Protein Binding and Astringent Taste of a Polymeric Procyanidin, 1,2,3,4,6-Penta-O-galloyl-β-D-glucopyranose, Castalagin and Grandinin

Abstract

The objective of the present investigation was to examine oral astringency and protein binding activity of four structurally well-defined tannins, namely procyanidin (epicatechin₁₆(4→8)catechin), pentagalloyl glucose (1,2,3,4,6-penta-O-galloyl-β-D-glucopyranose), castalagin, and grandinin, representing the three main structural categories of tannins, the proanthocyanidins, the gallotannins, and the ellagitannins. Astringency threshold and dose response were determined by the half-tongue test using a trained human panel. Protein binding stoichiometry and relative affinity were determined using radioiodinated bovine serum albumin in precipitation or competitive binding assays. Procyanidin and pentagalloyl glucose were perceived as highly astringent compounds and had relatively steep dose response curves but castalagin and grandinin had a lower mass threshold for detection. In vitro, procyanidin was the most effective protein precipitating agent, and grandinin the least. Increasing the temperature increased protein precipitation by the hydrolysable tannins, especially grandinin. All four polyphenols had higher relative affinity for proline-rich proteins than for bovine serum albumin.

INTRODUCTION

The defining characteristic of the high molecular weight polyphenols known as tannins is their ability to bind and precipitate proteins (1). The widespread distribution of tannins in plant-based foods and beverages has motivated decades of study of their interactions with proteins. Methods used to probe the interactions between polyphenol and protein include spectroscopic, thermodynamic, and chemical techniques (2–4) for examining the soluble or insoluble complexes formed. These studies suggest that the initial binding event between tannin and protein yields soluble complexes which upon subsequent cross linking are transformed into insoluble precipitates (3, 5). The interaction between polyphenols and proteins are the consequence of both hydrogen bonds between phenolic hydroxyl and peptide carbonyl, and hydrophobic “stacking” interactions between nonpolar amino acid residues and aromatic rings of the phenolic moiety (6). In addition, binding is clearly affected by protein characteristics including isoelectric point, secondary/tertiary structure, and amino acid composition, with proline-rich proteins having a particularly high relative affinity for tannins (7).

It is widely believed that the oral sensation of astringency is a consequence of interactions between ingested tannins and salivary proline-rich proteins (8–11). The high relative affinity of proline-rich proteins for polyphenols is a consequence of the open protein structure, the exposed polypeptide backbone, and strong hydrogen bonding properties of the tertiary amide in any amino acid-proline peptide bond (7, 12). Salivary proline-rich proteins may protect mammals from the nutritional consequences of consuming tannin-rich diets (13). Recent studies have demonstrated that in vitro, proline-rich proteins prevent uptake of tannins by gastrointestinal cells (14). In vivo, salivary proline-rich proteins diminish absorption and metabolism of dietary tannins (15).

Very few studies have been conducted to directly examine the role of polyphenol structure on protein binding, precipitation, or oral astringency. Broad structural features that distinguish condensed from hydrolysable tannins cannot be used to predict ability to precipitate protein. Among both condensed and hydrolysable tannins, the ability to precipitate protein increases as the number of catechol moieties on the polyphenol is increased (3, 9). It has been suggested that structurally flexible tannins bind protein more efficiently than more rigid tannins (16, 17). Polarity of the polyphenol may be important in the interaction, as suggested by the observation that the relatively hydrophobic α-anomer of 1,2,3,4,6-penta-O-galloyl-D-glucopyranose has a higher affinity for BSA than the slightly more polar β-anomer (10). More detailed comparisons of interactions between proteins and well defined tannins may ultimately allow reliable prediction of oral astringency and other bioactivities of tannins based on structural features.

The objective of the present investigation was to examine oral astringency and protein binding by four structurally well-defined tannins, namely procyanidin (epicatechin₁₆(4→8)catechin) (1) (Figure 1); pentagalloyl glucose (1,2,3,4,6-penta-O-galloyl-β-D-glucopyranose) (2), castalagin (3), and grandinin (4). These compounds represent the three main structural categories of tannins, the proanthocyanidins, the gallotannins and the ellagitannins. For each compound we quantitatively measured stoichiometry of protein precipitation and relative binding affinity for a proline-rich protein. In addition, we assessed the threshold and dose-response characteristics for astringency of the four compounds using a trained human taste panel.

Figure 1.

Chemical structures of procyanidin (epicatechin₁₆(4→8)catechin) (1), pentagalloyl glucose (1,2,3,4,6-penta-O-galloyl-β-D-glucopyranose) (2), castalagin (3), and grandinin (4).

MATERIALS AND METHODS

Chemicals

Caffeine, gallic acid, ellagic acid, epigallocatechin 3-gallate, chloramine T and bovine serum albumin (BSA) (Fraction V, fatty acid-free) were from Sigma-Aldrich (St. Louis, MO or Steinheim, Germany). Tannic acid and quercetin-3-O-β-D-galactopyranoside were obtained from Roth (Karlsruhe, Germany). Procyanidin (epicatechin₁₆(4→8)catechin (1) (Figure 1) was purified from Sorghum grain and its composition and average degree of polymerization determined by degradative cleavage (18). Pentagalloyl glucose (1,2,3,4,6-penta-O-galloyl-β-D-glucopyranose) (2) (Figure 1) was purified from tannic acid and its purity and identity confirmed by HPLC and mass spectrometry (5). Wood chips produced from oak (Quercus robur L. and Quercus alba L.), which was air dried for two years, were obtained from the cooperage industry (USA). Deionized water used for chromatography was purified by means of a Milli-Q Gradient A10 system (Millipore, Billerica, MA) or with a Nanopure system (Barnstead-Thermolyne, Dubuque, IA). For sensory analyses, bottled water (Evian) was adjusted to pH 4.5 with trace amounts of formic acid prior to use. All precipitation and competitive binding assays were carried out in 0.2 M acetate buffer containing 0.17 M NaCl, pH 4.9.

Bovine serum albumin was radioiodinated using chloramine T and Na¹²⁵iodine (Amersham Biosciences, Piscataway, NJ) (19). The labeled protein was stored at −20 °C in the acetate buffer. The protein was dialysed (12,000 MWCO) for 1–2 h at 4 °C immediately before use to ensure that all of the label was protein-bound. The specific radioactivity was adjusted to 20,000 cpm per 30 µg protein after dialysis. Calfskin gelatin (Eastman) was dissolved in the acetate buffer and diluted as needed for use in the assays.

Isolation and purification of castalagin and grandinin

Oak wood chips (500 g) were extracted with 1.5 L acetone/water (70/30 v/v) three times for 12 h while stirring. Acetone was removed and the extract was further separated by means of adsorption chromatography and preparative RP-HPLC as described recently (20) to obtain the pure ellagitannins castalagin (3) (Figure 1)and grandinin (4) (Figure 1). The purity of each ellagitannin was confirmed to be >99% by means of analytical HPLC, LC/MS and ¹H-NMR spectroscopy.

Protein binding and precipitation

Tannins were dissolved in water immediately before each experiment and concentrations were checked spectrophotometrically based on the following extinction coefficients at 280 nm: pentagalloyl glucose, 57.6 mL/mg/cm; castalagin, 22.7 mL/mg/cm; grandinin, 38.2 mL/mg/cm; procyanidin, 14.8 mL/mg/cm. The method described earlier was followed (4), with total reaction volumes of 400 µL for all determinations. Acetate buffer, protein, and tannin were dispensed into microfuge tubes with vortexing after each addition. The mixtures were incubated at room temperature (20 °C) or in a 40 °C water bath for 30 min, and were then centrifuged at room temperature at 12,000 × g for 10 min. Supernatants were removed by aspiration, and 100 µL of acetate buffer that was equilibrated at the appropriate temperature was added to each tube. Samples were not vortexed but were immediately centrifuged again for 3 min. Supernatants were aspirated and pellets were counted in a gamma counter (Packard Instruments, Downers Grove IL). Background binding of labeled protein to the tubes was always less than 10% of the total label added and was routinely subtracted during the calculations.

For stoichiometry and temperature dependence experiments, each reaction mixture contained 30 µg of the radiolabeled BSA and 0.5–60 µg of tannin. For the competitive binding experiments, each reaction mixture contained 30 µg of the radiolabeled BSA and either 10–100 µg of unlabeled BSA or 0.2–20 µg of unlabeled gelatin. The amount of tannin used in the competitive binding experiments was different for each tannin depending on the stoichiometry of binding for that compound. For experiments with procyanidin, 0.5 µg was used; for pentagalloyl glucose, 1.5 µg was used; for castalagin, 15 µg was used; for grandinin, 30 µg was used. Each point was replicated 3 times, and each experiment was performed at least two independent times. Data were fit and statistically analyzed using GraphPad Prism 4.03 (GraphPad Software, Inc. San Diego CA).

Sensory Analyses

Panel Training

In order to train the subjects to recognize and distinguish different qualities of oral sensations, 12 assessors with no history of known taste disorders (five women and seven men, age 24–38 years) participated for at least two years in weekly training sessions. Sensory analyses were performed in a sensory panel room at 19–22 °C in three different sessions. The subjects were trained to recognize the taste of aqueous solutions (5 mL each) of the following standard compounds dissolved in bottled water (Evian, low mineralization: 500 mg/L) adjusted to pH 4.5 with aqueous formic acid (0.1 %): sucrose (50 mmol/L) for sweet taste; lactic acid (20 mmol/L) for sour taste; NaCl (12 mmol/L) for salty taste; caffeine (1 mmol/L) for bitter taste, and sodium glutamate (3 mmol/L, pH 5.7) for umami taste. For puckering astringency and velvety-like astringency, the panel was trained by using gallustannic acid (0.05%) and quercetin-3-O-β-D-galactopyranoside (0.01 mmol/L), respectively, using the half-tongue test (21, 22).

Recognition Threshold Concentrations

Threshold concentrations of astringent compounds were determined in bottled water (pH 4.5) by means of the recently developed half-tongue test (21, 22) in order to overcome carry-over effects of astringent compounds. Serial 1:1 dilutions of the samples were presented in order of increasing concentrations to the trained panel of twelve persons in three different sessions, using the sip-and-spit method. At the start of the session and before each trial, the subject rinsed with water and expectorated. An aliquot (1 mL) of the aqueous solution containing the astringent compound was applied with a pipette on one side of the tongue, whereas pure water was applied on the other side of the tongue as the control. The sensory panelists were then asked to move their tongue forward and backward towards the palate for 15 s and to identify the place of astringent sensation by comparison of both sides. After indicating which part of the tongue showed the typical astringent sensation, the participant rinsed with water and, after 10 min, received another set of one blank and one taste-active sample. To prevent excessive fatigue, tasting began at a concentration level two steps below the threshold concentration that had been determined in a preliminary taste experiment. Whenever the panelist selected incorrectly, the next trial took place at the next higher concentration step. When the panelist selected correctly, the same concentration was presented again beside one blank as a confirmation of the initial response. The geometric mean of the two lowest concentrations was calculated and taken as the individual recognition threshold. The threshold value of the sensory group was approximated by averaging the threshold values of the individuals in three independent sessions. Values between individuals and separate sessions did not differ more than plus or minus one dilution step; that is, a threshold value of 1.1 µmol/L for castalagin represents a range of 0.55–2.2 µmol/L.

Recording of Human Dose-Response Functions

Serial 1:1 dilutions of the samples in water were prepared starting at the level of 256-fold above the recognition threshold concentration and ending at the concentration level two steps below the individual recognition threshold concentration. In order to fit the dose/response functions into a 5-point intensity scale, first, the taste intensity of the individual compounds was compared at the highest concentration level by means of the half-tongue tasting method, thus offering a direct comparison of the sensory impact and a reliable evaluation of the gustatory response of different compounds. To achieve this, the solutions of the individual compounds were applied in binary combinations to one side of the tongue and the assessors were asked to determine which side showed the stronger sensation (23). On a five-point scale with 0.25 scale subunits, a 10 mmol/L solution of epigallocatechin-3-gallate, used as the reference compound, was evaluated with the highest sensory intensity and set to the maximum score of 5.0. After the sensory intensity of each test compound at its maximum concentration had been rated, the sensory intensities of the other dilutions were determined by using the half-tongue tasting method. To achieve this, first, one dilution of an individual compound was rated against the intensity of the next lower as well as the next higher concentration of the same compound and the intensity of this solution was approximated by comparison to the taste intensity (scores given in brackets) of aqueous solutions containing the reference compound epigallocatechin 3-gallate in concentrations of 0.19 (0.5), 0.38 (1.0), 0.48 (1.5), 0.76 (2.0), 1.05 (2.5), 1.52 (3.0), 1.81 (3.5), 2.47 (4.0), 3.5 (4.5), and 10.0 mmol/L (5.0). Human response functions with dose-over-threshold factors on the x-axis and taste intensities on the y-axis were recorded for each individual subject in triplicates.

RESULTS AND DISCUSSION

Protein binding and precipitation

The four tannins examined in this study had different tendencies to precipitate protein (Figure 2, Table 1). As previously reported, protein precipitation by procyanidin was independent of temperature (Table 1) (4). There was a small but significant increase in protein precipitation by pentagalloyl glucose when the temperature was elevated to 40 °C (Figure 3, Table 1) (4). The hydrolysable tannins from oak precipitate protein more effectively at elevated temperatures than at room temperature (Table 1)… The effect of temperature was somewhat larger for castalagin and was substantial for grandinin (Table 1). Although as little as 1 µg of procyanidin or pentagalloyl glucose precipitated more than 10% of the available protein, at least 5 µg of the oak hydrolyxable tannins was required to precipitate detectable levels of protein under the conditions of this assay.

Figure 2.

Protein precipitation by procyanidin (1) (△- - -), pentagalloyl glucose (2) (), castalagin (3) () and grandinin (4) () at 20 °C. Points show means of three replicates, error bars indicate standard deviation, and lines are the fits obtained by analysis of the log transformed data in GraphPad.

Table 1.

Precipitation of BSA by Four Tannins at Two Temperatures^§.

Compound	PPT50 (µg tannin)		Apparent Hill slope		R2 for fit
	20 °C	40 °C	20 °C	40 °C	20 °C	40 °C
Procyanidin (1)	0.98 0.52 to 1.8	0.93 0.77 to 1.1	1.6 0.52 to 1.8	1.8 0.47 to 2.9	0.81	0.96
Pentagalloyl glucose (2)	2.6a 2.4 to 2.9	1.6b 1.5 to 1.8	2.6*x 2.1 to 3.0	2.4*x 1.8 to 3.0	0.91	0.87
Castalagin (3)	20c 18 to 22	12d 11 to 13	3.3* 2.6 to 4.1	2.4*y 2.0 to 2.8	0.97	0.99
Grandinin (4)	220e 74 to 670	25f 22 to 29	0.69z 0.27 to 1.12	1.0z 0.73 to 1.3	0.74	0.95

^§Different lower case letters indicate a statistically significant difference between temperature for each compound (p<0.001). An asterisk indicates Hill slope significantly different from 1.0 (p<0.01). The range of values indicates the 95% confidence limit for the PPT₅₀ or the apparent Hill slope.

Figure 3.

Precipitation of protein by pentagalloyl glucose at two temperatures. The amount of protein precipitated by various amounts of tannin was measured at 20 °C (□) and at 40 °C (▲). Points shown are the average of three determinations, and the lines are the fits to the log-transformed data.

Binding curves can be quantitatively compared by two coefficients, the EC₅₀ and the Hill slope (24). Since protein precipitation is more complex than simple binding, we defined PPT₅₀, a parameter analogous to EC₅₀, to describe the amount of tannin required to precipitate half of the protein that is present in the assay. Precipitation data for the four tannins was analyzed using GraphPad to obtain fits with acceptable correlation coefficients (Figure 3, Table 1). For procyanidin, precipitation was independent of temperature (data not shown). All three hydrolyzable tannins precipitate protein more efficiently at 40 °C than at room temperature, indicated by a lower EC₅₀ at the higher temperature (Table 1). For pentagalloyl glucose and castalagin, PPT₅₀ was about two-fold lower at 40 °C than at room temperature (Table 1). For grandinin, the difference was almost ten-fold (Table 1).

The Hill slope reflects the steepness of the binding curve. Ligands that bind to identical, independent sites on a protein yield a Hill slope equal to 1.0. A steeper binding curve (larger Hill slope) suggests that binding may be positively cooperative (24). By fitting our precipitation data we obtained values analogous to Hill slopes, and we report those as apparent Hill slopes. Castalagin and pentagalloyl glucose have apparent Hill slopes significantly larger than 1.0 (Table 1). Binding by procyanidin or grandinin fits a simple model with apparent Hill slope = 1.0 (Table 1).

Further comparisons of binding and relative affinity were conducted at 40 °C. In addition to promoting protein binding by the hydrolysable tannins, this temperature approaches physiological temperature. Competitive binding assays allow convenient comparison of relative binding affinity of tannins for various proteins, using a radiolabeled protein as the binding agent and other proteins as competitors. Since formation of either soluble or insoluble complexes between tannin and competitors inhibit precipitation of the tracer, this assay yields relative binding affinities (7). We expressed binding affinities as I₅₀ values, the amount of competitor required to inhibit precipitation of the radiolabeled binding agent by 50%. It is well established that procyanidin and pentagalloyl glucose have high relative affinity for proline-rich proteins including the salivary proline-rich proteins found in mammals (7, 25, 26). We used gelatin as a model proline-rich protein, unlabeled BSA as a control competitor and radiolabeled BSA as the tracer.

Because each tannin has a unique binding stoichiometry (Table 1), a different amount of each tannin was used in the competitive binding assays (Table 2). As expected when the competitor is identical to the binding agent,, the I₅₀ for unlabeled BSA was the same for all four tannins, and was equal to the amount of radiolabeled BSA used in each assay (30 µg). This confirms that these tannins do not discriminate between radiolabeled BSA and unlabeled BSA in the binding assay, and that differences in binding stoichiometry do not invalidate the method.

Table 2.

Competitive Binding Assays for Four Tannins^**.

Compound	Tannin (µg)	I50 (µg competing protein)
		BSA	Gelatin
Procyanidin (1)	0.5	29a 27 to 32	3.0x 2.6 to 3.6
Pentagalloyl glucose (2)	1.5	36a 28 to 46	6.3y 5.2 to 7.5
Castalagin (3)	15	26a 20 to 34	5.5y 4.9 to 6.1
Grandinin (4)	30	26a 24 to 29	4.3z 3.9 to 4.8

^**Different lower case letters indicate a statistically significant difference between compounds for a single protein (p<0.05). For all compounds, I₅₀ for gelatin was significantly less than I₅₀ for BSA (p<0.05). The range of values indicates the 95% confidence limit for the I₅₀.

All four tannins have a higher relative affinity for gelatin than for BSA (Table 2). Procyanidin has the highest relative affinity for gelatin. The relative affinity of grandinin for gelatin is about 30% lower than that of procyanidin, while pentagalloyl glucose and castalagin have relative affinities that are only about half the affinity of procyanidin for gelatin (Table 2).

Sensory Evaluation of Ellagitannins

To evaluate the sensory quality and sensory impact of these compounds, the oral recognition threshold concentrations were determined in water (pH 4.5) using the half-mouth test for astringency (Table 3). The oral sensation imparted by these compounds was described as astringent and was detectable at relatively low threshold concentrations ranging from 0.2 to 1.8 µmol/L. The lowest threshold concentration for oral astringency was obtained with the monomeric ellagitannin C-pentoside grandinin (Table 3). In contrast, the monomeric ellagitannin castalagin, which lacks the pentose moiety, was only detected at a five-fold higher threshold concentration, thus indicating that the C-glycosylation of the ellagitannin monomers enhances the astringent sensation. Pentagalloyl glucose, which is a key intermediate in ellagitannin biosynthesis (27), exhibited astringency at a threshold concentration very similar to the threshold found for castalagin (Table 3). Procyanidin had a lower perception threshold, very similar to that of grandinin.

Table 3.

Taste Threshold Concentrations for the Astringent Sensation Induced by Four Tannins in Aqueous Solution (pH 4.5)^††.

Compound	Oral threshold conc.
	[µmol/L]	[mg/L]
Procyanidin (1)	0.3	1.48
Pentagalloyl glucose (2)	1.8	1.69
Castalagin (3)	1.1	1.03
Grandinin (4)	0.2	0.21

^††Values are averages for 12 individuals each tested in three different session,, The values have a range of plus or minus one dilution step; that is, a threshold value of 1.1 µmol/L for castalagin represents a range of 0.55-2.2 µmol/L.

Human Dose/Response Functions

We recorded human dose/response functions for the four compounds to evaluate their sensory activity at different concentrations and to reveal differences in sensory behavior. Panellists often have difficulties in remembering the intensity of a taste compound for a long period of time, so the same solution of a given test compound tasted at different time intervals may be given different ratings (28). Consequently, recording dose/response functions with standard sensory methodologies usually leads to unreliable curves with very high error margins. To overcome this problem, we applied the recently reported half-tongue testing (21), offering the possibility of a direct comparison of the sensory impact of two samples. On a five-point numerical scale with 0.25 scale subunits, human dose/response functions were determined for each individual subject for pentagalloyl glucose, castalagin, grandinin and procyanidin using standard solutions of epigallocatechin 3-gallate as the reference to define the astringent intensity represented by the individual scores (Figure 3). After the taste intensity of each compound at its maximum solubility had been rated, the taste intensities of the other dilutions were determined by using the half-tongue tasting method so that one dilution of an individual compound was rated against the intensity of another dilution of the same compound and the intensity of this solution was approximated by comparison to the taste intensity of the reference compound epigallocatechin 3-gallate in defined concentrations. Human response functions with dose-over-threshold factors on the x-axis and taste intensities on the y-axis were recorded for each individual subject in triplicates. The intensity values between trained individuals and separate sessions did not differ more than plus or minus 0.4 units on the 5 point scale (Figure 3).

The results, shown in Figure 3, clearly demonstrated that the gustatory responses for the different compounds follow rather different dose/response functions. In particular, the perception of either pentagalloyl gluocse or procyanidin yields rather steep dose-response curves and high sensory intensities at higher concentration levels. The highest intensity of 5.0 was found for an aqueous solution of pentagalloyl glucose, at a concentration 256-fold higher than its threshold concentration. Procyanidin reached an intensity of 4.0 at a concentration 128-fold higher than its threshold, with testing at higher concentrations impossible due to limited solubility. Grandinin only reached a maximum bioresponse with a score of 3.0, whereas the monomeric ellagitannin castalagin did not reach the same taste intensity as found for the three other compounds, and was just perceived with an intensity score of 2.5 at a concentration 256-fold higher than the threshold concentration. Both ellagitannins exhibited a low slope for astringent intensity with ascending concentrations.

We probed the role of tannin structure in protein binding and astringency by examining four structurally defined tannins representing the three major classes of tannins found in terrestrial plants. Procyanidin (1) (Figure 1) is a simple B-1-type proanthocyanidin, with catechin terminal unit and (4→8)-linked epicatechin units. The average degree of polymerization of the polymer isolated from sorghum grain is 17 yielding a molecular weight of about 4900 Da (18). The extended random coil flavan-3-ol polymers (29) are extremely hydrophilic as indicated by their very small octanol water partition coefficients, P (30). The log P for procyanidin is −2.7 (4). Pentagalloyl glucose (2) is a lower molecular weight (940 Da), more hydrophobic compound (log P 2.2) (31) comprising 5 galloyl esters surrounding a core glucose. The more rigid ellagitannins form by oxidative coupling of adjacent galloyl groups in the parent compound pentagalloyl glucose. Castalagin (3) (934 Da) is an ellagitannin derived from pentagalloyl glucose by oxidation, glucose ring opening and galloyl group migration (32). C-glycosylation of castalagin on C-1 by the pentose lyxose yields grandinin (4) (1066 Da) (33, 34).

In our experiments, stoichiometry of binding reflects the amount of a given tannin required to precipitate a standard model protein, BSA, at pH 4.9, the optimum pH for precipitation of this protein (35). We have compared PPT₅₀ values on a mass basis, at two temperatures, for the four structurally distinct tannins (Table 1, Figure 2). Procyanidin is a very efficient protein precipitating agent, with the lowest PPT₅₀ among the compounds we examined. On a mass basis, over 1.5-times more pentagalloyl glucose than procyanidin is required to achieve PPT₅₀. On a molar basis, procyanidin is almost ten-times more effective than pentagalloyl glucose. Although pentagalloyl glucose is a more effective precipitating agent than some other simple galloyl glucoses (36) it is clearly a less efficient protein binding agent than tannins such as procyanidin. The rather poor precipitating efficacy of castalagin and grandinin, namely seven and fourteen times less effective than pentagalloyl glucose on a molar basis, may be a consequence of the rigid structures of the ellagitannins. Structural rigidity constrains cross-linking to a few specific geometries so higher concentrations of ligand are needed to achieve aggregation. In contrast, both procyanidin and pentagalloyl glucose have flexible structures and are free to form cross-links in many different conformations.

Our data suggest that entropy-driven mechanisms such as hydrophobic interactions may have a particularly important role in the interaction of grandinin with BSA. The PPT₅₀ for grandinin is ten times lower at 40 °C than at room temperature (Table 1, Figure 3), indicating substantially stronger binding at the higher temperature. Increased binding at higher temperatures is typical for entropy-driven mechanisms such as hydrophobic interactions, but not for enthalpy driven interactions such as hydrogen-bonding… Hydrophobic interactions may play a smaller role in precipitation of BSA by pentagalloyl glucose or castalagin, since temperature has a smaller effect with these compounds. The very polar, high molecular weight procyanidin must bind to BSA almost exclusively via hydrogen bonds since protein precipitation by procyanidin is temperature independent. The extreme temperature dependence of the grandinin-protein interaction is surprising based on structural considerations and measures of hydrophobicity. Grandinin is slightly more polar than castalagin (37) as expected since grandinin is the C-glycoside of castalagin. The estimated octanol water partition coefficient for castalagin, based on its chromatographic retention factor, is intermediate between pentagalloyl glucose and procyanidin (38). Clearly hydrophobicity is not the only characteristic that determines the tendency of a polyphenol to precipitate protein.

The temperature dependence of precipitation by gallotannins and ellagitannins is interesting in the context of foods, beverages and biological systems, as it suggests that the tannin may not complex to protein prior to consumption when the beverage is stored at or below room temperature. When the food or beverage is consumed, its tannins may interact strongly with food, salivary, or gastrointestinal tract proteins as the ingested polyphenols reach body temperature. An earlier report that affinities become weaker as temperature is increased was not confirmed here (3).

Models for precipitation of protein by tannin suggest that if tannin concentration is low, soluble tannin-protein complexes form, containing 1–3 mol tannin per mol protein (5). If tannin concentration is sufficiently high, multivalent complexes form. With purified tannins, saturation stoichiometries of 20–40 mol tannin per mole BSA have been reported (4) although stoichiometries as high as 175 mol tannin per mol BSA have been obtained in studies using unpurified mixtures of polyphenols (17). We used PPT₅₀ values (Table 1) to estimate the stoichiometry of tannin to protein in the precipitated complex, on a molar basis. The calculated stoichiometries for procyanidin, about 8 mol procyanidin per mol BSA, and pentagalloyl glucose, about 25 mol pentagalloyl glucose per mol BSA at room temperature, confirm data reported earlier (4). The calculated stoichiometries for castalagin (160 mol/mol) and grandinin (>5000 mol/mol) are very high, suggesting that in addition to stabilizing phenol-protein interactions, there must be numerous phenol-phenol interactions to form large colloidal aggregates (3).

All four polyphenols have five- to ten-times higher relative affinity for proline-rich proteins than for BSA (Table 2). Even polyphenols with rigid structures, such as the ellagitannins examined here, bind proline-rich proteins with high relative affinity because the structural flexibility of the protein compensates for structural rigidity of the phenolic. Affinity for proline-rich proteins is not a direct function of stoichiometry of binding. Procyanidin, the most efficient protein precipitating tannin, and grandinin, the least efficient protein precipitant, shared high relative affinities for proline-rich protein. Pentagalloyl glucose and castalagin had somewhat lower relative affinities for the proline-rich protein.

We propose that the astringent response is a combined function of the ability of a given tannin to bind soluble proteins such as BSA, and its tendency to interact with proteins via hydrophobic binding. The taste panel assigned a relatively high threshold concentration (Table 3) for detecting the astringency of procyanidin and pentagalloyl glucose, which are very effective protein precipitating agents. Castalagin and grandinin, which are less effective at precipitating BSA, were detected at lower levels by the taste panel. The entropically-driven component of binding which is characteristic of castalagin and grandinin suggests that these tannins may preferentially bind to hydrophobic constituents of the mouth rather than to soluble salivary proteins, and binding to these components may elicit the astringent response. Tannins like pentagalloyl glucose and procyanidin may selectively bind soluble proteins and only associate with membranes when present at high concentrations, resulting in a relatively large taste threshold. The steep taste dose response for procyanidin and pentagalloyl glucose (Figure 4) may reflect the ability of these tannins to saturate soluble proteins, and then bind membrane bound proteins very efficiently. Castalagin and grandinin may have flat dose-response curves because they bind and saturate the membrane required for the astringent sensation at relatively low concentration. Further structure/activity studies are needed in the future to validate and understand the relationship between the chemical structure of a polyphenol, its protein binding activity, and its oral astringency impact.

Figure 4.

Human dose/response functions recorded for procyanidin (1), pentagalloyl glucose (2), castalagin (3), and grandinin (4).