MS methods

Accepted Manuscript Analytical methods Differentiation of wines according to grape variety and geographical origin based on volatiles profiling using SPME-MS and SPME-GC/MS methods Angelika Zió łkowska, Erwin Wąsowicz, Henryk H. Jeleń PII: DOI: Reference:

S0308-8146(16)31022-6 http://dx.doi.org/10.1016/j.foodchem.2016.06.120 FOCH 19461

To appear in:

Food Chemistry

Received Date: Revised Date: Accepted Date:

20 March 2016 29 June 2016 30 June 2016

Please cite this article as: Zió łkowska, A., Wąsowicz, E., Jeleń, H.H., Differentiation of wines according to grape variety and geographical origin based on volatiles profiling using SPME-MS and SPME-GC/MS methods, Food Chemistry (2016), doi: http://dx.doi.org/10.1016/j.foodchem.2016.06.120

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1 2

Differentiation of wines according to grape variety and geographical

3

origin based on volatiles profiling

4

using SPME-MS and SPME-GC/MS methods

5

Angelika Ziółkowska1 Erwin Wąsowicz1,2 and Henryk H. Jeleń1 *

6 1

7

Faculty of Food Science and Nutrition, Poznań University of Life Sciences, Wojska Polskiego 31, 60-642 Poznań, Poland

8 9

2

Current affiliation: Department of Sports Medicine and Traumatology, Poznań University of Physical Education,

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

*Corresponding author: [email protected], Tel: +4861-8487273, Fax: +4861-

25

8487314 1

26 27

ABSTRACT

28

Among methods to detect wine adulteration, profiling volatiles is one with a great

29

potential regarding robustness, analysis time and abundance of information for

30

subsequent

31

microextraction with direct analysis by mass spectrometry without compounds

32

separation (SPME-MS) was used for differentiation of white as well as red wines. The

33

aim was to differentiate between varieties used for wine production and to also

34

differentiate wines by country of origin. The results obtained were compared to SPME-

35

GC/MS analysis in which compounds were resolved by gas chromatography. For both

36

approaches the same type of statistical procedure was used to compare samples:

37

principal component analysis (PCA) followed by linear discriminant analysis (LDA).

38

White wines (38) and red wines (41) representing different grape varieties and various

39

regions of origin were analysed. SPME-MS proved to be more convenient in use due to

40

higher sample throughput.

data

treatment.

Volatile

fraction

fingerprinting

by

solid-phase

41 42 43

Keywords: SPME-MS, SPME-GC/MS, wine, authentication, volatiles

44

45

1. INTRODUCTION

46

Wine flavour is one of the most complex in food products and is dependent on

47

aroma compounds originating from grapes, fermentation process, ageing and storage.

48

Wine distinct flavour come from specific groups of compounds characteristic of some

49

grapes (i.e. terpenes in Muscat wines), and compounds which originate from 2

50

fermentation and also

maturation processes. However, key odorants form only a

51

relatively small fraction of wine volatile compounds. There are as many as 800 volatiles

52

detected in wine, among them mainly alcohols, esters, aldehydes and ketones.. There

53

are also numerous compounds representing other classes (terpenoids, norisoprenoids,

54

thiols, methoxypyrazines; Ebeler, 2001). The profile of the volatile compounds in wine

55

depends on many factors, such as geographical origin, grape variety, vintage and

56

growing conditions (Setkova, Risticevic and Pawliszyn, 2007; Jurado et al., 2008,

57

Azozarena, Caps, Marín & Navarro, 2000; Cozzolino, Smyth, Cynkar, Dambergs &

58

Gishen, 2005; Marti, Busto & Guash, 2004; Alvarez, Aleixandre, García, Casp &

59

Zúnica, 2003).

60

Various methods have been used for the differentiation and classification of wines

61

based on the volatile compounds analysis. They involve gas chromatography and gas

62

chromatography-mass spectrometry (GC-MS) (Rebolo, Pena, Latorre, García, Botana

63

and Herrero, 2000; Diaz, Conde, Mendez, Perez and Trujillo, 2002; Falqué, Fernández

64

and Dubourdieu, 2002; Câmara, Alves, and Carlos Marques, 2007). A different

65

approach, the use of electronic noses (e-noses) based on chemical sensors (Buratti,

66

Benedetti, Scampichio and Pangerod, 2004; Pilar Marti, Boque, Busto and Guasch,

67

2005) or on mass spectrometers (Cynkar, Dambergs, Smith and Cozzolino, 2010; Martí

68

et al., 2004, Cozzolino et al., 2005), to fingerprint wine flavour has been used.

69

In the MS-based electronic nose the mass analyser (spectrometer) is used as an

70

“array of sensors”. Signals acquired by this type of electronic nose are ions of specific

71

m/z and their intensities as registered by MS. These ions originate from the TIC (total

72

ion current) acquired from a single peak after introduction of volatile compounds into

73

the MS without chromatographic separation. To achieve such a “spectrum”, the

74

analytical column is replaced by a restricting capillary with no stationary phase. There is 3

75

usually no information on particular compounds (as the compounds are not separated)

76

but provides a fingerprint of each sample based on the uniqueness of certain ions (Jeleń,

77

Mildner-Szkudlarz, Jasińska & Wąsowicz, 2007). The main advantages of MS-based

78

electronic noses over classic electronic noses (based on metal oxide sensors, conducting

79

polymers or quartz balance) are their resistance to moisture, high sensitivity and

80

selectivity and some qualitative information (Pavón & Sánchez, Pinto, Lespada,

81

Cordero & Peña, 2006). Extraction methods based on headspace analysis are used to

82

introduce the sample. Methods that include sample pre-concentration step, such as

83

solid-phase microextraction (SPME) or purge-and-trap (P&T) provide better

84

discrimination than these based on static headspace techniques (Lozano, Santos,

85

Gutiérrez, Horillo, 2007). SPME-MS has been already used for the analysis of food

86

products of various types, such as milk, cheeses, adulteration of olive oil and botanical

87

origin of spirits (Marsili, 1999, Pérès, Viallon & Berdagué, 2001, Mildner-Szkudlarz &

88

Jeleń, 2008, Jeleń et al., 2010, Majcher et al., 2015).

89

Linear discriminant analysis (LDA), principal component analysis (PCA),

90

canonical analysis, HJ-biplot, and artificial neural networks (ANN) have already been

91

employed to process data and enabled classification of wines in regard to varieties and

92

regions of origin (Márquez, Castro, Natera, & Garcia-Barroso, 2008; Cozzollino et al.,

93

2005; Setkova et al., 2007; Römisch, Jäger, Capron, Lanteri, Forina & Smeyers-

94

Verbeke, 2009, Díaz, Conde, Méndez & Pérez Trujillo, 2003, Saurina, 2010)

95

The aim of this study was to evaluate the usefulness of a method based on solid-

96

phase microextraction combined with mass spectrometry (SPME-MS) for the

97

differentiation of both white and red wines. The wines were classified based on their

98

grape varietal origin, which was the main goal and also with regard to country of

99

production. For varietal classification there was an abundant representation of 3 white 4

100

varieties (38) and 2 red varieties (41 bottles) used for wine production. The wine came

101

from fourteen different countries of origin; therefore the aim in this comparison was to

102

indicate the capabilities of the system to classify samples.

103

The SPME-MS method was compared for the majority of wines with SPME-GC-

104

MS, where compounds were resolved on a chromatographic column prior to

105

multivariate statistical analysis. The same extraction type (SPME) was used for both

106

approaches, however different types of data were collected – averaged mass spectra and

107

peaks of particular compounds for subsequent statistical analysis. The same multivariate

108

techniques (PCA and LDA) were used to process data from both techniques.

109

110

2. EXPERIMENTAL Wines

111

In this study a total of 79 monovarietal wines were analysed: 38 white wines

112

obtained from 3 different grape varieties (Chardonnay, Sauvignon Blanc, Muscat) and

113

41 red wines (Cabernet Sauvignon and Merlot) were studied. For a comparative study

114

in-laboratory mixtures (2) of Cabernet Sauvignon and Merlot were prepared

115

Commercially available white wines were produced in Chile (9), USA (California) (8),

116

France (6), Bulgaria (4), Moldova (5), Spain (3), Argentina (1), Australia (1) and South

117

Africa (1), and red wines in Chile (11), Bulgaria (9), California (8), France (6) and

118

Moldova (7). Wines were purchased in wine shops in Poznań. All the wines were

119

bottled at the place of production and were produced within two years to minimise the

120

influence of vintage on discrimination.

121

Analytical equipment

122

For the SPME-MS analysis a Hewlett-Packard HP5890 series II gas chromatograph

123

coupled to a Hewlett-Packard 5971 quadrupole mass spectrometer was used. Capillary 5

124

analytical column was replaced with fused silica capillary tubing with no phase coating

125

(5 m × 0.2 mm; Supelco, Bellefonte, PA). For the SPME-GC-MS analysis an Agilent

126

7890A gas chromatograph with a split/splitless injection port, coupled with Agilent

127

5975C VL MSD quadrupole mass spectrometer was used. Compounds were resolved on

128

a Supelcowax-10 (30 m × 0.25 mm × 0.25 µm, Supelco, Bellefonte) column. SPME

129

injections were performed manually in both methods.

130

Extraction methods of volatile compounds

131

Analyses of wine samples were performed in two independent experiments: SPME-

132

MS and SPME-GC-MS. Prior to these experiments extraction conditions were

133

optimised for each method independently. Wines (10 mL) were placed in a 20-mL vial

134

and the vial was then tightly capped with PTFE/silicone septum. Different SPME fibres

135

were tested for their extraction efficiency at the same time (20 min.) and temperature

136

(50

137

Carboxen/divinylbenzene/polydimethylsiloxane

(CAR/DVB/PDMS),

138

divinylbenzene/polydimethylsiloxane

and

139

polydimethylsiloxane (CAR/PDMS). All these fibres were 1 cm long except for the

140

CAR/DVB/PDMS, which was 2 cm. PA fibre and CAR/DVB/PDMS fibres were

141

chosen based on the highest total peak areas for SPME-MS and SPME-GC-MS

142

experiments, respectively. In the next step extraction times ranging from 2 to 30

143

minutes were compared. Volatile compounds were sampled at 50 0C after 5 min

144

preheating. For SPME-MS experiment the highest total area responses was obtained

145

after 2 minutes of extraction so this time was selected for the study. For SPME-GC-MS

146

experiment total peak area was highest for 20 min extraction and as total analysis time

147

is influenced mainly by the GC run time 20 min sampling time was selected.

148

Compounds were desorbed for 3 minutes at 280 °C in splitless mode, using a 0.75 mm

o

C):

polydimethylsiloxane

(PDMS),

(DVB/PDMS)

polyacrylate

(PA),

Carboxen/

6

149

dedicated SPME liner. In the SPME-MS experiment GC-MS operating conditions were

150

as follows: helium flow 0.4 mL min‒1, oven temperature 250 oC (isothermal). The

151

spectrometer operated in electron impact (EI) mode (70 eV). The ion source was

152

indirectly heated by transfer line set to 280 oC. Detection was carried out in scan mode

153

over a range of m/z 29–289. In the SPME-GC-MS experiment helium flow was 0.8 mL

154

min-1, initial oven temperature was 40 oC (1 min), then 15 oC min‒1 to 140 oC and 50oC

155

min‒1 to 260 oC (1.5 min). The spectrometer operated in electron impact mode (EI, 70

156

eV) at 240 0C. Detection was carried out in full scan mode over a range of m/z 29–289.

157

Each sample was run in 3‒5 replicates for SPME-MS and 2‒3 replicates for SPME-

158

GC/MS.

159

Statistical analysis

160

Statistical analyses were performed using Statistica 8.0 software (Statsoft, Tulsa,

161

OK) equipped with a multivariate statistics package. Principal component analysis

162

(PCA) and linear discriminant analysis (LDA) were used for sample discrimination and

163

classification. In these studies, external validation method was used, dividing each data

164

set into a training set (used to calculate the classification rules – approximately 80% of

165

samples) and an evaluation set (used to evaluate the prediction ability of rules and

166

models – 20% of samples), or with the leave-one-out method when the number of

167

samples was limited. To indicate clustering in LDA, prediction ellipses were created

168

using Statistica, based on normal distribution with a coefficient of 0.95.

169 170

3. RESULTS

171

Differentiation of white wines and red wines of different grape varieties using

172

SPME-MS and SPME-GC-MS

7

173

White wines (38) of three different grape varieties: Chardonnay, Sauvignon Blanc

174

and Muscat, and red wines (41) produced from Cabernet Sauvignon and Merlot grapes,

175

from different countries were analysed, in order to classify them according to grape

176

varieties. SPME-MS analysis produced a single peak for which an averaged spectrum

177

represents mixed spectra of compounds transferred from headspace to mass

178

spectrometer. Although the identification of particular compounds on a basis of an

179

“averaged” mass spectrum is usually not possible, the presence of certain ions can

180

provide some hints on the character of analysed compounds. The most abundant ions in

181

all spectra were the ion of m/z 31, which is base peak and characteristic ion for both

182

methanol and ethanol, and m/z 45, characteristic for ethanol. That is in accordance with

183

the presence of methanol and ethanol in the analysed wines. However, no ion intensities

184

vs amounts of methanol or ethanol in wines correlations can be expected as both ions

185

are of relatively low specificity. Other ions which could be observed on a spectrum

186

were ions m/z 44 which can be related to aldehydes, m/z 41, attributed among other

187

compounds to higher alcohols, m/z 43 which is characteristic for acetates and m/z 88

188

which is related to ethyl esters. One must bear in mind that low m/z fragments are

189

usually of limited usefulness in drawing conclusions on identities of particular

190

compounds or compound groups.

191

SPME-MS data were transferred from their tabulated form in MSD Chemstation to

192

Excel software and a table was built with intensities over m/z values. To facilitate

193

proper processing of samples, if ions of certain m/z were not observed (0 intensity) in

194

more than 50% of samples such ions were eliminated from the table. . Data acquired for

195

all the samples were processed using logarithmic transformation (logx+1) because of

196

significant differences in ions intensities and also to compensate for the occurrence of

8

197

zero values in matrix data. Then all variables were autoscaled (Berrueta, Alonso-Salces

198

& Herberger , 2007). Afterwards all data were subjected to multivariate analysis.

199

As a first step, principal component analysis (PCA) was carried out in order to

200

visualise eventual grouping of samples (Figure 1). White and red wines analysed by

201

SPME-MS yielded visible clusters on the PCA graph (Figure 1A and 1B) with the first

202

two components representing 67.61% and 91.98% variance for white and red wines,

203

respectively. Discriminant ions for white (m/z 31, 32, 35, 42, 43, 60, 72, 73) and red

204

wines (m/z 30, 31, 32, 40, 41, 42, 43, 44, 46) were similar. As can be seen in Figure 1A

205

Chardonnay formed a uniform cluster (with one Muscat outlier in it); similar grouping

206

was observed for Muscat. Sauvignon Blanc formed two clusters, one of them very close

207

to Muscat. Based on PCA, two clusters for two grape varieties were observed in the

208

case of red wines (Figure 1B). Cabernet Sauvignon/Merlot mixtures were not included

209

using SPME-MS and PCA.

210

SPME-GC-MS data were transferred from Agilent MSD Chemstation software to Excel

211

similarly to the above methodology, but the table was built with identified compounds

212

vs. their peak intensities (TIC areas). Then data were processed in the same way as for

213

the SPME-MS approach. For discrimination in PCA isoamyl acetate, ethyl hexanoate,

214

ethyl octanoate, ethyl dodecanoate, 2-phenylethanol and furfural were selected for white

215

wines and isobutanol, 1-hexanol, ethyl hexanoate, ethyl octanoate, ethyl decanoate and

216

ethyl dodecanoate for red wines. In the case of the SPME-GC-MS method the clustering

217

was less pronounced compared to SPME-MS, especially for white wines, where the first

218

two components represented only 45.84% variance (Figure 1C). For red wines analysed

219

by SPME-GC-MS, apart from Cabernet Sauvignon and Merlot monovarietal wines,

220

wines produced from a mixture of these varieties were also included (Figure 1D). The

9

221

first two PCs represented 64.27% of total variance. As indicated in Figure 1 PCA for

222

SPME-MS data presented better group discrimination than for SPME-GC-MS.

223

To improve classification of investigated samples, in a next step linear discriminant

224

analysis (LDA) was carried out. LDA is a supervised classification technique and is

225

based on the determination of linear discriminant functions, which maximise the ratio of

226

between-class variance and minimise the ratio of within-class variance. Features were

227

selected by forward stepwise selection, which chooses variables producing the highest

228

F-values and the highest decrease of Wilks’ λ. To assess the discriminant capacity of

229

variable F-tests and Wilks’ λ tests were performed. LDA was performed, not based on

230

ions selected in PCA, but using all ions from the matrix. For SPME-MS method ions

231

with the best discrimination power (highest F-values) were selected for each group of

232

wines. For white wines the following ions were chosen: m/z 30, 31, 32, 33, 35, 42, 43,

233

47, 57, 60, 72, 73, 74, 75, 90, 99, 120, 123 and 124. For classification of red wins the

234

following ions were selected: m/z 30, 32, 36, 39, 40, 43, 48, 56, 60, 64, 72, 77, 79, 82,

235

85, 92, 93, 102, 107 and 129. Figure 2 shows the clusters of examined white (Fig. 2A)

236

and red (Fig. 2B) wines after LDA analysis based on SPME-MS data. In both cases the

237

discrimination between sample clusters was very good. To assess the ability of the

238

model to identify wine authenticity and classify wine according to grape variety, all

239

samples were divided randomly into a learning set and a verification (testing) set.

240

Classification and prediction ability for the model built for white and for red wines

241

using SPME-MS-LDA was 100%.

242

When the same statistical approach was used to classify wine samples based on SPME-

243

GC-MS data the results were not resolved satisfactorily: Figures 2C and 2D show

244

classification of white and red wines, respectively, obtained after SPME-GC-MS data

245

analysis. Although it was possible to distinguish separate clusters for red wine, it was 10

246

not possible for white wines. The compounds with the highest discriminating power for

247

white wines in this method were isoamyl acetate, furfural, ethyl octanoate, ethyl

248

decanoate and ethyl dodecanoate. For red wines 1-hexanol, ethyl decanoate and 2-

249

phenylethanol were discriminant compounds in LDA. The ability of a model to classify

250

and predict red wines of different varieties correctly based on SPME-GC-MS was 95%.

251

The SPME-GC-MS-LDA method was applied also by Câmara et al. (2007) for the

252

differentiation of white wines of different grape varieties (Boal, Malvazia, Sercial and

253

Verdelho) using the same SPME fiber but the extraction lasted 120 min.

254

Differentiation of white wines of different geographical origin using SPME-MS

255

and SPME-GC-MS

256

To focus on the most informative approaches in samples discrimination for

257

geographical origin classification only LDA data are presented. Figure 3 shows data on

258

white wines analysis using SPME-MS (3A, 3C, 3E for Chardonnay, Sauvignon Blanc

259

and Muscat, respectively) and the same varieties analysed using SPME-GC-MS (3B,

260

3D, 3F, respectively). LDA evaluation of data of analyzed samples was carried out for

261

the white wines produced in various countries: Chile, USA (California), Bulgaria,

262

France, Moldova, Spain, Argentina, Australia and South Africa. Looking at data in

263

Figure 3 shows that differentiation using SPME-MS within white wines for all analysed

264

grape varieties (Chardonnay, Sauvignon Blanc and Muscat) was excellent when

265

geographical regions were compared. Similarly, classification of white wines from

266

different geographical origins was also satisfactory, not for Chardonnay, but for

267

Sauvignon Blanc and Muscat when SPME-GC/MS was used. LDA has been also

268

applied to the differentiation of wines by Falqué et al. (2002), Câmara et al. (2007), and

269

Bevin, Dambergs, Fergusson and Cozzolino (2008). Falqué used gas chromatography

270

with a flame ionisator detector (GC-FID) and GC–MS for the classification of white 11

271

wines Loureira, Dona Branca and Treixadura. GC-MS coupled with HS-SPME was also

272

adopted by Câmara to distinguish white wines of different grape varieties. Each of these

273

authors obtained good results using LDA but usually using longer extraction times.

274

The classification and prediction abilities of the model with regard to grape variety

275

for white wines for SPME-MS were 100%. LDA in the treatment of SPME-MS data is a

276

reliable tool for white wine samples recognition and prediction. SPME-MS has also

277

been used by other authors coupled with other statistical methods like PLS (Marsili,

278

2000) and SIMCA (Laguerre M, Mestres C, Davrieux F, Ringuet J, Boulanger R. 2007)

279

with satisfactory results.

280

For SPME-GC-MS of wine samples of particular varieties, country of origin

281

classification was also satisfactory in the majority of cases. The only unresolved clusters

282

were for Chardonnay wines from France and Chile (Fig. 3B). The most discriminant

283

variables were isoamyl acetate, furfural, decanoic acid ethyl ester, dodecanoic acid ethyl

284

ester and octanoic acid ethyl ester. These volatiles are mostly the same as in PCA;

285

therefore differences between results obtained using PCA and LDA depend only on the

286

different mathematical analysis used in these techniques. Most discriminating

287

compounds in white wines were simultaneously the most abundant when extracted by

288

SPME. Esters are the main volatile compounds in wines; therefore, they were expected

289

to influence data preparation for multivariate analysis. Moreover, ethyl esters of short

290

chain fatty acids (ethyl caproate, caprylate etc.) are usually very well adsorbed on the

291

SPME fibre as their partition constants favour their adsorption in the fibre.

292 293

Classification of red wines of different geographical origin using SPME-MS and

294

SPME-GC-MS

12

295

Linear Discriminant Analysis (LDA) was carried out, in order to obtain suitable

296

classification of red wines of various geographical origin. In this case ions with the best

297

discrimination power were those having the m/z values 30, 32, 36, 39, 40, 43, 48, 56,

298

60, 64, 72, 77, 79, 82, 85, 92, 93, 102, 107 and 129. Four of these ions were the same as

299

in PCA but the remaining ones were different. Figure 4 shows samples groupings based

300

on LDA showing differentiation of red wines according to varieties and regions of

301

origin. Good classification of the samples of Cabernet Sauvignon and Merlot was

302

achieved using SPME-MS, though for Cabernet Sauvignon originating from Chile and

303

Bulgaria their volatile profiles were so different from the remaining countries that the

304

rest of samples formed a single cluster separable only after extraction from the original

305

set of data (Fig. 4A1). For Merlot the cluster separation was satisfactory. The

306

classification ability for the model was 93%. Similar results were obtained by Bevin et

307

al. (2008) for differentiation of Cabernet Sauvignon, Merlot and Syrah based on mid-

308

infrared spectroscopy (MIR). Good differentiation of red wines according to

309

geographical origin was achieved by Rebolo et al. (2000), who classified Galician

310

Ribera Sacra wines using GC-FID data with pattern recognition analysis coupled also

311

with LDA.

312

Classification and prediction abilities of the model for SPME-MS, with regard to

313

different grape varieties, was 100%. Based on geographical origin, in the case of

314

Cabernet Sauvignon the classification and prediction ability of the model was also

315

100% (after extracting a subset of samples as in Fig. 4A1). However, in the case of

316

Merlot, the classification ability was 100% but the prediction ability was 90%.

317

When SPME-GC/MS was used, discrimination of volatile compounds using PCA

318

(results not shown) was based on the following compounds: isobutanol, 1-hexanol,

319

hexanoic acid ethyl ester, octanoic acid ethyl ester, decanoic acid ethyl ester, 13

320

dodecanoic acid ethyl ester. When LDA was carried out on the data, compounds with

321

the best discrimination power were 1-hexanol, decanoic acid ethyl ester, dodecanoic

322

acid ethyl ester and phenylethyl alcohol. As can be observed in Figure 4B and C,

323

sample discrimination using SPME-GC/MS was much worse than for SPME-MS.

324

4. Conclusions

325

Comparing obtained results for both approaches, SPME-MS provided much better

326

separation of clusters after LDA compared to SPME-GC/MS. As presented here SPME-

327

MS can provide reliable data especially for grape origin of wines. SPME-MS is a very

328

fast method, due to the lack of chromatographic separation, and as a consequence can

329

yield results in a short time, which is particularly important in analysing large sets of

330

samples and replicates required for statistical treatment. Analysis time using SPME-MS

331

is 10 times shorter than for SPME-GC/MS. Moreover, this method does not require a

332

dedicated instrument, but can be used with any GC/MS system after exchanging

333

analytical column for a restricting capillary.

334

5. Conflict of interest statement

335

Angelika Ziółkowska, Erwin Wąsowicz and Henryk Jeleń declare no conflict of

336

interest.

337 338

339

6. References

340

1. Ebeler S. E., (2001).Unlocking the secrets of wine flavor. Anal. Chem. 17: 45-64

341

2. Setkova L., Risticevic S., Pawliszyn J. (2007). Rapid headspace solid-phase

342

microextraction-gas chromatographic–time-of-flight mass spectrometric method for 14

343

qualitative profiling of ice wine volatile fraction II: Classification of Canadian and

344

Czech ice wines using statistical evaluation of the data. J. Chromatogr. A. 1147:

345

224–240.

346

3. Jurado J. M., Ballesteros O., Alcázar A., Pablos F., Martín M. J., Vilchez J. L.,

347

Navalón A. (2008). Differentiation of certified brands of origins of Spanish white

348

wines by HS-SPME-GC and chemometrics. Anal. Bioanal. Chem. 390: 961–970.

349

4. Arozarena I., Caps A., Marín R., Navarro M. (2000). Multivariate differentiation of

350

Spanish red wines according to region and variety. Journal of the Science of Food

351

and Agriculture 80: 1909-1917.

352

5. Cozzolino D., Smyth H. E, Cynkar W., Dambergs R. G., Gishen M. (2005).

353

Usefulness of chemometrics and mass spectrometry-based electronic nose to

354

classify Australian white wines by their varietal origin. Talanta 68: 382–387.

355

6. Martí M. P., Busto O., Guasch J. (2004). Application of a headspace mass

356

spectrometry system to the differentiation and classification of wines according to

357

their origin, variety and ageing. J. Chromatogr. A. 1057: 211–217.

358

7. Alvarez I., Aleixandre J. L., García M. J., Casp A., Zúnica L. (2003). Geographical

359

differentiation of white wines from three subzones of the designation of origin

360

Valencia. Eur. Food. Res. Technol. 217: 173–179.

361

8. Rebolo S., Peña R. M., Latorre M. J., García S., Botana A. M., Herrero C. (2000).

362

Characterisation of Galician (NW Spain) Ribeira Sacra wines using pattern

363

recognition analysis. Anal. Chim. Acta 417: 211–220.

364

9. Díaz C., Conde J. E., Méndez J. J., Pérez Trujillo J. P. (2002). Chemometric studies

365

of bottled wines with denomination of origin from the Canary Islands (Spain). Eur.

366

Food. Res. Technol. 215: 83–90.

15

367

10. Saurina J. (2010). Characterization of wines using compositional profiles and chemometrics. Trends Anal. Chem. 29: 234-245.

368 369

11. Falqué E., Fernández E., Dubourdieu D. (2002). Volatile Components of Loureira, Dona Branca, and Treixadura Wines. J. Agric. Food Chem. 50: 538-543.

370 371

12. Câmara J. S., Alves M. A., Carlos Marques J. C. (2007). Classification of Boal,

372

Malvazia, Sercial and Verdelho wines based on terpenoid patterns. Food Chem.

373

101: 475–484.

374

13. Buratti S., Benedetti S., Scampicchio M., Pangerod E. C. (2004). Characterization

375

and classification of Italian Barbera wines by using an electronic nose and an

376

amperometric electronic tongue. Analytica Chim. Acta 525: 133–139..

377

14. Pilar Marti M., Boque R., Busto O., Guasch J. (2005). Electronic noses in the quality control of alcoholic beverages. Trends in Anal. Chem. 24: 57-66.

378 379

15. Cynkar W., Dambergs R., Smith P., Cozzolino D. (2010). Classification of

380

Tempranillo wines according to geographic origin: Combination of mass

381

spectrometry based electronic nose and chemometrics. Analytica Chim. Acta 660:

382

227–231.

383

16.

Jeleń H. H., Mildner-Szkudlarz S., Jasińska I., Wąsowicz E. (2007). A headspace-

384

SPME-MS metod for monitoring rapeseed oil autooxidation. J. Amer. Oil Chem.

385

Soc. 84: 509-571.

386 387

17. Marsili R. T., (1999). SPME-MS-MVA as an electronic nose for the study of offflavors in Milk. J. Agric. Food Chem. 47: 648-654.

388

18. Pérès C., Viallon C., Berdagué J-L. (2001). Solid phase microextraction – mas

389

spectrometry: A New approach to the rapid characterization of Cheeses. Anal.

390

Chem. 73: 1030 – 1036.

16

391

19. Mildner-Szkudlarz S., Jeleń H. H. (2008). The potential of different techniques for

392

volatile compounds analysis coupled with PCA for the detection of the adulteration

393

of olive oil with hazelnut oil. Food Chem 110: 751-761.

394

20. Jeleń H. H., Ziółkowska A., Kaczmarek A. (2010). Identification of the botanical

395

origin of raw spirits produced from rye, potato, and corn based on volatile

396

compounds analysis using a SPME-MS method. J. Agric. Food Chem. 58: 12585-

397

12591.

398

21. Majcher M., Kaczmarek A., Klensporf-Pawlik D., Pikul J., Jeleń H. (2015) SPME-

399

MS-based electronic nose as a tool for determination of authenticity of PDO

400

cheese, Oscypek. Food Anal. Methods, 8: 2211-2217.

401

22. Pavón J. L. P., Sánchez M. de Nogal., Pinto C. G., Laespada E. F., Cordero B. M.,

402

Peña A. G. (2006). Strategies for qualitative and quantitative analyses with mass

403

spectrometry-based electronic noses. Trends in Anal. Chem. 25: 257-266.

404

23. Lozano J., Santos J. P., Gutierrez J, Horillo M. C. (2007). Comparative study of

405

sampling systems combined with gas sensors for wine discrimination. Sensors and

406

Actuators B. 126: 616-623.

407

24. Márquez R., Castro R., Natera R., García-Barroso C. (2008). Characterisation of

408

the volatile fraction of Andalusian sweet wines. Eur. Food. Res. Technol. 226:

409

1479–1484.

410

25. Römisch U., Jäger H., Capron X., Lanteri S., Forina M., Smeyers-Verbeke J.

411

(2009). Characterization and determination of the geographical origin of wines.

412

Part III: multivariate discrimination and classification methods. Eur. Food. Res.

413

Technol. 230: 31–45.

17

414

26. Díaz C., Conde J. E., Méndez J. J., Pérez Trujillo J. P. (2003). Volatile compounds

415

of bottled wines with Denomination of Origin from the Canary Islands (Spain).

416

Food Chem. 81: 447–452.

417 418

27. Berrueta L. A., Alonso-Salces R. M., Heberger K. (2007). Supervised patter recognition In food analysis. J. Chromatogr. A. 1158: 196-214.

419

28. Bevin C. J., Dambergs R. G., Fergusson A. J., Cozzolino D. (2008). Varietal

420

discrimination of Australian wines by means of mid-infrared spectroscopy and

421

multivariate analysis. Anal. Chim. Acta 621: 19–23.

422

29. Marsili R. T. (2000). Shelf-life prediction of processed milk by solid phase

423

microextraction, mass spectrometry and multivariate analysis. J. Agric. Food

424

Chem. 48: 3470-3475.

425

30. Laguerre M., Mestres C., Davrieux F., Ringuet J., Boulanger R. (2007). Rapid

426

discrimination of scented rice by solid-phase microextraction, mass spectrometry,

427

and multivariate analysis used as mass sensor. J. Agric. Food Chem. 55: 1077-

428

1083..

429 430 431 432 433 434 435 436

18

437 438

Figure Captions

439

Figure 1. PCA projection of white (A, C) and red (B, D) wines made from different

440

grape varieties based on correlation analysis obtained using SPME-MS (A, B) and

441

SPME-GC-MS (C, D) methods

442

Figure 2. LDA classification of white (A, C) and red (B, D) wines made from different

443

grape varieties based on correlation analysis obtained using SPME-MS (A, B) and

444

SPME-GC-MS (C, D) methods

445

Figure 3. SPME-MS-LDA (A, C, E) and SPME-GC-MS-LDA (B, D, F) analysis of

446

volatile fraction of white wines of different regional origin; Chardonnay (A, B),

447

Sauvignon Blanc (C, D), Muscat (E, F).

448

Figure 4. SPME-MS-LDA (A1, A2, B) and SPME-GC-MS-LDA (C, D) analysis of

449

volatile fraction of red wines of different regional origin; Cabernet-Sauvignon (A, A1,

450

C), Merlot (B, D).

451 452 453 454

19

455 5

A

4 3

Component 2: 16,19%

B

SB SB SB SB SB SB SB SB SB SB SB

2 M M M M M M MM MM M M M M M M M MM M M M M M M M M M M M MSB M SB M MMM M M M SB M SB SB SB SB SB SB SB SB M SBSB SB SB SB SB SB SB SB SB SB SB SB SB SB

1 0 -1 -2

SB

SB Ch ChCh Ch Ch Ch Ch Ch Ch Ch Ch ChCh Ch CCh h MCh Ch Ch Ch Ch Ch Ch Ch Ch Ch Ch Ch Ch

-3 -4 -6

-5

-4

-3

-2

-1

0

1

2

3

4

5

6

7

Component 1: 51,42%

456

Ch - Chardonay SB - Sauvignon Blanc M - Muscat

50000 40000 CS 30000

CS CS CS

20000 Component 2: 12,14%

D

C

CS

CS CS CS CS CS CS CS CS CS CS CS CS CS CSCS CS CS CS CS CS CSCS CS CS CS CS CS CS CSCS CS M CSCS CS CS CS CS M M CS CS CS CSCS M M CS M CS CS M MM CSCS CS CS MM MMMM M M M M MMM CS CS M MMM M M CS MM MMMM M MM M MM M

10000 0 -10000 -20000 -30000 -40000 -50000 -1,5E5

457

-1E5

-50000

0

50000

Component 1: 79,84%

1E5

1,5E5 CS - Cabernet-Sauvignon M - Merlot

20

3

2 Ch Ch Ch

SB Ch SB SB SBSB Ch Ch CCh h M Ch Ch Ch Ch SB Ch Ch Ch M MCh SB Ch SB Ch Ch Ch SB M SB Ch SB M Ch M ChSB Ch SB Ch Ch SBSB MM SB M M SBM MM M SB MMM Ch M M M Ch

Component 2: 20,85%

1

0

M

-1 SB SB M

-2

M M

-3

-4

-5 -3,5 -3,0 -2,5 -2,0 -1,5 -1,0 -0,5 0,0

0,5

1,0

1,5

2,0

Component 1: 24,99%

458

2,5

3,0

Ch - Chardoany SB - Sauvignon Blanc M - Muscat

1,2 1,0 0,8

CS

0,6

CS-M CS CSCS CS CS CS CS CS CS CS CS CS-M

Component 2: 17,41%

0,4 0,2

M M M M MM

CS-M 0,0

CS CS CS CS CS

-0,2

M

CS-M

-0,4

M MM CS CS M M CS M M M M MM M

-0,6

M

M

-0,8 -1,0 -1,2 -1,4 -2,5

-2,0

-1,0

-0,5

0,0

0,5

Component 1: 46,86%

459 460 461

-1,5

1,0

1,5

2,0 CS - Cabernet Sauvignon M - Merl ot CS-M - Cabernet S.-Merlot

Figure 1

462 463 464

21

A

B

465

466

C

D

467

468 469 470

Figure 2

22

A

B

C

D

471

472

473

474

23

E

F

475

476 477 478

Figure 3.

479 480 481

24

A1

B

482

483

A2

C

484

485

25

D

486 487 488

Figure 4

489 490 491 492 493

26

494

495

Highlights • 38 white and 41 red wine samples from different grape varieties and countries

496

were analyzed

497

• SPME-MS-LDA

498

differentiation

499

•

and

SPME-GC/MS-LDA

approaches

were

used

for

SPME-MS-LDA is preferred as a rapid tool to classify wines

500

27