Abstract
A protein precipitation method was used to measure the tannin concentration of 1325 commercial red wines made from Vitis vinifera cultivars: Cabernet Sauvignon, Merlot, Pinot noir, Syrah, Zinfandel, and several blends. Samples were taken from the United States (California, Oregon, Washington), Australia, and France. The concentration of tannins ranged from 30 to 1895 mg/L catechin equivalents (CE) among all wines measured, with a mean concentration of 544 mg/L CE and a standard deviation of 293. Within a single variety the variation in the concentration of tannins was larger than an order of magnitude and in two varieties (Cabernet Sauvignon and Pinot noir) the variation was 32-fold. Because of the wide variation in concentration range, there was overlap among the different populations. However, differences were found (p < 0.05) among the varieties at the population mean value: Cabernet Sauvignon (672 mg/L CE) ≥ Zinfandel (652 mg/L CE) > Merlot (559 mg/L CE) > Syrah (455 mg/L CE) > Pinot noir (348 mg/L CE). Pinot noir and Syrah wines from different states were compared and revealed that the tannin concentration in Oregon and California Pinot noir wines were not statistically different, while California Syrah was higher in tannin than both Washington State wines and those from several states in Australia.
Condensed tannins are the most abundant class of phenolic substances found in the skin and seeds of the winegrape Vitis vinifera (Singleton 1992). They are polymers composed of different flavan-3-ol subunits: catechin, epicatechin, epigallocatechin, and epicatechin gallate. Interflavan bonds link together the polymers primarily, between the C4–C8 and to a lesser degree the C6–C8 positions of the polymer (Prieur et al. 1994, Souquet et al. 1996). Epigallocatechin is unique to the skin, while epicatechin gallate that was previously thought an exclusive terminal subunit in the seed is also found in the skin (Prieur et al. 1994, Souquet et al. 1996, Downey et al. 2003). The average number of polymer subunits (reported as mean degree of polymerization; mDP) of seed and skin tannins is also measurably different: ~32 mDP for skin and ~10 mDP for seed (Souquet et al. 1996, Kennedy et al. 2001). The concentration of tannin in the fruit can be altered by environmental factors, including vine water status, heat, sunlight, and vine vigor. (Kennedy et al. 2000, 2002, Cortell et al. 2005). Tannins are extracted into wine from the skins and seeds during alcoholic fermentation; their ultimate concentration in wine can be enhanced or underplayed by winemaking style and technique (Sacchi et al. 2005).
The major interest in wine tannins is derived from their ability to interact with salivary proteins, and it is this mechanism that is thought partially responsible for the tactile sensation of astringency (Gawel 1998). Tannins are also important because they form covalent adducts with anthocyanins that retain some of the original color of the anthocyanins (Remy et al. 2000). Tannins contribute to the ability of red wines to age and have been implicated for potential health-related benefits (Corder et al. 2006).
An established protein precipitation assay (Hagerman and Butler 1978) has been adapted for use in grapes and wines and has been combined with bisulfite bleaching to measure polymeric pigments (Harbertson et al. 2002, 2003). This method has demonstrated a strong correlation with the sensory perception astringency (Kennedy et al. 2006) and has been extensively compared to other modern chromatographic techniques for red wine tannin content. It correlates well to the high molecular weight polymers and total polymer content accounted for by normal-phase HPLC and correlates strongly to the results of phloroglucinolysis and gel permeation chromatography (DeBeer et al. 2004, Kennedy et al. 2006). The method can be modified for use with a microplate reader to improve throughput and reduce reagent volumes for more efficient use in a winery laboratory (Heredia et al. 2006).
Wineries in the United States (California, Washington, and Oregon) and Australia have begun to use this method to measure tannins in red wines for use in pressing and blending decisions. The purpose of this study was to establish normative ranges for tannin in commercially available red wines using the protein precipitation method, a key undertaking because no single winery laboratory would be likely to perform such work and make the results widely available. This work records the tannin concentration of over 1300 commercially available red wines from Vitis vinifera cultivars (Cabernet Sauvignon, Merlot, Pinot noir, Syrah, Zinfandel) and some blends (from Bordeaux).
Materials and Methods
Chemicals.
Bovine serum albumin (BSA, fraction V, lyophilized powder), sodium dodecyl sulfate (SDS; lauryl sulfate, sodium salt, 95%), triethanolamine (TEA, 98%), ferric chloride hexahydrate (98%), and (+)-catechin hydrate (98%, powder) were purchased from Sigma (St. Louis, MO), as were materials for preparing buffers used in analyses. Reagents were prepared (Harbertson et al. 2003) and stored (Heredia et al. 2006) as described elsewhere.
Wine samples.
1325 commercial wines were analyzed for tannin concentration (Table 1⇓): 55 from the Bordeaux region of France, 364 Cabernet Sauvignon wines from Washington and California, 197 Merlot wines from Washington, 261 Pinot noir wines from Oregon and California, 266 Syrah wines from California, Washington, and Australia, and 182 Zinfandel wines from California. All U.S. wines were commercial products for sale and subject to standard Alcohol and Tobacco Tax and Trade Bureau regulation, while French and Australian wines were subject to each country’s respective regulations. Apart from Pinot noir, which is rarely blended, many of the wines were blends of multiple grapes but met the legal requirement for use of the grape variety indicated on the label.
Tannin measurements.
Tannins were measured as described elsewhere (Hagerman and Butler 1978) and modified for wine (Harbertson et al. 2002). Briefly, 500-μL aliquots of wine diluted (up to 5-fold) into a model wine buffer containing 5 g/L potassium bitartrate adjusted to pH 3.3 with HCl were added to 1 mL of pH 4.9, 200 mM acetic acid, 170 mM NaCl containing 1 mg/mL BSA and incubated at room temperature for 15 min. Samples were then centrifuged at 13,500 g for 5 min to form a pellet with a clear supernatant. The supernatant was discarded, and the remaining pellet was incubated for 10 min after adding 875 μL TEA buffer containing 5% TEA (v/v) and 10% SDS (w/v) adjusted to pH 9.4 with HCl. After the incubation period the sample was mixed mechanically to dissolve the tannin-protein pellet. To each sample, a 125-μL aliquot of ferric chloride reagent containing 10 mM FeCl3 in 0.01 N HCl was added to the tube and allowed to stand at room temperature for 10 min. After the incubation period the absorbance at 510 nm was determined in either a Shimadzu UV-160 spectrophotometer (Kyoto, Japan) or Beckman DU 640 spectrophotometer (Fullerton, CA) using the TEA buffer as a blank.
Data processing.
Microsoft Excel (Redmond, WA) was used for data entry and storage. Data analysis was performed with Statistica (StatSoft Tulsa, OK).
Results
The amount of tannin in all red wines was exceptionally variable. The mean red wine contained 544 mg/L CE with a standard deviation that was more than half the mean (Figure 1A⇓). The variation was larger than an order of magnitude with the range varying 33.6-fold. Distribution was tested using a Shapiro–Wilk test (Shapiro and Wilk 1965) for normality and was nonnormal at p < 0.01. The data showed a positive skew (+1.02) and kurtosis value of 1.58, meaning the distribution had a larger proportion of data on the right side of the distribution than would be expected in a normal distribution.
The average tannin concentration of Cabernet Sauvignon was 672 mg/L ± 303 CE. The average Pinot noir wine contained 348 ± 200 mg/L CE. Merlot wines averaged 559 ± 200 mg/L CE (Figure 1B⇑). Both Cabernet Sauvignon wines and Pinot noir wines had very wide ranges, varying about 32.6- and 33.3-fold, respectively, while Merlot only varied 12.4-fold. The average of Zinfandel and Syrah wines was 652 ± 342 mg/L CE and 455 ± 205 mg/L CE, respectively (Figure 1C⇑). The range of Zinfandel varied 13.1-fold while Syrah varied 19.1-fold. Despite the tremendous variation in the different wine types, they display concentration differences at the population mean level (Figure 2⇓). A comparison of means using an unequal N Tukey honestly significant difference (HSD) showed differences at the p < 0.05 level between each of the wine types except Zinfandel and Cabernet Sauvignon, which were similar to each other but were statistically different from the other wine types: Cabernet Sauvignon (672 mg/L CE) ≥ Zinfandel (652 mg/L CE) > Merlot (559 mg/L CE) > Syrah (455 mg/L CE) > Pinot noir (348 mg/L CE).
Pinot noir samples from California and Oregon demonstrated that those from Oregon were slightly more tannic but not enough to be statistically significant (Table 2⇓). Washington and California Cabernet Sauvignon were compared and were of similar magnitude. Syrah wines from Australia, California, and Washington were compared with ANOVA and a comparison of means using an unequal N Tukey HSD test ( p < 0.05). California wines had significantly higher tannin than Washington and Australia wines, which were not different from each other. Bordeaux wines, although blends of different grape cultivars, were included to determine their similarity to U.S. Cabernet Sauvignon and Merlot. The Bordeaux appellation and its subappellations are dominated by Merlot and Cabernet Sauvignon and have served as a model for New World wine production; thus, it is not surprising the average tannin concentration is similar to U.S. Merlot and Cabernet Sauvignon.
Discussion
In 2002 we found that total tannin per berry varied almost 2-fold while that of finished wines varied nearly 10-fold (Harbertson et al. 2002), a finding that led us to further investigate the variability in tannin concentration in over 1300 commercially available red wines from five grape varieties, three continents, and three countries. Each cultivar had a sample from several different vintages. Results showed that the tannin concentration of red wines was 3.3 times more variable than our original study indicated, likely because of increased sample size, inclusion of additional wine types, and range of geographic location. Other factors likely to have increased the individual population distributions include varied enological and viticultural techniques and conditions that alter the tannin content of fruit and wine (Downey et al. 2006, Sacchi et al. 2005, Cortell et al. 2005).
Variability of tannin concentration within a single wine type was up to 33-fold, and usually the standard deviation was very large (at least half of the mean value). Despite the wide variation, the population means of the different wine types were statistically different from each other except Cabernet Sauvignon and Zinfandel. That does not imply it is possible to identify wine type by measuring tannin concentration alone; there is too much overlap between the populations for such identification. The tannin concentrations of a wine type made in different regions were generally quite similar, such as Cabernet Sauvignon from Washington and California and Pinot noir from California and Oregon. There may be so much variability within the tannin concentrations of a wine type that it is difficult to discern a difference between regions. However, Syrah wines from Washington and several states in Australia were different from those from California, perhaps because of the smaller sample sizes from Australia and Washington State or the different winemaking styles and grape compositions. Further research is necessary to discover the origin of these apparent differences.
There are several methods developed to measure the tannin concentration in grapes and wine, and they have been reviewed (Herderich and Smith 2005). Tannins are of variable length and composition (Harbertson et al. 2002), and because of this complex chemical composition, the methods for their measurement are generally incomparable, although often they correlate (Sarneckis et al. 2006, Kennedy et al. 2006). Here we focus on the BSA protein precipitation methodology. One study reported the tannin concentration and several other phenolic components of five different wine types from Marche, Italy with six examples of each (Bosseli et al. 2004). The variation of all samples was 11.4-fold, while within a variety the greatest variation was only 3.4-fold. Another study assessed the tannin concentration of 40 Syrah, Merlot, and Cabernet Sauvignon wines by several different methods (including the same protein precipitation method in this paper) and compared them to sensory scores for astringency (Kennedy et al. 2006). The reported tannin values ranged from 387 to 1655 mg/L CE, which is only 4.3-fold variation. Much like previous studies (Bosseli et al. 2004), the number of samples evaluated was small compared to the data compiled in the work reported here. The values of the wines recorded easily fit into the distributions shown here.
An examination of tannin content of grape skins from 38 grape cultivars using various methods found only a 3.4-fold variation using protein precipitation with BSA (Seddon and Downey 2008), similar to results where skin tannin per berry only varied 2-fold (Harbertson et al. 2002). In both studies where grape tannins have been evaluated, the variation had a maximum of 3.4-fold in the skin and 2-fold for the entire berry. There is more variation in the recorded tannin concentrations of wine than previously thought and significantly more variation than the tannin concentration of fruit, suggesting that winemaking may play the larger role in a wine’s final tannin concentration.
Sensory scientists have recently created language for the description of astringency (Gawel et al. 2000). Some work has been done to show different astringency descriptions for different tannin polymer lengths (Vidal et al. 2003). The language developed and used is qualitative, but given the variation we observe it may be possible to connect the quantitative data and other wine chemistry components to sensory language to provide a more complete description of astringency. A recent study using qualitative sensory descriptions and chemical composition data has shown that the textural properties of Shiraz wines were associated with tannins, anthocyanins, and acidity (Gawel et al. 2007). This undertaking will initiate more research trying to associate the astringent properties of wine with quantitative values and qualitative terms.
The need for rapid high-throughput determination of the tannin concentration of red wines has lead to the development of predictive functions based upon the UV-visible spectra (Skogerson et al. 2007) and Fourier transform mid-infrared spectrum (FT-MIR) (Fernandez and Agosin 2007). In both studies, the BSA protein precipitation method used here was the reference method to compare different spectra to determine whether a model could be built to alleviate lab work. Results were successful; however, both studies evaluated approximately half of the variability of tannin concentrations as were evaluated here. The UV-visible work measured 200 red wine samples with a maximum value of 769 mg/L CE tannin (Skogerson et al. 2007). In the FT-MIR study, 68 samples were used with a maximum concentration of 900 mg/L CE tannin. Seventeen percent of the samples in our study were greater than 800 mg/L CE; therefore it may be necessary to refine these models to take the full variation into account.
Conclusion
The variability in the tannin concentration of commercially available red wine examined in this study was exceptionally great. Wines made from some grape varieties could be differentiated from each other based on the average tannin concentration, but there was too much overlap between the populations for it to be used as a sole identifier. The sources of red wine tannin variation are viticultural practices and winemaking techniques, with winemaking techniques likely playing the greater role.
Footnotes
Acknowledgments: A portion of this research was made possible by a grant provided by the Washington Wine Grape Funds.
The authors thank the various wineries in Washington, California, and Oregon that donated wine for this project.
- Received November 2007.
- Revision received February 2008.
- Copyright © 2008 by the American Society for Enology and Viticulture