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/copyright
ContentslistsavailableatScienceDirect
Talanta
j ourna l h o me p a g e :w w w . e l s e v i e r . c o m / l o c a t e / t a l a n t a
Multivariate analysis of the polyphenol composition of Tempranillo and Graciano red wines
Matilde García-Marino, José Miguel Hernández-Hierro, Celestino Santos-Buelga, Julián C. Rivas-Gonzalo, M. Teresa Escribano-Bailón
∗GrupodeInvestigaciónenPolifenoles,UnidaddeNutriciónyBromatología,FacultaddeFarmacia,UniversidaddeSalamanca,CampusMigueldeUnamuno,E37007Salamanca,Spain
a r t i c l e i n f o
Articlehistory:
Received8March2011
Receivedinrevisedform7July2011 Accepted9July2011
Available online 23 July 2011
Keywords:
Wine Colour Chemometrics Tempranillo Graciano Polyphenols
a b s t r a c t
VitisviniferaL.cvGracianoisoftenusedasablendingpartnerofTempranillobasedwinesbecauseitis consideredtocontributesignificantlytothequality.Theaimofthisstudyistodiscriminatebetween TempranilloandGracianomonovarietalwines,andthosemadebytheincorporationof20%ofGraciano varietyintwodifferentstages(attheendofmalolacticfermentationandmixingthetwograpevarieties inthepre-fermentativemacerationstage)ofthewinemakingprocessoftheTempranillovariety.To achievethis,supervisedandunsupervisedpatternrecognitiontoolswereappliedtothedataobtained inthestudyofthedetailedpolyphenoliccomposition,colourandotheroenologicalparameters(143 variables).Patternsrelatedtostagesinthewinemakingandageingprocess,differentwinesandvintages canbeobservedusingprincipalcomponentanalyses.Furthermore,lineardiscriminantanalysishasbeen appliedinordertocharacterisethewinesamples.Fromthe143variables,flavan-3-olshaveexerteda profoundinfluenceonwinedifferentiation.
© 2011 Elsevier B.V. All rights reserved.
1. Introduction
VitisviniferaL.cvGracianorepresentsonlyasmallproportion ofthetotalgrapesgrowninSpain.IntheRiojaregionthisvarietyis commonlyusedasablendingpartnerofTempranillo-basedwines towhichitaddsbrightness,aroma,tanninsandacidity.Although lessusualsomewineriesalsoproducevarietalGracianowines.Dif- ferencesinthephenolicprofilesbetweenGracianoandTempranillo havebeenstudied[1–3]andithasbeenshownthatflavanolsfrom grapeskinsofGracianoaremoreeffectivecopigmentsofantho- cyaninsthanthoseofTempranillo,andtheoppositeoccursforseed flavanols,whereTempranillowasmoreeffective[2].Ontheother hand,ithasbeenproposedthatcopigmentationcouldassistthe extractionofpolyphenolicsfromgrapes[4]andthatitcouldalsoact asafirststageintheformationofnewpigmentsthatdeterminethe colourofagedredwines[5].Takingintoconsiderationtheafore- mentioned,itcouldbereasonablyexpectedthatifthecontactof thefermentingmustwiththeskinsandseedsofbothvarietieswas facilitated,thefinalwinecouldresultinhighercontentsofsome polyphenolics,whichwouldaffectnotonlythesensorialproperties butalsoitsageingpotential.
NowadaystheuseoftechniquessuchasHPLC-DADcoupledto ESI/MSnallowsthedeterminationofalargequantityofphenolic
∗ Correspondingauthor.Tel.:+34923294537;fax:+34923294515.
E-mailaddress:[email protected](M.T.Escribano-Bailón).
compoundsinredwines[6].Thehugeamountofdatathatcan beobtainedmakestheuseofchemometryanecessaryapproach tohelpselectingrelevantinformation[7–9].Chemometrictools havebeenusedina widerangeoffoods[10,11]includingwine [12].Specifically,thephenoliccomposition,colourandoenological parametersofwinehavebeenstudiedusingdifferentprocedures for classificatory purposes and for recognising grape cultivars [13–20]andalsoforstudyingtherelationshipbetweenthecolour andphenoliccompositionofredwinesobtainedunderdifferent winemakingconditions[3,21,22].
TheaimofthisstudyistodiscriminatebetweenTempranillo and Graciano varietal wines,and those madeby both blending thesewinesattheendofmalolacticfermentationandmixingthe twograpevarietiesinthepre-fermentativemacerationstage.To achievethis,supervisedandunsupervisedpatternrecognitiontools were appliedtothe dataobtained inthe studyof the detailed polyphenoliccomposition,colourandotheroenologicalparame- ters(143variables).Also,thevariablesthatcanexplainmostofthe varianceobservedinthedatasethavebeeninvestigated.
2. Experimental 2.1. Samples
Thewinesusedinthisstudywereobtainedfromredgrapesof V.viniferaL.oftheTempranilloandGracianovarieties,harvested in2005and2006andprocessedbyBodegasRoda(Haro,LaRioja, 0039-9140/$–seefrontmatter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.talanta.2011.07.039
Table1
StagesofsamplingofwinesT,G,MandWduringwinemakingandageing.
Stages Time
1 Endofalcoholicfermentation 16days
2 Middleofpost-fermentativemaceration 20days 3 Beginningofmalolacticfermentation 22days
4 Middleofmalolacticfermentation 29days
5a Endofmalolacticfermentation 39days
6 Before1strack 3moths
7 Before2ndrack 6moths
8 Before3rdrack 12moths
9 Clarificationandbottling 14moths
10 Ageinginbottleduring5months 602days
11 Ageinginbottleduring9months 736days
12 Ageinginbottleduring12months 827days
aInthecaseofWwinessamples,werecollectedfromstage05.
Spain).InthestudiedsamplesnotationsTandGwereusedtodesign TempranilloandGracianovarietalwinesrespectively;Wwasthe wineobtainedbytheblendofthesewines(proportion80:20)atthe endofmalolacticfermentation,andMthewineresultingoftheco- vinificationofthetwograpevarietiesmixedinproportion80:20in thepre-fermentativemacerationstage.Samplesofeachwinewere collectedintriplicateduringwinemakingandageing(Table1).The totalnumberofwinesamplescollectedwas264,132pervintage.
Analysisofeachsamplewasalsoperformedintriplicateandthe meanvalueobtainedwasusedinthestatisticaltreatment.
2.2. Winepolyphenolanalysis
Fortheanthocyaninsandflavonolsanalysis1mLofwinesam- pleswerediluted(1:1)with0.1NHCl(Panreac®Barcelona,Espa ˜na), filteredthrough0.45mMillex®HVsyringedrivenfilterunitsand injectedintothechromatographicsystem.Fortheflavan-3-olsand phenolicacidsanalysis2mLofeachwinesamplediluted(1:1)with 0.1NHClwereelutedthroughOasis®MCX3cm3(WatersCorpo- rationMildford,MA,USA)cartridgespreviouslyconditionedwith 2mLmethanoland 2mLwater, withtheobjective ofeliminat- ingtheredpigments[23].Afterwashingwith4mLofultrapure water,flavan-3-olsandthephenolicacidswereelutedwith8mL methanol,whereasanthocyaninsandtheflavonolswereretained inthecartridges.Asmallvolumeofwaterwasaddedtotheelu- ate and concentrated under vacuum at lower than 30◦C until complete elimination of methanol. The volume of theaqueous residuewasadjustedto0.5mLwithultrapurewater(DirectQ3, Millpore-Waters,Bedford,MA),filtered(0.45m)andanalysedby HPLC-DAD–MS.Allanalyseswereperformedintriplicate.
2.3. HPLC-DAD–MSanalysis
TheHPLC-DADanalysiswasperformedina Hewlett-Packard 1100seriesliquidchromatograph.TheLCsystemwasconnectedto theprobeofthemassspectrometerviatheUVcelloutlet.Themass analyseswereperformedusingaFinniganTMLCQiontrapdetec- tor(Thermoquest,SanJose,CA,USA)equippedwithanAPIsource, usinganelectrosprayionisation(ESI)interface.TheHPLC-DAD–MS analysisconditionsofredpigmentsandflavonolswerecarriedout inaccordancewithGarcía-Marinoetal.[3]selectinganadditional wavelengthat360nmtoachievetheanalysisofflavonols.Analy- sesofflavan-3-olsandphenolicacidswerecarriedoutasdescribed byGarcía-Marinoetal.[24]selectinganadditionalwavelengthat 330nmtoachievetheanalysisofphenolicacids.
2.4. Quantification
Forthequantitativeanalysis,calibrationcurveswereobtained usingstandards of anthocyanins(delphinidin, cyanidin, petuni-
Table2
Pigmentsanalysed.
Pigments
1 Directcondensationproductbetween malvidin-3-O-glucosideandgallocatechin
2 Delphinidin-3,5-diglucoside
3 Directcondensationproductbetween petunidin-3-O-glucosideandcatechin
4 Petunidin-3,5-diglucoside
5 Delphinidin-3,7-diglucoside
6 VitisinAofdelphinidin-3-O-glucoside 7 Directcondensationproductbetween peonidin-3-O-glucosideandcatechin 8 Directcondensationproductbetween
malvidin-3-O-glucosideandcatechin
9 Delphinidin-3-O-glucoside
10 Petunidin-3,7-diglucoside
11 Cyanidin-3-O-glucoside
12 VitisinAofpetunidin-3-O-glucoside
13 Petunidin-3-O-glucoside
14 Malvidin-3,7-diglucoside
15 VitisinAofpeonidin-3-O-glucoside
16 Peonidin-3-O-glucoside
17 VitisinAofmalvidin-3-O-glucoside
18 Malvidin-3-O-glucoside
19 Delphinidin-3-O-(6-acetyl)-glucoside
20 Malvidin-3-O-hexose
21 VitisinBofmalvidin-3-O-glucoside
22 Malvidin-3-O-pentoside
23 Malvidin-3-O-glucoside-8-ethyl-catechin 24 Petunidin-3-O-(6-acetyl)glucoside
25 Malvidin-3-O-glucoside-8-ethyl-gallocatechin 26 Malvidin-3-O-glucoside-8-ethyl-epicatechin
27 Cyanidin-3-O-hexoside
28 Malvidin-3-O-glucoside-8-ethyl-metoxicatechin 29 Malvidin-3-O-(6-p-coumaroyl)glucoside-8-ethyl-
catechin
30 Peonidin-3-O-(6-acetyl)glucoside
31 VitisinAofmalvidin-3-O-(6-p-coumaroyl)glucoside 32 Malvidin-3-O-(6-acetyl)glucoside
33 Delphinidin-3-O-(6-p-coumaroyl)glucoside 34 Petunidin-3-O-(6-p-coumaroyl)glucoside(cis) 35 Malvidin-3-O-(6-p-caffeoyl)glucoside 36 Cyanidin-3-O-(6-p-coumaroyl)glucoside 37 Petunidin-3-O-(6-p-coumaroyl)glucoside(trans) 38 Malvidin-3-O-(6-p-coumaroyl)glucoside(cis) 39 Peonidin-3-O-glucoside-4-vinylcatechol 40 Peonidin-3-O-(6-p-coumaroyl)glucoside(trans) 41 Malvidin-3-O-(6-p-coumaroyl)glucoside(trans) 42 Malvidin-3-O-glucoside-4-vinylphenoladduct 43 Malvidin-3-O-glucoside-4-vinylguaiacoladduct 44 Malvidin-3-O-(6-acetyl)glucoside-4-vinylphenol
adduct
45 Malvidin-3-O-(6-p-coumaroyl)glucoside-4- vinylphenol
46 Malvidin-3-O-(6-p-coumaroyl)glucoside-4- vinylguaiacol
47 Totalanthocyanidin-monoglucosides 48 Totalanthocyanidin-diglucosides
49 Totalacetylanthocyanins
50 Totalcaffeoylanthocyanins
51 Totalp-coumaroylanthocyanins
52 Totalvitisins
53 Totalvinylanthocyaninadducts
54 Totalpyranoanthocyanidins
55 Acetaldehyde-mediatedflavanol-anthocynidin condensationproducts
56 Directflavanol-anthocyanincondensationproducts
57 Totalcondensationproducts
58 Totalpigments
59 Totalanthocyanins
60 Totalacylatedanthocyanins
61 Totalderivedpigments
62 A-typevitisins
63 B-typevitisins
64 Anthocyanin-4-vinylphenoladducts 65 Anthocyanin-4-vinylcatecholadducts 66 Anthocyanin-4-vinylguaiacoladducts
Table3
Flavonols,flavan-3-ols,phenolicacids,chromaticparametersandothervariablesanalysed.
Flavonols Flavan-3-ols Phenolicacids Colorimetric
parameters
Othervariables
1 Myricetin-3-O-galactoside 1 Procyanidintetramer 1 Gallicacid 1 L* 1 pH
2 Myricetin-3-O-glucoside 2 Procyanidinpentamer 2 Protocatechuicacid 2 a* 2 Meandegreeof
polymerisation (mDP)
3 Myricetinaglycone 3 (+)-gallocatechin 3 Vanillicacid 3 b* 3 %Copigmentation
4 Quercetin-3-O-galactoside 4 Procyanidintetramer 4 Syringicacid 4 C*ab
5 Quercetin-3-O-glucoside 5 Procyanidindimer(B1) 5 trans-Caftaricacid 5 hab
6 Quercetinglucuronide 6 Procyanidindimer(B3) 6 cis-Cutaricacid 6 s*uv
7 Quercetin-7-O- neohesperidoside-3-O- rutinoside
7 Prodelphinidindimer 7 trans-Cutaricacid
8 Quercetinaglycone 8 (−)-Epigallocatechin 8 trans-Fertaricacid 9 Kaempferol-3-O-galactoside 9 Procyanidintrimer 9 p-Coumaroylhexose(1) 10 Kaempferol-3-O-glucoside 10 Procyanidintrimer 10 trans-Caffeicacid
11 Kaempferolaglycone 11 (+)-Catechin 11 p-Coumaroylhexose(2)
12 Isorhamnetinaglycone 12 Procyanidintetramer 12 p-Coumaricacid
13 Totalflavonols 13 Procyanidintrimer 13 Totalhydroxybenzoicacids
14 Totalmyricetinderivatives 14 Procyanidindimer(B4) 14 Totalhydroxycinnamicacids 15 Totalquercetinderivatives 15 Procyanidindimer(B6) 15 Totalphenolicacids 16 Totalkaempferolderivatives 16 Procyanidindimer(B2)
17 Totalisorhamentinderivatives 17 (−)-Epicatechin 18 Totalmyricetinglycosides 18 Procyanidintrimer 19 Totalmyricetinaglycons 19 Galloyldimer 20 Totalquercetinglycosides 20 Procyanidintetramer 21 Totalquercetinaglycons 21 Totalflavan-3-ols 22 Totalkaempferolglycosides 22 Totalprocyanidins 23 Totalkaempferolaglycons 23 Totalprodelphinidins 24 Totalisorhamentinaglycons 24 Totalprocyanidinmonomers
25 Totalprocyanidindimers 26 Totalprocyanidintrimers 27 Totalprocyanidintetramers 28 Totalprocyanidinpentamers 29 Totalgalloylatedflavan-3-ols
din,peonidinandmalvidin3-O-glucosides),flavonols(myricetin, quercetinandkaempferol),flavanols(catechin,gallocatechin,epi- catechingallate,dimerB2andtrimerepicatechin-4,8-epicatechin- 4,8-catechin)andphenolicacids(3,4-dyhydroxybenzoicacidand 4-hydroxycinnamic acid). Anthocyanins were purchased from PolyphenolsLabs.(Sandnes,Norway).Myricetin,kaempferol,gal- locatechinandepicatechingallatewerepurchasedfromExtrasyn- thèse(Genay,France).Quercetin,catechin,3,4-dyhydroxybenzoic acid and 4-hydroxycinnamic acid were purchased from Sigma (Steinheim, Germany). Procyanidin dimer and trimer were obtainedinourlaboratoryin accordance withEscribano-Bailón etal.[25].
Thetotalcontentofthedifferentgroupsofphenoliccompounds studiedwascalculatedfromthesumoftheindividualconcentra- tionsobtainedforeachindividualcompound,expressedinmgL−1 ofwine.
2.5. Meandegreeofpolymerisation(mDP)
The mean degree of polymerisation of wine flavanols was calculatedinaccordancewiththemethoddescribedbyGonzález- Manzanoetal.[26].
2.6. Copigmentation
Thecontribution of thecopigmentationphenomenon to the absorbanceat520nmofthewines(%copigmentation)wasdeter- minedinaccordance withthemethodproposedbyBoulton[4]
usingaHewlettPackardUV–visHP-3853spectrophotometer.
2.7. Analysisofthechromaticparameters
A Hewlett Packard UV–vis HP3853 was used for scanning between 380and 770nmat 2nm intervals witha 2mm path- lengthquartz cell,andCIE 196410◦ standardobserverandthe CIE D65 illuminantobserver, asreferencestocalculate thetris- timulusvaluesrecommendedbythe“ComissionInternationalede l’Éclairage”[27].TheCIELABandCIELUVcolourspaceswereused andparametersmeasuredincluded:lightness(L*),hue(hab),red- green coordinate(a*,−a*)andyellow-bluecoordinate (b*,−b*), Chroma(C*ab)andsaturation(s*uv).Calculationsweremadeusing theCromaLab®software[28].
2.8. Chemometricanalysis
Chemometrictoolscouldbedividedintoquantitativeandqual- itative methods. Thelattercouldbesubdivided intosupervised andunsupervisedpatternrecognitiontools.Unsupervisedmeth- odsareappliedasapreliminarystagepriortosupervisemodelling, suchasclassification,soastoobservetrendsinthedataindicating relationshipsbetweensamplesandbetweenvariables.Supervised patternrecognitionmethodsusuallyindicatewhethersamplesfall intopre-definedgroups,howwell,andwhatcausesthisseparation.
Thehumancapabilityofvisuallyrecognisingregularitiesindatais stillunsurpassedbythesecomputermethods[8,9].
The SPSS 13.0 for Windows software package (SPSS, Inc., Chicago,IL)wasusedfordataprocessing.Theunsupervisedpattern recognition method used for data analysis was principal com- ponents analysis(PCA),which wasappliedfromthecorrelation matrixoftheoriginalvariables.Thesupervisedpatternrecognition methodwaslineardiscriminantanalysis(LDA).Stepwisefeature selectionwasemployedtoselectthemostsignificantvariablesfor
Fig.1.Representationofthewinesamples(2005vintage)ontheplanedefinedbythefirstandthirdprincipalcomponents,PC1(41.6%)andPC3(9.4%)(AandB)and representationofthewinesamplesontheplanedefinedbyepicatechinandanthocyanidin-monoglucosides(CandD).Differencesbetweenwines(AandC).Differences betweenstagesinthewinemakingprocess(BandD).
thediscriminationbetweenclassesusingF-statistictotestthesig- nificanceofthechangeinWilks’Lambdabyaddingorremovinga variable.Thepredictionabilitywasestimatedconsideringtheper- centageofsamplesadequately classifiedbytherulesdeveloped withthetrainingsetusingaleave-one-outcross-validationproce- dure.Thetotalvariablesusedwere143,correspondingtodetailed phenoliccomposition,colourparameters,mDP,%copigmentation andpH.TheyaresummarisedinTable2(pigments)andTable3 (othervariables).
3. Resultsanddiscussion 3.1. Principalcomponentanalysis
Principalcomponentanalysiswasusedasanunsupervisedpat- ternrecognitionmethodwiththedatasetfromthe2005vintage.
Fig.1AandBshowstheprojectionofthewinesamplesontheplane definedbythefirstandthirdprincipalcomponents.Thefirstprin- cipalcomponent(PC1)describes41.6%ofthevariabilityinthedata andthethird(PC3)9.4%.Althoughthesecondcomponent(PC2) accountsfor14.6%ofthevariabilityinthedata,itisnotrepresented sincenoeasy toexplain tendencywasfoundfor it. Thismight bedue tothis componentisinfluenced byfactorsforwhich no relatedinformationwasavailable.PC3allowsdistinctionbetween T(placedbelow)andG(placedabove),whereasMandWwinesare locatedbetweenthemandtheyarehardlydistinguishable(Fig.1A).
PC1allowsthevisualisationofdifferencesbetweenstagesofthe
winemaking process (Fig. 1B). From the obtainedloadings, the variablesmostrelatedwithPC3andPC1werethecontentsofepi- catechinandanthocyanidin-monoglucosides,respectively.Fig.1C and Dshows theplots of winesamplesconsideringonly these variables.Inthisplot,TandGcanalsobedistinguishedwhereas Mand Wwines arelocatedbetweenthem in a similarwayto that shown in Fig.1A and B, which indicatesthe highrelation betweentheaforementionedvariablesandPC1andPC3,respec- tively.
Similarresultswereobtainedwiththedatasetfromthe2006 vintage,inthiscase,theprojectionofthewinesamplesontheplane definedbythefirstandsecondprincipalcomponentsPC1(38.3%) andPC2(19.9%) (Fig.2Aand B)showsasimilarpatterntothe 2005vintageplot,asstatedabovethevariablemostrelatedwith PC1wasthecontentofanthocyanidin-monoglucosideswhichis relatedtostagesofthewinemakingprocess(Fig.2B).PC2allows distinctionbetweenwines(Fig.2A)andthevariablemostrelated tothisPCwasthecontentintheflavonolquercetin.Theplotsusing theanthocyanidin-monoglucosidesandquercetincontentsarealso presented(Fig.2CandD);theseplotsshowasimilardistribution patterntothoseinFig.2AandBandtheyarealsosimilartothe plotsobtainedforthe2005vintage.Itisnoticeablethatin2006 thedifferencesbetweenwinesarerelatedtoPC2,thusagreater amountofdatavariabilitythaninvintage2005wasrelatedtothe typeofwine.
Inordertoincreasethevariability,thesamplesfrombothvin- tages,2005and2006,wereusedinanewPCAanalysis.Theplotof
Fig.2.Representationofthewinesamples(2006vintage)ontheplanedefinedbythefirstandthirdprincipalcomponents,PC1(38.3%)andPC2(19.9%)(AandB)and representationofthewinesamplesontheplanedefinedbyquercetinaglyconeandanthocyanidin-monoglucosides(CandD).Differencesbetweenwines(AandC).Differences betweenstagesinthewinemakingprocess(BandD).
thewinesamplesinthespacedefinedbyPC1(31.3%),PC2(18.6%) andPC3(11.7%)showsdifferencesbetweenthewinesandstages inthewinemakingprocess(Fig.3AandB).However,inthiscase, anotherandmoreintensepatternwasalsoobservedrelatedtothe vintagethatallowedacleardistinctionbetweenthewinesof2005 and2006vintages.Fig.3C emphasisesthis patternshowingthe planedefinedbyPC1andPC2.Ingeneral,theprincipalcompo- nentsPC1,PC2andPC3standforthedifferencesbetweenstages ofthewinemakingprocess,vintagesandtypesofwines,respec- tively.Fromtheobtainedloadingsthevariables mostrelated to thesePCswerethecontentofmalvidin-3-O-glucosidefollowedby totalanthocyanidin-monoglucosides’contents,forPC1,whichalso madeagreatcontributioninthemodelswherethe2005and2006 vintageswereconsideredseparately.ForPC3themostrelatedvari- ablewasthetotalcontentofpyranoanthocyanidins,andforPC2 thetotalcontentofanthocyanidin-diglucosides. Anthocyanidin- diglucosides are minor anthocyanins in V. vinifera although in recentyearstheyhavebeenextensivelyfoundingrapesandwine fromthisspecies[3,6,29,30].Thefactthatinourstudytheircon- tentsallowthedistinctionbetweenvintagesseemstoindicatethat theirvariationsingrapes(andconsequentlyinwines)arerelated toweatherconditions.
InthethreePCAanalysesperformed,thevariationsinthecon- tents of anthocyanidin-monoglucosides play a decisive role for distinguishingbetweendifferentstages ofthewinemaking.The changesintheamountofthesecompoundsduringthewinemaking
processcanberelatedtotheirinvolvementintheformationofnew pigments(includingpyranoanthocyanidins)ortotheirdegradation byoxidativeorenzymaticprocesses[31–33].
Polyphenols whose variations allowthedistinctionbetween thedifferentwineswerenotthesameinallofthePCAanalyses performed(i.e.epicatechinin2005,quercetinin2006andpyra- noanthocyanidinsin2005+2006).Thiscouldbeattributedtothe differentphenoliccompositionofthegrapeineachvintage.Differ- encesbetweenthepolyphenolicprofilesofagivencultivarreflects toagreatextentitsgeneticpotentialbutenvironmentalstimuli alsoplaycriticalrolesinpolyphenols’biosynthesisandthiseffect isreflectedinthepolyphenolicprofiles[18].
3.2. Lineardiscriminantanalysis
Lineardiscriminantanalysiswasperformedasasupervisedpat- ternrecognitionmethodwiththedatasetfromthe2005vintage soas toallocatethewinesamplestotheirwinegroup.Thirty- one variables were retainedthat allowed100% of thesamples werecorrectlyclassifiedin boththeinternal andleave-one-out cross-validation procedures. The most represented family was flavan-3-olswithelevenvariablesthatcorrespondtothisfamilyof phenoliccompounds.AnewLDAmodelwasdevelopedusingonly theseelevenvariablesinordertouseasinglefamilyofcompounds asclassifiers.Usingthisapproach100%ofsampleswerecorrectly classified in internal validation and 96.2% in theleave-one-out
Fig.3.Representationofthewinesamplesinthespacedefinedbythefirst,secondandthirdprincipalcomponents,PC1(31.3%),PC2(18.6%)andPC3(11.7%)(AandB)and intheplanedefinedbyPC1andPC2(C)(2005and2006vintagestogether).(A)Differencesbetweenwines.(B)Differencesbetweenstagesinthewinemakingprocess.(C) Differencesbetweenvintages.
cross-validationprocedure.Usingthevariablescorrespondingto otherphenolicfamilies,thepercentageofsamplescorrectlyclassi- fiedwaslowerinallcases.Thesameprocedurewasappliedtothe datasetfromthe2006vintage.Forty-onevariableswereretained, whichallowed100%ofsampleswerecorrectlyclassifiedinboth theinternaland leave-one-outcross-validationprocedures.The mostrepresentedfamilywasalsoflavan-3-olswithtwelvevari- ablesthatwereusedtodevelopanewLDAmodel,inwhich91.7%
ofsampleswerecorrectlyclassifiedintheinternalvalidationand 85.6%intheleave-one-outcross-validationprocedure.Similarly, whenvariablesoftherestofthefamilieswereusedthepercent- ageofsamplescorrectlyclassifiedwaslowerinallcases.Inthe caseofthe2005and2006vintagestogether,fifty-fourvariables wereretained,and100%ofsampleswerecorrectlyclassifiedinthe internalvalidationand98.5%intheleave-one-outcross-validation procedure(Fig.4).Themostrepresentedfamilywasagainflavan- 3-olswithsixteen variables(Table3,variables 1,2,5,7–10,12, 14–17,19,20,23and24).AnLDAmodelwasdevelopedusingonly thesevariables:82.4%ofsampleswerecorrectlyclassifiedinthe internalvalidationand79.0%intheleave-one-outcross-validation procedure.Again,worseclassificationresultswereobtainedusing
variablesofotherphenolicfamilies. Fig.4. Representationofthewinesamplesinthespacedefinedbythefirst,second andthirddiscriminantfunctions(2005and2006vintagestogether).
Itappearsthatflavan-3-olsarepredominantfactorsregarding differentiationbetweenT,G,MandWwines.Flavan-3-olcompo- sitionisdependentontheripenessstageandgrowthconditions, butit is alsogreatlyaffectedby genetics[18,34].Thisgroupof flavonoidswasalsoutilisedinordertoclassifywinesaccording totheirgeographicalorigin[17],cultivars[20]orboth[18].Inthe presentstudythegeneratedmodelshavepermitteddiscrimina- tionnotonlybetweendifferentcultivars(TandGwines),butalso betweendifferentoenologicalpractices(MandWwines).There- fore,thedifferentproceduresusedforthewinemakingofMandW wines(mixinggrapesorblendingwines)wouldimplyimportant differencesrelatedtothequalitativeand/orquantitativephenolic profile,whicharemainlyreflectedintheflavanolscomposition.
Althoughother families ofcompounds or moreeasily obtained parameters,suchascolourparameters,havebeenseparatelyeval- uated,inallcasestheresultsobtainedwerenotasgoodasusing flavan-3-olcompounds.
4. Conclusions
Basedontheresultsobtaineditcanbeconcludedthatpatterns relatedtostagesof thewinemakingprocess,type ofwinesand vintagescanbedifferentiatedusingprincipalcomponentanalyses.
Moreover,thevariablesmostrelatedtotheprincipalcomponents obtainedintheseanalyseshavebeenidentified,whichprovidea chemicalexplanationoftheobservedtendencies.
Furthermore,lineardiscriminantanalysishasbeenappliedin ordertoclassifythesamplesofthedifferentstudiedwines(T,G,M andW).Thegeneratedmodelshavepermitteddiscriminationnot onlybetweendifferentcultivars(TandGwines)wines,butalso betweendifferentoenologicalpractices (MandWwines).From the143consideredvariables,flavan-3-olshaveprovedtobethe mostprofoundlyinfluentialforwinedifferentiation.
Acknowledgements
Thanksare due totheSpanish MICINN and FEDER(Projects ref.AGL2005-07245-C03-01andAGL2008-05569-C02-01)andto the Junta de Castilla y León (Grupo de Investigación de Exce- lencia (GIP/USAL), GR133) and to the Consolider-Ingenio 2010 Programme(FUN-C-FOOD,CSD2007-00063)forfinancialsupport.
TheauthorsalsothankBodegasRODAS.A.(Haro,LaRioja,Spain) forsupplyingthewinesamples.
References
[1] M.Monagas,V.Nú ˜nez,B.Bartolomé,C.Gómez-Cordoves,Am.J.Enol.Viticult.
54(2003)163–169.
[2]S.González-Manzano,M.Due ˜nas,J.C.Rivas-Gonzalo,M.T.Escribano-Bailón,C.
Santos-Buelga,FoodChem.114(2009)649–656.
[3]M.García-Marino,J.M.Hernández-Hierro,J.C.Rivas-Gonzalo,M.T.Escribano- Bailón,Anal.Chim.Acta660(2010)134–142.
[4]R.Boulton,Am.J.Enol.Viticult.52(2001)67–87.
[5]R.Brouillard,O.Dangles,FoodChem.51(1994)365–371.
[6]C.Alcalde-Eón,M.T.Escribano-Bailón,C.Santos-Buelga, J.C.Rivas-Gonzalo, Anal.Chim.Acta563(2006)238–254.
[7]D.L.Massart,B.G.M.Vandeginste,L.M.C.Buydens,S.D.Jong,P.J.Lewi,J.Smeyers- Verbeke,HandbookofChemometricsandQualimetrics,ElsevierScienceInc., 1998.
[8] R.G.Brereton,Chemometrics:DataAnalysisfortheLaboratoryandChemical Plant,J.Wiley,Chichester,WestSussex,England,2003.
[9] B.G.M.Vandeginste,D.L.Massart,L.M.C.Buydens,HandbookofChemometrics andQualimetrics:PartB,ElsevierScienceInc.,1998.
[10] D.M.A.M.Luykx,S.M.VanRuth,FoodChem.107(2008)897–911.
[11]L.A.Berrueta,R.M.Alonso-Salces,K.Héberger,J.Chromatogr.A1158(2007) 196–214.
[12]J.Saurina,Trac-TrendsAnal.Chem.29(2010)234–245.
[13] L.Almela,S.Javaloy,J.A.Fernández-López,J.M.López-Roca,J.Sci.FoodAgric.
70(1996)173–180.
[14]B.Berente,D.DelaCalleGarcía,M.Reichenbächer,K.Danzer,J.Chromatogr.A 871(2000)95–103.
[15] C.Santos-Buelga,S.S.Mu ˜noz,Y.Gutierrez,E.Hebrero,J.L.Vicente,P.Galindo, J.C.Rivas-Gonzalo,J.Agric.FoodChem.39(1991)1086–1090.
[16]S.Pérez-Magari ˜no,M.Ortega-Heras,M.L.González-SanJosé,Z.Boger,Talanta 62(2004)983–990.
[17] M.A. Rodríguez-Delgado,G. González-Hernández,J.E. Conde-González, J.P.
Pérez-Trujillo,FoodChem.78(2002)523–532.
[18]D.P.Makris,S.Kallithraka,A.Mamalos,Talanta70(2006)1143–1152.
[19]S. Kallithraka,A. Mamalos, D.P. Makris,J. Agric. Food Chem. 55 (2007) 3233–3239.
[20] A.deVilliers,P.Majek,F.Lynen,A.Crouch,H.Lauer,P.Sandra,Eur.FoodRes.
Technol.221(2005)520–528.
[21]E.Gomez-Plaza,R.Gil-Mu ˜noz,J.M.Lopez-Roca,A.Martínez,FoodRes.Int.32 (1999)503–507.
[22]C.Gómez-Cordovés,M.L.Gónzalez-SanJosé,J.Agric.FoodChem.43(1995) 557–561.
[23]M.García-Marino,C.Santos-Buelga,J.C.Rivas-Gonzalo,M.T.Escribano-Bailón, SemanaVitivinícola3248(2009)6–14.
[24] M.García-Marino,J.C.Rivas-Gonzalo,E.Ibanez,C.Garcia-Moreno,Anal.Chim.
Acta563(2006)44–50.
[25]M.T.Escribano-Bailón,Y.Gutierrez-Fernandez,J.C.Rivas-Gonzalo,C.Santos- Buelga,J.Agric.FoodChem.40(1992)1794–1799.
[26]S.González-Manzano,C.Santos-Buelga,J.J.Pérez-Alonso,J.C.Rivas-Gonzalo, M.T.Escribano-Bailón,J.Agric.FoodChem.54(2006)4326–4332.
[27]M.Melgosa,E.Hita,A.J.Poza,D.H.Alman,R.S.Berns,ColorRes.Appl.22(1997) 148–155.
[28] F.J.Heredia,C.Álvarez,M.L.Gonzalez-Miret,A.Ramirez,in:CromaLab,análisis decolor.RegistroGeneraldelaPropiedadIntelectualSE-1052-04,Sevilla,Spain, 2004.
[29]A.Baldi,A.Romani,N.Mulinacci,F.F.Vincieri,B.Casetta,J.Agric.FoodChem.
43(1995)2104–2109.
[30] A.Heier,W.Blaas,A.Dro,R.Wittkowski,Am.J.Enol.Viticult.53(2002)78–86.
[31]M.V.Moreno-Arribas,C.Polo,WineChemistryandBiochemistry,Springer, NewYork,NY,2009.
[32]R.L.Jackman,R.Y.Yada,M.A.Tung,R.A.Speers,J.FoodBiochem.11(1987) 201–247.
[33]H.Li,A.Guo,H.Wang,FoodChem.108(2008)1–13.
[34]J.A.Kennedy,Y.Hayasaka,S.Vidal,E.J.Waters,G.P.Jones,J.Agric.FoodChem.
49(2001)5348–5355.