), 949e962.

Ivanova, V., Stefova, M., Vojnoski, B., D€ornyei, �A., M�ark, L., Dimovska, V., Stafilov, T.,Kil�ar, F., 2011. Identification of polyphenolic compounds in red and white grape

A. Degu et al. / Plant Physiology and Biochemistry 86 (2015) 24e33 33

varieties grown in R. Macedonia and changes of their content during ripening.Food Res. Int. 44 (9), 2851e2860.

Katajamaa, M., Ore�si�c, M., 2007. Data processing for mass spectrometry-basedmetabolomics. J. Chromatogr. A 1158 (1), 318e328.

Kennedy, J.A., Matthews, M.A., Waterhouse, A.L., 2002. Effect of maturity and vinewater status on grape skin and wine flavonoids. Am. J. Enol. Vitic. 53 (4),268e274.

Koes, R.E., Quattrocchio, F., Mol, J.N., 1994. The flavonoid biosynthetic pathway inplants: function and evolution. Bioessays 16 (2), 123e132.

Laucou, V., Lacombe, T., Dechesne, F., Siret, R., Bruno, J., Dessup, M., Dessup, T.,Ortigosa, P., Parra, P., Roux, C., 2011. High throughput analysis of grape geneticdiversity as a tool for germplasm collection management. Theor. Appl. Genet.122 (6), 1233e1245.

Li, S., Zhang, Z., Cain, A., Wang, B., Long, M., Taylor, J., 2005. Antifungal activity ofcamptothecin, trifolin, and hyperoside isolated from Camptotheca acuminata.J. Agric. Food Chem. 53 (1), 32e37.

Liang, Z., Wu, B., Fan, P., Yang, C., Duan, W., Zheng, X., Liu, C., Li, S., 2008. Antho-cyanin composition and content in grape berry skin in Vitis germplasm. FoodChem. 111 (4), 837e844.

Luo, S., He, P., 2004. The inheritances of fruit skin and must colors in a series ofinterspecific and intraspecific crosses between V. vinifera and the wild grapespecies native to China. Sci. Hortic. 99 (1), 29e40.

Mateo, J., Jim�enez, M., 2000. Monoterpenes in grape juice and wines. J. Chromatogr.A 881 (1), 557e567.

Mattivi, F., Guzzon, R., Vrhovsek, U., Stefanini, M., Velasco, R., 2006. Metaboliteprofiling of grape: flavonols and anthocyanins. J. Agric. Food Chem. 54 (20),7692e7702.

Moco, S., Bino, R.J., Vorst, O., Verhoeven, H.A., de Groot, J., van Beek, T.A., Vervoort, J.,De Vos, C.R., 2006. A liquid chromatography-mass spectrometry-basedmetabolome database for tomato. Plant Physiol. 141 (4), 1205e1218.

Monagas, M., Nú~nez, V., Bartolom�e, B., G�omez-Cordov�es, C., 2003. Anthocyanin-derived pigments in Graciano, Tempranillo, and Cabernet Sauvignon winesproduced in Spain. Am. J. Enol. Vitic. 54 (3), 163e169.

Monagas, M., Su�arez, R., G�omez-Cordov�es, C., Bartolom�e, B., 2005. Simultaneousdetermination of nonanthocyanin phenolic compounds in red wines by HPLC-DAD/ESI-MS. Am. J. Enol. Vitic. 56 (2), 139e147.

Rib�ereau-Gayon, P., Boidron, J., Terrier, A., 1975. Aroma of Muscat grape varieties.J. Agric. Food Chem. 23 (6), 1042e1047.

Rodriguez Montealegre, R., Romero Peces, R., Chac�on Vozmediano, J., MartinezGascuena, J., Garcia Romero, E., 2006. Phenolic compounds in skins and seeds often grape Vitis vinifera varieties grown in a warm climate. J. Food Compos. Anal.19 (6), 687e693.

Roubelakis-Angelakis, K., Kliewer, W., 1986. Effects of exogenous factors onphenylalanine ammonia-lyase activity and accumulation of anthocyanins andtotal phenolics in grape berries. Am. J. Enol. Vitic. 37 (4), 275e280.

Saito, K., Kasai, Z., 1978. Conversion of labeled substrates to sugars, cell wall poly-saccharides, and tartaric acid in grape berries. Plant Physiol. 62 (2), 215e219.

S�anchez-Rabaneda, F., Jauregui, O., Lamuela-Ravent�os, R.M., Viladomat, F.,Bastida, J., Codina, C., 2004. Qualitative analysis of phenolic compounds in applepomace using liquid chromatography coupled to mass spectrometry in tandemmode. Rapid Commun. Mass Spectrom. 18 (5), 553e563.

Schauer, N., Zamir, D., Fernie, A.R., 2005. Metabolic profiling of leaves and fruit ofwild species tomato: a survey of the Solanum lycopersicum complex. J. Exp. Bot.56 (410), 297e307.

Schauer, N., Semel, Y., Roessner, U., Gur, A., Balbo, I., Carrari, F., Pleban, T., Perez-Melis, A., Bruedigam, C., Kopka, J., 2006. Comprehensive metabolic profiling andphenotyping of interspecific introgression lines for tomato improvement. Nat.Biotechnol. 24 (4), 447e454.

Sneath, P.H., Sokal, R.R., 1973. Numerical Taxonomy. The Principles and Practice ofNumerical Classification.

Steinhauser, D., Krall, L., Müssig, C., Büssis, D., Usadel, B., 2008. Correlation Net-works. Analysis of Biological Networks, pp. 305e333.

Strack, D., 1997. 10 phenolic metabolism. Plant Biochem. 387.Sweetman, C., Deluc, L.G., Cramer, G.R., Ford, C.M., Soole, K.L., 2009. Regulation of

malate metabolism in grape berry and other developing fruits. Phytochemistry70 (11), 1329e1344.

Teusch, M., Forkmann, G., Seyffert, W., 1987. Genetic control of hydroxycinnamoyl-coenzyme a: Anthocyanidin 3-glycoside-hydroxycinnamoyltransferase frompetals of Matthiola incana. Phytochemistry 26 (4), 991e994.

This, P., Lacombe, T., Thomas, M.R., 2006. Historical origins and genetic diversity ofwine grapes. Trends Genet. 22 (9), 511e519.

Toubiana, D., Semel, Y., Tohge, T., Beleggia, R., Cattivelli, L., Rosental, L., Nikoloski, Z.,Zamir, D., Fernie, A.R., Fait, A., 2012. Metabolic profiling of a mapping popula-tion exposes new insights in the regulation of seed metabolism and seed, fruit,and plant relations. PLoS Genet. 8 (3), e1002612.

Toubiana, D., Fernie, A.R., Nikoloski, Z., Fait, A., 31.1.2013. Network analysis: tacklingcomplex data to study plant metabolism. Trends Biotechnol. 29e36.

Tzin, V., Galili, G., 2010. New insights into the shikimate and aromatic amino acidsbiosynthesis pathways in plants. Mol. Plant 3 (6), 956e972.

Urbanczyk-Wochniak, E., Willmitzer, L., Fernie, A.R., 2007. Integrating profiling data.In: Anonymous (Ed.), Metabolomics. Springer, pp. 77e85.

Vezzulli, S., Troggio, M., Coppola, G., Jermakow, A., Cartwright, D., Zharkikh, A.,Stefanini, M., Grando, M.S., Viola, R., Adam-Blondon, A., 2008. A reference in-tegrated map for cultivated grapevine (Vitis vinifera L.) from three crosses, basedon 283 SSR and 501 SNP-based markers. Theor. Appl. Genet. 117 (4), 499e511.

Vincent, D., Ergul, A., Bohlman, M.C., Tattersall, E.A.R., Tillett, R.L., Wheatley, M.D.,Woolsey, R., Quilici, D.R., Joets, J., Schlauch, K., 2007. Proteomic analysis revealsdifferences between Vitis vinifera L. cv. Chardonnay and cv. Cabernet Sauvignonand their responses to water deficit and salinity. J. Exp. Bot. 58 (7), 1873e1892.

Weckwerth, W., Wenzel, K., Fiehn, O., 2004. Process for the integrated extraction,identification and quantification of metabolites, proteins and RNA to revealtheir co-regulation in biochemical networks. Proteomics 4 (1), 78e83.

Wulf, L.W., Nagel, C.W., 1978. High-pressure liquid chromatographic separation ofanthocyanins of Vitis vinifera. Am. J. Enol. Vitic. 29 (1), 42e49.