Langmuir 2008, 24, 11027-11035 11027

Modeling Procyanidin Self-Association Processes and Understanding

Their Micellar Organization: A Study by Diffusion NMR and
Molecular Mechanics
Isabelle Pianet,*,†,‡ Yann André,†,‡ Marie-Agnès Ducasse,‡ Isabelle Tarascou,†
Jean-Claude Lartigue,† Noël Pinaud,† Eric Fouquet,† Erick J. Dufourc,‡ and
Michel Laguerre*,‡
Institut des Sciences Moléculaires, UMR 5255 CNRS - UniVersité Bordeaux 1, Talence, France, and
CBMN, UMR 5248 CNRS - UniVersité Bordeaux 1 - ENITAB, Pessac, France

ReceiVed May 23, 2008. ReVised Manuscript ReceiVed July 2, 2008

The colloidal behavior of eight synthetic procyanidins (three monomers, four dimers, and a trimer) has been
investigated in water or in a winelike medium using DOSY NMR spectroscopy and molecular dynamics simulations.
Different behavior was observed for monomers and oligomers. Monomers self-associate with a high affinity constant
(37-53 M-1) to form micelles at low cmc (critical micelle concentration) values (1-5 g · L-1). These micelles undergo
a time-dependent coalescence process to form hazes and precipitates. As for dimers and the trimer, self-association
also occurs but with a lower affinity (∼6 M-1) and at higher cmc values (10-20 g · L-1) to form small micelles (<5
nm) that remain stable throughout the experiment. The presence of 10% ethanol does not significantly affect the
self-association constant for monomers and oligomers but increases their cmc values by approximately 50% and
decreases the micelle size by a factor 2. However, the presence of 20 mM NaCl appears to negate the effect of ethanol.
This study helps to clarify the role of procyanidin monomers versus oligomers in wine turbidity and demonstrates
that procyanidin oligomers are fully available to interact with saliva proteins.

Introduction able to stack with itself or with caffeine.10 Studying different

Among the large family of tannins,1 procyanidins are
widespread in human food (e.g., in grapes where they are located
in the solid part of the berry, seeds, and skin). As a consequence,
they are present in wine, their concentration depending on the
variety of vine, vintage, climate, soil, and winemaking procedure.2
Besides their well-known antioxidant properties in human health,3
they participate in wine quality in two different ways. First, they
play an important gustative role because of their ability to induce
astringency and/or bitterness,4 where astringency is known to be
the consequence of a nonspecific interaction occurring between
tannins and salivary proteins5 and bitterness comes from their
specific interaction with receptors present in all areas of the
tongue.6,7 Second, they have been proposed to play a key role
in wine limpidity and stability,8,9 owing to their ability to self-
associate in a nonspecific way and to form colloids, hazes, and
even precipitates. This behavior was early recognized by Baxter
et al., who showed that the very simple tannin methyl gallate was
* To whom correspondence should be sent. Fax: 33 540002623. Tel: 33
540006448. E-mail: i.pianet@ism.u-bordeaux1.fr.
†
UMR 5255 CNRS - Université Bordeaux 1.
‡
UMR 5248 CNRS - Université Bordeaux 1 - ENITAB.
(1) Haslam, E. J. Nat. Prod. 1996, 59, 205.
(2) Ribéreau-Gayon, P. Traité d’Oenologie. 2. Chimie du Vin, Stabilisation
et Traitement ISBN 2100073028 ed. Dunod, La Vigne: Paris, 1998; Vol 2, p 520.
(3) Chung, K. T.; Wong, T. Y.; Wei, C. I.; Huang, Y. W.; Lin, Y. Crit. ReV.
Food Sci. Nutr. 1998, 38, 421.
(4) AdVances in Food Research. Astringency of fruit and fruit products in
relation to phenolic content; Joslyn, M. A.; Goldstein, J. L., Mrack, M. E. Stewart,
G. F., Eds. Academic Press: New York, 1964; p 179.
(5) Hagerman, A. E.; Butler, L. G. J. Biol. Chem. 1981, 256, 4494.
(6) Noble, A. C. Physiol. BehaV. 1994, 56, 1251.
(7) Chandrashekar, J.; Hoon, M. A.; Ryba, N. J. P.; Zuker, C. S. Nature 2006,
444, 288.
(8) Poncet-Legrand, C.; Cartalade, D.; Putaux, J. L.; Cheynier, V.; Vernhet,
A. Langmuir 2003, 119, 10563.
(9) Riou, V.; Vernhet, A.; Doco, T.; Moutounet, M. Food Hydrocolloids 2002,
16, 17.
tannin fractions (differing by the average degree of polymerization
(mPD), the degree of galloylation, or the stereochemistry of the
flavan-3-ol) by diffusion light scattering, Poncet-Legrand et al.
reported the formation of aggregates but did not find any clear
relationship between the tannin structure and their tendency to
self-assemble.8 Nevertheless, they demonstrated the strong
influence of ethanol content and ionic strength on this aggregation
process. In a recent study and working again with a mixture of
tannins of mDP ) 11, Zanchi et al. showed that only 2% of this
fraction behaves as a colloid whereas the remaining 98% acted
as hydrophilic single molecules.11 However, they were not able
to attribute the behavior of both populations to structural
differences because the fraction on which they worked was very
disperse in size and in chemical structure.12
Beyond its evident role in wine turbidity, tannin self-association
must also be considered for tannin-protein interaction purposes
because it governs their bioavailability (i.e., the effective
concentration of tannins able to interact with proteins).
In the present work, the colloidal properties of procyanidin
monomers, dimers, and a trimer (Scheme 1) in water and in a
winelike medium are investigated. Individual molecules have
been synthesized in our laboratory,13 and their 3D structures
have been previously determined by using both multidimentional
NMR and molecular modeling.14,15 Notably, we demonstrated
that their chemical structure triggers their degree of “compactness”
and internal dynamics. For dimers, the compact form with an
equatorial conformation of the two heterocyclic rings is the
(10) Baxter, N. J.; Williamson, M. P.; Lilley, T. H.; Haslam, E. J. Chem. Soc.,
Faraday Trans. 1996, 92, 231.
(11) Zanchi, D.; Vernhet, A.; Poncet-Legrand, C.; Cartalade, D.; Tribet, C.;
Schweins, R.; Cabane, B. Langmuir 2007, 23, 9949.
(12) Mané, C.; Sommerer, N.; Yalcin, T.; Cheynier, V.; Cole, R. B.; Fulcrand,
H. Anal. Chem. 2007, 79, 2239.
(13) Tarascou, I.; Barathieu, K.; André, Y.; Pianet, I.; Dufourc, E. J.; Fouquet,
E. Eur. J. Org. Chem. 2006, 23, 5367.

10.1021/la8015904 CCC: $40.75  2008 American Chemical Society

Published on Web 09/04/2008
11028 Langmuir, Vol. 24, No. 19, 2008 Pianet et al.

Scheme 1. Planar Structures and Numbering of Tannins Used in This Studya

a
(A) Monomers: (1) catechin (Cat); (2) epicatechin (Epi); (3) epicatechin gallate (EGCG,)3. (B) Dimers: (4) B1, (5) B2, (6) B3, (7) B4 C) trimer (8) C2.

Scheme 2. Tannin Colloidal Behaviora

conformation largely preferred in water or the winelike medium
when the lower unit is catechin (B1, B3, Scheme 1) whereas the
proportion of “extended” rotamer is enhanced for dimers with
an epicatechin as a lower unit (B2, B4, Scheme 1), suggesting
that the stereochemistry of carbon 3 of the lower unit plays an
important role in the rotamer ratio. Procyanidin trimer behavior
is not as clear. We demonstrated that the two trimers catechin
(4R-8) catechin (4R-8) catechin and catechin (4R-8) catechin
(4R-8) epicatechin exist as four rotameric forms.15 One rotameric
form is predominant and preferentially adopts a conformation
in which stacking between different phenolic rings is favored.
Herein, we report on the colloidal behavior of these individual
molecules using both diffusion NMR and molecular modeling
in light of their particular 3D structures. The main results obtained
clearly show that the colloidal behavior of procyanidins is rather
more influenced by their degree of polymerization than by their a
(A) Self-association constant, Ka, according to an indefinite noncoopera-
overall 3D structure: monomers are the only species able to form tive model.10 (B) Critical micelle concentration, above which nanometer-
colloidal particles at the concentration observed in wine (up to sized micelles are formed. (C) Time course, ta, of micelle aggregation to
2.9 mM, in Gers’ department in southwest France16). All of the form visible (µM/mm) precipitates.
data recorded are consistent with the model described in Scheme
2 in which (i) the self-association of procyanidins occurs at low model);10,17 (ii) above a certain concentration (cmc, critical
concentration (Scheme 2A) in an indefinite noncooperative way micelle concentration, or even cac, critical aggregation concen-
to form stacks (with 2, 3, 4... molecules) where the equilibrium tration) a micellization process takes place, leading to the
constant Ka is always the same for each step (isodesmic formation of disperse, nanometer-sized colloids (Scheme 2B);
and (iii) exclusively for the monomers catechin and epicatechin,
(14) Tarascou, I.; Barathieu, K.; Simon, C.; Ducasse, M.-A.; André, Y.; Fouquet,
a time-dependent aggregation process occurs, leading to the
E.; Dufourc, E. J.; Freitas, V. d.; Laguerre, M. Magn. Reson. Chem. 2006, 44, formation of larger, visible aggregates in the form of either a
868. cloudy solution or even precipitates (Scheme 2C).
(15) Tarascou, I.; Ducasse, M.-A.; Dufourc, E. J.; Moskau, D.; Fouquet, E.;
Laguerre, M.; Pianet, I. Magn. Reson. Chem. 2007, 45, 157.
(16) Corder, R.; Mullen, W.; Khan, N. Q.; Marks, S. C.; Wood, E. G.; Carrier, (17) Davies, D. B.; Djimant, L. N.; Veselkov, A. N. J. Chem. Soc., Faraday
M. J.; Crozier, A. Nature 2006, 444, 566. Trans. 1996, 92, 383.
Colloidal BehaVior of Synthetic Condensed Tannins
Figure 1. (A) 1H NMR spectra of B4 in D2O at 298 K (3.9-4.5 ppm region, showing resonances for protons 2C, 3C, 4C, and 3F; see Scheme 1
for numbering). Concentrations are given on the right-hand side of the spectra. (B) Chemical shift variations of the 3F proton (δobs, ppm) vs B3
concentration (M) in D2O (9) or D2O/(10%)EtOD (b). Solid lines represent the best fit obtained using eq 2 with Ka ) 6.5 and 5.5 M-1, ∆δmax )
0.26 and 0.1; χ2 ) 1.6 × 10-6 and 3.1 × 10-6 for D2O and D2O/EtOD mixtures, respectively.
Materials
(+) Catechin (1, cat), epicatechin (2, epi), and epigallocatechin
gallate (3, EGCG) were purchased from Sigma-Aldrich. Procyanidins
B1 (4), B2 (5), B3 (6), B4 (7) and C2 (8) were synthesized as
previously described.13 Planar structures with appropriate numbering
are shown in Scheme 1.
Methods
Sample Preparation. Tannins were dispersed in three different
solution mixtures: (1) D2O, pH 7.0; (2) D2O/EtOD (90:10 v/v), pH
7.0; and (3) D2O/EtOD (90:10 v/v), 10 mM perdeuterated acetic
acid, 20 mM NaCl adjusted to pH 4.0.
NMR Experiments. NMR experiments were performed at 298
K on a Bruker DPX 400 equipped with a 5 mm gradient inverse
broadband probe. All proton chemical shifts are given with respect
to TMS in CDCl3 as an external reference. The self-association of
tannins (1-8, Scheme 1) was characterized by following the evolution
of both the chemical shift and diffusion coefficient with respect to
tannin concentration (between 2 to 40 mM). Single-pulse experiments
were used to record 1D spectra, and diffusion measurements were
performed using 1H NMR pulsed gradients as described by Stejskal
and Tanner.18 In the Hahn echo sequence, two pulse field gradients,
G, of the same duration and magnitude are inserted during both echo
time periods, τ. Experimental parameters: spectral width ) 10 ppm,
number of acquisitions ) 64, recycling delay ) 3 s, intergradient
delay ∆ ) 150 ms; gradient pulse duration δ ) 1.3 ms; 16 different
values of G, the gradient magnitude, varying between 0.01 and 0.4
G/m.
NMR Data Analysis. The diffusion coefficient, D, may be
obtained from fitting the line area, I, of a given resonance as a
function of G using the equation
δ
ln(I) ) -γ2G2Dδ2 ∆ - ( 3 ) (1)
where γ is the proton gyromagnetic ratio.19 Alternatively, D can be
obtained from a 2D-like treatment (2D-DOSY), where the second
dimension is obtained from the classical Fourier transformation of
the spectral series obtained with different G values and the first
dimension is obtained from a Laplace transformation leading to a
diffusionlike dimension from which D values are directly obtained.
Diffusion coefficients obtained for different concentrations and
calibrated with respect to the diffusion coefficient of water at 298
K (2.23 × 10-9 m2s-1)18 were used to obtain the cmc value. As seen
(18) Stejskal, E. O.; Tanner, J. E. J. Chem. Phys. 1965, 42, 288.
(19) Price, W. S. Concepts Magn. Reson. 1997, 9, 299.
Langmuir, Vol. 24, No. 19, 2008 11029
in Figure 2A, plotting D against the inverse concentration leads to
two straight lines with different slopes whose intercept gives the
cmc. The self-association constant, Ka, of tannins is obtained by
following the evolution of either the chemical shift or the diffusion
coefficient versus concentration. The observed chemical shift (or
diffusion coefficient) changes were least-squares fitted to the
following equation10
Aobs)Afree- [(|Amicelle- Afree|)KaC0{2/[1 + ({KaC0+1)1/2]}2]
(2)
where Aobs is the observed proton chemical shift (or diffusion
coefficient), Afree is the proton chemical shift (or diffusion coefficient)
of the nonassociated (free) tannin, Amicelle is the proton chemical
shift (or diffusion coefficient) of the tannin embedded in a micelle,
and C0 is the total tannin concentration. This equation can be used
to determine Ka from D values assuming that the T2 relaxation is
negligible during the intergradient delay duration ∆. From the Dmicelle
value obtained, the micelle hydrodynamic radius rH can be calculated
using the Stokes-Einstein equation
kBT
D) (3)
6πηrH
with η ) 1.095 cP, the viscosity of D2O at 298 K.20 η was taken
equal to 1.9 cP in a mixture composed of 10% EtOD in D2O according
to ref 21, which accounts for the replacement of hydrogen by
deuterium (η ) 0.89 cP for H2O at 298 K and η ) 1.56 cP for a
mixture of 10% EtOH in H2O).
Dynamic Light Scattering. DLS experiments were performed
on an ALV/CGS-3 goniometer coupled with ALV/LSE-5003
electronics. Hydrodynamic radii were calculated using ALV-
correlator software version 3.0 fitting routines. Hardware and software
are both from ALV-GmbH (Langen, Germany). All DLS experiments
were performed at room temperature (∼296 K), and samples used
were the same as those used for NMR except that they were diluted
100-fold to be in the elastic collision range.
Molecular Modeling. Tannins structures were obtained from
previous publications14,15 on an SGI Octane platform running
Macromodel version 6.5 (Schrödinger Inc.). The minimized con-
formations were exported in PDB format to Discover software
(Accelrys Inc.) and were then included in a water box and subjected,
after full minimization, to molecular dynamics simulations at 300
(20) Cooper, J. R.; Dooley, R. B. http://www.iapws.org, 2007.
(21) Nagasawa, Y.; Nakagawa, Y. N., A.; Okada, T.; Miyasaka, H. J. Mol.
Struct. 2005, 735, 217.
11030 Langmuir, Vol. 24, No. 19, 2008
Figure 2. (A) Determination of cmc’s from diffusion coefficients. The intercept of slopes obtained for concentrated (colloidal state) and diluted (soluble
state) regions of D versus inverse tannin concentration gives the cmc (shown as arrows). Examples shown: monomer catechin (9), dimer B1 (b),
and trimer C2 (2) in water. Sketches as insets depict colloidal versus soluble states. (B) Variation of D vs B3 concentration (M) in D2O (b) or
D2O/(10%) EtOD (9). Solid lines represent the best fit obtained using eq 2 with Ka ) 10 and 12 M-1, Dmicelle ) 2.0 and 1.3 10-10 m2 s-1 and C2
) 0.059 and 0.00536 for D2O and D2O/EtOD mixtures, respectively.
K and constant volume. The simulation was performed on an Origin
200 quadriprocessor SGI server. For the dynamics calculations, a
time step of 1.5 fs was used as well as a group-based cutoff of 13
Å that can be considered to be an atom-based cutoff of 20 Å because
of the very large groups. Calculations were first performed for isolated
tannins. Boxes of (30 × 30 × 30), (32 × 32 × 32), and (35 × 35
× 35) Å3 size were used for monomers (catechin and epicatechin),
dimers (B1 to B4), and the trimer (C2), respectively. In all cases,
the simulation was run for 1 ns, with the last 500 ps being used for
calculations. Micelle formation was monitored by placing several
tannins in simulation boxes. Sixteen catechin (epicatechin) molecules
were randomly located within a (55 × 55 × 55) Å3 box containing
5492(5218) water molecules for a total of 17 020(16 214) atoms
(concentrations of 0.16 and 0.17 M · L-1, respectively). Eight B3
procyanidins (7 compact and 1 extended conformer as observed in
water 14) were located within a (55 × 55 × 55) Å3 box containing
5492 water molecules for a total of 17 020 atoms (concentration
0.08 M · L-1). Twelve B2 procyanidins (5 compact and 7 extended
conformers as observed in water14) were located within a (63 × 63
× 63) Å3 box containing 8291 water molecules for a total of 25 689
atoms (concentration 0.08 M · L-1). Molecular dynamic simulations
were performed for 3 ns.
D, Radii of Gyration, and Hydration Parameters from
Molecular Modeling. Self-diffusion coefficients are obtained within
the DeCipher module of the Insight II computing package (Accelerys
Inc.) using the Einstein equation: one-sixth of the slope of the mean
square displacement versus time.22 This equation is derived from
isotropic motions and is valid only if the correlation time for a given
diffusive motion is smaller than the time frame of the entire molecular
dynamics calculation. To fulfill these conditions, at least a 1 ns
molecular dynamics simulation was performed, and the last 500 ps
was used for calculations. The average radius of gyration (RG) can
also be obtained from these simulations using the DeCipher module.
It represents the root-mean-square average distance of all atoms in
the particle from the center of mass. The hydration spheres and
number of associated water molecules can be calculated using the
DeCipher module by evaluating and integrating the radial distribution
function by atom pairing between water molecules and (i) each
oxygen atom of the considered tannin (first hydration sphere) and
(ii) between water molecules and the entire tannin or micelle moiety
(22) Allen, M. P.; Tildesley, D. J. Computer Simulation of Liquids; Clarendon
Press: Oxford, England, 1989.
Pianet et al.
(cumulative number of water molecules within a sphere of known
size, as otherwise obtained from experimental RH). As usual, the last
500 ps of each simulation was used and averaged for this purpose.
For the pair distribution function, the maximum radius considered
was 25 Å for one monomer and 50 Å for micelles, and the interval
was 0.1 Å.
Results
1H
NMR Chemical Shift Variation. 1D-NMR experiments
were performed as a function of tannin concentration both in
water and in hydroalcoholic media representative of wine (ca.
10% alcohol). Concentrations ranged from 0-50 mM as regularly
found in wine media (i.e., a few g · L-1). As shown in Figure 1A,
in the case of the B4 dimer, proton chemical shifts from the
heterocyclic ring (H2, H3, and H4) are very responsive to
concentration. Figure 1B shows the concentration dependence
of 3F protons of procyanidin B3 both in water and in
hydroalcoholic media. Fitting the experimental chemical shift
variations using eq 2 (Aobs) δobs) provides a measurement of Ka
with high accuracy (Figure 1B, solid lines), indicating that the
isodesmic model that is used describes the autoassociation process
well. From such chemical shift variations and data analysis, Ka
values were obtained for at least three monomers, four dimers,
and one trimer of procyanidins under different experimental
conditions (water alone, water-ethanol and water-ethanol
buffered at pH 4 with 10 mM acetic acid, 20 mM NaCl, Table
1). Association constants for monomers are found to be close to
40 M-1; nonetheless, it must be mentioned that Ka values obtained
for catechin or epicatechin in water should be taken with care
because of the solubility issue. Indeed, it was very difficult to
obtain a clear, stable solution of catechin or epicatechin at a
concentration of up to 35 mM (>10 g · L-1). However, the results
are in accordance with the literature (Ka ) 27 ( 7 M-1 for
epicatechin in water23). Association constants for all dimers and
the trimer fall in the 5-7 M-1 range, apparently independent,
within experimental error, of the presence of ethanol, pH, and
ionic strength. It is noteworthy that Ka values decrease by 8- to
10-fold on comparison with monomers.
(23) Charlton, A. J.; Baxter, N. J.; Lilley, T. H.; Haslam, E.; McDonald, C. J.;
Williamson, M. P. FEBS Lett. 1996, 382, 289.
Colloidal BehaVior of Synthetic Condensed Tannins Langmuir, Vol. 24, No. 19, 2008 11031

Table 1. Association Constants, Ka, Measured at 298 K on Table 3. cmc Values Measured at 298 K for Different
Monomers, Dimers, and a Trimer Procyanidin Monomers, Dimers, and a Trimer in Water or in a
Hydroalcoholic Solution (90:10 D2O/EtOD)
D2O D2O/EtOD (90:10)
D2O D2O/EtOD
Cat 53 ( 10 45 ( 5
Epi 39 ( 8 37 ( 2 tannin/media mM g.L-1 mM g · L-1
EGCG 44 ( 5(44)a 39 ( 8(39)
Epi <5 1.5 <5 1.5
B1 7 ( 2(8) 7 ( 2(7)
Cat 9(2 2.6 ( 0.6 16 ( 1 5.5 ( 0.3
B2 7 ( 1(12) 6 ( 1(10)
EGCG 6(2 2.7 ( 0.9 10 ( 2 4.5 ( 0.9
B2 buffer 5 ( 2b
B1 17 ( 3 10.0 ((1.5 27 ( 5 16 ( 3
B3 6.5 ( 1(10) 6 ( 1(12)
B2 18 ( 3 10.5(( 1.5 28 ( 8 16 ( 4
B4 7 ( 2(10) 6 ( 1(9)
17 ( 3 10 ( 1.5a
C2 5 ( 1(11) 5.5 ( 1(10)
B3 24(( 4 14 ( 2 19 ( 3 11.0 ( 1.5
a
Ka (M-1) values are obtained from fitting the concentration dependence B4 17 ( 4 10 ( 2 26 ( 5 15 ( 3
of chemical shifts (protons H2, H3, and H4; See Tables S1 and S2) or C2 15 ( 2 13.0 ((1.5 26 ( 6 22 ( 5
diffusion coefficients (accuracy is estimated to approximately 15%; see Tables a
S3 and S4) using eq 2. All proton spectra were recorded at 298 K, pH 7. Measured in the presence of 10 mM perdeuterated acetic acid, pH 4,
b
Obtained with 10 mM perdeuterated acetic acid, pH 4, 20 mM NaCl. 20 mM NaCl.

Table 2. Diffusion Coefficients (10-6 cm2 · s-1) of Tannins in ethanol appear to decrease the solubility to the level detected in
Aggregated and Free States at 25 °C in Water and Water/ water.
Ethanol Mediaa As described in the Materials and Methods sections, fitting the
Dfree Dmicelle diffusion coefficient variation using eq 2 (and, in this case, Aobs
) Dobs), leads to the determination of Ka and Dmicelle (Figure 2B).
tannins D2O H2Ocalc D2O/EtOD D2O H2Ocalc D2O/EtOD
The association constant can be compared to Ka values obtained
Cat 3.2 3.7 2.7 nm 1.2, 15 Catd nm from the chemical shifts variation fitting (Figure 1B). Values (in
Epi 3.4 3.1 2.7 nm nm
EGCG 3.0 2.7 2.0 1.8 0.8 parenthesis) are reported in Table 1 and are of the same order
B1 2.5 2.4 1.9 0.7 1.0 of magnitude within experimental error (close to 15%). Second,
B2 2.6 2.6/2.7b 1.9 (1.7c) 0.9 0.9 (0.7c) it gives the diffusion coefficient value of the tannin in its free
B3 2.7 2.5 2.1 0.9 0.94, 5 B3d 0.9
B4 2.8 3.3 2.0 0.9 0.85
form (Dfree) as well as the smallest value that tannins can reach
C2 2.3 2.6 1.6 0.4 0.4 (Dmicelle in Table 2). This latter value, which decreases by a
a
Values are obtained from fitting, using eq 2, the concentration dependence factor of 2-6, reflects the diffusion coefficient of the tannin
of diffusion coefficients as measured from NMR. The H2Ocalc columns report when embedded in the micelle. From these values, we can estimate
on diffusion coefficients estimated from molecular dynamics simulations the corresponding hydrodynamic radius, RH, obtained using the
using the DeCipher module of Insight II software.(See the text.) b Calculated
for the compact/extended form. c Measured in the presence of 10 mM Stokes-Einstein relationship3 (Table 4). RH in the free forms
perdeuterated acetic acid, pH 4, 20 mM NaCl. d Indicates the number of vary from 6 Å for monomers to ca. 8 Å for dimers and to ca.
molecules in the micelle. nm: not measurable. Accuracy is estimated to 9 Å for the C2 trimer. In the micelle form, RH values that had
approximately 15%.
not been measured for monomers owing to their insolubility or
quick precipitation reach ca. 25 Å for the dimers and 50 Å for
Diffusion Coefficients Measured from 1H NMR. As for the trimer. The addition of 10% ethanol involves a 20-30%
chemical shifts, diffusion coefficients, measured as described in decrease in Dfree (Table 2) that is not associated with an increase
the Materials and Methods sections, vary with increasing in RH (Table 4). On the contrary, a general decrease in RH is
concentration. Figure 2A depicts the evolution of D as a function observed in the presence of EtOH for the free forms of monomers
of inverse tannins concentration for a selection of monomer and dimers and the trimer (ca. 4, 6, and 7 Å, respectively). This
(catechin), dimer (B1), and trimer (C2) in water. Values of (3.2 is due to a large difference in viscosity between a solution of
( 0.2) × 10-10, (2.5 ( 0.2) × 10-10, and (2.2 ( 0.2) × 10-10 pure deuterated water (η ) 1.095 cP at 298 K) and a solution
m2 s-1 are found at low concentration. They correspond to tannin containing 10% ethanol (η ) 1.9 cP at 298 K). Concerning
values in their free form (Dfree) in solution and thus reflect the tannins in their micellar form, it appears that Dmicelle does not
size increase. At high concentration (ca. 50 mM), D respectively vary within experimental error when 10% ethanol is added to
decreases to (2.8 ( 0.1) × 10-10, (2.0 ( 0.1) × 10-10, and (1.8 the water solution (Table 2). RH, when measurable, decreases by
( 0.1) × 10-10 m2 s-1. It is interesting that the variation of D half (ca. 13 and 25 Å for dimers and the trimer, respectively,
versus inverse concentration undergoes a sudden decrease, Table 4). The size of micelles was also investigated by DLS.
suggesting that a micellization process takes place at this particular Experiments were performed on a solution of B1 and B2 (in an
concentration, the critical micelle concentration (cmc). An acetic acid solution buffered at pH 4, NaCl 10 mM, H2O/EtOH
accurate measure of the cmc can be obtained from the intercept 90:10) at different concentrations: 15.7 and 29.6 mM for B2 and
of the two linear variations obtained at low and high concentration 51 and 113 mM for B1. Results indicated that sizes were below
(arrows in Figure 2A). Values measured for different procyanidins the detection limit of the apparatus, which is RH < 50 Å. Although
and in different media are reported in Table 3. In D2O, they the DLS experiments could not give an accurate size, they indicate
range from 2 to 3 g · L-1 for Cat and EGCG to 10 to 14 g · L-1 that the formed micelles are smaller than 5 nm, a result consistent
for dimers and the trimer. Values could not be measured for Epi with NMR measurements. It is noteworthy that larger micelles
because it readily precipitates at high concentration. A value had not been observed using DLS.
below 1.5 g · L-1 could be estimated only for the cmc. It is very Kinetics of Tannin-Tannin Association and NMR Visibility.
interesting that in 10% alcohol the cmc values increase to 5 Using an external reference (1.17 mM cyclohexane in an external
g · L-1 for monomers, ca. 14 g · L-1 for dimers, and 22 g · L-1 for capillary), it is possible to determine the concentration of tannins
the C2 trimer. This indicates that the presence of ethanol increases observed by NMR: NMR lines are integrated, and the area
the procyanidin oligomers solubility (i.e., increases the critical decrease versus time will reflect the presence of large aggregates
concentration above which micelles are detected). It is also (>µM) that can no longer be detected under conditions of solution
noteworthy that acidic media and high ionic strength with 10% NMR. A clear solution of every procyanidin dimer or trimer is
11032 Langmuir, Vol. 24, No. 19, 2008 Pianet et al.

Table 4. Hydrodynamic Radius (RH), Radius of Gyration (RG), and Water Molecules in the First Hydration Shell
free micelles
a b c
RH Å RG Å n H2O/tannin RH Å RG Å
tannin D2O D2O/Ed H2O H2Oemeasured H2Of first shell H2O/Egmeasured D2O D2O/Ed H2O
6.4 (6 Cat)i
6.9 (7 Cat)
Cat 6.2 4.3 3.95 10 18 2.4 nm nm
7.1 (8 Cat)
8.6 (15 Cat)
6.4 (6 Epi)
Epi 5.9 4.4 3.91 9 14 3 nm nm
7.3 (9 Epi)
EGCG 6.7 5.7 4.53 14 25.6 8 11.0 14.0
B1 8.0 6.0 4.5 22 24 8 28.5 11.5
B2 7.7 6.0 4.9 (5.0)h 20 (22.3)h 27 (30)h 8.3 (9.3)h 22.1 12.8 7.6 (3 B2)
6.2 (3 B3)
B3 7.4 5.5 4.6 17.3 28 6 22.1 12.8
8.5 (5 B3)
B4 7.7 5.7 5.0 20.7 27 7.3 24.9 13.5
C2 8.7 7.2 5.2 32.3 45 16.3 49.8 23
a
Calculated from Dfree as measured from NMR using eq 3 and correcting for the isotopic D2O/H2O difference; see the text. b Calculated using the DeCipher
module; see the text. c Number of water molecules per tannin molecule. d D2O containing 10% deuterated ethanol. e Obtained from the pairwise distribution
function of water and experimental RH; see the text. f Calculated in the first hydration sphere using the DeCipher module (see the text) by calculating the
water molecules present in the first hydration shell of each oxygen atoms. g As in footnote e with the experimental RH obtained in D2O/ethanol. h Values
are given for the two rotamers: compact (extended). i Number of molecules in the micelle, estimated because calculation is clearly not completed “nm” indicates
not measured.

is more easily dissolved in water or a winelike medium at high

Figure 3. Aggregation kinetic of tannins: (() 42 mM catechin in D2O,
(9) 54 mM catechin in 90:10 D2O/EtOD, (b) 31 mM epigallocatechin
gallate in D2O;(2) 29 mM epicatechin in D2O, and (×) 35 mM dimer
B4 in a solution mixture composed of 90:10 D2O/EtOD in a perdeuterated
acetate buffer (10 mM, pH 4.4). The experimental error is estimated to
be ca. 10%.
readily obtained whatever the concentration (even above the
cmc) and the solvent used: the mixture is stable during the
experiment (up to 3 days) as illustrated by the unchanged area
versus time of protons resonances of B4 dissolved in a winelike
medium in (Figure 3, ×). On the contrary, it is necessary to heat
the solution of procyanidin monomers for a few minutes to obtain
a clear solution at a concentration of 10 g · L-1. Then, the solutions
evolved quite differently depending on the monomer and the
solvent used. When dispersed at high concentration in water, the
catechin mixture takes a fleecy aspect as previously described,24
and this phenomenon is directly related to a decrease in the
NMR signal ((, Figure 3). When dispersed in a solution containing
at least 10% ethanol, the solution (54 mM, 15 g · L-1) remains
clear for at least 3 days, and the total amount of catechin put into
the sample is detected (9, Figure 3). It was not possible to dissolve
more than 29 mM (8.5 g · L-1) epicatechin in water. Following
the behavior of this solution saturated in epicatechin for 3 days,
part of the epicatechin NMR signal disappears (around 20%)
when a precipitate at the bottom of the tube is formed. EGCG
(24) Mirabel, M.; Glories, Y.; Pianet, I.; Dufourc, E. J. Enologie 99. 6e
Symposium International d’Oenologie; Technique & Documentation; Paris, 1999;
p 519.
concentration (slight heating is sufficient), after which the solution
remains clear during the 3 days experiment and the total amount
of EGCG is detected by NMR, within experimental error.
Molecular Mechanics of Isolated Tannins. Molecular
dynamics measurements were performed for 1 ns on tannins
placed in a simulated box full of water molecules, as described
in the Materials and Methods sections. Using the last 500 ps, one
could calculate the self-diffusion coefficient, Dfree (Table 2).
Although they have been obtained with a totally different method,
the values are close to those measured with NMR within the
15% experimental error. The radius of gyration, RG, can also be
obtained from the simulation (Table 4). RG values are of interest
because they reflect the mean volume of the tannin alone and,
when compared to RH, can lead to the degree of hydration of all
tannins (see the Materials and Methods sections), expressed by
the number of water molecules per tannin molecule (nH2O/tannin).
In pure water, this degree of hydration corresponds to ap-
proximately 10 water molecules for a monomer, around 20 for
dimers, and close to 30 for a trimer. This represents a large
increase in the degree of hydration whereas the volume of the
“dry” tannin skeleton barely doubles on going from monomers
to trimer (RG increases from 3.95 to 5.2 Å). This indicates that
trimers are able to hydrate more easily than dimers, which in turn
are more hydrated than monomers. Simulations carried out for
the free forms of the monomers, dimers and trimer also allow
an accurate description of hydration at the atomic level. Figure
4 shows that only oxygen atoms are hydrated, with the volume
of the red CPK sphere being proportional to the number of water
molecules found in the first hydration sphere (see Materials and
Methods sections, Table 4 for the first shell of H2O, and Table
S5 for details in Supporting Information). This Figure confirms
that trimers are more hydrated than dimers and monomers.
However, differences appear within a tannin class. For example,
both monomers Cat and Epi present quite similar hydration
profiles (Table S5 in Supporting Information) with well-
distributed hydration on the 3′, 4′, 5, and 7 OH groups (Figure
4). However, their respective level of hydration (first shell) is
different: 18 water molecules for catechin versus 14 for
epicatechin. This demonstrates that catechin can be slightly more
hydrated than epicatechin (Table 4). The four dimers investigated
have about the same number of water molecules bound in the
Colloidal BehaVior of Synthetic Condensed Tannins
Figure 4. Relative hydration of oxygen atoms for monomer Cat, dimer
B3, and trimer C2. Oxygens are represent as red CPK spheres with a
diameter proportional to the number of water molecules present in the
first hydration sphere. The scale (2 Å) is the same for all frames.
first hydration zone (26 ( 2), which is comparable to the total
number of water molecules bound that ranges from 17.3 for B3
to 22 for B1, showing marked differences in hydration capability
among dimers. The C2 trimer also shows large differences
between the number of water molecules bound in the first shell
(45) and the total number of bound molecules (32.3). Water
molecules show a large tendency to group in some part of the
structure, leaving the remaining parts almost completely dry.
This appears in Figure 4 for the Epi monomer, but this
phenomenon is obviously present for the B3 dimer, where entire
parts of the structure are almost completely dry. Even if it is hard
to detect this tendency for the C2 trimer because the global
pattern is much more intricate as a result of the more complex
3D structure of the molecule, large parts of the structure are
completely dry, strongly contrasting with largely hydrated
surfaces. It is noteworthy that the presence of 10% ethanol induces
a dramatic decrease in the level of hydration of every tannin.
Molecular Mechanics of Tannins above the cmc. Calcula-
tions were performed above the cmc for catechin, epicatechin,
and dimers B2 and B3. To achieve calculations on reasonable
time scales, runs were performed on only select tannins. Initially,
molecules are placed far from each other in the water box (Figure
Langmuir, Vol. 24, No. 19, 2008 11033
Figure 5. (Bottom, A) View of a micelle composed of 15 catechins
obtained after 2 dynamic simulations of 1 ns each. (B) View of a micelle
composed of 5 B3’s obtained after a 2 ns simulation. The starting
conditions are shown in the upper panels. Molecules are in stick
representations with carbon in green and oxygen in red. Hydrogen atoms
as well as H bonds have been omitted for clarity.
5), and molecular dynamics simulations are run for at least 1 ns.
At the end of the first calculation, two micelles of seven and
eight catechins were formed (Figure 5 A), leaving only one
molecule of catechin alone. When put together in a new, smaller
box (60 × 40 × 40) Å3 to diminish the calculation time and
submitted to another 1 ns simulation process, these two micelles
are able to coalesce and give a larger micelle composed of 15
catechins (Figure 5A), suggesting their capacity to self-associate
very quickly, that is, on the time scale of the calculation.
Epicatechin seems to behave quite similarly because 2 micelles
of 9 and 5 epicatechins are formed at the end of the 1 ns
dynamic calculation (not shown). When 8 B3 procyanidins are
put in a box full of water with respect to their specific rotameric
ratio generally observed14 (7 B3’s in a compact form and 1 in
an extended form), after a 2 ns molecular dynamics simulation,
2 micelles are formed: one is composed of 5 B3’s and the other
is composed of the 3 remaining B3’s (Figure 5B). The 12 B2
procyanidins (7 compact and 5 extended conformers) embedded
in a box full of water and submitted to a 2.7 ns simulation process
seem to self-associate more slowly and give rise to the formation
of the smallest micelles composed of 3 B2’s.
Radii of gyration calculated on micelles ranged from 6 to 9
Å for Epi, Cat, B2, or B3 (i.e., much smaller than RH as measured
from NMR). We believe that this is due to the short time of the
calculation: to obtain the measured size of the micelles, one
would have to calculate for hundreds of nanoseconds, which was
clearly beyond the scope of the present study.
Discussion
Several findings were obtained through the current study of
the tannin self-association process. First, we demonstrated that
procyanidin monomers behave quite differently than oligomers.
Monomers are able (i) to self-associate with a 10-fold-higher
affinity than dimers or trimers and (ii) to form micelles at a lower
cmc value. At a concentration up to their cmc, monomers may
11034 Langmuir, Vol. 24, No. 19, 2008
also lead to visible hazes or precipitates at longer times. However,
if oligomers are able to self-associate, then they get organized
into the smallest micelles of nanometer size, which are not
responsible for wine turbidity. These self-association properties
have been linked to their hydration threshold, which depends
both on their degree of polymerization and internal dynamics.
All of these findings will be discussed below, together with
possible implications for biology.
Modeling the Tannin Self-Association Process. The self-
association of tannins in general and procyanidins in particular
has already been described in winelike media.2,25-27 Tannins
are supposed to form stacks, with π-π stacking being the major
driving force involved.28 This behavior should be followed using
different complementary techniques: DLS gives insight into the
size and the polydispersity of the formed aggregates but is still
a macroscopic technique. CD spectroscopy also gives a mac-
roscopic idea of the aggregation process. NMR associated with
molecular dynamics calculations gives the same kind of
information at the molecular or even atomic level. Diffusion
coefficient values can be obtained from DOSY NMR experiments,
which are commonly described as a noninvasive “chromatog-
raphy” method because each component of a mixture can in
principle be separated owing to its own diffusion coefficient.
However, DOSY NMR limitations are rapidly reached when the
mixture is composed of compounds with small differences in
diffusion coefficients and/or chemical shifts. Such a limitation
makes the use of DOSY NMR impossible to determine D values
of the procyanidins present in tannin extracts. Therefore, the
strategy we propose in this article in order to follow procyanidin
self-association is (i) to work on pure synthesized tannins, thus
allowing a look at their own behavior that leads to an interpretation
of data by taking into account both their stereochemical structures
and their degree of polymerization (up to 3) and (ii) to use both
NMR and molecular dynamics to interpret the data at an atomic
level. The fact that the diffusion coefficient values decrease with
increasing tannin concentration (Figure 2) clearly demonstrates
that tannins are able to be organized in larger molecular objects.
Such behavior is confirmed by molecular dynamics simulation
showing that a recognition process instantaneously occurs between
them. This phenomenon can be fitted using an isodesmic model
in which the association constant Ka is the same in each step of
the recognition process. It is noteworthy that the same Ka value
is obtained whatever the variations monitored (chemical shift or
diffusion coefficient). This suggests that the chemical shift changes
observed also reflect the self-association process and confirm
the hydrophobic π-π stacking nature of this association (Figure
5), with this interaction in turn being obvious in molecular
dynamics runs. Interestingly, Ka differences are clearly observed
to differentiate the behavior of monomers (∼50 M-1) from that
of the dimers/trimer (∼6 M-1), showing the higher efficiency of
monomer self-association compared to that of oligomers.
However, every tannin studied exhibits a micellization process
above a certain concentration (cmc, Figure 2A). The largest
differences are observed between monomers for which cmc values
are less than 3 g · L-1 in water and 5 g · L-1 in the presence of
ethanol and dimers (10 g · L-1 in water and 16 g · L-1 with ethanol,
except for B3) or the trimer (13 g · L-1 in water and 22 g · L-1
with ethanol). The fact that this micellization process occurs
with the same Ka in every step suggests that the formed micelles
(25) Charlton, A. J.; Baxter, N. J.; Khan, M. L.; Moir, A. J. G.; Haslam, E.;
Davies, A. P.; Williamson, M. P. J. Agric. Food Chem. 2002, 50, 1593.
(26) Monteiro, C.; Maechling, C.; Du Penhoat, C. H. Magn. Reson. Chem.
2002, 40, S110.
(27) Saucier, C.; Bourgeois, G.; Vitry, C.; Roux, D.; Glories, Y. J. Agric. Food
Chem. 1997, 45, 1045.
(28) Bax, A.; Davis, D. G. J. Magn. Reson. 1985, 63, 207.
Pianet et al.
have polydisperse sizes. Indeed, a particular evolution of Ka is
required with respect to concentration to obtain a narrow
distribution of micelles sizes: at lower concentrations, Ka must
increase up to a plateau value and then decrease at higher
concentrations. The average sizes have been measured to reach
only a few nanometers, as estimated from experimental fittings
(Figure 2B) and molecular dynamics (Figure 5).
Following further aggregation as a function of time and at a
concentration of tannins chosen to be largely above the cmc, we
do not observe large aggregates over at least 72 h for EGCG,
dimers, or the trimer whatever the absence/presence of 10%
ethanol (Figure 3). Once again, the behavior of catechin or
epicatechin monomers is quite different. Dissolving epicatechin
in water at a concentration of up to 10 g · L-1 (34 mM) is
impossible even if the solution is heated a few minutes at a
temperature higher than 330 K. Part of the epicatechin is dissolved
(<10 g · L-1), but the rest remains at the bottom of the tube. This
saturation threshold increases to around 15 g · L-1 in the presence
of ethanol. Dissolving catechin in water above a concentration
of up to 12 g · L-1 readily leads to a clear solution when the
solution is heated to around 330 K. However, large, visible
(micrometer-sized) aggregates appears rapidly, leading to the
progressive disappearance of the NMR signal (Figure 3). This
threshold increases to a concentration of ∼30 g · L-1 (ca. 100
mM) in the presence of 10% ethanol as previously described.29
The differences observed between procyanidin monomers
(mainly catechin and epicatechin with the case of EGCG, which
is clearly neither a monomer nor a dimer, being taken aside at
the moment) and oligomers should be explained by both their
3D structures and levels of hydration. Monomers, which have
a lower level of hydration than do dimers/trimers, have a relatively
planar structure favoring intermolecular stacking, affording the
possibility to form larger and larger micelles that are able to
coalesce to form hazes or even precipitates. On the contrary, the
structure of all four C4-C8 dimers, with a general U-shape
owing to stable intramolecular stacking between aromatic groups,
is disadvantageous to intermolecular stacking. In addition, there
is intrinsic dynamics associated with dimer molecules: confor-
mational exchanges have been evidenced from NMR measure-
ments.14 This is even amplified in the trimer, where there are two
C4-C8 rotameric degrees of freedom, hence hiding the
hydrophobic surfaces from water. To summarize, the higher the
degree of polymerization, the greater the number of structural
degrees of freedom that favor hydration and prevent intermo-
lecular hydrophobic stacking. This property explains their great
solubility in water as identified by Bate-Smith and Swain: “tannins
are water-soluble phenolic compounds having molecular weights
between 500 and 3000”.30 Molecular dynamics calculations
confirm this hypothesis and show that monomers micelles are
more compact than dimers micelles (Figure 5). Therefore, our
findings clearly point out the following result: only monomers
are able to form colloids or precipitates at concentrations found
in wines (ca. 1-4 g · L-1). If this result is contradictory to
preliminary work performed on fractions containing a range of
different tannins characterized by a specific average degree of
polymerization,8,9 then it is in perfect agreement with a recent
report by Zanchi et al.11
Micelle Organization. Micelle intrinsic organization is
interesting. The number of hydrogen bonds may be estimated:
10 for the micelle made of 15 catechins but only 6 for the one
(29) Mirabel, M.; Glories, Y.; Pianet, I.; Dufourc, E. J. J. Chim. Phys. Phys.-
Chim. Biol. 1999, 96, 1629.
(30) Bate-Smith, E. C.; Swain, T. Flavonoid Compounds. In: ComparatiVe
Biochemistry; Forklin, M., Mason, H. S., Eds.; Academic Press: New York, 1962;
Vol. III, pp 75-809.
Colloidal BehaVior of Synthetic Condensed Tannins
made of 5 B3. The low content of H bonds in micelles suggests
that π-π stacking, rather than hydrogen bonding, acts as the
major cohesion force leading to micelle formation. It is also
noteworthy that the volume of a free catechin molecule is 256
Å3 and that it decreases to 180 Å3 in a micelle of 15 catechins.
Hydrophobic interactions thus drastically decrease the micelle
volume. On the contrary, an increase in the apparent volume of
the B3 dimer is observed when it is embedded inside a micelle:
535 Å3 in a micelle of 5 B3 as compared to 400 Å3 when free
in solution. As shown in Figure 5A, the micelle of 15 catechins
appears largely more compact than the micelle of 5 B3.
Except for EGCG, the presence of 10% ethanol leads to a
2-fold decrease in micelle size. Owing to the amphiphilic nature
of ethanol, this result clearly strengthens the hydrophobic nature
of the cohesion forces that permit the micelles to be formed. This
observation is of interest because it indicates that tannin particles
in wine are small enough not to intervene alone in wine turbidity.
Biological Implications. Our findings contradict axioms
concerning a correlation between increasing tannin insolubility
and the degree of polymerization. It should be pointed out that
earlier work supporting the latter was carried out on tannin extracts
with a mean degree of polymerization whereas we are working
on pure, structurally well characterized molecules. Only a minor
part of the unavoidable mixture obtained with mean isolated
fractions may lead to colloidal behavior, whereas the rest of the
molecules are soluble.11 A 4 g · L-1 tannin mixture, as in wines,
Langmuir, Vol. 24, No. 19, 2008 11035
would therefore lead to free dimers and trimers in solution and
colloids or precipitates for the monomers under investigation.
(We do not know the colloidal properties of polyphenols with
a higher degree of polymerization.) Preliminary experiments with
a synthetic tetramer (to be published) indicate even greater
solubility, which is not surprising according to the increase in
the number of degrees of freedom within the molecule, as
discussed above, which hides hydrophobic parts of the molecule
and therefore reduces π-π intermolecular interactions and hence
the appearance of micelles and precipitates. It is hard to predict
whether this behavior can be extended to tannins with degrees
of polymerization of up to 10 or 20; however, our study indicates
that at least up to tetramers the molecules are available for
interaction with proteins as free molecules in solution.
Acknowledgment. This work has been supported by the
Conseil Interprofessionnel du Vin de Bordeaux (CIVB, Bordeaux,
France), the Région Aquitaine, The Centre National de la
Recherche Scientifique (CNRS, France), and the Université
Bordeaux1 (Talence, France). We also thank Nathan McClena-
ghan for helping us with the English.
Supporting Information Available: Results of chemical shifts
and diffusion and fittings and hydration values for each OH group as
obtained from molecular mechanics. This material is available free of
charge via the Internet at http://pubs.acs.org.
LA8015904