Abstract
The polyphenols of a 2005 Cabernet Sauvignon were fractionated by gel permeation chromatography. The obtained fractions were reconstituted in a model wine medium for analysis. The chemical properties of the fractions were investigated by HPLC-DAD, HPLC-ESI-MSn, colorimetric analysis, and additional chemical assays. Application of these different techniques and combination with the elution pattern of the gel permeation material revealed some valuable structural information. The early eluting fractions, which were larger in molecular size and/or more polar than late eluting fractions, contained higher proportions of pigmented polymers and higher amounts of bound anthocyanins. These anthocyanins were responsible for the color properties of the early eluting fractions, which could be separated according to CIELab analysis. The later eluting fractions exhibited lighter, more yellowish color, provoked by the higher tannin concentration, and contained low molecular weight proanthocyanidins. Descriptive sensory analysis was carried out, assessing bitterness, sourness, and attributes describing different subqualities of astringency. Dry tannin intensity was rated lower for the early eluting anthocyanin-rich polymers compared to the later eluting fractions, which were rich in procyanidins and/or oligomerized flavan-3-ols. Accordingly, the attenuation of the astringent perception, as it is generally observed during red wine aging, may be attributed to increasing anthocyanin incorporation into the polymers. In contrast to astringency, the intensity of bitterness was not sufficiently correlated with any chemical parameter. This observation suggests that very specific phenolic structures are responsible for the bitter perception in red wines. As the high molecular pigmented polymers evoked lowest astringency ratings, the results of this study indicate that the increase of astringency due to higher degrees of polymerization can be substantially prevailed by the incorporation of anthocyanins into polymeric structures.
The analysis of polymerized phenolics in red wine bears numerous difficulties and hence information about the sensory characteristics of these polymers is rare. While monomers and low molecular weight polyphenols can be easily evaluated by HPLC-DAD analysis, red wine polymers elute mainly as a polydisperse hump and thus further analysis provides limited results (Bakker and Timberlake 1985, Revilla et al. 1999). There are a few methods that allow a partial characterization of red wine polymers with regard to their chain length and chemical composition. This inter alia includes precipitation assays using either proteins (Hagerman and Butler 1978) or polysaccharides (Sarneckis et al. 2006). In the case of pigmented polymers, their resistance against SO2 bleaching is determined (Harbertson et al. 2003). For homogenous proanthocyanidins, acid-catalyzed cleavage and subsequent reaction with a nucleophile (e.g., phloroglucinol) allows the determination of the mean degree of polymerization (Kennedy and Jones 2001). In the case of complex phenolic polymers from red wine, however, this procedure yields few meaningful results (Vidal et al. 2004c), most likely because of the higher number of possible monomers and different interflavonoid bond energies. Accordingly, the composition and the molecular size of phenolic polymers in red wines remain largely unknown.
The color of red wines is mainly formed by anthocyanins and anthocyanin derivatives, as well as phenolic reaction products, which are extracted during maceration and/or are formed during wine aging. Most of these pigmented phenolic compounds are thought to contribute to the taste and tactile sensations of red wines (Noble 1998). The bitter taste of wines is evoked by an unknown number of compounds and the contribution of phenols to the overall bitterness is still unclear (Noble 2002). On the other hand, astringency in red wines is exclusively induced by polyphenols that react with salivary proteins and eventually precipitate in the oral cavity. As a result, the saliva loses its lubrication properties and the typical puckering mouthfeel arises, which is commonly known as astringency (Noble 1998, Bajec and Pickering 2008). Astringency is known to be induced by monomeric flavan-3-ols and their condensed derivatives, the proanthocyanidins, also called condensed tannins (Brossaud et al. 2001). While the astringency in homogenous proanthocyanidins is reported to increase with the degree of polymerization (Goldstein and Swain 1963, Lea and Arnold 1978), the perceived bitterness of smaller oligomeric proanthocyanidins is thought to be higher than the bitterness of larger polymerized ones (Noble 2002). More recent findings have shown that astringency depends not only on the degree of polymerization but also on the composition of the polymers and the kind of linkage between the different monomers. Vidal et al. (2003) reported that the degree of galloylation plays an important role in the bitterness and astringency of equally sized tannins. Tannins extracted from the skins of grapes were found to be less astringent than seed-derived tannins, which show a higher degree of galloylation (Vidal et al. 2003). Anthocyanins also play a crucial role in astringency. It was assumed that adducts between proanthocyanidins and anthocyanins, which are designated as polymeric pigments and basically found in aged red wines, are less astringent than homogenous proanthocyanidins (Vidal et al. 2004c). Accordingly, the authors concluded that the attenuation and smoothing of mouthfeel, which is generally observed during red wine aging, may be explained by polymerization processes between proanthocyanidins and anthocyanins. A possible explanation of this phenomenon is given by Gawel (1998), who assumed that phenolic polymers consisting of tannins and anthocyanins are more polar than the consistently structured proanthocyanidins. As the formation of astringency is governed by hydrophobic interactions between saliva proteins and polyphenols, the astringency of more polar polymers is thought to be lower than the astringency of nonpolar proanthocyanidins (McRae and Kennedy 2011).
Although the molecular factors are crucial for understanding the changing astringency perception during red wine-making and aging, only a few publications have focused on a combined chemical and sensory characterization of polymeric red wine phenolics (Cheynier 2006). Among these are studies using isolated anthocyanins and proanthocyanidins from grapes and apples (Vidal et al. 2003, 2004c) and a study that emulates red wine aging by adding isolated anthocyanins to white wines (Oberholster et al. 2009). To our knowledge there are no published data showing a combined chemical and sensory approach on isolated and fractionated polymeric red wine polyphenols. Accordingly, this was attempted in the present study. As the isolation of polymeric red wine polyphenols was considered as a key element, we decided to apply a comprehensive extraction procedure to finally obtain 14 Sephadex fractions containing the polymeric red wine polyphenols. Different chemical techniques were applied to explore the structural and compositional complexity of these fractions. Results of the chemical analysis were substantiated by the size/polarity information given by the Sephadex separation. Furthermore, a descriptive sensory analysis using the descriptors bitterness and sourness and attributes describing different subqualities of astringency was conducted to characterize the 14 fractions. Finally, the obtained sensory and compositional data are brought in context, thus enabling a better understanding of how the composition of the polymers is likely to influence the astringency of a red wine.
Materials and Methods
Wine. The red wine was a 2005 Cabernet Sauvignon from the winery Winzer eG Herrenberg Honigsäckel, Rhineland-Palatinate, Germany. This wine was primarily chosen because of its high content of polyphenolic compounds and its maturity level suggested the presence of a sufficient amount of polymers. In 2009, the wine composition reflected the following: 12.5% ethanol, 5.0 g/L glycerin, pH 3.8, titratable acidity as 4.5 g tartaric acid equivalents/L, total phenolics as 3850 mg gallic acid equivalents/L, 0.4 g/L residual sugars, and 54 mg/L total SO2, determined using Fourier-transform midinfrared spectroscopy, including the appropriate calibration method (WineScan FT120 Basic; Foss, Hillerød, Denmark). Total phenols were determined using the Folin–Ciocalteu method (Möbius and Görtges 1974).
Chemicals. All solvents used for analyses were of HPLC quality. Other solvents and acids used for extraction were of analytical grade. Tartaric acid, potassium bitartrate, tannic acid, gallic acid, caffeine, aluminium sulfate, (+)-catechin, and (-)-epicatechin were purchased from Sigma Chemical (St. Louis, MO). Acetonitrile, ethanol, and monopotassium phosphate were purchased from Merck (Darmstadt, Germany). Malvidin 3-glucoside was purchased from Extrasynthese (Genay, France). Cyanidin 3-glucoside was isolated from a blackberry extract using the procedure described by Schwarz et al. (2003). Deionized water was purified to HPLC grade using a Milli-Q water purification system (Millipore, Bedford, MA).
Phenol extraction and fractionation. A volume of 8 L wine was dealcoholized using a rotary evaporator (40°C), and applied to an Amberlite XAD-7 column (1900 × 100 mm) (Sigma Chemical), which was first rinsed with 15 L methanol and then preconditioned with 20 L water. After elution of the wine, the column was washed with 40 L water to remove sugars, organic acids, and other nonphenolic compounds. The polyphenolic compounds were subsequently eluted with methanol acidified with 3% acetic acid at a gravity flow rate of ~5 mL/min. After complete elution, the solvent was evaporated and the extract was lyophilized. To remove medium polar substances like monomeric flavan-3-ols, benzoic acids, and hydroxycinnamic acids, the lyophilized XAD7-extract was dissolved in water (portions of 5 g in 500 mL) and extracted three times with 500 mL ethyl acetate. The aqueous layers were combined and lyophilized (yield 32.5 g). This procedure was adopted from an earlier study (Schwarz et al. 2003). 30 g of the so-obtained polar extract were dissolved in methanol/water (75/25 v/v) and applied to a Sephadex LH-20 column (1000 × 50 mm) (Sigma Chemical) that was preconditioned with methanol/water (75/25). The column was eluted with methanol/water (75/25) for 10 days and subsequently with pure methanol for additional 5 days at a gravity flow rate of ~0.5 mL/min. The fractionation process was monitored at 500 nm with a Knauer Variable Wavelength Detector (Knauer, Berlin, Germany) and Pharmacia LKB REC 102 writer, and fractions were collected by a LKB SuperFrac (Pharmacia, Uppsala, Sweden) according to the obtained chromatogram at 500 nm (Figure 1). As the concentration of the eluate was very high, the separation was monitored beside the expected absorption maximum at 520 and/or 280 nm to prevent problems that would have occurred by measuring out of the detector range. The eluate was stored at -20°C under N2 until evaporation and lyophilization. The 14 obtained fractions were additionally evaporated under high vacuum (10-4 bar) to remove any residual organic solvents. Thereafter, the original concentrations of the fractions were calculated by dividing the yield of each fraction by the volume of wine initially used (Table 1). Prior to chemical characterization, the lyophilized fractions were dissolved in a winelike solution (2.5 g/L potassium bitartrate, 12% vol ethanol, pH 3.6) at concentrations of 1 g/L.
Chromatogram and fractionation pattern of the Sephadex LH-20 separation recorded at 500 nm.
Yield of fractions after Sephadex LH-20 separation and recalculated concentrations of individual fractions in the Cabernet Sauvignon wine.
HPLC-ESI-MSn. A Bruker HCT Ultra Iontrap with electrospray ionization (ESI) was connected to an Agilent HPLC station consisting of a binary pump HP series 1100 and HP1200 series autosampling unit (Agilent, Santa Clara, CA). The column was a Kinetex RP-18 endcapped 150 × 4.6 mm, 2.6 μm (Phenomenex, Torrance, CA). The following gradient was used with a flow rate of 0.5 mL/min: 0 min, 95% A; 12 min, 80% A; 22 min, 60% A; 30 min, 40% A; 35 min, 10% A; 40 min, 10% A; 41 min, 95% A, 45 min, 95% A, with A being water/acetonitrile/formic acid (87/3/2.5; v/v/v) and B being water/acetonitrile/formic acid (40/50/2.5; v/v/v); ambient temperature. This method was adapted from one previously described (Rentzsch et al. 2009).
Quantification of monomeric flavan-3-ols and anthocyanins. HPLC-DAD analysis was performed on a modular system equipped with a low-pressure gradient pump (PU 980; JASCO, Groß-Umstadt, Germany), a ternary gradient unit (LG 20880-02; JASCO), a degasser (SCM 1000, Thermo Fisher Scientific, Waltham, MA), an autosampler (Basic Marathon, Spark, AJ Emmen, Netherlands), a column oven (Thermasphere TS-130; Phenomenex), and a multiwavelength detector (MD 2010 Plus; JASCO). A RP-18 endcapped 250 × 4.6 mm, 5 μm column (YMC Pack ODS AM, YMC Europe, Dinslaken, Germany) and a SecurityGuard precolumn (Phenomenex) of the same material were used for separation and kept at 40°C. After injecting 10 μL of the sample, the separation was carried out at a flow rate of 1 mL/min. The gradient consisted of two eluents: (A) 10 mM monopotassium phosphate in water with 5% vol acetonitrile at pH 1.5 and (B) 10 mM monopotassium phosphate in water with 50% vol acetonitrile at pH 1.5. The gradient program was as follows: 0 min, 92% A; 55 min, 73% A; 60 min, 73% A; 65 min, 30% A; 70 min, 30% A; 75 min, 92% A; 90 min, 92% A. The detection wavelengths were 280 nm for flavan-3-ols and 520 nm for anthocyanins. Major monomeric anthocyanins were quantified as malvidin 3-glucoside equivalents. The limit of detection (LOD) was set to a signal to noise ratio of 3 to 1 and reached a value of 1 mg/L.
Spectrophotometric analysis. Absorbance spectra were recorded in undiluted samples between 370 and 830 nm by a Cary 100 double-beam spectrophotometer (Varian, Palo Alto, CA), using a 1 mm path-length cell. A resolution of 1 nm and a slit width of 2.0 nm were used for all recordings. After values were corrected to a 10 mm path length, absorbance spectra were converted to transmission spectra. Thereafter, rectangular coordinates for lightness L*, red/green-plane a*, and yellow/blue-plane b* as well as cylindrical coordinates chroma C* and hue h* were calculated using the Commission Internationale de l′Eclairage method (CIE 2004) with the 10° standard observer and the illuminant D65, according to OIV recommendations (OIV 2006).
Chemical characterization. The methylcellulose precipitation (MCP) assay was performed according to a published procedure (Mercurio and Smith 2006). Polymeric pigments were analyzed using the modified Harbertson–Adams assay (Harbertson et al. 2003). The amount of incorporated anthocyanins in the polymeric fractions was determined by quantifying glucose after acid-catalyzed hydrolysis with sulfuric acid (glycosyl-glucose assay). The assay is primarily used to determine glycosidic aroma precursors in wine. However, it can also be used to determine the glucose moiety in nonvolatile phenolics (Williams et al. 1995). As Vitis vinifera contain only anthocyanin-3-O-monoglucosides, there is an equimolar relation between anthocyanin and glucose. The analytical procedure applied was adopted from a published method (Williams et al. 1995). After acidic hydrolysis and separation of phenolic compounds by solid-phase extraction, the glycosyl-glucose was quantified enzymatically using the Enzytec fluid test combination for the determination of dglucose (Product code 5140; Scil Diagnostics, Viernheim, Germany). For quantification, a cyanidin-3-glucoside standard was also treated according to the method of Williams et al. (1995) and the amount of glycosyl-glucose was expressed as mmol/g anthocyanin. A Cary 100 double-beam spectrophotometer (Varian) was used.
Sensory analysis. The sensory panel was composed of 18 (13 male and 5 female) volunteer judges from the Wine Campus in Neustadt/Weinstrasse. The age of the panelists ranged from 20 to 36 years with an average age of 25. All panelists had previously participated in descriptive analysis panels and reported to consume red wine at least twice a month. Prior to sensory sessions, standard attributes and concentrations were developed by an expert panel according to a previously developed attribute list (Durner et al. 2010). To familiarize panelists with the quantitative evaluation of the attributes, all panelists participated in four training sessions rating standards (Table 2) in three different concentrations: undiluted (high intensity), 1/2 diluted (medium intensity), and 1/3 diluted (low intensity). Quantitative training was conducted three times with single attribute standards and three times with mixed attribute standards (Table 2). For the administration of sensory training and data collection, FIZZ for Windows (ver. 2.00 D; Biosystèmes, Couternon, France) was used. Judges were asked to rate low intensity of descriptors in the first third on a 15-cm unstructured line scale shown on the screen, medium intensity in the middle third, and high intensity in the last third. The scales were anchored with “low intensity” at 1.5 cm and “high intensity” at 13.5 cm. All samples were evaluated in isolated booths under standard conditions (Meilgaard et al. 1999) using black DIN 10960 wine testing glasses to prevent discrimination due to different colors (Schott, Mainz, Germany).
Sensory descriptors, their definitions, and standard recipes for standards. All standards were prepared in Evian mineral water at pH 4.0.
Descriptive analysis of reconstituted wine and single fractions. Prior to sensory evaluation lyophilized fractions were dissolved either in a model wine (100 g/L ethanol, 3.0 g/L lactic acid, 1.5 g/L tartaric acid, 1.0 g/L glucose, 5.0 g/L glycerin, 200 mg/L catechin, 90 mg/L potassium bisulfite in Evian water; pH 3.8) adjusted to pH 4.0 with tartaric acid at concentrations as calculated for the initial wine (Table 1). Reconstituted wines and the original red wine were presented in one flight to all judges in volumes of 10 mL at room temperature. The three samples were evaluated in triplicate by all judges with at least 24 hr between repetitions. After evaluating the reconstituted wine, the 14 lyophilized fractions were individually dissolved in Evian water adjusted to pH 4.0 with tartaric acid at iso-concentrations of 500 mg/L. The dissolved fractions were also presented to judges in 10 mL volumes at room temperature. In each session four samples were evaluated in duplicate by all judges with at least 24 hr between repetitions. Irrespective of whether reconstituted wine or single fractions were evaluated, panelists had been instructed to take the total 10 mL sample volume in their mouth and to rinse thoroughly with water between samples because of the in-mouth persistence of the wines. In order to improve performance, panelists had access to a set of standards (Table 2) in each individual booth during session. In each of the three sessions, the three-digit coded samples were presented in a completely randomized order. All descriptors were evaluated monadically with a mandatory 3-min rest between each sample as part of the computerized evaluation session.
Statistical analyses. Sensory data was analyzed using a three-way mixed model ANOVA, treating the panelists as a random effect and wines and replications as fixed effects. The least significant difference (LSD) test was used to determine statistically different values at á ≤ 0.05. The principal component analysis (PCA) was conducted on means for judges and repetitions using Spearman’s correlation matrix with no rotation. Since sensory ratings showed non-normal distribution, the Spearman’s correlation coefficient was used for linear regression analyses. Statistical analyses were performed using XLSTAT (ver. 2008.7.03; Addinsoft, Paris, France).
Results and Discussion
Isolation and fractionation of polymeric pigments. The separation with Sephadex LH-20 material is based on a combination of size exclusion and adsorption mechanisms (Kantz and Singleton 1990, Henke 1999). Therefore, large molecules elute before smaller molecules and polar compounds before nonpolar compounds. HPLC-ESI-MSn analyses were conducted to evaluate the low molecular composition of the freeze-dried fractions. Main monomeric anthocyanins, catechin, and small oligomeric procyanidins up to trimers were detected (Flamini 2003) (Table 3). The analysis showed that catechin, epicatechin, and the procyanidin dimers and trimers eluted generally later on the Sephadex column than the anthocyanins. However, dimers and trimers of catechin were also detected in fraction 1, suggesting that some instable polymers degraded during Sephadex separation or MS detection. The presence of malvidin-3-glucoside in fractions 1 to 10 might also occur from cleavage of larger molecules as well as from entrainment effects on the Sephadex column. However, the overall elution pattern can be substantiated by the Sephadex separation theory as described earlier. Starting with anthocyanins from fraction 6 and progressing with catechin, its dimers and trimers from fraction 11, low molecular weight molecules were predominantly found in the second half of the Sephadex separation. The absence of low molecular weight compounds in the early eluting fractions indicated that these fractions mainly consisted of polymerized pigments. Nevertheless, no pigmented oligomers from anthocyanins and proanthocyanidins could be observed, which have been reported previously by several authors (Hayasaka and Kenney 2003, Vidal et al. 2004a), suggesting that they were either present only in very low concentration or incorporated into larger structures that cannot be detected by HPLC-MS. The absence of detectable amounts of low molecular compounds among others also suggests that the main proportion of the different fractions consists of high molecular compounds, which are barely ionizable and thus defying detection with conventional LC-MS methods (Fulcrand et al. 2008).
HPLC-ESI-MSn detection pattern of several phenolic compounds in fractions 1 to 14 (x indicates the existence of a signal).
In addition to HPLC-ESI-MSn analyses, monomeric flavan-3-ols and anthocyanins were quantified by HPLC-DAD using external calibration. Anthocyanins could be quantified only in fraction 7 (2.9 mg/g) and monomeric flavanols could be quantified in fractions 8 (3.3 mg/g) and 11 (1.5 mg/g). The original wine contained 36 mg/L catechin and 102 mg/L epicatechin. Anthocyanin content in the wine was below LOD. The absence of monomeric anthocyanins in the wine confirmed that they were largely consumed during the aging process. Anthocyanins found in fraction 7 may be explained by the concentration step due to the Sephadex chromatography. Compared to the anthocyanins, catechin and epicatechin were found in higher concentrations in the XAD-7 extract; however, the concentrations of these compounds were considerably lower in the individual fractions. This finding gives evidence that the ethyl acetate extraction was sufficient to deplete flavan-3-ol monomers.
Chemical analysis. Three well-established chemical assays were applied to assess the chemical composition of the 14 fractions. The MCP assay, which aims to quantify tannins by methylcellulose precipitation, showed that tannin concentration increased with a progressing retention time of Sephadex separation. Fractions 13 and 14 contained more than twice as much tannins as fractions 2 and 3 (Figure 2). However, the MCP assay provides no basis for the differentiation between tannins of different polymerization degrees, as the precipitate may be formed either with many smaller oligomers or with fewer large polymers and both will result in similar readings. Moreover, the early eluting fractions contain more anthocyanins, which certainly will effect precipitation with methylcellulose and may lead to lower readings. Nevertheless, it can be assumed that the high tannin readings in the late eluting fractions are caused by oligomers with a relatively low mDP. This assumption is based on the principle of the Sephadex separation, that is, small and/or nonpolar polymers elute after large and/or polar polymers. Furthermore, HPLC-ESI-MSn analysis revealed dimeric and trimeric proanthocyanidins only in late eluting fractions 11 to 14, which implies higher concentrations of these low molecular oligomers in the later fractions.
Tannin concentration, as determined by the MCP assay, in fractions F1 to F14 (error bars = 2SD calculated on analytical replicates).
In contrast to the MCP assay, the Harbertson–Adams assay uses a different precipitating agent and a different detection wavelength. Characterizing the fractions by this method revealed that large polymer pigments (LPP) were present in small proportions solely in the first two fractions, suggesting that the investigated red wine contained only low amounts of high molecular weight pigments assessed as LPP (Figure 3). In contrast to the first two fractions, fractions 3 to 14 contained only small polymer pigments (SPP). According to Landon et al. (2008), this finding suggests that the analyzed pigments in those fractions were composed of four or less monomeric units. Disregarding the size of the polymeric pigments in all 14 fractions, decreasing pigmentation was observed until fraction 7. In contrast, pigmentation increased again from fraction 8 to 13, suggesting that the analyzed pigments changed their intramolecular composition with a progressing Sephadex separation. However, similar readings in the Harbertson–Adams assay do not necessarily indicate a similar intramolecular composition of these polymeric pigments, owing to the principle of the assay. As the measurement is based solely on the absorption at 520 nm, the assay provides information on pigmentation but not necessarily on pigment concentration because different pigments may possess different concentrations of incorporated anthocyanins and other molecular diversities. Distinct polymers with an alternating amount of incorporated anthocyanins may result in comparable results, because the assay cannot differentiate between a high concentration of polymeric pigments with a low amount of incorporated anthocyanins and a low concentration of polymeric pigments with a high amount of incorporated anthocyanins, which will result in similar readings. Accordingly, no conclusions on the intramolecular composition can be drawn from the assay results itself. Nevertheless, considering the separation of polymers on the Sephadex column, the differences given by the Harbertson–Adams assay for the fractions suggest that the later eluting fractions contain higher concentrations of weakly pigmented polymers. These molecules are smaller and less polar and will therefore elute later than larger molecules and also later than molecules with a higher amount of incorporated anthocyanins.
Absorption of small and large polymeric pigments, as determined by the Harbertson–Adams assay, in fractions F1 to F14 (error bars = 2SD calculated on analytical replicates).
As the total amount of anthocyanins could not be specified by the Harbertson–Adams assay, the glycosyl-glucose assay was used to quantify the anthocyanin portion incorporated in the polymer fractions. Since free anthocyanins occur only in trace amounts in the fractions, it can be assumed that the released glucose determined in this assay has its origin in those anthocyanins that are incorporated in the polymers. Other sources of glycosyl-glucose-like flavanol-glucosides have been removed during the extraction. Their absence was proved by HPLC-MS, which could detect flavanol-glucosides in the organic layer of the liquid/liquid extraction but not in the water layer. Our results showed the highest glucoside concentrations in fractions 3 to 7, suggesting high concentrations of anthocyanin-rich polymers with high color potential (Figure 4). The decreasing glucoside content from fraction 7 to 14 may be explained by the decreasing polarity of the later eluting compounds and hence a smaller proportion of anthocyanins in the polymers. Conversely, the increase from fraction 1 to 3 is presumably induced by the decreasing size of the polymers and a proportionately smaller amount of anthocyanins in the molecules. However, comparable with the MCP assay, the glycosyl-glucose assay does not differentiate between the origin of the glycosyl glucose. A certain amount of anthocyanins will result in similar readings regardless of whether they were incorporated in one large molecule or in many small molecules. The assay provides information on the degree of the entity of compounds contained in the different fractions but not the degree of pigmentation of single oligomeric molecules. The findings of all three assays have to be considered in close relation to each other in order to draw structural and compositional conclusions about the fractions.
Glucoside concentration, as determined by the glycosyl-glucose assay, in fractions F1 to F14 (error bars = 2SD calculated on analytical replicates).
The Harbertson–Adams assay revealed that LPP were present in the first two fractions only and that the main proportion of the polymeric pigments was found in the first seven fractions (Figure 3). However, the results of the glycosyl-glucose assay showed that the main part of the anthocyanin-rich polymers eluted in the fractions 3 to 7 (Figure 4). Consequently, fractions 1 and 2 presumably comprised large polymers with low anthocyanin content. The subsequent fractions 3 to 7 most likely consisted of polymers with higher anthocyanin content. The MCP assay showed considerably lower readings in the first fractions and increasing tannin concentration from fraction 8 to 14 (Figure 2). The detailed structure of the polymers in the first fractions is completely unclear and, hence, also the mechanisms of complex formation with methylcellulose. In contrast, the content of incorporated anthocyanins, which was analyzed by the glycosyl-glucose assay, decreased with a progressing Sephadex separation, suggesting that fractions 8 to 14 presumably consisted of low molecular proanthocyanidin-like oligomers with a decreasing proportion of incorporated anthocyanins from fraction 8 to 14 (Figure 4). The increase in polymeric pigments from fraction 8 to 14 (Figure 3) and the contemporaneously observed decrease in anthocyanin content in the polymers (Figure 4) may be explained by the increasing tannin content (Figure 2), which contributes to color that is not explained by anthocyanins. In conclusion, three major groups of phenolic fractions were classified: fractions 1 to 3, consisting of large polymers with low anthocyanin content; fractions 4 to 7, containing anthocyanin-rich highly pigmented polymers; and fractions 8 to 14, consisting of small proanthocyanidin-like oligomers weak in anthocyanins.
Color analysis. CIELab analysis revealed that the 14 separated fractions formed two major groups: fractions 1 to 6, clustering in one group with lower yellow coloring, and fractions 7 to 14, clustering in another group with higher yellow tonality (Figure 5A). Except for fractions 1 and 7, the red tonality of all fractions was similar. Expressed by a very low chroma, fraction 1 was grey to brown (Figure 5B). Fraction 7 exhibited the lowest lightness; however, its high chroma and its red tonality identified this fraction as the one with the highest red color intensity. This finding is consistent with the results of the glycosyl-glucose assay (Figure 4), which revealed that fraction 7 had the highest proportion of anthocyanins incorporated in the polymer. With chroma values between 32 and 37 and a color hue of ~25, fractions 2 to 6 exhibited a clear red color. Interestingly, the lightness was consistently increasing from fraction 2 to fraction 6, indicating an increasing proportion of anthocyanins in the polymers (Liang et al. 2011). Due to the high b* values and chroma values between 47 and 52 (Figure 5), a higher orange tonality was recorded for fractions 8 to 13, suggesting greater proportions of proanthocyanidin condensation products in the polymer molecules (Jurd and Somers 1970). Considering the findings from the chemical assays, these proanthocyanidin condensation products appear to increasingly contribute to the perceived color while the proportion of anthocyanins decreased with a progressing Sephadex separation.
Chromatic characteristics a*, b* (plot A), and L*, C* (plot B) of fractions F1 to F14 (numbers in plots indicate fraction numbers).
Sensory analysis. In order to verify the sensory integrity of the polyphenolic isolate and to develop an appropriate reconstitute solution prior to sensory evaluation of the single fractions, reconstituted wines containing all fractions at concentrations as calculated for the initial wine were tasted (Table 1). In all assessed descriptors, the model-wine-reconstitute and the original wine were rated similarly, suggesting that the investigated taste active compounds were fully included in the fractions (Figure 6). The sensorial comparison of the Evian-reconstitute and the original red wine showed that the sour tannin and the dry tannin intensity of the Evian-reconstitute were rated significantly lower. Both differences may be explained by the pH, which was 4.0 in the Evian reconstitute compared to 3.8 in the original wine and in the model-wine-reconstitute (Fontoin et al. 2008). The higher bitter rating in the Evian-reconstitute may be a result of the absence of mollifying compounds (such as glycerin and sugar) in the original wine and in the model-wine reconstitute. Based on the objective to sensitively distinguish between different sensory properties, Evian water at pH 4.0 was chosen as an appropriate reconstitute solution to assess the mouthfeel and taste attributes of the single fractions at iso-concentrations of 500 mg/L.
Sensory intensity ratings for Cabernet Sauvignon wine and phenolic fractions combined and reconstituted in model wine and Evian water (* indicates p < 0.05).
As expected, all 14 fractions were rated similar in the sour intensity (results not shown), providing a valuable basis for the discrimination of the mouthfeel and taste attributes. Astringency ratings increased continuously with a progressing fraction number (Figure 7A), suggesting a systematic relation between the changing chemical composition of the individual fractions and their induced astringency. Due to the principle of Sephadex separations and the findings from Goldstein and Swain (1963), stating that astringency increases with an increasing polymerization of proanthocyanidins, different results have been expected. However, Goldstein and Swain (1963) focused on homogenously structured proanthocyanidins and our chemical analysis showed major variations not only in molecular size but also in composition (Figures 2, 3, and 4). Anthocyanins that are incorporated in polymeric polyphenols may be responsible for the attenuation of astringency in the early eluting fractions. This finding is in accordance with others (Vidal et al. 2004a); however, the argumentation did not cover the low astringency ratings in fractions 1 to 3, which presumably consisted of large polymers with relatively low anthocyanin content. Other intramolecular mollifying factors, such as trihydroxylation of the phenolic Brings (Vidal et al. 2003) or covalently bonded polysaccharides and/or peptides (Vidal et al. 2004b), may be responsible for this phenomenon. Another explanation for the increase in perceived astringency with progressing Sephadex separation may be the increasing amount of tannic material in the late eluting fractions, as discovered by the MCP assay (Figure 2) and HPLC-ESI-MSn analyses (Table 3).
Sensory intensity ratings for astringency (A), green tannins (B), dry tannins (C), and bitter (D) for phenolic fractions individually reconstituted in Evian water (values sharing the same letter are not significantly different at p < 0.05).
A further discrimination of astringency into astringent subqualities showed predominance of green tannins in fractions 3, 6, and 12 (Figure 7B). In contrast, dry tannins were more prevalent in fractions 4, 7, and 13 (Figure 7C). These results suggest that very individual structures contributed to dryness and greenness and cannot be correlated to parameters such as molecular size, polarity, or anthocyanin proportion in the polymeric polyphenols. Although not investigated as components incorporated in polymeric polyphenols, specific phenolic structures such as phenolic acids (Lawless et al. 1994) or the galloylation or trihydroxylation of proanthocyanidins (Vidal et al. 2003) are known to contribute to different astringent subqualities such as dryness, roughness, or coarseness.
Compared to astringency, the increasing intensities in green tannins and dry tannins were not as distinct, but also evident with a progressing Sephadex separation (Figures 7B, 7C). To verify the integrity of these sensory descriptors, the descriptive analysis ratings of the individual attributes were correlated to the results of the MCP assay by linear regression analysis. The highest correlation between tannin concentration and descriptive analysis ratings (r2 = 0.74) was observed for the astringency attribute (Figure 8A), confirming the efficiency of the MCP assay to model red wine astringency with reasonable confidence (Mercurio and Smith 2008). In contrast, the green tannin character (Figure 8B) showed only a weak correlation with the tannin content (r2 = 0.21), suggesting sensory information different from tannin-induced astringency. This finding confirms the hypothesis, discussed earlier, that specific structures in the fractions contribute to an astringent subquality described as an association between astringency and green/unripe flavor. The sensory descriptor dry tannin (Figure 8C) correlated quite well with the tannin concentration (r2 = 0.64), revealing a close relation to the astringency attribute, as noted previously (Mercurio and Smith 2008).
Regression analysis between sensory scorings and tannin concentration, as determined by the MCP assay.
Somehow puzzling is the protruding bitterness observed in fraction 9 (Figure 7D), suggesting no relation to molecular size and polarity of the fractions. This assumption initially seems to be in contradiction with the findings of Peleg et al. (1999), who stated that the bitterness declines with increasing degree of flavanoid polymerization. Similar to Goldstein and Swain (1963) with astringency, the bitter tastings of Peleg et al. (1999) focused on consistently structured proanthocyanidins; however, sensory properties of polymeric polyphenols, which are of heterogeneous nature, apparently do not follow any systematic mechanisms. Noble (2002) stated that even small differences in the molecule configuration can produce significant differences in bitterness, implying that fraction 9 contains very specific phenolic structures. These structures may be responsible for the observed bitterness and should be considered elsewhere for closer examination.
The PCA for the 14 fractions explained 71% of the variance among the sensory attributes, which consisted of astringency, dry tannins, green tannins, and bitterness (Figure 9). PC1 (47%) distinguished the fractions with different intensities in astringency and dry tannins, suggesting that these descriptors were affected most by the Sephadex fractionation. On closer examination, fraction 2 showed the lowest astringency/dry tannins intensity, which may be explained by covalently bonded polysaccharides and/or peptides incorporated in the fractions (Vidal et al. 2004b). The same fraction 2 also exhibited low green tannins and bitter ratings, indicating that other sensory attributes such as smoothness or suppleness may have expediently contributed to the sensory profiling of the fractions (Gawel et al. 2000). As apparent on PC2, fractions 9 and 10 were identified as the most bitter. At the same time, fraction 10 was weak in green tannin intensity, implying that unripeness of tannic structures was not necessarily related to bitterness. On the contrary, fractions 12 and 14 exhibited a high green tannin character, but low bitterness. To summarize the PCA results, the Sephadex fractionation, which yields small and/or nonpolar polymers after large and/or polar polymers, was strongly reflected in the systematic separation of polymeric polyphenols that increased in astringency with progressing elution time. The bitterness character, but also the green tannin character, could not be attributed to the Sephadex elution pattern, suggesting that very specific molecule structures are responsible for these sensations (Noble 2002).
Principal component analysis (PCA) biplot for phenolic fractions individually reconstituted in Evian water. PCA space was calculated on the four displayed sensory attributes.
Conclusions
The combination of different analytical techniques and sensory analysis revealed structural and sensory information of polymerized and pigmented polyphenols in an aged red wine. Very polar and large polymeric pigments exhibited a fairly low proportion of incorporated anthocyanins and hence a weaker chroma. With a decreasing molecular size, the content of incorporated anthocyanins increased, causing a systematic increase in lightness. After half-time of the Sephadex separation, the anthocyanin proportion in the fractions declined again and MCP tannins started to increase. As revealed by LC-MS analysis, these late eluting fractions inherited a number of low molecular proanthocyanidins, which contribute to the high MCP tannin readings. By alignment of all chemical data, three groups of fractions were observed that were differentiated by their size and molecular composition. Large and weakly pigmented polymers eluted before small and highly pigmented polymers and these prior to small proanthocyanidin-like oligomers. The sensory data showed interesting differences between the fractions. The early eluting fractions exhibited low astringency intensity. Fraction 2 was rated significantly lower in astringency, suggesting that smaller pigments with a higher amount of anthocyanins are less astringent than larger pigments with a higher amount of anthocyanins. The perceived astringency continuously decreased with a rising fraction number, suggesting an adverse interrelation between molecular size and astringency. Nevertheless the fractions with a low amount of incorporated anthocyanins showed higher astringency ratings, consistent with the decrease of astringency by incorporation of anthocyanins into polymers during wine aging, already proposed by several authors. Consequently, the key finding of this study is that the attenuating effect of incorporation of anthocyanins regarding astringency exceeds the intensification of the astringency due to higher polymerization degrees. Nevertheless, the high bitterness perceived for fraction 9 could not be explained in this study. Further studies have to be conducted to reveal the molecular size of the polymers and to add more insight into the composition.
Acknowledgments
The authors thank Martina Sokolowsky for support in conducting sensory descriptive analysis and the students of the Wine Campus in Neustadt/Weinstrasse for their participation in the sensory panel.
Footnotes
-
Publication costs of this article defrayed in part by page fees.
- Received May 2012.
- Revision received September 2012.
- Accepted September 2012.
- ©2013 by the American Society for Enology and Viticulture
Literature Cited
Vol 64 Issue 1
