Multivariate statistical analysis of botrytised wines of different origin

Available online at www.sciencedirect.com

www.elsevier.com/locate/foodchem

Food Chemistry 110 (2008) 742–750

Analytical Methods

Multivariate statistical analysis of botrytised wines of different origin

Agnes Sass-Kiss *, Judit Kiss, Bence Havadi, Nora Adanyi

Unit of Analytics, Central Food Research Institute, Herman Otto ut 15, 1022 Budapest, Hungary

Received 23 November 2007; received in revised form 15 February 2008; accepted 18 February 2008

Abstract

The study examined several types of compounds can be suitable to characterise wines made from botrytised grapes and to determinetheir origin and authenticity. Amines, acids, macro- and microelements of botrytised sweet wine specialities, coming from Hungary anddifferent countries, were analysed. Measured values of twenty-one Tokaji aszu wines and twenty-three foreign botrytised wines were com-pared by multivariate statistical methods. Characterising the effect of Botrytis cinerea and the winemaking technology, amines were themost suitable components for determination of authenticity and origin of wines from the three types of compounds studied. However, inacids and elemental composition, differentiation of wine samples by principal component analysis was not complete but a tendency canbe observed for separation according to origin. The knowledge on composition of acids and elements can support the results of amineanalysis in reaching the goal to determine the origin of wines.� 2008 Elsevier Ltd. All rights reserved.

Keywords: Botrytis cinerea; Tokaji aszu wines; Botrytised wines; Amines; Acids; Micro-, Macro-elements; Origin; Authenticity; HPLC; Atomic absorption

1. Introduction

Botrytis cinerea can cause a destructive grey mould rot ora so-called noble rot in certain conditions on grape berries. Inthe latter instance, the rotting process is slowed down by theeffect of dry weather and sunshine. Piercing and weakeningthe grape skin, B. cinerea as noble mould alters the composi-tion of grapes by converting it to a raisin-like form with anunusually pleasant, special taste and delicious flavour. Theseberries are called as aszu grapes. During this natural processof noble rot, water content decreases and the compounds areconcentrated (sugar, acids, aroma compounds, etc.). Usingaszu grapes in wine making process, it yields wines of specialquality that are highly prized, sweet, smooth and full-bodiedwith a pleasant bouquet (Haraszti, 2002; Ribereau-Gayon,Ribereau-Gayon, & Seguin, 1980).

Tokaj wines are made from white grape varieties such asFurmint, Yellow Muscat and Linden Leaf. To produceaszu wines, the shrivelled, raisin-like aszu grapes are har-vested in October and November into wooden butts with

0308-8146/$ - see front matter � 2008 Elsevier Ltd. All rights reserved.

doi:10.1016/j.foodchem.2008.02.059

* Corresponding author. Tel.: +361 355 8838; fax: +361 214 2247.E-mail address: [email protected] (A. Sass-Kiss).

a capacity of 20–25 kg. Three, four, five or six butts ofaszu-paste is added to newly fermented dry wine of thesame year in a ‘Gonci’ oak barrel (136 l) mixed and soakedfor one or two days in order to extract the natural sugarcontent and flavours. The wine is then drawn off to fer-ment. The quantity of aszu grapes above is specified onthe label of the bottles of aszu wine. Eszencia is the firstrun juice of the aszu grapes, which seeps from the pressunder the own weight of grapes (Haraszti, 2002).

Wine is a widely consumed beverage in the world withthousands of years of tradition. Determination of itsauthenticity is one of the most important aspects in foodquality and safety. Many successful studies have shownthat it is possible to distinguish grape variety, vintage yearsor geographical zones on the basis of chemical parameters.Several papers have been published about classification ofwines (Arvanitoyannis, Katsota, Psarra, Soufleros, & Kal-lithroka, 1999; Csomos, Heberger, & Simon-Sarkadi, 2002;Day, Zhang, & Martin, 1995; Etievant, Schlich, Cantagrel,Bertrand, & Bouvier, 1989; Heberger, Csomos, & Simon-Sarkadi, 2003; Kiss & Sass-Kiss, 2005; Latorre, Garcıa-Jares, Medina, & Herrero, 1994; Muranyi & Kovacs,2000; Vasconcelos & Dasneves, 1989) but only few studies

mailto:[email protected]

A. Sass-Kiss et al. / Food Chemistry 110 (2008) 742–750 743

are available about classification of botrytised wines (Hav-adi, Kiss, Sass-Kiss, Adanyi, & Varadi, 2006; Kiss & Sass-Kiss, 2005). Recently, pattern recognition has become auseful and often applied method in food analysis. (Berrue-ta, Alonso-Salces, & Heberger, 2007).

Putrescine, spermidine, spermine from biogenic aminesare the major cellular polyamines in living organisms. Thesebiogenic amines are involved in cellular growth, regulationof nucleic acids and protein synthesis, stabilisation of lipids,brain development, nerve growth and regeneration (Her-nandez, Sanchez, & de Tarlovsky, 2006; Igarashi, 2006; I-garashi & Kashiwagi, 2000; Schreiber, Boeshore, Laube,Veh, & Zigmond, 2004). Other biogenic amines such as his-tamine, tyramine, and phenylethylamine etc. are formed pri-marily by decarboxylation of amino acids by the action ofmicroorganisms. Numerous research works have appearedin the literature dealing with formation of biogenic aminesby the effect of yeast fermentation in foods and beverages,including wine (Hyotylainen, Savola, Lehtonen, & Riekk-ola, 2001; Kallay & Nyitraine, 2003; Lehtonen, 1996; Lei-tao, Marques, & San Romao, 2005; Vidal-Carou, Lahoz-Portoles, Bover-Cid, & Marine-Font, 2003). Only a fewpublications appeared about aliphatic primer amines, whichare formed by the action of bacteria (Bast, 1971, 1972) orcould be found in wines produced from grapes infected bymoulds (B. cinerea, Penicillium expansum) (Eder, Brandes,& Paar, 2002; Hajos, Sass-Kiss, Szerdahelyi, & Bardocz,2000; Kiss, Korbasz, & Sass-Kiss, 2006; Sass-Kiss, Szerdah-elyi, & Hajos, 2000). It was established that the content ofprimer aliphatic amines in grape berries increased mostlyas a result of B. cinerea infection. During wine-making pro-cess, amine content continued to increase in wines, e.g. pri-mer aliphatic amines and tyramine. As a consequence, theamine composition of botrytised wines obtains a specificcharacter that makes possible to distinguish them from nor-mal wines (Sass-Kiss & Hajos, 2005; Sass-Kiss et al., 2000).

The most important acids always present in wines are tar-taric acid, malic acid, citric acid, succinic acid and lacticacids. Tartaric, malic and citric acids come from the grape,the others form during the fermentation. Tartaric acid isthe most important and most acidic component. Its highquantity gives for wine sharp and unpleasant hard tastetherefore the excellent quality wines contain usually less tar-taric acid. Among organic acids of biological origin, themost relevant one is lactic acid, which originates fromalcoholic or malolactic fermentation. Quantitative determi-nation of organic acids can corroborate sensorial, microbio-logical quality assessment (Kordis-Krapez, Abram, Kac, &Ferjancic, 2001) and authentication (Etievant, Schlich, Can-tagrel, Bertrand, & Bouvier, 1989) of wines. During infec-tion of grapes with B. cinerea under ideal conditions,water is lost from the berry and the acid constituents are con-centrated. However, at the same time, the Botrytis fungusmetabolizes these organic acids for use as an energy source(Eperjesi, Kallay, & Magyar, 1998).

Daily consumption of wine in low quantities contributessignificantly to the needs of the human organism for essen-

tial elements such as K, Ca, Mg, Cr, Co, Fe, F, I, Cu, Mn,Mo, Ni, Se, Zn (Day et al., 1995). At the same time, theanalysis of certain elements in wines is of special interestdue to their toxicity in case of excessive intake, and theyalso seem to have an effect on organoleptic properties ofwine (Lara, Cerutti, Salonia, Olsina, & Martinez, 2005).In other aspects, the analysis of elements has an importantrole in characterisation and classification of wines fordetermination of authenticity and geographical origin(Baxter, Crews, Dennis, Godall, & Anderson, 1997; Mar-engo & Aceto, 2003; Martin et al., 1999).

The goal of our work was to study several types of com-pounds such as amines, acids and elemental composition ofwines with regard to their suitability to characterise botry-tised wines and to determine their origin and authenticitymade from botrytised grapes. Using multivariate statisticalmethods, we analysed Tokaji aszu and foreign botrytisedwines to determine differences between them and to estab-lish which class of compounds is the best in determining theauthenticity and origin of wines.

2. Materials and methods

2.1. Reagents and chemicals

All reagents and authentic compounds were of analyti-cal reagent grade or HPLC grade as required. Acids usedfor sample digestion were hyper pure grade. Acetonitrileand methanol (HPLC grade) were obtained from Merck,standard solutions for elemental analysis were purchasedfrom Carlo Erba, Merck and Pancreac. Ultra pure watergenerated by the Milli-Q System (Millipore) was used.Anhydrous sodium acetate, boric acid, potassium hydrox-ide, acetic acid, Brij 35 and 2-mercaptoethanol were fromReanal (Budapest, Hungary); o-phthaldialdehyde was fromFluka and sodium octane sulfonate was obtained from Ro-mil (Cambridge, UK). Authentic amines – putrescine (Put),i-butyl amine (iBa), cadaverine (Cad), tyramine (Tyr), his-tamine (His), 2-methyl-butyl amine (2MeBa), agmatine(Agm), 3-methyl-butyl amine (3MeBa), n-pentyl-amine(Pa), spermidine (Spd), phenylethylamine (Phe), and hexylamine (internal standard, Istd) – were purchased fromSigma.

2.2. Wine samples

Studied wines took part in the VIth International WineCompetition VinAgora organised 2004 in Budapest. Afteropening the bottles, wine samples were taken and frozenat �20 �C until analysis.

Twenty-six Tokaji aszu wines and twenty-four foreignwines produced from botrytised grapes were analysed.The foreign wines came from nine countries (Portugal: P/1, P/2; Italy: I/1, I/2; Spain: E/1–E/4; Austria: A/1–A/7;Slovakia: SK/1, SK/2; Switzerland: CH/1, CH/2; France:F/1–F/3; Germany: G; United States of America: USA).

744 A. Sass-Kiss et al. / Food Chemistry 110 (2008) 742–750

2.3. HPLC analysis and chromatographic conditions

2.3.1. Amines

Chromatography was performed with Alliance Waters2690 HPLC chromatograph equipped with a Waters 474fluorimetric detector (kex = 345 nm, kem = 455 nm).

Separation of amines was performed as reported in ourprevious work (Kiss & Sass-Kiss, 2005) with ion pair forma-tion (octanesulfonic acid) on reverse phase column (lBond-apak C18, 300 � 3.9 mm, 10 lm; from Waters) usingpost-column derivatisation with OPA (o-phthaldialdehyde,2-mercaptoethanol). Compounds were eluted with a gradi-ent prepared from three solutions as follows: A: 0.165 Msodium acetate, pH 5.25, containing 10 mM octane sulfo-nate, B: 0.2 M sodium acetate, pH 4.5, containing acetoni-trile:water (66:34) and 10 mM octane sulfonate and C:0.01 M sodium acetate, pH 5.25, containing 10 mM octanesulfonate. The flow rate of the mobile phase was 1 ml min�1

and the flow rate of OPA was 0.8 ml min�1.Standard solutions of amines in the concentration range

of 0.1–10 mg l�1 were used for the calibration. Peak areaswere recorded and calculated using the Waters MilleniumSoftware package. The gradient elution program and thevalidation of the present method can be found in our pre-vious study (Kiss & Sass-Kiss, 2005). Wines were injecteddirectly into the column after filtration without any furthersample preparation.

2.3.2. Acids

Separation of acids was performed as reported in our pre-vious work (Kiss & Sass-Kiss, 2005) with ODS-AQ (YMCEuropean GMB) column (250 � 4.6 mm i.d., S-5lm) usingprecolumn ODS-AQ cartridge (YMC European GMB).For elution, 0.02 M phosphate buffer (2.75 g l�1 KH2PO4,

pH 2.7) was used under isocratic condition at 0.7 ml min�1

flow rate at room temperature. Peaks were identified withtartaric, malic, citric, shikimic, fumaric acids and ascorbicacid. Compounds were detected at 214 nm.

Validation of the present method can be found in ourprevious study (Kiss & Sass-Kiss, 2005). Standard solutionof acids in the concentration range between 0.04 g l�1 and1.0 g l�1, except for fumaric acid, (0.08–2 mg l�1) was usedfor the calibration. Ten microliters was injected from allsamples after 10-fold dilution with 2% meta-phosphoricacid.

2.4. Atomic absorption spectroscopy: sample preparationand instrumentation

Wine samples (4 ml) were placed in digestion tubes andevaporated to ca. 1 ml for elimination of ethanol. Ten mil-liliters of concentrated nitric acid was added to each sampleand left in a fume cupboard for digestion overnight. Thefollowing day, tubes were heated to 120 �C for 30 min, thento 160 �C for about 4 h. The digestion was completed whenapproximately 0.5 ml of the solution remained. After cool-ing, samples were diluted to 20 ml with bi-distilled water.

Measurements were carried out using Solaar M5(Thermo Elemental, Waltham, MA, USA) double-beamatomic absorption spectrometer. Potassium and Na con-tent of the samples was measured by flame photometry.Calcium, Mg, Fe, Cu, Zn, and Mn were determined byflame atomic absorption technique (AOAC 975-03 1990).Cadmium was measured by electrothermal atomic absorp-tion spectroscopy with standard addition.

2.5. Statistics

Concentration of compounds of three compound classeswas evaluated by statistical methods. Descriptive statisticalanalysis and correlation analysis were performed withMicrosoft Excel. For classification of Hungarian and for-eign wines, principal component analysis (PCA) and lineardiscriminant analysis (LDA) were used, applying Minitabstatistical program.

Principal component analysis (PCA) is a powerful visu-alisation tool for data evaluation, which can graphicallyrepresent intersample and intervariable relationships andprovides a way to reduce the dimensionality of the data.PCA is an unsupervised method of pattern recognition inthe sense that no grouping of the data has to be knownbefore the analyses. Using PCA, class membership is easyto indicate on a score plot.

Linear discriminant analysis (LDA) is a method to dis-criminate between two or more groups of samples. LDAis a supervised pattern recognition, which means that theclass membership has to be known prior to the analysis.The groups to be discriminated can be defined either natu-rally by the problem under investigation, or by some pre-ceding analysis such as cluster analysis or PCA.

3. Results and discussion

3.1. Amines

The amine components of twenty-one Hungarian andtwenty-three foreign botrytised wine samples were evalu-ated by statistical analysis. Results of descriptive statisticalanalysis used for evaluation of measured data can be seenin Table 1.

Relatively high concentration of putrescine(1.6–3.6 mg l�1, 0.3–14.4 mg l�1) tyramine (0.7–2.9 mg l�1,0.0–10.4 mg l�1) and phenylethylamine (0.04–20.2 mg l�1,1.6–6.2 mg l�1) characterised Hungarian and most of theforeign wine samples, respectively, when comparing normalwines. Among aliphatic primer amines, similarly to previousresults (Kiss & Sass-Kiss, 2005), 3-methyl-butyl amine (0.1–23.9 mg l�1) was found in the highest concentration in allwine samples studied. It was followed by 2-methyl-butylamine (0.0–13.8 mg l�1) and i-butylamine (0.1–6.2 mg l�1).The concentration of two unknown compounds, presum-ably also amines (unknown 1 and unknown 2) was calcu-lated on the basis of the calibration curve of i-butyl amine.

Tab

le1

Des

crip

tive

stat

isti

csfo

ram

ines

of

win

esa

mp

les

Pu

tres

cin

eT

yram

ine

Agm

atin

eC

adav

erin

eH

ista

min

eS

per

mid

ine

Ph

enet

yl-

amin

eis

o-

Bu

tyla

min

e2-

Met

hyl

bu

tyla

min

e3-

Met

hyl

bu

tyla

min

eP

enty

lam

ine

Un

kn

ow

n1

Un

kn

ow

n2

TF

TF

TF

TF

TF

TF

TF

TF

TF

TF

TF

TF

TF

Mea

nva

lue

2.28

4.30

1.78

1.79

0.36

0.12

0.15

0.21

0.07

2.14

0.13

2.97

14.7

43.

522.

660.

486.

251.

3919

.19

5.26

0.06

0.02

2.57

1.25

13.8

010

.42

Sta

nd

ard

dev

iati

on

0.64

4.69

0.64

3.09

0.24

0.19

0.05

0.21

0.04

4.95

0.28

5.16

2.46

5.56

1.27

0.63

2.48

3.12

2.25

6.86

0.08

0.04

0.76

1.48

3.12

5.98

Med

ian

2.00

2.31

1.64

0.15

0.30

0.05

0.15

0.12

0.08

0.09

0.04

0.79

15.4

91.

112.

190.

235.

370.

1719

.23

2.01

0.03

0.00

2.59

0.74

14.4

89.

50M

inim

um

1.57

0.25

0.68

0.00

0.07

0.00

0.08

0.04

0.00

0.00

0.00

0.03

9.63

0.04

1.06

0.11

3.51

0.00

15.7

90.

060.

000.

001.

040.

059.

310.

61M

axim

um

3.64

14.3

82.

8710

.04

0.93

0.78

0.27

0.81

0.12

15.4

00.

9620

.23

19.0

620

.23

6.21

2.52

13.7

613

.82

23.9

422

.24

0.30

0.21

4.14

6.70

20.8

022

.84

Ran

ge2.

0714

.13

2.19

10.0

40.

860.

780.

190.

770.

1215

.40

0.96

20.2

09.

4320

.19

5.15

2.41

10.2

513

.82

8.15

22.1

80.

300.

213.

106.

6511

.49

22.2

3S

um

4810

737

458

33

51

513

7430

988

5612

131

3540

313

21

154

3129

026

1K

urt

osi

s�

0.48

0.53

�1.

021.

950.

326.

500.

643.

01�

1.19

3.75

6.02

5.57

0.02

3.02

1.68

4.83

2.97

10.6

50.

181.

153.

1715

.79

0.08

7.20

�0.

29�

0.76

Sk

ewn

ess

0.87

1.45

0.22

1.79

1.12

2.62

0.90

2.00

�0.

652.

292.

652.

47�

0.54

1.99

1.30

2.33

1.54

3.12

0.45

1.55

2.00

3.69

0.23

2.51

0.32

0.27

Nu

mb

ero

fsa

mp

les

2124

2124

2124

2124

2124

2124

2124

2124

2124

2125

2124

2124

2124

Lev

elo

fco

nfi

den

ce(9

5.0%

)

0.29

1.94

0.29

1.27

0.11

0.08

0.02

0.09

0.02

2.09

0.13

2.13

1.12

2.30

0.58

0.26

1.13

1.29

1.02

2.83

0.04

0.02

0.35

0.61

1.42

2.47

T:

To

kaj

ias

zuw

ines

;F

:b

otr

ytis

edw

ines

fro

mfo

reig

nco

un

trie

s.


These unknown compounds appeared in high concentrationin botrytised wines, as well.

Variability of amine concentration of all Hungarian sam-ples was relatively small (±12% up to ±23%), which isshown by the standard deviation around the means. Themean concentration with standard deviation of putrescine(2.3 ± 0.6 mg l�1), phenylethyl amine (14.7 ± 2.5 mg l�1),unknown 2 (13.8 ± 3.1), unknown 1 (2.6 ± 0.8 mg l�1),and 3-methyl-butyl amine (19.2 ± 2.3 mg l�1) indicates thesimilarity in wine making technology of the region.

The question was whether botrytised wines from differ-ent origin can be distinguished or not. Former resultsshowed when seven foreign wine samples were comparedto Hungarian aszu wines that Tokaji aszu samples couldbe differentiated from some foreign wines with multivariatemethods (Kiss & Sass-Kiss, 2005). In our recent work,twenty-three wines originating from nine countries havebeen compared to twenty-one Tokaji aszu wines andessences. To visualise the differences among wine samplesstudied, principal component analysis was used. The resultof statistical evaluation is shown in Fig. 1.

PCA allowed 83.0% of the total variance to be explainedby the first three principal components. The separation offirst and second PC scores (Fig. 1A) of aszu wines from for-eign botrytised wines was significant. Only three Austrianwines have fallen in the range of Tokaji aszu wines fromwhich two samples came from the same producers. In thethird dimension of principal components (Fig. 1C andD), the scores of the three Austrian wines are scattered tothe left from the origin on the axe of third principal com-ponent. The score (third PC �1.97) of one Austrian wineseparated significantly from Hungarian wines in this direc-tion, while the third PC score of another wine (third PC�0.29) felt among those of Hungarian samples. The sepa-ration of third PC score of third Austrian wine was not sig-nificant. In a former study (Kiss & Sass-Kiss, 2005), closesimilarity of certain Austrian wine samples to Hungarianones has been observed as well.

Scores of other foreign wines are scattered to the leftfrom the origin on the first principal component and alongthe axis of second and third principal components. In mostcases, the PC scores of wines, coming from foreign coun-tries, are clustered together forming a tight group, exceptfor a few Austrian, Spanish and one Italian wine.

In the Fig. 1B, the loading plot of variables shows thatprimarily phenetylamine and primer aliphatic amines (firstPC > 0.83) take part in the formation of the first principalcomponent (PC). There is a high correlation among thesevariables (Kiss & Sass-Kiss, 2005) as there are small angelsbetween the loading vectors of these variables. The coordi-nates of PC scores of Hungarian wines (Fig. 1A) along theaxis of first principal component is determined mostly bythese compounds. Higher primer amine content of winespresents higher values to the right from the origin for firstprincipal component of individual samples in the scoreplot. Some samples are characterised by high tyramine con-tent such as Spanish wines and one Italian wine.

210-1

2

1

0

-1

-2

-3

First PC (58.2 %)

Seco

nd P

C (

12.4

%)

1.00.50.0

0.5

0.0

-0.5

First PC (58.2 %)

Seco

nd P

C (

12.4

%)

Unk2

Unk1Pa

3MeBa

2MeBaiBa

Phe

Agm

Tyr

-3 -2 -1 2

-2

-1

0

1

2

3

Second PC (12.4%)

Thi

rd P

C (

9.8%

)

-1

-2

-1

0

1

2

3

First PC (58.2%)

Thi

rd P

C (

9.8%

)

0 1 20 1

A B

DC

Fig. 1. Score plot (A) and loading plot (B) of first two principal components (PC), score plot of first and third PC (C) and second and third PC (D) for clas-sification of botrytised wines originating from Tokaj region of Hungary and other countries. Variables: i-butyl amine, unknown 1, tyramine, unknown 2,2-methyl-butyl amine, agmatine, 3-methyl-butyl amine, n-pentyl-amine, phenylethylamine; s: Tokaji aszu, Tokaji aszu-essence, d: Austrian wines, j: Span-ish wines, N: Portuguese wines, h: Italian wines,4: Slovakian wines, : Swiss wines, +: French wines, : German wine, : USA wine.


Linear discriminant analysis was used for the differenti-ation of aszu and foreign botrytised wines. Using Tyr,Agm, 3MeBa as variables, good classification of wine sam-ples was achieved. The percentage of correctly classifiedwines, using the above variables, was 93%. From 45 sam-ples 42 samples were correctly classified. The classificationof Tokaji aszu wines was 100%. From 24 foreign wine sam-ples, three Austrian samples were misclassified to Tokajiaszu wines; accordingly only 88% was obtained for classifi-cation of foreign wines. Comparing to our earlier results(Kiss & Sass-Kiss, 2005), it was interesting that two winesfrom Austria were perfectly similar to Hungarian Tokajiaszu wines in that case also.

3.2. Acids

Concentration values measured for organic acids ofwines are shown in Table 2.

As regards acid content, tartaric acid (2.14 ± 0.31 g l�1

and 2.20 ± 0.59 g l�1) and acetic acid (1.20 ± 0.26 g l�1

and 1.18 ± 0.49 g l�1) were the most balanced and were sim-ilar in all Hungarian and foreign samples, respectively. Themean concentration of malic acid in Hungarian wines(4.26 ± 1.37 g l�1 was significantly higher (p < 0.01) than

that in other wines (1.99 ± 1.52 g l�1). The mean valueof citric acid (0.82 ± 0.39 mg l�1) and fumaric acid(1.30 ± 0.75 mg l�1) was also significantly higher (p < 0.01)in Hungarian wines than in foreign ones (0.47 ± 0.35 mg l�1

and 0.52 ± 0.76 mg l�1, respectively).Principal component analysis was used for evaluation of

acids. Fig. 2 shows the score plot of the first and secondprincipal component of samples and loading plot ofvariables.

Correlation analysis shows only a slight correlationbetween malic acid and fumaric acid (0.654, p < 0.001) ormalic acid and citric acid (0.732, p < 0.001), which is indi-cated by the small angle between loadings of variables(Fig. 2B). PCA allowed more than 61.4% of the total vari-ance to be explained by the first two principal componentscomputed from the correlation matrix of the acid com-pounds as variables. The score plot of acids (Fig. 2A) showsonly some tendency for separation of foreign and Hungar-ian wines. Total separation, in contrast to amines, couldnot be observed. Principal component scores of most Hun-garian botrytised wines are located on the left from the ori-gin on the first principal component axis that can beexplained by their relatively higher malic acid, fumaric acidand citric acid content. The high value of PC scores of wines

Table 2Descriptive statistics for acids of wine samples

Tartaric acid(g l�1)

Malic acid(g l�1)

Shicimic acid(mg l�1)

Acetic acid(mg l�1)

Citric acid(g l�1)

Fumaric acid(mg l�1)

T F T F T F T F T F T F

Mean 2.14 2.20 4.26 1.99 35.74 24.42 1.20 1.18 0.82 0.47 1.30 0.52Standard deviation 0.31 0.59 1.37 1.52 13.37 20.69 0.26 0.49 0.39 0.35 0.75 0.76Median 2.11 2.13 4.77 1.93 35.07 17.21 1.14 1.18 0.76 0.37 1.30 0.28Kurtosis 1.33 1.08 �1.01 2.28 0.16 3.38 0.02 �0.58 �0.59 �0.37 �0.93 4.26Skewness 1.20 0.82 0.01 1.14 0.29 1.92 0.84 0.10 0.53 0.82 �0.22 2.19Range 1.12 2.56 4.66 6.49 52.60 77.83 0.91 1.79 1.39 1.23 2.50 2.83Minimum 1.79 1.22 1.97 0.13 13.21 5.11 0.87 0.33 0.32 0.00 0.00 0.00Maximum 2.91 3.78 6.63 6.62 65.81 82.94 1.78 2.12 1.71 1.23 2.50 2.83Sum 44.93 52.90 89.43 47.72 750.44 586.04 25.12 28.34 17.30 11.23 27.30 12.57Number of samples 21.00 24.00 21.00 24.00 21.00 24.00 21.00 24.00 21.00 24.00 21.00 24.00Confidence interval (95.0%) 0.14 0.25 0.62 0.64 6.09 8.73 0.12 0.20 0.18 0.15 0.34 0.32

T: Tokaji aszu wines; F: botrytised wines of foreign countries.

-2 -1 0 1 2

-3

-2

-1

0

1

2

3

First PC (39.6%)

Seco

nd P

C (

21.8

%)

0.0-0.4-0.8

0.5

0.0

-0.5

First PC (39.6 %)

Seco

nd P

C (

21.8

%)

fumaric acid

citric acid

acetic acid

shicimic acid

malic acid

tartaric acid

A

B

Fig. 2. Score plot (A) and loading plot (B) of first two principal compo-nents for classification of botrytised wines originating from Tokaj regionof Hungary and other countries. Variables: tartaric acid, malic acid,fumaric acid, citric acid, shicimic acid; s: Tokaji aszu, Tokaji aszu-essence, d: Foreign wines.


(mostly in foreign botrytised) to both directions from theorigin along the second PC axis (Fig. 2B) indicates the rela-tively high concentration of acetic acid or shicimic acid,respectively. The concentration of acetic acid (>1.76 mg l�1)in German and two Austrian wines or that of shicimic acid(>89.6 mg l�1) was higher in one Austrian and one Frenchwine than in all others.

3.3. Elements

Eleven Hungarian and 20 foreign botrytised wine sam-ples were analysed for elemental composition. Results ofdescriptive statistical analysis can be seen in Table 3.

The average sodium level in Hungarian botrytised and for-eign botrytised wines was 54.9 ± 18.7 mg l�1 and 21.2 ±8.1 mg l�1, respectively. Concentration of sodium wasslightly higher in Hungarian wine samples than in foreignwines. Potassium is one of the most abundant elementalconstituents in wines. The highest levels of potassium werefound in Swiss (2226 ± 168 mg l�1) and Austrian(1681 ± 794 mg l�1) samples. Hungarian botrytised winescontained 920 ± 320 mg l�1 potassium. Calcium content ofwines was similar in all samples, irrespectively of origin, andthe measured levels were in the range of 112.8 ± 26.2 mg l�1.Wines contain magnesium in relatively high level with anaverage concentration of 129.1 ± 33.4 mg l�1. The highestamounts of magnesium were detected in Hungarian(150.7 ± 19.3 mg l�1) and Austrian (140.3 ± 41.2 mg l�1)botrytised wines.

From microelements, iron is a general component in allgrape and wine varieties. Iron concentration depends onseveral factors, mostly on the soil, however iron levelsmay increase in wines due to the usage of steel devices dur-ing production. In the present study, 3.06 ± 1.66 mg l�1

iron concentration was detected on average. Copper con-tent of wines may be mainly originated from residues ofcopper-based pesticides in addition to transport from thesoil. The levels of copper found in the studied wines rangedbetween 0.48 ± 0.62 mg l�1. Presence of zinc in wines maybe originated from zinc-containing pesticides in addition totransport from the soil (Anwar, Farh, & Friedrich, 2002).The average zinc content of Hungarian botrytised and for-eign wines was 0.95 ± 0.97 mg l�1, and 1.31 ± 038 mg l�1,respectively. Manganese in small amount is a natural con-stituent of grape and wine. The measured values werebetween 1.23 ± 0.81 mg l�1 and 1.32 ± 0.81 mg l�1 for To-

kai Aszu and foreign wines, respectively. Hungarian winesamples had slightly lower amounts of manganese.

Tab

le3

Des

crip

tive

stat

isti

csfo

rel

emen

tso

fw

ine

sam

ple

s

Zn

(mg

l�1)

Mn

(mg

l�1)

Cu

(mg

l�1)

Fe

(mg

l�1)

Na

(mg

l�1)

K(m

gl�

1)

Mg

(mg

l�1)

Ca

(mg

l�1)

Cd

(lg

l�1)

TF

TF

TF

TF

TF

TF

TF

TF

TF

Mea

nva

lue

0.95

1.53

1.23

1.32

0.49

0.47

3.92

2.88

54.8

821

.24

920

1435

150.

711

7.2

123.

310

5.4

1.28

1.41

Sta

nd

ard

dev

iati

on

0.97

0.88

0.81

0.67

0.70

0.59

1.23

2.39

18.7

18.

0532

075

419

.332

.331

.635

.90.

640.

86M

edia

n0.

471.

380.

971.

190.

230.

193.

662.

1653

.32

20.8

181

910

9315

5.7

107.

112

7.1

94.7

1.13

1.13

Min

imu

m0.

080.

630.

120.

270.

100.

072.

030.

3818

.63

6.06

489

300

107.

982

.153

.272

.70.

140.

25M

axim

um

2.68

4.90

2.64

2.78

2.57

2.12

5.59

8.71

81.0

936

.98

1512

2843

174.

018

6.8

164.

422

7.3

2.31

4.03

Ran

ge2.

604.

272.

522.

512.

462.

053.

568.

3362

.46

30.9

210

2325

4366

.110

4.7

111.

215

4.5

2.17

3.78

Su

m10

.530

.513

.626

.55.

49.

543

.157

.760

442

510

125

2870

615

0723

4512

3321

0914

.128

.2K

urt

osi

s�

0.31

�0.

030.

839.

742.

14�

1.37

1.32

�0.

180.

32�

0.67

�1

2�

0.2

1.8

6.4

�0.

53.

473.

47S

kew

nes

s1.

140.

801.

093.

061.

76�

0.12

1.44

�0.

330.

240.

631

�1

0.9

�1.

12.

30.

01.

621.

62N

um

ber

of

sam

ple

s11

2011

2011

2011

2011

2011

2010

2010

2011

20L

evel

of

con

fid

ence

(95.

0%)

0.65

0.41

0.55

0.31

0.47

0.28

0.83

1.12

12.5

73.

7721

535

313

.815

.122

.616

.80.

430.

40

T:

To

kaj

ias

zuw

ines

;F

:b

otr

ytis

edw

ines

fro

mfo

reig

nco

un

trie

s.


Cadmium levels in wines may come from environmentalcontamination caused by industrial activity or by residuesof agrochemical products. The measured values of cad-mium were quite low, but the presence of this elementhas been detected in all samples. The average of resultswas in the range of 1.35 ± 1.02 lg l�1.

Fig. 3 shows the score plot and loading plot of the firstand second principal components and variables.

PCA allowed 51.1% of the total variance to be explainedby the first two principal components. Differentiation ofscores (Fig. 3A) of aszu wines and foreign botrytised wineswas not significant using eight elements as variables, how-ever a good tendency for separation of scores betweenHungarian and foreign wines samples can be observed. Itwas interesting that two wines from Austria were stronglysimilar to Hungarian Tokaji aszu wines, similarly to resultsof amines.

In Fig. 3B, the loading plot of variables shows that pri-marily Na and Zn (first PC > �0.78) took part in forma-tion of first principal component, while Fe (�0.91) wasthe main constituent of second principal component.Higher Na and lower Zn content of most Hungarian winesthan that of foreign wines took their first principal compo-

-0.5 0.0 0.5

-0.8

-0.4

0.0

First PC (26.5 %)

Seco

nd P

C (

24,6

%)

Zn

Mn

Cu

Fe

Na

K

Mg

Ca

Cd

3210-1-2

1

0

-1

-2

First PC (26.5 %)

Seco

nd P

C (

24.6

%)

Fig. 3. Score plot (A) and loading plot (B) of first two principal compo-nents for classification of botrytised wines originating from Tokaj regionof Hungary and other countries. Variables: Ca, Cd, Cu, Fe, K, Mg, Mnand Zn; s: Tokaji aszu, Tokaji aszu-essence, d: Austrian wines, j: Span-ish wines, N: Portuguese wines, h: Italian wines, 4: Slovakian wines, :Swiss wines, +: French wines, : German wine, : USA wine.


nent score to the left direction from origin on the first PCaxis. Scores of samples containing high Fe (>3 mg l�1)are characterised by high value to the left direction fromorigin on the second PC, just like one German and one Ital-ian and two Portuguese, Hungarian, and Austrian samples.

Using joined data matrix (amines, acid, elements), thethree Austrian wines above could not be also separatedfrom Hungarian wines using all compounds of compoundclasses as variables. This close similarity to Hungarianwines is not typical for all Austrian ones.

As a conclusion it can be established from the threetypes of compounds investigated, the amines, characteris-ing the effect of B. cinerea and the winemaking technology,are the most suitable component type for determination ofauthenticity and origin of botrytised wines. The informa-tion from composition of acids and elements can supportthe results of amine analysis in reaching the goal to deter-mine the origin of wines.

Acknowledgement

The authors thank Mr. Zoltan Zilai (president of Vinag-ora 2004) for his help in selecting samples from the winecompetition.

References

Anwar, G. J., Farh, M. A. N., & Friedrich, B. (2002). Pesticide and heavy

metals residue in wastewater, soil and plants in wastewater disposal

site near Al-Lajoun Valley, Karak/Jordan. Water, Air, and Soil Pollu-

tion, 133(1–4), 97–107.Arvanitoyannis, I. S., Katsota, M. N., Psarra, E. P., Soufleros, E. H., &

Kallithroka, S. (1999). Application of quality control methods for

assessing wine authenticity: Use of multivariate analysis (chemomet-

rics) – Review. Trends in Food Science and Technology, 10(10),321–336.

Bast, E. (1971). Occurrence and origin of volatile primary amines in bac-

teria. Archives of Microbiology, 79(1), 7–11.Bast, E. (1972). Biosynthesis of primary aliphatic monoamines by amino

acid-aldehyde transamination in Sarcina lutea. Archives of Microbiol-

ogy, 86(3), 203–210.Baxter, M. J., Crews, H. M., Dennis, M. J., Godall, I., & Anderson, D.

(1997). The determination of the authenticity of wine from its trace ele-

ment composition. Food Chemistry, 60(3), 443–450.Berrueta, L. A., Alonso-Salces, R. M., & Heberger, K. (2007). Supervised

pattern recognition in food analysis. Journal of Chromatography A,

1158(2), 196–214.Csomos, E., Heberger, K., & Simon-Sarkadi, L. (2002). Principal compo-

nent analysis of biogenic amines and polyphenols in Hungarian wines.Journal of Agriculture and Food Chemistry, 50(13), 3768–3774.

Day, M. P., Zhang, B. L., & Martin, G. J. (1995). Determination of the

geographical origin of wine using joint analysis of elemental and isoto-

pic composition. II – Differentiation of the principal production zones

in France for 1990 vintage. Journal of the Science of Food and Agricul-

ture, 67(1), 113–123.Eder, R., Brandes, W., & Paar, E. (2002). Influence of grape rot and fining

agents on the contents of biogenic amines in musts and wines. Mitteil-

ungen Klosterneuburg, Rebe und Wein, Obstbau und Fruchteverwertung,

52(5–6), 204–217.Eperjesi, I., Kallay, M., & Magyar, I. (1998). Boraszat (Winemaking) (3rd

ed.). Budapest: Mez}ogazda Kiado.

Etievant, P., Schlich, P., Cantagrel, R., Bertrand, A., & Bouvier, J. C.(1989). Varietal and geographic classification of French red wines in

terms of major acids. Journal of the Science of Food and Agriculture,

46(4), 421–438.Hajos, G., Sass-Kiss, A., Szerdahelyi, E., & Bardocz, G. S. (2000).

Changes in biogenic amine content of Tokaj grapes, wines and aszu-

wines. Journal of Food Science, 65(7), 1142–1144.Haraszti, Gy. (2002). Tokaj wines. Budapest: Kossuth Kiado (in English).Havadi, B., Kiss, J., Sass-Kiss, A., Adanyi, N., & Varadi, M. (2006). Clas-

sification of wine samples on the basis of elemental and amine compo-sition. In Proceedings of environmental aspects of trace element

research-water, soil, microorganisms, plants congress, Budapest (pp.246–250).

Heberger, K., Csomos, E., & Simon-Sarkadi, L. (2003). Principal compo-

nent and linear discriminant analyses of free amino acids and biogenic

amines in Hungarian wines. Journal of Agriculture and Food Chemistry,

51(27), 8055–8060.Hernandez, S. M., Sanchez, M. S., & de Tarlovsky, M. N. S. (2006). Poly-

amines as a defense mechanism against lipoperoxidation in Trypano-

soma cruzi. Acta Tropica, 98(1), 94–102.Hyotylainen, T., Savola, N., Lehtonen, P., & Riekkola, M. L. (2001).

Determination of biogenic amines in wine by multidimensional liquid

chromatography with online derivatization. Analyst, 126(12),2124–2127.

Igarashi, K. (2006). Physiological functions of polyamines and regulation

of polyamine content in cells. Yakugaku Zasshi – Journal of the Phar-

maceutical Society of Japan, 126(7), 455–471.Igarashi, K., & Kashiwagi, K. (2000). Polyamines: Mysterious modulators

of cellular functions. Biochemical and Biophysical Research Communi-

cations, 271(3), 559–564.Kallay, M., & Nyitraine, S. D. (2003). Tokaji borkulonlegessegek biogen-

amin-tartalmanak vizsgalata (Determination of the biogenic amine

content of dessert wines from Tokaj region). Boraszati Fuzetek, 1,16–20.

Kiss, J., Korbasz, M., & Sass-Kiss, A. (2006). Study of amine composition

of botrytized grape berries. Journal of Agriculture and Food Chemistry,

54(23), 8909–8918.Kiss, J., & Sass-Kiss, A. (2005). Protection of originality of Tokaji aszu.

Amines and organic acids in botrytised wines by HPLC. Journal of

Agriculture and Food Chemistry, 53(26), 10042–10050.Kordis-Krapez, M., Abram, V., Kac, M., & Ferjancic, S. (2001). Determi-

nation of organic acids in white wines by RP-HPLC. Food Technology

and Biotechnology, 39(2), 93–99.Lara, R., Cerutti, S., Salonia, J. A., Olsina, R. A., & Martinez, L. D.

(2005). Trace element determination of Argentine wines using ETAAS

and USN-ICP-OES. Food and Chemical Toxicology, 43(2), 293–297.Latorre, M. J., Garcıa-Jares, C., Medina, B., & Herrero, C. (1994). Pat-

tern- recognition analysis applied to classification of wines from Gali-

cia (Northwestern Spain) with certified brand of origin. Journal of

Agriculture and Food Chemistry, 42(7), 1451–1455.Lehtonen, P. (1996). Determination of amines and amino acids in wine –

A review. American Journal of Enology and Viticulture, 47(2), 127–133.Leitao, M. C., Marques, A. P., & San Romao, M. V. (2005). A survey of

biogenic amines in commercial Portuguese wines. Food Control, 16(3),199–204.

Marengo, E., & Aceto, M. (2003). Statistical investigation of the differ-

ences in the distribution of metals in Nebbiolo-based wines. Food

Chemistry, 81(4), 621–630.Martin, G. J., Mazure, M., Jouitteau, C., Martin, Y. L., Aguile, L., & Al-

lain, P. (1999). Characterization of the geographic origin of Bordeaux

wines by a combined use of isotopic and trace element measurements.American Journal of Enology and Viticulture, 50(4), 409–417.

Muranyi, Z., & Kovacs, Zs. (2000). Statistical evaluation of aroma and

metal content in Tokay wines. Microchemical Journal, 67(1–3), 91–96.Ribereau-Gayon, J., Ribereau-Gayon, P., & Seguin, G. (1980). Botrytis

cinerea in enology. In J. R. Coley-Smith, K. Verhoeff, & W. R. Jarvis(Eds.), The biology of botrytis (pp. 251–274). London: Academic PressInc.


Sass-Kiss, A., & Hajos, G. (2005). Characteristic biogenic amine compo-

sition of Tokaj Aszu-wines. Acta Alimentaria, 34(3), 227–235.Sass-Kiss, A., Szerdahelyi, E., & Hajos, G. (2000). Study of biologically

active amines in grapes and wines by HPLC. Chromatographia,

51(Suppl.), 316–320.Schreiber, R. C., Boeshore, K. L., Laube, G., Veh, R. W., & Zigmond, R.

E. (2004). Polyamines increase in sympathetic neurons and now neuro-

nal cells after axotomy and enhance neurite outgrowth in nerve growth

factor-primed PC12 cells. Neuroscience, 128(4), 741–749.

Vasconcelos, A. M. P., & Dasneves, H. J. C. (1989). Characterization of

elementary wines of Vitis vinifera varieties by pattern recognition of

free amino acid profiles. Journal of Agriculture and Food Chemistry,

37(4), 931–937.Vidal-Carou, M. C., Lahoz-Portoles, F., Bover-Cid, S., & Marine-Font,

A. (2003). Ion-pair high-performance liquid chromatographic determi-

nation of biogenic amines and polyamines in wine and other alcoholic

beverages. Journal of Chromatography A, 998(1–2), 235–241.

Documents

Multivariate statistical analysis of botrytised wines of different origin