Abstract
A reported analytical method for tannin quantification relies on selective precipitation of tannins with bovine serum albumin. The reliability of tannin analysis by protein precipitation on wines having variable tannin levels was evaluated by measuring the tannin concentration of various dilutions of five commercial red wines. Tannin concentrations of both very diluted and concentrated samples were systematically underestimated, which could be explained by a precipitation threshold and insufficient protein for precipitation, respectively. Based on these findings, we have defined a valid range of the tannin response in the protein precipitation-tannin assay, which suffers minimally from these problems.
Tannins play an important role in the mouthfeel properties and color stability of red wines and are therefore related to wine quality (Kennedy et al. 2006, Singleton and Trousdale 1992). However, reliable quantitative analysis of wine tannins is challenged by the chemical diversity of tannins. Various analytical methods for tannin analysis have been described and reviewed elsewhere (Herderich and Smith 2005, Makkar 1989, Schofield et al. 2001). Tannin analysis by protein precipitation was recently reintroduced as a fast and precise tool for measuring tannins in grapes and wines (Harbertson et al. 2003) and has been recommended for applications within winery settings (Harbertson and Spayd 2006). The method relies on tannins being separated by precipitation with bovine serum albumin (BSA), redissolved, and measured by a color reaction with ferric chloride (Hagerman and Butler 1978, Harbertson et al. 2003).
Tannins determined by protein precipitation have a particularly good correlation with astringency, as compared with some of the other available methods for measuring tannins or related polyphenols (Kennedy et al. 2006). While the mechanism for the precipitation is not fully understood, the existence of threshold levels for tannin precipitation to occur have been reported (Hagerman and Butler 1978, Hagerman and Robbins 1987). The purpose of this study was to investigate the influence of dilution degree on the reliability of tannin analysis by protein precipitation.
Materials and Methods
Chemicals and wines.
Five commercial wines with both low and high tannin concentrations were purchased locally: wine 1, Atacama, Cabernet Sauvignon 2005, Chile; wine 2, Cahors Cuvée Prestige, Malbec 2003, France; wine 3, Cecchi–Chianti Classico, Sangiovese 2004, Italy; wine 4, Argento–Mendoza, Malbec 2005, Argentina; and wine 5, Valpolicella Ripasso–Cantina de Soave, Corvina 2004, Italy. Chemicals for tannin analysis were all of analytical grade and purchased from Sigma-Aldrich (St. Louis, MO): bovine serum albumin (BSA, fraction V powder), tartaric acid, sodium dodecyl sulfate (SDS), acetic acid, sodium chloride, triethanolamine, (+)-catechin hydrate, and ferric chloride hexahydrate
Tannin analysis by protein precipitation.
All wines were analyzed in duplicates by the precipitation method as described by Harbertson and colleagues (Harbertson et al. 2003) with a few modifications. Prior to analysis the wines were diluted in a model wine solution of 12% v/v ethanol containing 5 g/L tartaric acid, which had been adjusted to pH 3.3 with NaOH. The tannin protein precipitate was formed by mixing 0.5 mL diluted wine and 1 mL BSA solution (containing 1 mg BSA/mL dissolved in a buffer of aqueous 0.2 M acetic acid and 0.17 M sodium chloride adjusted to pH 4.9) for 30 min. Adding the more acidic wines to the BSA solution only slightly decreased the pH, consistently giving final pH values ≥4.8. The precipitate was centrifuged 5 min at 14,000 g (Centrifuge 5415D, Eppendorf AG, Hamburg, Germany) to form a pellet and the supernatant was discarded. The pellet was washed twice: each time with 0.25 mL of the pH 4.9 buffer, discarding the supernatant after centrifugation for 3 min at 14,000 g. The rinsed pellet was redissolved in 1.5 mL buffer (containing 5% w/v SDS and 5% v/v triethanolamine and adjusted to pH 9.4 with HCl) by constant mixing for 20 min. The background (ABG) was measured in a spectrophotometer (Cary 300, Varian, St. Helens, Australia) as the absorbance at 510 nm of 1 mL redissolved solution in a 10-mm semimicro cuvette (Brand, Wertheim, Germany), which was subsequently mixed with 125 μL ferric chloride solution (11.4 mM ferric chloride in 11.4 mM aqueous HCl). The final absorbance at 510 nm (AFeCl3) was measured after 10 min, and the tannin response was calculated as 1.125 times the final absorbance minus the background absorbance. Accounting for dilutions, the tannin concentration was calculated and expressed as mg catechin equivalents (CE) per L from a standard curve of the color reaction between catechin and ferric chloride.
Results and Discussion
The linearity of the tannin response was evaluated by analyzing the tannin concentration of various dilutions of five wines (ranging from undiluted to 10 times dilution) and plotting the tannin response against the inverse dilution factor (Figure 1⇓). Within parts of the dilution ranges, there were good linear relationships (r > 0.999) between the inverse dilution factor and the tannin response. However, in some cases at low dilutions (i.e., the concentrated wine samples), the tannin response did not increase proportionally with the inverse dilution factor. This deviation from linearity was likely caused by insufficient protein for the precipitation step. Furthermore, all five wines caused negative y-intercepts, which indicated the existence of a threshold level for the precipitation to occur. The variation in the y-intercepts values did not allow assigning these to a constant value and thereby correcting for this systematic error. At low tannin responses, the y-intercept amounted to a high percentage of the response and hence caused tannin estimations that were too low.
The calculated tannin concentration of both very diluted and concentrated samples were systematically underestimated (Figure 2⇓), probably due to a precipitation threshold and insufficient protein for precipitation, respectively. The maximum tannin concentration was determined for each of the five wines (wine 1, 298 mg CE/L; wine 2, 546 mg CE/L; wine 3, 703 mg CE/L; wine 4, 457 mg CE/L; and wine 5, 391 mg CE/L). These concentrations were used as benchmarks for expressing the calculated tannin values as percentages of maximum concentration. The concentrations, ranging from ∼300 to 700 mg CE/L, only covered a part of the known high variation in tannin concentration of wines (Harbertson et al. 2003, Heredia et al. 2006). The relation between the tannin response and the measured tannin concentration in percentage of the maximum demonstrated that at both low and high tannin responses the tannin concentrations were underestimated (Figure 3⇓). Considering the substantial impact of the intercept for highly diluted wines and the need for sufficient protein for the precipitation step, we defined a valid range of the tannin response, where the tannin precipitation suffered minimally from the described problems. By allowing a 5% underestimation, we recommend that the tannin response lies between 0.3 and 0.75 abs under the given conditions. When the original volumes of the protocol are used (Harbertson et al. 2003), the range must be multiplied by a factor of 1.5, which gives a valid range of the tannin response between 0.45 and 1.125 (calculated as AFeCl3 – 0.875·ABG). If the tannin response falls outside this range, then there is a risk that the tannin level is underestimated. For example, when wine 3 was measured undiluted, tannin concentration was underestimated by 22% compared with the maximum determined concentration. Likewise, when wine 5 was diluted 10 times, the tannin content was underestimated by 27%. Since these results are obtained using only wine samples, it is advised to check the linearity for other sample types, for example, grape extracts which have much smaller background readings than wines.
Setting a minimum tannin response of 0.3 absorbance units limits the level of tannin that can be reliably quantified to ∼140 mg CE/L without prior concentration of the sample. Samples with tannin concentrations less than 140 mg CE/L will most likely be underestimated because of the impact of the precipitation threshold. Even though the saturation stoichiometries of mole tannin per mole BSA are known to vary for different tannins (Hagerman et al. 1998), we recalculated the data to tannin:protein ratios (expressed as mole CE per mole BSA in the precipitation step) and related these to the percent measured tannin levels (Figure 4⇓). From this, the valid range of tannin (in mole CE) to protein (in mole BSA) was between 17 and 45.
Conclusions
The reliability of tannin quantification was hampered by nonlinearity at higher tannin concentrations and the existence of threshold levels for protein precipitation to occur. These two phenomena caused underestimated tannin contrations in samples with either low or high tannin responses. To ensure reliable tannin quantification, we recommend that sample dilutions are carefully carried out to give a tannin response between 0.3 and 0.75 absorbance units under the given conditions.
Footnotes
Acknowledgments: The Danish Ministry of Science and Technology is gratefully acknowledged for their financial support for the project.
- Received July 2007.
- Revision received September 2007.
- Copyright © 2008 by the American Society for Enology and Viticulture