Introduction
Astringency is not a taste but a tactile sensation felt in the entire mouth [1]. This feeling is the result of a strong interaction between tannins and saliva proteins
leading to the formation of a supramolecular colloidal complex that can precipitate
and, consequently, modify the palate lubrication [2]. Numerous specific terms have been used to describe the intricate sensation of astringency
of alcoholic beverages [3], especially red wine, this vocabulary being associated to either gustative qualities
or defaults. One possible way to explain the complex oral perceptual phenomenon induced
by tannins might be their high level of structural polymorphism [4]. Wine tannins are mainly derived from the solid part of the grape and are transferred
to wine during the maceration process. The major tannins present in wine come from
the proanthocyanidin family, especially procyanidins. They are polymers of 2 basic
units that can be distinguished from each other by the stereochemistry on carbon 3
([Fig. 1]). The polymerization process takes place from the C4 of an upper unit and the C6
or C8 of a lower unit, leading to potentially 8 different dimers, 32 trimers, 128
tetramers, and so on. It is commonly acknowledged that the concentration of tannin
in wine influences wine taste and is dependent on factors including soil, vintage,
wine-making process, vine, or weather [5]. However, little is known about the influence of procyanidin structure and colloidal
state on astringency.
Fig. 1 Procyanidin structures. The two monomers catechin and epicatechin (upper) and C4–C8
procyanidin polymers (bottom).
In the present work, we try to decipher some facets of astringency by adopting the
following strategy. First, a new way to synthesize procyanidins by controlling both
the stereo- and regiochemistry of the interflavan link as well as their polymerization
degree was developed [6]. Second, structural preferences [7], [8] and colloidal state of defined procyanidins [9] were investigated in a wine-like medium using NMR and molecular modeling. And third,
the interaction of different procyanidins with a peptide representative of a human
saliva proline-rich protein (PRP, [Fig. 2]) was investigated. This protein plays a key role in astringency owing to its strong
affinity for polyphenols [10]. This study, using NMR and molecular dynamics [11], sheds new light on the understanding of astringency at a molecular level.
Fig. 2 PRP structures showing their common repeated region.
Materials and Methods
Procyanidin synthesis
Procyanidins dimers and trimers were synthesized using a general way to obtain procyanidins,
in quantity, regio- and stereocontrolled at the level of the interflavan bond [6].
IB7-14, a 14 residues peptide representative of human saliva PRPs, was synthesized
using a solid phase Fmoc synthesis strategy as previously described [12].
NMR experiments
NMR was used for the following purposes:
(i) To determine the 3D structure of tannin or peptide alone or for the tannin/peptide
complex: in this case, we used the classical 1D- 2D NMR experiments such as COSY,
TOCSY, NOESY, HSQC, and HMBC.
(ii) To measure physicochemical parameters governing the self-association of tannins.
To do so diffusion coefficients (D) variations were followed at various concentrations
of tannins. D was obtained from DOSY NMR as previously described [13]. The change in D with respect to tannin concentrations was fit to the relationship
shown below to obtain the association constant value Ka [14]:
ΔD = |Dobs-Dfree| = (Dmax − Dfree) Ka[T0]{2/[1 + ({Ka[T0] + 1)1/2]}1 (Eq. 1)
Dobs is the observed D; Dfree is the D of the non-associate tannin; Dmax is the maximal D of the tannin T present in an aggregated form; and [T0] is the total concentration of tannins. The CMC value (critical micelle concentration)
of the tannins could also be deduced: when D values were plotted against the inverse
of the concentration, two straight lines with different slopes were obtained, and
the x-coordinate of the intersection point between the two straight lines gave the
CMC values.
(iii) To measure physicochemical parameters characterizing the interaction between
the peptide and the tannins [11]. Notably, the stoichiometry and the dissociation constant of the complex were obtained
by fitting the chemical shift variations of characteristic protons of the peptide
at different tannin/peptide ratios using Equation 2 [15]:
Δδ = œΔδ
max [(1 + Kd/n[P0] + [Ti]/n[P0]) − {(1 + Kd/n[P0] + [Ti]/n[P0])2 − 4[Ti]/n[P0]}œ] (Eq. 2)
Kd is the dissociation constant of the tannin peptide complex; n is the number of
tannins able to bind to the peptide; [Ti] is the tannin concentration able to bind
the peptide; and [P0] is the peptide concentration. Following the size of the complex through diffusion
measurement by DOSY NMR gives rise to the same parameters. In this case, equation
2 was used but with D instead of Δδ.
Molecular modeling
Molecular modeling and dynamic calculations were performed on a SGI Octane R10 K work
station using different force fields depending on the systems considered: MM3 was
used for all 3D structure determination of tannins [7], [8] whereas the AMBER force field was used to obtain the 3D structure of the peptide
IB7-14 or the IB7-14/B3 complex [11].
Results and Discussion
The general approach for synthesis of 4–8 linked procyanidins was based on the stoichiometric
coupling of 2 tetrabenzylated monomeric units with a TiCl4 catalyst: the nucleophilic partner was the tetrabenzylated monomer and the electrophilic
partner was the C4-activated and C8-protected tetrabenzylated monomer ([Fig. 3]). This strategy was inspired by two different works: Tückmantell and coworkers,
who developed the coupling strategy between nucleophilic and electrophilic partners
[16] and Saito and coworkers who protected the C8 of the electrophilic partner to control
the degree of polymerization [17]. Using this general strategy, we produced all the 4–8 procyanidin dimers with relatively
good overall yields: 29 % for B1, 27 % for B2, 38 % for B3, and 30 % for B4.
Fig. 3 General strategy to synthesize procyanidin 4–8 dimers.
The synthesis of all the 8 (4–8) trimers was also possible by using the same strategy.
Two ways are possible ([Fig. 4]):
Fig. 4 Ways to access procyanidins with higher degrees of polymerization illustrated for
trimers.
Method 1: The compound resulting from the coupling step (the benzylated and C8-brominated
dimer) could be used as the nucleophilic partner after C-8 debromination, while the
activated monomer was used as the electrophilic partner. In this case, an upper extension
occurred with a total overall yield of 27 %.
Method 2: The octobenzylated C8-protected dimer could be C4-oxidized to form the electrophilic
partner followed by coupling to the tetrabenzylated monomer to form a trimer. In this
case a lower extension occurred with a total overall yield of 18 %.
Procyanidins occur in a conformational mixture in solution owing to two distinct conformational
mechanisms:
(i) The heterocyclic ring (C, F, I) oscillates between two states where the 2-aryl
group is in a pseudo-equatorial (Eq) or ‐axial (Ax) position. This phenomenon is extremely
rapid with regard to the NMR time scale so that the measurement of 1H coupling constants between the H2 and the H3 of each heterocyclic ring gives rise
to the (Eq)/(Ax) ratio [18].
(ii) The interflavanoid bond decreases the rotational rate due to steric hindrance.
Two rotameric forms (compact and extended) are expected at each interflavanoid link
[19].
(iii) The systematic study of the four 4–8 procyanidin dimers and the trimer C2 in
water or in a wine-like medium shows that these two mechanisms are responsible for
the 3D‐structural differences. These differences are not really predictable from one
procyanidin to the other and can greatly influence their overall 3D structure.
The 3D structures of synthesized procyanidin dimers [7] and trimers [8] in water were determined using both NMR and molecular modeling. The preferred 3D
structures adopted in a wine-like medium are displayed in [Fig. 5]. For dimers, the compact form always dominates, but in very different proportions
from one dimer to the next (95 % for B1, 55 % for B2, 98 % for B3, and 76 % for B4).
The heterocyclic rings (C/F) always adopt a conformation in which the catechol rings
(B/E) are in the equatorial position. The 3D structure of trimers in a hydroalcoholic
solution has also been reported previously for Cat-Cat-Cat and Cat-Cat-Epi [8]. Four rotameric forms coexist, one of them being predominant. For the Cat-Cat-Cat
trimer a compact-compact conformer predominates, in which the catechol rings B and
E adopt an equatorial position when the H ring adopts an axial position (60 %, [Fig. 5]).
Fig. 5 3D structure preferences in a wine-like medium for the five different procyanidin
dimers and one trimer.
In light of these findings, it is clear that the complexity of the 3D structure of
procyanidins increases with their degree of polymerization and that these molecules
have to be considered as dynamic mixtures.
The self-association that occurs when procyanidins are dispersed in water or hydroalcoholic
solutions has to be investigated in order to evaluate their bioavailability towards
saliva proteins, their real contribution in wine turbidity, and their probable influence
for tannins/proteins interactions. This can be followed by measuring diffusion coefficient
with diffusion NMR spectroscopy (DOSY) [9]. Under certain conditions, the diffusion coefficient of a molecular species depends
on its molecular weight, size, and shape [20], but it can also be used as a probe to follow molecular association leading to varied
“colloidal mixtures” [13], [21], [22].
First, measurement of diffusion coefficient values at different tannin concentrations
provides access to the association constant. The simplest way to characterize tannin
self-association is to consider that all the stepwise association constants, Ka, are
the same with respect to an isodesmic model [14]. In this case, Ka should be deduced from Eq. 1 by fitting the experimental data
[9]. The values measured are close to 7 M−1 for dimers and 5 M−1 for trimers ([Fig. 6 A]).
Fig. 6 Colloidal behavior of procyanidin B3. The two plots represent two ways of observing
the evolution of D with respect to tannin concentration. A A direct representation showing the decrease of D that has been fitted using Eq. 1.
Experimental points: symbol, calculated points: line. B An indirect representation showing the evolution of D as a function of the inverse
concentration of tannin. The intersection of the two lines gives the CMC value.
Second, the optimal conditions for micelle formation occur above the critical micelle
concentration (CMC). This value is obtained by plotting the D value against the inverse
of the tannin concentration. As shown in [Fig. 6 B], two straight lines are obtained: one representative of the free state and the other
of the aggregated state. The intersection of the two lines gives the inverse of the
CMC value. It appears that the CMC value increases when the degree of polymerization
of the tannin increases: from ∼ 10 g/L for dimers to 13 g/L for the trimer C2. Above
this CMC value, micelles of polydisperse size are formed. The average size could be
estimated from both the measurement of Dmicelle deduced from Equation 1 and [Fig. 6 B, ] and the following Stokes-Einstein relationship:
D = kBT/6πηrH (Eq. 3)
RH is the hydrodynamic radius of the formed micelle. While Ka and CMC values are of
key importance to determine the “active” proportion of tannin able to fix proteins,
Dmicelle gives access to the mean size of the formed micelles and, thus, to the contribution
of tannins in wine turbidity. It is noteworthy that significant differences are observed
depending on the solvent used, for example, 10 % ethanol increases tannin solubility.
Finally, it is of interest to highlight that the average size that tannins micelles
can reach decreases in the presence of 10 % ethanol. Such small micelles cannot play
a role in wine turbidity (around 25/13 Å for dimers and 50/23 Å for trimers in water
alone/with 10 % ethanol).
A study of tannin-saliva proteins interaction has been undertaken in light of both
structural and dynamical data obtained for procyanidin dimers and trimers. In the
first step, a representative saliva peptide was synthesized [11]: IB7-14, as shown in [Fig. 2]. It represents a model containing the repetitive sequence found two to five times
in almost all basic PRPs [23]. The interaction between different procyanidins (B1, B2, B3, B4, and C2) and the
IB7-14 fragment was monitored by using both chemical shift variations of selected
peptide protons ([Fig. 7 A]) and DOSY‐NMR experiments. Plotting these variations as a function of tannin concentration
shows that the process is saturable binding, in accordance with a specific binding.
That conclusion can be confirmed using DOSY‐NMR experiments: collecting D values at
various tannin concentrations ([Fig. 7 B]) clearly shows their progressive decrease. The data suggest that the molecular object
formed diffuses more and more slowly, as expected if procyanidin binds the peptide.
By fitting the experimental chemical shift or D data points using Eq. 2 and considering
a multisite peptide-tannin interaction, where all the binding sites exhibit the same
affinity [24], physicochemical parameters characterizing the complex formation were obtained ([Table 1]). For all the tannins studied, the number of binding sites remains the same and
is approximately three dimers or trimers for one peptide. These sites have been located
at the level of the P2, P9-P10, and G13–G14 residues in the peptide [11] by using both ROESY experiments, chemical shift variations amplitude, and molecular
dynamics (vide infra). However, great differences are observed between Kd values leading to an affinity
scale in which C2 is 20 times more efficient in binding to the peptide than its dimer
counterpart B3. These affinity differences can be correlated with structural preferences
in solution as well as their ability to induce the peptide-tannin complex aggregation.
The tannins that are most potent for inducing peptide precipitation are the dimer
B2 and the trimer C2. These are also the procyanidins that appear to have structures
in which the phenolic moieties are exposed so that one tannin can bind two peptides
and initiate network formation and subsequent precipitation.
Fig. 7 Monitoring procyanidin-PRP interactions by using either chemical shift variation
of the peptide NH resonances (A) or diffusion variation (B) as reported in [Table 1].
Table 1 Binding, diffusion data, and sizes of soluble tannin-PRP complexes. Dissociation
constants (Kd) and number of tannin binding sites (n) were obtained from the fit of experimental
chemical shift variations of G13 and G14 NH, P2, P9, and P10 Hα and from the diffusion coefficients variation for the peptides using Eq. 2. The different
Kd and n values obtained for one tannin were averaged and are reported ± SD. The hydrodynamic
radius RH was obtained with the Stokes-Einstein Equation (Eq. 3).
Procyanidins
|
Kd (mM)
|
n
|
Dmax
|
RH
|
(10−10 m2 · s−1)
|
(Å)
|
B1
|
2.9 ± 1.4
|
3.0 ± 0.4
|
2.2 ± 0.1
|
9 ± 1
|
B2
|
1.1 ± 0.4
|
3.2 ± 0.5
|
1.9 ± 0.1
|
11 ± 1
|
B3
|
8.0 ± 0.9
|
3.0 ± 0.5
|
2.1 ± 0.1
|
9 ± 1
|
B4
|
2.5 ± 0.4
|
3.5 ± 0.3
|
1.8 ± 0.1
|
11 ± 1
|
C2
|
0.4 ± 0.1
|
3.0 ± 0.2
|
1.3 ± 0.1
|
15 ± 1
|
IB714
|
|
|
3.0 ± 0.1
|
7 ± 1
|
Molecular dynamic calculations were run in order to test whether the network initiation
process could occur. In the first simulation, two peptides and three trimers were
randomly dispersed in a full box of water to give a tannin concentration below their
CMC (5 mM). At the end of the 60 ns calculation, one C2 was able to link two peptides
([Fig. 8 A]). The binding sites were the same as those previously observed for B3 [11] and were located in the hydrophilic parts of the two peptides. The second run was
performed with nine C2 trimers corresponding to a 15 mM tannin concentration, i.e.,
at its CMC value. In the first few ns of this simulation, micelles of two to three
C2 trimers are formed. The association between a peptide and a C2 micelle preceded
formation of a more intricate supramolecule formed of at least six C2 trimers and
two peptides ([Fig. 8 B]). In this particular case, random hydrophobic interactions and not specific stacking
occur between tannin and protein.
Fig. 8 Molecular lipophilic calculation of the complexes formed of 2 peptides (green ribbon)
with 1 (A) or 6 C2 trimers (B). The hydrophilic part is shown in blue, the hydrophobic part in red and the interface
in white.
All these results shed new light on the molecular explanation of tannin astringency.
Two cases have to be taken into account depending on the colloidal state of tannins
([Fig. 9]). Below the CMC, tannins interact specifically with proline-rich peptides, with
three specific binding locations. It was clearly established that the affinities of
different tannins towards proline-rich peptides depend on their structural features:
tannins presenting phenolic rings free of any intramolecular stacking are able to
bind up to two PRPs with high affinity and initiate precipitation of the complex.
Tannins that adopt a compact conformation bind only one peptide with a lower affinity.
Above the CMC, tannins interact with PRPs in a micellar state: even if the first PRP
appears to bind in a specific way (the same peptide sites are occupied initially as
at lower tannin concentration), a more complicated complex is formed in fine in which both hydrophobic and hydrophilic forces are involved.
Fig. 9 Tannin-PRP interaction depends on the tannins 3D‐structural preferences and the colloidal
state of the tannin.
Acknowledgements
We thank the Conseil Interprofessionnel des Vins de Bordeaux (CIVB), the Centre National
de la Recherche Scientifique (CNRS), the Aquitaine Government, and the Université
de Bordeaux for financial support, Nathan Mc Clenaghan for the review of the English
language, and Paul de Boissel for the artwork.