Characterization and Classification of Romanian Wines by Origin A chemometric approach based on some metals and phenolic composition

There is a growing trend in food industry in combining safety and quality attributes of products with their distinct regional identity. In this study we used data on several elements (Cd, Cu, Cr, Zn, Pb and Ni) and phenolic compounds (gallic acid, syringic acid, p-coumaric acid, (+)-catechin, (-)-epicatechin, trans-resveratrol, rutin, quercetin and vanillin) as complementary markers in the attempt to characterise and differentiate wines from three winegrowing areas (Murfatlar, Recas and Jidvei) of Romania, according to their geographical and varietal origin. Wines from south-eastern region (Murfatlar) were characterized by their lower content in Cu, Cr and Ni, compared with wines from western (Recas) and central (Jidvei) region, while the content of phenolic compounds in the wine samples showed a visible variation, depending both of the grape variety and geographical origin. The statistical processing tools applied to our results allowed robust diûerentiation between regions, wine varieties, and vintage year. The phenolic compounds (gallic acid, (+)-catechin, (-)-epicatechin, rutin, quercetin and resveratrol) showed the best correlations for the varietal discrimination of wines. Using discriminant analysis, a correct classification of the wines by variety was achieved in proportion of 88.89% (based only by phenolic compounds) and 92.59 % (based both on phenolic compounds and metals), revealing that the content of Cu, Cr and Ni is significant for discrimination. Moreover, a 100% successful classification of wines by region of origin was accomplished.

The guaranteed quality (e.g. taste, texture, flavor, etc.) of many foodstuffs, including wine, is owed to the specific conditions encountered in the cultivation area. The local characteristics of climate and soil combine to produce crops with certain specific features. Sometimes, people are tempted to confer the consumption of products, mainly for certain foods such as wine, an aura of romance and tradition, reason for many consumers to demand for a guarantee of the product in terms of naturalness and authenticity.
Wine is a complex matrix which contains many classes of compounds such as sugars, alcohols, acids, flavonoids (tannins), non-flavonoids, minerals and proteins among others. Its composition and distinctive character are basically the result of interactions between several factors, including soil, climate, and environment, grape varieties, and local viticulture practices [1,2]. Therefore, a thorough investigation of the wine regarding its chemical composition can provide essential information for both authentication and quality certification, but also data concerning the vintage year and area of origin [3].
Stable isotopes are primary origin markers used to authenticate wines [4], both in terms of geographic [5,6] and variety [7] differentiation. Combined with other suitable indicators, as organic compounds [8], a better classification and assessment of origin may be provided. However, the possibility to distinguish between near small regions of appellation remains a challenging issue.
When considering the discrimination of the geographical origin of wine and wine authentication, another valuable tool for verifying the origin has been proved to be the quantification of trace elements in wine [9][10][11][12] and some metals isotope ratio, such as 87 Sr/ 86 Sr [13][14][15].
The recent research studies revealed the existence of a good correlation between the content of cations (Zn, W, Mn, Fe, Mg, Cu, Ca, Al, Sr, B, Na, K) in wines with the ones in soils from the area/vineyard of origin [16]. Basically, the multielemental composition of wine will reflect the geochemistry of provenience soil. However, there are several factors such as environmental contamination, agricultural practices, climate change or the winemaking processes that can significantly affect the multi-element composition of wines, endangering thus the direct correlation between the soil and wine composition. As a result, the applicability of multi-element analysis of the wine as fingerprinting technique could be narrowed only for characterization of those high-quality monovarietal wines for which the influence of winemaking processes against wine pattern has been studied previously and is permanently controlled. In Romania, a multielemental analysis performed on benchmark wines originated from three major regions producing wine (Dobrogea, Moldova and Valea Calugareasca) highlighted that the elements Mn, Cr, Sr, Ag and Co can be useful indicators for differentiating region of origin [12].
Wines were also classified based on the phenolic compounds [17,18] which are significant secondary metabolites of plants concentrated in the husks, seeds and pulp of beans grape and wine extracts [2,19] contributing to the formation of specific characteristics of wine such as colour, aroma, mouth-feel (astringency) and flavour, the most abundant and functional polyphenols being flavan-3-ol monomers, -catechin and (-)-epicatechin. [20]. These compounds are responsible for different wine types [21]. The structure of these compounds differs depending on the fraction of the grape, through their mean degree of polymerization. [20].
In the attempt to find reliable methods of authentication, several other possibilities to differentiate wines were provided over time, based on the aromatic compounds [21,22], amino acids, combinations between trace elements and stable isotopes or elemental composition and phenolic compounds [23].
The main goal of this work was to assess the usefulness of common investigated parameters for wine characterisation, like metals and phenolics, in classifying and differentiating wines in terms of their geographical and variety origin. Thus, a characterisation of three selected wine types from three different renowned wine-growing areas (Murfatlar, Recas and Jidvei) of Romania was performed based on data regarding several elements and phenolic compounds. For an accurate classification of the samples, the data were modelled by chemometrics.

Experimental part Materials and methods Samples
Twenty-seven samples of wine (crops 2010, 2011 and 2012) obtained from grapes of three different Vitis vinifera varieties, Chardonnay, Muscat Ottonel and Pinot Noir, grown in Murfatlar (Dobrogea region, south-eastern Romania), Recas (Banat region, western Romania) and Jidvei (Tarnave-Transylvania region, central Romania) vineyards were used in our work. In the experimental investigation, for each type of wine were selected samples from each vineyard and vintage year. Samples obtained directly from wineries were collected in 750 mL glass bottles and stored prior to analysis at 15 o C and protected from light. All the wine samples were analyzed in two replications.

Characterization of the selected vineyards
Three vineyards situated in the central, western and southeastern regions of Romania were selected for this work, due to their differentiation in climatic conditions and relief variability and diversity. Jidvei vineyard lies on the gentle hills of southern Transylvania, on the Tarnave's Valley, in central Romania. The land of the vineyard with its brown soil and the continental climate, not especially warm, with average temperatures of +9 °C, favours during the maturation of grapes special conditions for slow oxidation reactions which helps to continuously accumulates sugars and aromas and prevents the accumulation of acids. These natural conditions seem to be especially favourable to the cultivation of white grape varieties.
Located in western Romania, the Recas vineyard is situated on top of a hill, with gentle slopes. The climate is temperate continental with slight Mediterranean influences. Here the winters are generally mild, the summers are warm, the autumns are long and the transition from winter temperatures to summer ones is rather sudden. The oenoclimatic conditions in the area are favourably for the production of high quality red wines.
Murfatlar vineyard is situated in the south-eastern part of Romania, on hilly terrains situated between the Danube River and the Black Sea. The continental climate, the sheltering effect of the relief against the cold currents coming from the north-east, and the unique and extremely favourable influence of the Black Sea that results in less aggressive summer heat and less frosty winters, with early springs and late autumns, induce here longer vine vegetation offering thus the perfect microclimate to produce high-quality grapes and wines.

Reagents and standards
Unless specified, all the used reagents were of analytical grade.
Gallic acid, syringic acid, p-coumaric acid, (+) -catechin, (-)-epicatechin, trans-resveratrol, rutin, quercetin and vanillin were purchased from Sigma-Aldrich (Steinheim, Germany). Stock solutions of all these standards were prepared in methanol. Working standards were made by diluting the stock solutions with mixture methanol: water (50:50, v/v). The linearity of the method was between 0 -50 mg/L, for each compound. Both stock and working standards were stored at 4 o C until further use. Formic acid, acetonitrile and methanol were of LC grade, purchased from Merck. (Darmstadt, Germany). Twice distilled and demineralised water produced by a Milli-Q Millipore system (Bedford, MA, USA) was used for preparation of the aqueous solutions.
A multi-element standard solution XVI CertiPUR (mixture of 21 elements in diluted nitric acid), with a certified value of 100±3 mg/L, purchased from Merck (Darmstadt, Germany) was used for calibration curve in the quantitative analysis of metals (Cd, Cu, Cr, Zn, Pb and Ni). Solutions of varying concentrations were prepared for all the metals by diluting the standard solution. Nitric acid 65 % from Merck and ultrapure water, with a maximum resistivity of 18.2 MΩ/cm, obtained from a Milli-Q Millipore water purification system (Bedford, MA, USA), were used for sample treatment and dilution.

Phenolic compounds
Phenolic compounds were evaluated by reversed phasehigh performance liquid chromatography (RP-HPLC) with direct injection. Chromatographic analysis was carried out with a Thermo Finnigan Surveyor Plus HPLC (Thermo Scientific, USA) equipped with a Surveyor Photodiode Array Detector (PDA), Surveyor autosampler, Surveyor LC Pump (Quaternary gradient) and Chrome Quest Chromatography Workstation.
Separation of phenolic compounds was carried out at 30°C [19], on an Accuacore PFP (2.6 µm, 100 x 2.1 mm) column. The flow rate was 0.4 ml/min and injection volume 1 µL. Gradient elution of two solvents was used: solvent A consisted of water with 0.1% formic acid and solvent B: acetonitrile with Table 1 SOLVENT GRADIENT CONDITIONS WITH LINEAR GRADIENT 0.1% formic acid. The used gradient programme is given in Table 1. Detection was made at 280 nm. The wines were injected into HPLC system after filtering through a 0.45 µm pour size membrane filter. The amount of phenolic compounds in the extracts were calculated as mg/ L wine using external calibration curves, which were obtained for each phenolic standard. Each determination was carried out in duplicate and the mean value was reported. Blank solution and control samples were analyzed in order to monitor performance related to variable factors or random error.

Metals
A series of six elements (Cu, Zn, Pb, Ni, Cd and Cr) were determined in wine samples according to the OIV official methods [24], using a novAA 300 (Analytic Jena, Germany) flame atomic absorption spectrophotometer, equipped with deuterium ark background corrector, in an air-acetylene flame (for Cu) and respectively a Zeenit 650 (Analytic Jena, Germany) atomic absorption spectrophotometer equipped with a tubetype pyrolytic graphite furnace. The light sources were hollow cathode lamps from Analytik Jena. Hollow cathode lamps (Analytik Jena, Germany) were used as radiation sources, and the analytical measurements were based on time average absorbance. The analytical lines: 324.8, 213.9, 283.3, 232, 228.8 and 357.9 nm were used for registration of integral atomic absorbance values of Cu, Zn, Pb, Ni, Cd and Cr, respectively.
Previously, due to the high content of organic compounds in wine, the samples have undergone a pre-treatment procedure, a microwave assisted digestion, using a Mars 5 Microwave (CEM Microwave Technology Ltd, UK) system operated at max 1600W and equipped with fiber optic sensor for automatic temperature control and pressure sensor. Thus, the wine samples (2.5 mL) were introduced in Teflon digestion vessels, followed by addition of 2.5 mL of concentrated nitric acid. After sealing, the vessels were placed in the microwave oven set to run the following program: step 1-20 min to reach 180°C and step 2 -15 min cooling. After cooling to ambient temperature, the reactors were opened and sonicated to eliminate the nitrous vapors. The resultant solution (5 mL) was quantitatively transferred into a 50-mL volumetric flask and brought to the volume with ultrapure water.
Each sample was analyzed in duplicate, and each analysis was repeated five times. Calibration standards were prepared from multi-element standard solution XVI CertiPUR.

Data analysis
Considering the various factors (e.g. geographical origin, vine variety, vintage year and thus climatic conditions, and winemaking) that could influence the composition in metals and the phenolic profile of wine, a factorial analysis was applied to the obtained results for a better understanding of the studied variables considered the most appropriate to characterize and differentiate the wine samples. All the mathematical and statistical calculations (mean, standard deviation, median, range of variation) were performed using Microsoft Excel 2010 and XLSTAT software.

Results and discussions
The experimental results (minimum, maximum and average values) obtained for the analyzed wine samples are detailed further, for both metals and phenolic compounds, grouped by region of origin (vineyard) and wine variety.

Metals concentration
The results for metals measurement shown in Table 2 reveal that in some cases Cd and Pb concentration remained below the quantification limit and could not be detected.
A comparative assessment of the abundance of elements in wine according with the region of origin was performed; their tendency of breakdown is as follows: for Murfatlar area: Cd < Pb < Ni < Cr < Cu < Zn for Jidvei and Recas areas:Cd < Pb < Ni < Cr < Zn< Cu If the variety of wines is the comparing term, then the trend is: for Chardonnay: Cd < Pb < Ni < Cr < Zn < Cu for Muscat Ottonel: Cd < Cr < Pb < Ni < Zn < Cu for Pinot Noir: Cd < Pb < Ni < Zn < Cr < Cu This classification is specific to this study not being generalized since the concentrations of elements in wine are influenced by several factors including among others the area of origin, type of soil, grape variety, climate (terroir) and wine and oenological practices. That is why the content of an element in wine could be detected in a wide range of values, even if it was made from the same grape variety, which can be observed in Figure 1.
Another Copper (Cu), which is both an essential element and potentially toxic (when it is in excess), is normally associated with the use of pesticides and the production process. During the winemaking process, the copper content can vary. It decreases due to the formation of insoluble precipitates or increase as a result of corrosion processes that may affect the equipment used. Variation of Cu content in the analyzed wine samples is relatively high (Figure 1), ranging from 10.19 µg/L to 2733 µg/L. It may predominantly originate from the residual accumulation of Cu in soil due to the old farming practices of using copper sulphate or other copper-based fungicides to control the vine downy mildew. The results correspond to those reported in a previous study on Romanian wines from the Regarding the content of zinc (Zn) in wine, it can be also related with the antifungal treatments of the vine. In all the analyzed samples, its variation was between 309.30 µg/L and 1733 µg/L, respectively with an average of 816.13 µg/L. Cadmium is one of the natural components of the wine derived directly from the grapes, the maximum OIV admitted values being of 10 µg/L. In large quantities, it is harmful for the human body, affecting the calcium in the bone structure and the kidneys. In our study, for Chardonnay the Cd content ranged from 8.46 µg/L to n.d. (not detected), with a mean value of 2.80 µg/L and for Muscat Ottonel from 5.93 µg/L to 0.02 µg/L (mean value 1.88µg/L), while Pinot Noir had values from 6.53 µg/L to n.d., with a mean value of 2.53 µg/L.
All our results were in agreement with the literature, no significant difference being observed [1,12,25].

Phenolic compounds
The content of phenolic compounds in the wine samples showed a visible variation, depending both of grape variety and geographical origin ( Table 3) Another important flavonoid, the (+)-catechin, has recorded values between 0.14 mg/L and 5.33 mg/L, with an average of 1.79 mg/L. A significantly higher content of (+)-catechin was found in the red wine, Pinot Noir (from 2.07 mg/L to 5.33 mg/ L, with an average of 3.45 mg/ L), compared to the white ones (from 0.14 mg/L to 2.47 mg/ L and a mean of 1.20 mg/L for Chardonnay and between 0.71 mg/L and 1.22 mg/ L, averaging 0.72 mg/L for Muscat Ottonel). In terms of geographical origin,  [23] and Croatia [26].
From the hydroxy cinnamic acids, the p-coumaric acid was analyzed, which recorded values between n.d. and 2.34 mg/ L, in agreement with other previously published data [23]. In the Pinot Noir samples, the mean content of p-coumaric acid was 0.

Wines differentiation by variety
The main factor affecting the phenolic content in wine is the sun exposure of grapes during the growing/maturation season [26]. Moreover, the terroir effect leaves its mark on both the metals content and the phenolic compounds of wines.
The results achieved in this work highlighted the several variations in wine composition, both due to the grape variety, area of origin and year of harvest. Thus, the discrimination capability of the measurement results was verified by using a multivariate statistical analysis, based on discriminant analysis. In a first phase the definition of classes was made starting from the grape variety originating the analyzed wines ( Figure  2). A very good differentiation was observed for red wine (Pinot Noir) versus white wines (Chardonnay and Muscat Ottonel). When both phenolic compounds and metals are used as discriminant parameters (Figure 2b) an improvement of the discrimination function can be observed.
Efficiency of classification is given by the grading matrix (Table 4). Therefore, the characterization data of a wine sample (considered unknown) are removed in the calculation model and then the origin of that wine is predicted. Chemometric processing tools applied to our data were able to correctly classify the wines by variety in a proportion of 88.89% (when only the phenolic compounds are used for analysis) and 92.59% (when both phenolic compounds and metals are used for analysis). The influence of the metals is mainly observed to white wines, Chardonnay variety.
Elemental and phenolics data were also examined according to the variety using one-way ANOVA. The results indicate that the content of gallic acid, (+)-catechin, (-)epicatechin, rutin, quercetin and resveratrol showed the best correlations with the grapes variety of wines origin (Table 5), being significant for differentiation.

Wines differentiation by geographical origin and harvest year
The ability of some elements like Ni, Ag, Cr, Sr, Zn, Cu, Rb, Mn, Pb, Ba, Li, Mg, and Na to discriminate the geographical origin of wines was already demonstrated in some studies on Romanian wines [6,12,23]. In our work, commonly used parameters for wine characterisation as heavy metals and phenolic compounds were assessed to see their usefulness in providing preliminary information on wines origin.
Based on the elemental contents, the cross-validation technique provided a 100 % percentage of predicted membership according to the wine geographic origin (71.79 % F1 and 28.21 % F2) (Figure 3a), but the efficiency of classification given by the confusion matrix for samples prediction was low (correct classification of wines by origin of provenance in proportion of 62.96%). A great improvement in the correct classification of wine geographical origin (predictive rate of 100%) was obtained by combining the elemental profile with the phenolic composition (82.69% F1; 17.31 % F2) (Figure 3). Among the investigated parameters, Cu and Pb were identified as the most significant for geographic differentiation of the wines (Figure 4).
Wines originating from Recas region can be well characterized based on parameters like Cd, Ni, Cu, and syringic acid, while wines from Murfatlar can be defined by the content in rutin, epicatechin and vanillin, and wines from Jidvei by  parameters like Zn, Cr, Pb, resveratrol, catechin, quercitin, gallic acid and p-coumaric acid.
The same set of elements that we used to classify the wine samples according to their geographical origin, namely Cd, Cu, Cr, Zn, Pb, Ni, gallic acid, (+)-cathechin, syringic acid, vanillin, (-)-epicatechin, p-coumaric acid, rutin, resveratrol and quercitin, was applied to discriminate between 2010, 2011 and 2012 vintages. Figure 5 represents the correlation chart of loadings for the selected variables of the stepwise process, in the plane designed by the first two discriminant functions (F1 and F2). Function 1 express 65.99 % of the variance which provides the main separation between the vintage years and has a strong positive correlation with Cr, Ni, and Cd while function 2 (34.01 % of the variance) has a strong correlation with syringic acid, p-coumaric acid, vanillin and Pb concentrations. The efficiency of classification provided by  (Table 6).
Concluding, the set of markers, six inorganic (Cd, Cu Cr, Zn, Pb, and Ni) and nine organic ((+)-catechin, (-)-epicatechin, p-coumaric acid, gallic acid, syringic acid, rutin, quercitin, resveratrol and vanillin), that we used to characterize the wines, provide us a good classification of the samples in terms of grape variety and geographical origin, even vintage years. Our results confirm the powerful link between the wine composition and terroir, being in agreement with those reported by other authors [6,23].

Conclusions
The results of this study report that classical parameters, like phenolic compounds and metals, commonly used for characterization of wines may provide a picture of the oenoclimatic conditions variability across different regions and years, subsequently applicable to achieve reliable information for classifying wines by variety, origin of provenance and vintage.
The content of gallic acid, (+)-catechin, (-)-epicatechin, rutin, quercetin and resveratrol revealed the best ability for discriminating the wine samples according to the grapes variety of origin. By using variance analysis by average comparative (ANOVA) and linear discriminant analysis (LDA) based on the concentration of the selected elements a correct classification of the wines by variety was achieved in proportion of 88.89% (based only by phenolic compounds) and 92.59 % (based both on phenolic compounds and metals), revealing that the content of Cu, Cr and Ni is significant for discrimination. Moreover, a 100% successful classification of wines by region of origin was accomplished, while a differentiation by year of production was achieved with a correct prediction rate of 81.48%.
The most important conclusion is the fact that based on wine characterization parameters, like phenolics compounds coupled with multi-element, a separation of wine samples in terms of variety and geographical origin can be achieved.