Prediction of the functionality of young South American red wines based on chemical parameters

Authors


Abstract

Background and Aims

Wine functionality is an emerging parameter that may affect consumers' decision to purchase a wine. Thus, our objective was to classify young South American red wines according to their functionality.

Methods and Results

Four factors were considered for sample selection: vintage (2009/10), cultivar (Cabernet Sauvignon, Malbec, Carménère, Merlot, Syrah and Tannat), country (Argentina, Brazil, Chile and Uruguay) and price ($US1.0–50.0/bottle). Functionality of the wines was defined by their antioxidant activity, phenolic substances (TP), anthocyanins (TA) and colour. After applying multivariate analysis, wines were separated into three clusters. Wines grouped into the highest functionality cluster presented values of TP, TA, antioxidant activity, darker and purple colour higher than those of wines classified as intermediate or of low functionality. More than 96% of the wines were correctly classified by discriminant analysis according to their functionality using only seven markers.

Conclusions

Applying multivariate analysis, 666 South American wines were classified according to their functionality based on chemical markers, colour and price. Malbec and Tannat wines produced in Argentina, priced above $US15.00/bottle, showed the highest functionality.

Significance of the Study

The present study provides relevant information that may help wine consumers on their choices regarding functionality based on cultivar, origin and price of young South American red wines.

Introduction

Daily moderate consumption of red wine has been widely associated with a reduced risk of degenerative diseases. This association was first reported by Renaud and De Lorgeril (1992), who observed that citizens from France and Switzerland, despite a high consumption of saturated fat and the prevalence of other risk factors, presented a lower rate of cardiovascular diseases. This condition, known as the ‘French Paradox’, was partially attributed to the antioxidant activity of the phenolic substances present naturally in red wines.

Grapes contain two major groups of phenolic substances, the non-flavonoids (hydroxybenzoates, hydroxycinnamates and stilbenes) and the flavonoids [flavan-3-ols, flavonols and anthocyanins (TA)], located in the skins, seeds and stems (Yang et al. 2009). A proportion of these compounds is transferred to wine during the winemaking process, which contributes to its colour, taste and health properties (Fanzone et al. 2012). The phenolic substances profile is genetically determined, but it can be altered by several factors, including climate, soil, vine age and winery practices (Jaitz et al. 2010, Yoo et al. 2011). The two most abundant classes of phenolic substances found in grapes are the anthocyanins and the condensed tannins (Jensen et al. 2008). Individual anthocyanins have proven to be helpful for differentiating red wines according to grape cultivar and winegrowing region (González-San José et al. 1990). In addition, anthocyanins contribute to the antioxidant activity of red wines, leading to interest in their safety and health benefits (Yang et al. 2009). Although South American wines are increasingly being consumed worldwide, information concerning their antioxidant activity, phenolic substances profile and variation in these parameters depending on the vintage, grape cultivar or region of wine production remains scarce.

Nowadays, many factors, such as cultivar, vintage, brand, region of origin, price and packaging, are involved at the moment of purchasing a wine (Lockshin et al. 2006). The antioxidant activity of wine is an emerging parameter that affects consumers' decisions to purchase a wine (Yoo et al. 2011). According to Barreiro-Hurlé et al. (2008), wine consumers would be willing to pay more for an antioxidant (resveratrol)-enhanced wine. In a recent study, Macedo et al. (2013) showed a significant correlation between wines and plasma antioxidant activity, suggesting that the potential health benefits of wines are transferred to consumers. The ‘functionality’ term applied in our study was based on the fact that many health benefits promoted by the intake of phenolic substances are explained by a multitude of non-antioxidant activities (Frankel and Finley 2008). Taking resveratrol as an example, Ramprasath and Jones (2010) suggested that this stilbene exerts several health benefits including anti-atherogenic, anti-inflammatory and anti-cancer effects; it may also prevent lipid oxidation, platelet aggregation and arterial vasodilation, and may also modulate the level of lipids and lipoproteins. Thus, the term ‘functionality’ was applied in our study because it appears to be a concept wider than antioxidant activity, and it is associated with the concentration of phenolic substances and the colour of wine.

Considering these facts, consumer access to information related to wine functionality is becoming essential. Such access could be achieved either by including the functionality information on the bottle label or by associating functionality with cultivar, region or price; as a result, it is critical to determine how these characteristics change between vintages.

Thus, the main objective of this study was to classify South American red wines according to their functionality based on their antioxidant activity, concentration of phenolic substances, concentration of anthocyanins and of individual anthocyanins, and colour. In addition, the influence of the vintage, origin, cultivar and price on the functionality was also evaluated. Finally, discriminant analysis was applied to estimate the functionality of young South American red wines based on the proposed chemical markers and price.

Materials and methods

Wine samples

A group of 666 red monovarietal wines from Vitis vinifera (Cabernet Sauvignon, Carménère, Malbec, Merlot, Tannat and Syrah) produced during 2009 (333 wines) and 2010 (333 wines) in Chile, Argentina, Uruguay and Brazil, bottled in 750-mL bottles and with retail prices of $US1–50 was chosen for the present study. All samples were purchased from local markets and wine distributors in São Paulo (Brazil) and Montevideo (Uruguay). Wines from the 2009 and 2010 vintages were purchased at the beginning of 2011 and 2012, respectively. Thus, wines from both vintages underwent an approximately equal storage period. Based on official statistics describing the wine production in each country, 324 wines from Chile, 236 from Argentina, 64 from Uruguay and 42 from Brazil were selected. Following the same criteria, the sampling was composed of 221 Cabernet Sauvignon, 107 Carménère, 108 Malbec, 106 Merlot, 76 Syrah and 48 Tannat wines. Wine bottles were opened, and the colour was determined immediately. The wine was then stored at −80°C in 1.5-mL microtubes (Axygen Scientific, Union City, CA, USA) until chemical analysis.

Reagents

The 2,2-diphenyl-1-picrylhydrazyl (DPPH), 2,2′-azobis(2-methylpropionamidine) dihydrochloride, Folin-Ciocalteu reagent, gallic acid, trifluoroacetic acid (TFA) and malvidin-3-glucoside standard were supplied by Sigma Chemical Co. (St. Louis, MO, USA), high-performance liquid chromatography (HPLC)-grade acetonitrile was purchased from Merck & Co., Inc. (Darmstadt, Germany), and the anthocyanin standards (cyanidin-3-glucoside, delphinidin-3-glucoside, peonidin-3-glucoside, petunidin-3-glucoside) were purchased from Polyphenols Laboratories AS (Sandnes, Norway).

Colour determination

Colour was measured by the transmittance in a ColorQuest XE colorimeter (Hunter Associates Laboratory, Inc., Reston, VA, USA) using the Commission Internationale d'Eclairage (CIE) 1964 standard observer (10° visual field) and the CIE standard illuminant D65 as reference. The three CIELAB coordinates a* (red-green; +a*, −a*), b* (yellow-blue; +b*, −b*) and lightness L (white-black, 0–100) were determined in triplicate using the software EasyMatch QC (Hunter Associates Laboratory, Inc.).

Phenolic substances

Phenolic substances were determined in triplicate by the Folin-Ciocalteu colorimetric method (Singleton and Rossi 1965) adapted for a microplate reader. First, the wine was diluted 25 times in water, and 25 μL of this solution was mixed with 25 μL of twofold-diluted Folin-Ciocalteu phenol reagent. Then, 200 μL of water was added, and after 5 min, 25 μL of a 10 g/100 g sodium carbonate solution was added to the mixture and shaken thoroughly. The mixture was incubated for 60 min in the dark at 25°C. The blue colour formed in the mixture was measured at a wavelength of 725 nm with a multidetection microplate reader (BioTek Instruments, Inc., Winooski, VT, USA). A standard curve of gallic acid (ranging from 0 to 100 mg/L) was prepared, and the results were expressed as mg gallic acid equivalents per litre (mg GAE/L).

Anthocyanins

Anthocyanins (TA) were quantified as described by Fuleki and Francis (1968) with some modifications. Briefly, 2.25 mL of potassium chloride buffer, pH 1.0 (0.025 mol/L) and 2.25 mL of sodium acetate buffer, pH 4.5 (0.40 mol/L), were each mixed with 0.25 mL of wine. The absorbance of both solutions was measured at 510 and 700 nm in a 1-cm path length cell using a UVmini-1240 UV-Vis spectrophotometer (Shimadzu Corporation, Kyoto, Japan). The experiments were performed in duplicate. Anthocyanins were calculated using Equation (1) and expressed as mg of cyanidin-3-glucoside per litre (mg cyanidin-3-glu/L).

display math(1)

where A510 and A700 are the absorbance at 510 and 700 nm, respectively, MM is the molecular mass of cyanidin-3-glucoside (449.2), DF is the dilution factor (10) and ε is the molar absorptivity of cyanidin-3-glucoside (26 900).

Individual anthocyanins

HPLC-diode array detection

Individual anthocyanins were determined by HPLC according to the method of Boido et al. (2006) with some modifications. The chromatographic system was an Agilent 1200 series liquid chromatograph (Agilent Technologies, Santa Clara, CA, USA) with a diode array detector (DAD). A Synergi Hydro-RP column, 4 μm, 80 Å, 150 mm × 4.6 mm (Phenomenex, Torrance, CA, USA) was maintained at a constant oven temperature of 35°C. The DAD was set to measure at 520 nm as the preferred wavelength, although spectra were recorded from 220 to 600 nm. The elution solvents were A (an aqueous solution of 0.1% TFA) and B (100% acetonitrile). The flow rate was 0.5 mL/min, and the gradient was as follows: isocratic at 10% B for 5 min, a linear gradient to 15% B in 15 min, isocratic at 15% B for 5 min, a linear gradient to 18% B in 5 min, then a linear gradient to 28% B in 20 min and a linear gradient to 30% B in 5 min, completing 55 min of analysis. Before directly injecting the wines into the chromatograph, they were filtered with a 0.45 μm polytetrafluoroethylene-modified membrane (Millipore Co., Billerica, MA, USA). Solvents A and B were filtered in an identical manner. The injection volume was 30 μL, and analyses were in duplicate. The anthocyanins were identified by comparing the retention time of each peak with that of the available standards and with values obtained from the literature. To confirm these results, some of the samples were analysed by HPLC-mass spectroscopy (MS), as described later. The individual anthocyanins were quantified using the areas obtained in the HPLC-DAD analysis and two calibration curves of malvidin-3-glucoside (0.1–7.0 and 7.1–200 mg/L). The results were expressed as mg malvidin-3-glu/L.

HPLC-DAD-MS

Individual anthocyanins were measured with a Shimadzu Prominence SPD-M20A liquid chromatograph (Shimadzu Corporation) that was connected via the UV cell outlet to a Bruker Esquire HCT ion trap mass spectrometer (Bruker Daltonics, Inc., Billerica, MA, USA). The HPLC-DAD conditions were as described in the previous section. The MS contained an electrospray ionisation interface and used nitrogen as the drying gas at a flow rate of 6.0 L/min. The pressure of the nebuliser was set at 172 mPa. The capillary temperature was 280°C. Spectra were recorded in positive ion mode between m/z 50 and 1100. The mass spectrometer was programmed to undertake a series of three consecutive scans: a full mass scan, an MS2 scan of the most abundant ions in the full mass scan and an MS3 scan of the most abundant ions in the MS2 scan. The data obtained were analysed using the Bruker Compass DataAnalysis 4.0 software (Bruker Daltonik GmbH, Bremen, Germany).

Free-radical scavenging capacity

The free-radical scavenging capacity measured by the DPPH radical, known as the DPPH method, was described by Arnous et al. (2001) and applied with some modifications. Samples were diluted 1:10 with an ethanolic solution (12% in a 0.2 mol/L KCl-HCl buffer, pH 2.0). An aliquot of 0.075 mL of the diluted sample was added to 2.925 mL of DPPH solution (60 μmol/L in methanol), vortexed and incubated in the dark for 150 min (the time required to reach the stationary state). A control was prepared by substituting the sample aliquot with methanol. The absorbance at 515 nm was read with a UVmini-1240 UV-VIS spectrophotometer (Shimadzu Corporation). The antiradical activity (AAR) of each sample, which was expressed in millimolar Trolox (water soluble analog of tocopherol) equivalents (mmol/L TRE), was calculated using Equation (2):

display math(2)

where (%ΔA515) = [(A515,control − A515,t=150)/A515,control] × 100.

Equation (2) was determined by linear regression (r2 = 0.992) after plotting ln(%ΔA515) of known Trolox solutions (0.25–1.50 mmol/L) against concentration. The analyses were performed in duplicate.

Oxygen radical absorbance capacity

The oxygen radical absorbance capacity (ORAC) assay was performed according to the method described by Huang et al. (2002) with slight modifications. A volume of 25 μL of diluted wine (900 times in 75 mmol/L phosphate buffer, pH 7.1) was transferred to a microtitre plate, filling the wells surrounding the samples with 250 μL of Milli-Q water (to maintain the temperature). Then, 75 μL of a solution of fluorescein (40 nmol/L in phosphate buffer) was added to each well. After incubation at 37°C for 30 min, 25 μL of AAPH● (153 nmol/L in phosphate buffer) was added. Immediately afterwards, the fluorescence was measured every minute for 1 h in a multidetection microplate reader (BioTek Instruments, Inc., Potton, England). The blank was prepared with 25 μL of phosphate buffer instead of the sample. The area under each curve was integrated using the Gen5 software (BioTek Instruments, Inc.). The area of the blank was subtracted from the area obtained for each wine. Each sample was analysed in quadruplicate. A calibration curve was prepared with known Trolox solutions (6.25–100 μmol/L). The results are expressed as micromolar Trolox equivalents (μmol/L TRE).

Statistical analysis

The effect of each factor (vintage, cultivar, country and price) and their first-order interactions on antioxidant activity was first evaluated by multiple factors analysis of variance (ANOVA) followed by the Tukey honestly significance test comparison post-hoc test. Linear correlations were determined by the Pearson coefficient. Multivariate principal component analysis (PCA) was applied to reduce the original number of variables in the new orthogonal PC and also to project the wine samples according to the plot generated by the selected PC. Because of the large number of samples (n = 666), cluster analysis (K-means) was applied to the normalised data to group the wines according to the seven original variables, the colour parameters L*, a* and b*; TP; TA; ORAC; and DPPH values). Euclidean distance and Ward's method were used to separate the groups. Here, one-way ANOVA was applied to identify differences among the groups formed by the cluster analysis. A linear discriminant analysis was performed to classify the wines according to the three clusters. To validate the predictions, a repeated random subsampling cross-validation was performed, and the prediction power was estimated. For all statistical treatments, normality and homogeneity of variance were initially checked by the graphs of residual dispersion, and variable transformation or equivalent nonparametric analysis were used when necessary. Initially, price was considered a qualitative factor. For this reason, this numerical variable was categorised as low (from $US0 to 15), medium (from $US15.1 to 25) and high (from $US25.1 to 50). For the other analyses, price was taken as a quantitative variable. It was given an alpha value of 0.05 to reject the null hypothesis. All calculations were performed, and graphs generated using the software STATISTICA version 9.0 (StatSoft, Inc., Tulsa, OK, USA) and the SAS (Statistical Analysis Software) version 9.2 (SAS Institute, Inc., Cary, NC, USA).

Results

In our study, two methods that differ in terms of assay principle were applied to measure the antioxidant activity: ORAC and DPPH. Although the linear correlation between these methods was significant, the coefficient of correlation was low (r = +0.43, P < 0.01). Thus, data obtained from both methods (ORAC and DPPH) were retained in our analysis. Next, each factor, including vintage (2009 and 2010), cultivar (Cabernet Sauvignon, Carménère, Malbec, Merlot, Syrah and Tannat), price (low, medium and high) and country (Argentina, Chile, Uruguay and Brazil), as well as their first-order interactions, were evaluated in terms of the antioxidant activity using DPPH and ORAC methods. All the factors promoted significant change (P < 0.05) to the antioxidant activity as measured by both methods. No interaction among the factors was verified using the DPPH method, but a significant interaction between vintage and country was found using the ORAC method. Differences among the levels of each factor are shown in Figure 1. For vintage (Figure 1a), the antioxidant activity increased from 2009 to 2010 when DPPH was applied to measure this parameter. The vintage effect on the ORAC response must be interpreted by considering the country of production because of the first-order interaction between these two factors (vintage × country). Figure 1b shows that the ORAC values decreased from 2009 to 2010 in the wines produced in Argentina, Brazil and Chile but increased in the wines produced in Uruguay. Regarding the country of production (Figure 1c), the results suggested that when the antioxidant activity is analysed using the DPPH method, wines produced in Argentina and Uruguay presented a higher antioxidant activity than those produced in Chile and Brazil. Figure 1d shows the results of the antioxidant activity when cultivar is used as a grouping factor. Using the DPPH method, wines produced from Carménère demonstrated lower antioxidant activity than that of the other cultivars. When analysed by the ORAC method, Carménère wines showed antioxidant activity lower than that of Cabernet Sauvignon and Tannat. The antioxidant activity of Tannat wines demonstrated a value higher than that of all other wines using DPPH and a higher value than that of Carménère wines using ORAC. When wines were grouped by price range (Figure 1e), it was observed that those available in the market place with prices below $US15.00 (low) demonstrated lower antioxidant activity than those priced between $US15.10–50.00 (medium and high).

Figure 1.

The antioxidant activity of the 666 young red wines from Argentina, Brazil, Uruguay and Chile measured with (a) the 2,2-diphenyl-1-picrylhydrazyl (DPPH) method according to vintage, (b) the oxygen radical absorbance capacity (ORAC) method according to country and vintage [(2009, image) and (2010 image)], (c) the DDPH method by country, (d) the DPPH (image) and (ORAC) (image) methods by cultivar, and (e) the DPPH (image) and ORAC/5 (image) methods by price. The data represent the means with errors bars (standard error of the mean). *P < 0.05, **P < 0.01.

Subsequently, a multivariate analysis was performed aiming to reduce the seven original variables and project the wine samples in a three-dimensional plot according to the new principal components (PCs) generated by the original variables. Table 1 presents the eigenvalues and the correlation of each of the seven variables selected in our study with the PCs. From these data, three PCs were retained in our analysis. The first PC was designated as ‘colour + DPPH + TP’ and explained 53.62% of the variation, the second PC representing ‘DPPH, ORAC, TP and TA’ explained 20.00% of the variation, and the third PC, which represented ‘TA’ alone, explained 11.59% of the variation; the three PCs explained 85.21% of the variation. By applying cluster analysis, joining (tree clustering) confirmed the correlation observed by PCA (Figure 2a) in which the colour parameters (L, a* and b*) were highly associated (first PC) followed by an association between TP, DPPH and ORAC (second PC), whereas TA appeared more isolated from the other variables (third PC). Tree clustering for the 666 samples (Figure 2b) suggested that the samples could be adequately divided into three clusters. Because of the large number of samples, K-means clustering was applied to identify the wines included into each cluster. The three clusters were designated here as ‘low, intermediate and high functionality’ according to their colour, antioxidant activity, and the concentration of phenolic substances and of anthocyanins (Figure 3). Table 2 presents the differences of each cluster in terms of the seven variables evaluated in our study. The cluster designated as high functionality (HF) contained the wines that presented the highest antioxidant activity and highest concentration of phenolic substances and anthocyanins, and were darker and purple in colour. This HF cluster was mainly composed of Malbec and Tannat wines from Argentina with medium-to-high prices. The intermediate functionality (IF) cluster was composed of all cultivars except Tannat, with a predominance of Chilean wines of medium-to-low price, while the low functionality (LF) cluster contained wines from Brazil and Uruguay, independent of the cultivar, with a prevalence of low-priced wines. Because anthocyanins were not associated with colour and antioxidant activity in our samples, and they had demonstrated low linear correlation coefficients with these variables (r values <0.35), it was investigated whether the individual anthocyanins differed among the clusters (Table 3). In terms of individual anthocyanins, wines from Cluster HF demonstrated a higher concentration of cyandin-3-glucoside, delphinidin-3-acetylglucoside, delphinidin-3-glucoside, peonidin-3-acetylglucoside, peonidin-3-coumarylglucoside, petunidin-3-coumarylglucoside, petunidin-3-acetylglucoside, petunidin-3-glucoside and vitisin A, suggesting that these individual anthocyanins could be more associated with wine functionality than the other anthocyanins.

Table 1. Individual and cumulative (%) eigenvalues of the correlation matrix and the contribution of each variable to the 1st, 2nd and 3rd principal components (factor loadings)
Eigenvalues and variables1st Principal component2nd Principal component3rd Principal component
  1. L, a* and b*, CIELAB coordinates; DPPH, 2,2-diphenyl-1-picrylhydrazyl; ORAC, oxygen radical absorbance capacity.
Eigenvalue3.751.400.81
Explained variation (cumulative %)53.6273.6285.21
L0.93−0.27−0.20
a*0.92−0.27−0.20
b*0.93−0.29−0.19
Phenolic substances−0.63−0.53−0.33
Anthocyanins−0.380.50−0.69
ORAC method−0.44−0.580.23
DPPH method−0.67−0.55−0.25
Figure 2.

(a) Tree diagram for the following seven variables measured on 666 young red wines from Argentina, Brazil, Uruguay and Chile: the three CIELAB coordinates, a* (red-green; +a*, −a*), b* (yellow-blue; +b*, −b*) and lightness L (white-black, 0–100); phenolic substances concentration (TP); anthocyanins concentration (TA); oxygen radical absorbance capacity (ORAC) and 2,2-diphenyl-1-picrylhydrazyl (DPPH) values. (b) Tree diagram for same wines illustrating the separation of the wines into three clusters.

Figure 3.

Representation of the 666 young red wines from Argentina, Brazil, Uruguay and Chile on the plane generated by the first and second principal components (PCs) according to the following clusters: high functionality (image), intermediate functionality (image) and low functionality (image). The first PC correlates positively with the three CIELAB coordinates L, a* and b*, and negatively with the phenolic substances concentration (TP) and the 2,2-diphenyl-1-picrylhydrazyl (DPPH) method, whereas the second PC correlates negatively with oxygen radical absorbance capacity, DPPH and TP, and positively with anthocyanins concentration.

Table 2. Number and proportion of 666 young red wines from Argentina, Brazil, Chile and Uruguay by vintage, cultivar, country and price range observed in each of the three functionality clusters and mean values (±standard error of mean) of the antioxidant activity, concentration of phenolic substances, and of anthocyanins and colour parameters observed in each cluster
Wine characteristicsNumber and proportion (%)
Low functionalityIntermediate functionalityHigh functionalityP
n = 133n = 291n = 242
  1. †Number of wines in each of the three clusters. ‡Probability values obtained by one-way analysis of variance. Values in the same row followed by different superscript letters present a significant difference (P < 0.05). DPPH, 2,2-diphenyl-1-picrylhydrazyl; ORAC, oxygen radical absorbance capacity.
Vintage   
 200963 (19%)149 (45%)121 (36%)
 201070 (21%)142 (43%)121 (36%) 
Cultivar   
 Cabernet Sauvignon43 (19%)100 (45%)78 (35%)
 Carménère16 (15%)53 (50%)38 (36%)
 Malbec18 (17%)44 (41%)46 (43%)
 Merlot31 (29%)44 (42%)31 (29%)
 Syrah11 (14%)37 (49%)28 (37%)
 Tannat14 (29%)13 (27%)21 (44%)
Country
 Argentina45 (19%)90 (38%)101 (43%)
 Brazil17 (40%)13 (31%)12 (29%)
 Chile43 (13%)169 (52%)112 (35%)
 Uruguay28 (44%)19 (30%)17 (27%)
Price
 Low107 (32%)157 (48%)66 (20%)
 Medium19 (8%)116 (46%)118 (47%)
 High7 (8%)18 (22%)58 (70%)
L22.73 ± 0.31a15.40 ± 0.14b11.18 ± 0.15c<0.001
a*52.24 ± 0.15a46.86 ± 0.14b41.79 ± 0.23c<0.001
b*36.76 ± 0.36a26.19 ± 0.23b18.96 ± 0.27c<0.001
Phenolic substances (mg GAE/L)1645.28 ± 26.14a1827.48 ± 17.57b2199.95 ± 18.79c<0.001
Anthocyanins (mg cyan-3-glu/L)100.65 ± 3.13a114.24 ± 2.13b133.81 ± 2.65c<0.001
ORAC (mmol/L TRE)36.96 ± 5.96a40.24 ± 4.35b45.35 ± 4.92c<0.001
DPPH (mmol/L TRE)8.44 ± 0.07a9.02 ± 0.03b9.54 ± 0.03c<0.001
Table 3. Concentration of the individual anthocyanins in each of the three functionality clusters for 666 young red wines from Argentina, Chile, Uruguay and Brazil
AnthocyaninsConcentration (mg malvidin-3-glu/L)
Low functionalityIntermediate functionalityHigh functionalityP
n = 133n = 291n = 242
  1. †Number of wines in each functionality cluster. ‡Probability values obtained by one-way analysis of variance. Values in the same row followed by a different superscript letter present a significant difference (P < 0.05).
Cyanidin-3-glucoside0.20 ± 0.01a0.30 ± 0.02b0.31 ± 0.01c<0.001
Delphinidin-3-acetylglucoside0.70 ± 0.03a0.87 ± 0.02b1.10 ± 0.03c<0.001
Delphinidin-3-glucoside2.65 ± 0.16a3.25 ± 0.14b3.75 ± 0.16c<0.001
Malvidin-3-(coumaryl)glucoside4.90 ± 0.26a5.03 ± 0.19a5.45 ± 0.22a0.192
Malvidin-3-acetylglucoside10.55 ± 0.63a11.28 ± 0.4a11.91 ± 0.40a0.174
Malvidin-3-glucoside37.53 ± 1.90a39.92 ± 1.30a42.05 ± 1.44a0.162
Peonidin-3-(coumaryl) glucoside1.28 ± 0.07a,b1.25 ± 0.05a1.50 ± 0.08b0.034
Peonidin-3-acetylglucoside1.28 ± 0.06a1.33 ± 0.04a1.59 ± 0.05b<0.001
Peonidin-3-glucoside3.04 ± 0.232.88 ± 0.133.03 ± 0.140.688
Petunidin-3-(coumaryl)glucoside0.23 ± 0.01a0.29 ± 0.01b0.37 ± 0.02c<0.001
Petunidin-3-acetylglucoside0.74 ± 0.06a0.76 ± 0.04a0.96 ± 0.07b0.002
Petunidin-3-glucoside4.24 ± 0.25a4.74 ± 0.18a5.91 ± 0.25b<0.001
Vitisin A9.42 ± 0.57a9.35 ± 0.38a11.81 ± 0.54b0.038

The main objective of this study was to estimate wine functionality based on simple markers. A discriminant analysis was applied considering the three clusters presented in Figure 3. The following three discriminant functions were obtained considering L, a*, b*, TP, TA, DPPH and price as independent variables:

display math
display math
display math

The cross-validation classified a total of 644 wines from the original 666 (Table 4), representing more than 96% of the samples.

Table 4. Classification of 666 young red wines from Argentina, Chile, Uruguay and Brazil using linear discriminant analysis and the prediction ability using cross-validation
ClustersNumber of wines
Low functionalityIntermediate functionalityHigh functionality
  1. HF, high functionality; IF, intermediate functionality; LF, low functionality.
Cluster 1 (LF)13040
Cluster 2 (IF)32808
Cluster 3 (HF)07234
Total133291242
Correct classification130280234
Prediction ability (%)97.796.296.7

Finally, we plotted the antioxidant activity determined using the DPPH and ORAC methods against wine price, and based on Figure 1e, a polynomial model was adjusted to the data (Figure 4). By derivation, it was suggested that $US37.00 and 27.00 were the plateau prices above which no additional gain in terms of antioxidant activity could be achieved when DPPH (Figure 4a) and ORAC (Figure 4b) methods were applied to measure this parameter, respectively.

Figure 4.

(a) Scatterplot of the correlation between the 2,2-diphenyl-1-picrylhydrazyl (DPPH) method results against price and (b) the oxygen radical absorbance capacity (ORAC) results against price illustrating the price relative to the plateau estimated by the derivatisation of the respective regressions.

Discussion

Functional or sensory classification of red wines based on chemical parameters or categorical factors has been a great challenge for researchers, winemakers and regulatory agencies. In our study, using multivariate analysis, young South American red wines were separated according to their ‘functionality’ based on simple chemical markers, retail prices and colour parameters.

Vintage appeared to be the least effective of the categorical factors for separating the wines. It was observed in Table 2 that the distribution of the wines produced in 2009 and 2010 inside the clusters was similar. In other studies that analysed the effect of the vintage on the concentration of phenolic substances and antioxidant activity of wines, the analyses were generally performed during the same period of time. It implies that vintage confounds with aging time of the wines. In nearly all of the studies involving more than two vintages, vintage and age are mixed in the same interpretation. This mistake was avoided in our study because the wines were analysed immediately after acquisition. The reduced effect of the vintage on wine functionality could in part be explained by the similar weather conditions observed during the 2 years (2009 and 2010), except for the rainfall in Uruguay (Dirección Metereológica de Chile 2013, Dirección Nacional de Meteorologia 2013, Instituto Nacional de Meteorologia 2013, Servicio Meteorológico Nacional 2013). In addition, irrigation programs are a common management practice in some wine production regions of South America, where the rainfall is scarce. Therefore, in these regions, the water availability of the vines is quite independent of the vintage.

Although in 2010 wines have shown a decrease in the antioxidant activity measured by the ORAC method, Uruguayan wines showed an opposite behaviour, increasing their ORAC value from 2009 to 2010. As previously mentioned, Uruguay underwent the greatest climatic variation among the countries studied, with the rainfall more than doubling in 2010 compared with that in 2009 (Dirección Nacional de Meteorologia 2013). Water stress, either deficit or excess, strongly influences the chemical and particularly the composition of the phenolic substances of grapes and the wines produced from them (Jackson and Lombard 1993). Therefore, the variation in the composition of the phenolic substances of Uruguayan wines likely differed from that of wines from the other countries, in which the weather changes were not so drastic. This difference could affect the ORAC value and lead to a behaviour opposite that of wines from the other countries in the study. This hypothesis, however, must be further evaluated.

Price was the most effective factor for the separation of the wines (Table 2). A possible explanation for this behaviour is that more expensive wines are usually obtained from older vines containing grapes with a higher concentration of phenolic substances and other chemical compounds. Moreover, more expensive wines are aged longer in oak barrels, and the winemaking process can differ among wines of different prices, directly affecting their concentration of phenolic substances and antioxidant activity (Villaño et al. 2006). Fanzone et al. (2012) found similar results for Argentine red wines, observing an increase in the phenolic substances of Malbec and Cabernet Sauvignon wines when their prices increased. The correlation between price and functionality, however, is controversial. Yoo et al. (2011) found a negative correlation between price and antioxidant activity using the DPPH method. In addition, in a previous study performed by our group including 72 South American wines (Granato et al. 2011), price was not considered a suitable factor to discriminate wines according to their antioxidant activity. In this particular case, the results could have been influenced by the inclusion of blends from Vitis labrusca grapes mixed with V. vinifera. Conversely, in another previous study performed by our group with 29 Brazilian red wines from only V. vinifera grapes (Granato et al. 2010), a trend was observed of higher prices associated with higher antioxidant activity.

Two additional aspects deserve attention in our study: the low correlation between the antioxidant activity measured with the DPPH and ORAC methods, and the low correlation among anthocyanins, antioxidant activity and colour. Although the DPPH and ORAC methods measure the ability of a compound or a mixture of compounds to transfer an electron or hydrogen to a radical molecule (DPPH and AAPH, respectively), the assay conditions are different. In addition to the stereochemical structure of the molecules involved in the reactions, other factors such as the bond dissociation energy, pH, reduction potential and mobility of the radical according to the solvents used in each method can also influence the results (Cao et al. 2007). Therefore, the different solvent used in the two methods (methanol in the DPPH assay and phosphate buffer in the ORAC assay) could have contributed to the observed differences. Another factor that may have influenced the results was the composition of the individual phenolic substances of each wine. Tabart et al. (2009) evaluated the antioxidant activity of phenolic substances using five methods and did not observe a correlation between DPPH and ORAC results. The authors concluded that the specific structure of each compound contributed to this discrepancy and suggested the use of a weighted mean of the five methods to express the results. Villaño et al. (2005) determined the antioxidant activity of the major individual phenolic substances present in wines and observed that antioxidant activity, measured with the DPPH method, increased with the number of hydroxyl groups present in the phenolic substances molecule, whereas antioxidant activity measured using the ORAC method was enhanced in phenolic substances containing a catechol group. Similar conclusions were reported by Capitani et al. (2009) who evaluated the antioxidant activity of 22 compounds using five methods. Likewise, only compounds that present an oxidation potential above 0.56 V would be able to thermodynamically oxidise the DPPH radical (Arteaga et al. 2012). This restriction limits the antioxidant activity of some phenolic substances present in the wine towards the DPPH radical.

Based on this fact, different results could be expected depending more on the individual phenolic substances present in the wine than on the concentration of phenolic substances. In our study, the concentration of phenolic substances correlated much better with the DPPH (r > + 0.69) than with the ORAC (r > + 0.35) values. This tendency was also observed by other authors (Fernández-Pachón et al. 2004), suggesting that phenolic substances containing hydroxyl groups could contribute more to the antioxidant activity of wines than phenolic substances containing a catechol structure. As no single method is adequate for evaluating the antioxidant activity of foods (Frankel and Meyer 2000), we suggest that multivariate statistical techniques, such as those applied in this study, should be used instead of results obtained from individual methods.

Several researchers have found an elevated correlation between anthocyanins content and wine colour (Boido et al. 2006, Gómez-Gallego et al. 2012, McRae et al. 2012). The colour–anthocyanins correlation, however, observed in our study was low for both the anthocyanins concentration measured by spectrophotometry (r < + 0.35) and for individual anthocyanins evaluated by HPLC/MS (r < +0.31). A possible explanation for this divergence is that contrary to other studies, the pH of the wines was not normalised prior to the measurement of the colour in our study. The pH of the wine affects the distribution of the free anthocyanins forms and consequently affects the colour in young red wines. As the objective of our study, however, was to evaluate commercial samples, we opted to measure the colour of the wines at their original pH, which represented the original colour, as it was presented to the consumers. In addition to pH, the formation of copigmentation complexes between anthocyanins and other phenolic substances, which form anthocyanin-derived pigments, is also responsible for red wine colour and colour changes (Gómez-Gallego et al. 2012). Copigmentation is the major mechanism for colour stabilisation and is affected by the pH, winemaking process and storage conditions of the wines (Castañeda-Ovando et al. 2009). Although for young red wines, the concentration of these compounds is typically quite low (Boido et al. 2006, McRae et al. 2012), it can account for up to 50% of their colour (Boulton 2001). The diversity of our wine sampling, in terms of winemaking and storage condition, likely influenced the concentration of copigments. This factor could affect wine colour independently of the concentration of individual anthocyanins.

Although health claims for alcoholic beverages are forbidden in many countries, consumers have sought wines with higher functionality. There is no information on the bottle labels, however, concerning the functionality of the wine. Including this information could lead consumers to the best choice with regard to health claims, price and sensory expectations. From our study, which was limited to young red wines from South America and from only two vintages, the wines included in Cluster HF appeared to demonstrate the highest functionality. This cluster contained a higher proportion of Tannat and Malbec wines from Argentina. But, it also contained 70% of the wines with a price in the higher range.

Conclusions

Applying cluster and discriminant statistical analysis, it was possible to classify young South American red wines according to their functionality based on simple chemical markers, colour and price parameters. In summary, Malbec and Tannat wines produced in Argentina, priced above $US15.00/bottle showed the best functionality. Including this kind of information on the bottle label would provide an objective tool to aid consumers that consider functionality as a factor of their purchase decision.

Acknowledgement

The authors wish to thank the São Paulo Research Foundation – FAPESP – (process numbers: 2010/19845-2 and 2010/18765-5) for their financial support of the present work.

Ancillary