Application of a molecular sensory science approach to alkalized cocoa (Theobroma cacao): Structure determination and sensory activity of nonenzymatically C-glycosylated flavan-3-ols

Article abstract of DOI:10.1021/jf062403+

Application of comparative taste dilution analyses on nonalkalized and alkalized cocoa powder revealed the detection of a velvety, smoothly astringent tasting fraction, which was predominantly present in the alkalized sample. LC-MS/MS analysis, 1D- and 2D-NMR, and CD spectroscopy as well as model alkalization reactions led to the unequivocal identification of the velvety, smoothly astringent molecules as a series of catechin- and epicatechin-C- glycopyranosides. Besides the previously reported (-)-epicatechin-8-C-β-D- galactopyranoside, additional flavan-3-ol-C-glycosides, namely, (-)-epicatechin-8-C-β-D-glucopyranoside, (-)-catechin-8-C-β-D- glucopyranoside, (-)-catechin-6-C-β-D-glucopyranoside, (-)-epicatechin-6-C- β-D-glucopyranoside, (-)-catechin-8-C-β-D-galactopyranoside, (-)-catechin-6-C-β-D-galactopyranoside, (-)-catechin-6-C,8-C-β-D- diglucopyranoside, (-)-epicatechin-6-C,8-C-β-D-digalactopyranoside, (-)-catechin-6-C,8-C-β-D-digalactopyranoside, and epicatechin-6-C,8-C- β-D-diglucopyranoside, were identified for the first time in cocoa. Most surprisingly, these phenol glycoconjugates were demonstrated by model experiments to be formed via a novel nonenzymatic C-glycosylation of flavan-3-ols. Using the recently developed half-tongue test, human recognition thresholds for the astringent and mouth-drying oral sensation were determined to be between 1.1 and 99.5 μmol/L (water) depending on the sugar and the intramolecular binding position as well as the aglycone.

Full text of DOI:10.1021/jf062403+

9510 J. Agric. Food Chem. 2006, 54, 9510−9521
Application of a Molecular Sensory Science Approach to
Alkalized Cocoa (Theobroma cacao): Structure Determination
and Sensory Activity of Nonenzymatically C-Glycosylated
Flavan-3-ols
TIMO STARK AND THOMAS HOFMANN*
Institut fu¨r Lebensmittelchemie, Universita¨t Mu¨nster, Corrensstrasse 45, D-48149 Mu¨nster, Germany
Application of comparative taste dilution analyses on nonalkalized and alkalized cocoa powder revealed
the detection of a velvety, smoothly astringent tasting fraction, which was predominantly present in
the alkalized sample. LC-MS/MS analysis, 1D- and 2D-NMR, and CD spectroscopy as well as model
alkalization reactions led to the unequivocal identification of the velvety, smoothly astringent molecules
as a series of catechin- and epicatechin-C-glycopyranosides. Besides the previously reported (-)-
epicatechin-8-C-â-D-galactopyranoside, additional flavan-3-ol-C-glycosides, namely, (-)-epicatechin-
8-C-â-D-glucopyranoside, (-)-catechin-8-C-â-D-glucopyranoside, (-)-catechin-6-C-â-D-glucopyrano-
side, (-)-epicatechin-6-C-â-D-glucopyranoside, (-)-catechin-8-C-â-D-galactopyranoside, (-)-catechin-
6-C-â-D-galactopyranoside, (-)-catechin-6-C,8-C-â-D-diglucopyranoside, (-)-epicatechin-6-C,8-C-â-
D-digalactopyranoside, (-)-catechin-6-C,8-C-â-D-digalactopyranoside, and epicatechin-6-C,8-C-â-D-
diglucopyranoside, were identified for the first time in cocoa. Most surprisingly, these phenol
glycoconjugates were demonstrated by model experiments to be formed via a novel nonenzymatic
C-glycosylation of flavan-3-ols. Using the recently developed half-tongue test, human recognition
thresholds for the astringent and mouth-drying oral sensation were determined to be between 1.1
and 99.5 µmol/L (water) depending on the sugar and the intramolecular binding position as well as
the aglycone.
KEYWORDS: Cocoa; C-glycosylation; alkalization; astringency; taste dilution analysis; half-tongue test;
flavan-3-ol-C-glycosides
INTRODUCTION
To bridge the gap between pure structural chemistry and
human taste perception, taste reconstitution and omission
experiments were performed on the basis of quantitative data
demonstrating the bitter-tasting alkaloids theobromine and
caffeine, seven bitter-tasting diketopiperazines, seven bitter- and
astringent-tasting flavan-3-ols, six puckering astringent N-
phenylpropenoyl-L-amino acids, four velvety astringent flavonol
glycosides, γ-aminobutyric acid, â-aminoisobutyric acid, and
six organic acids as the key organoleptics of the roasted cocoa
nibs (3, 5). Sensory analyses of a cocktail of these taste
compounds, each in “natural” concentration, revealed that the
taste profile of this artificial mixture was very close to the taste
profile of an aqueous suspension of roasted cocoa nibs (5).
Due to its pleasant aroma as well as its typical bitter and
astringent taste, altogether imparting rich mouthfeel, complexity,
and palatability, the fermented and roasted seeds of the cocoa
tree, Theobroma cacao, are widely enjoyed by consumers as
the desirable tasty ingredient in cocoa-based products.
Very recently, application of taste dilution analysis (TDA)
on roasted cocoa nibs (T. cacao) revealed that, besides the
alkaloids theobromine and caffeine, a series of bitter-tasting 2,5-
diketopiperazines and monomeric and oligomeric flavan-3-ols
were the key inducers of the bitter taste imparted upon
consumption of roasted cocoa (1-3). In addition, the flavan-
3-ols were found to contribute to the astringency of cocoa nibs.
Moreover, a number of velvety mouth-coating O-â-D-glycopy-
ranosides of quercetin, naringenin, luteolin, and apigenin as well
as a series of puckering astringent N-phenylpropenoyl-L-amino
acids have been identified among the key taste compounds of
roasted cocoa (1-4).
For the manufacturing of cocoa powder used as the basis for
cocoa-containing beverages, the roasted cocoa nibs are subjected
to an alkalization process involving the use of potassium
hydroxide or magnesium oxide. This alkalization process is
needed to mellow the flavor, to partially neutralize free organic
acids, to enhance the color of the product, and to improve
dispersability and lengthen suspension-holding ability of cocoa,
thus preventing rapid sedimentation of the cocoa drink. In a
typical process, the cocoa nibs are treated with a dilute 2.0-
* Author to whom correspondence should be addressed (telephone +49-
251-83-33-391; fax +49-251-83-33-396; e-mail thomas.hofmann@uni-
muenster.de).
10.1021/jf062403+ CCC: $33.50
© 2006 American Chemical Society
Published on Web 11/14/2006

Structure Determination of Flavan-3-ol-C-hexosides in Cocoa
J. Agric. Food Chem., Vol. 54, No. 25, 2006 9511
2.5% alkali solution at an ambient temperature of 75-100 °C,
then neutralized, and dried to a moisture content of about 2%
in a vacuum dryer or by further kneading of the mass at a
temperature above 100 °C. Finally, the alkalized cocoa material
is ground using fine roller mills. Although the taste of alkalized
cocoa is well-known to be less astringent and milder in bitterness
compared to the sensory profile of a nonalkalized cocoa, there
are no data available on the differences in the key organoleptics
in both products.
Aimed at defining the influence of the alkalization process
on the key taste compounds in cocoa powder, the objectives of
the present investigation were, therefore, to screen the taste
compounds in nonalkalized as well as alkalized cocoa powder
by means of a “molecular sensory science” approach: after
screening for the cocoa components inducing the most intense
human taste response using a comparative TDA, the target
compounds differing in both types of cocoa were isolated and
their chemical structures determined. To evaluate their taste
impact, the human oral recognition threshold concentrations of
the sensory active compounds were determined.
Table 1. Yield and Sensorial Evaluation of Fractions I−V Isolated from
Nonalkalized (NC) and Alkalized Cocoa (AC) Powder
intensitya perceived for
sampleb
NC
fraction I
fraction II
fraction III
fraction IV
fraction V
yieldc (g/100 g)
bitterness
astringency
sourness
4.7
0
3.0
0
0
1.0
2.5
<0.5
2.1
0
2.2
0
0
0.5
1.5
<0.5
1.1
0
13.6
2.0
1.6
16.7
66.1
3.0
1.8
4.0
<0.5
3.1
0
2.0
1.1
2.5
<0.5
AC
fraction I
fraction II
fraction III
fraction IV
fraction V
15.7
1.5
0.9
14.2
67.7
0
0
0.8
2.0
<0.5
0.4
1.0
<0.5
a The taste intensity of aqueous mixtures of cocoa powder (10 g) or the individual
fractions isolated from cocoa powder (10 g) in bottled water (100 mL; pH 6.0) was
evaluated on a scale from 0 (not detectable) to 5.0 (strongly detectable). b Fractions
I−V are the pentane solubles (I), dichloromethane extractables (II), ethyl acetate
extractables (III), water solubles (IV), and nonsoluble residue (V) isolated from the
corresponding cocoa powder. c Yields were determined by weight.
MATERIALS AND METHODS
Chemicals. The following compounds were obtained commer-
cially: (-)-epicatechin, (+)-catechin (Sigma, Steinheim, Germany);
potassium carbonate, D-glucose, D-galactose (Merck, Darmstadt, Ger-
many); apigenin-8-C-â-D-glucopyranoside, apigenin-6-C-â-D-glucopy-
ranoside (Roth, Karlsruhe, Germany). Solvents were of HPLC grade
(Merck). Deuterated solvents were obtained from Euriso-Top (Gif-Sur-
Yvette, France). Bottled water (Evian) adjusted to pH 6.0 with aqueous
formic acid (1%) was used for sensory evaluation. Alkalized cocoa
(AC) and nonalkalized cocoa (NC) prepared from the same blend of
cocoa were obtained from the German food industry. For the alkal-
ization process, cocoa nibs were mixed with an aqueous solution of
potassium carbonate, heated for 1 h at 80 °C, then dried for 24 h at 40
°C, and, finally, roasted.
Sensory Analyses. Training of the Sensory Panel. Twelve subjects
(seven women and five men, 25-38 years of age), who gave informed
consent to participate in the sensory tests of the present investigation
and have no history of known taste disorders, were trained to evaluate
the taste of aqueous solutions (3 mL each) of the following standard
taste compounds by using a triangle test as described in the literature
(6): sucrose (12.5 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) for umami taste.
For the puckering astringency and the velvety astringent, mouth-drying
oral sensation, the panel was trained by using gallotannic acid (0.05%)
and quercetin-3-O-â-D-glucopyranoside (0.002 mmol/L), respectively,
using the half-tongue test (7). Sensory analyses were performed in a
sensory panel room at 22-25 °C in three different sessions.
Pretreatment of Fractions. Prior to sensory analysis, the fractions
or compounds isolated were suspended in water, and, after removal of
the volatiles in high vacuum (<5 mPa), were freeze-dried twice. GC-
MS and ion chromatographic analyses revealed that food fractions
treated by that procedure are essentially free of the solvents and buffer
compounds used.
Half-Tongue Test. To overcome carry-over effects of astringent
compounds, threshold concentrations of astringent compounds were
determined in bottled water by means of the recently developed half-
tongue test (7). Serial 1:1 dilutions of the samples were presented in
order of increasing concentrations to a trained panel of 12 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 pipet 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 toward 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 induced by the tastant, 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 proof for the
correctness of the data. The geometric mean of the last and the second
to last concentrations was calculated and taken as the individual
recognition threshold. Values between individuals and three separate
sessions differed by not more than plus or minus one dilution step;
that is, a threshold value of 3.1 µmol/L for (-)-epicatecin-8-C-â-D-
glucopyranoside represents a range from 1.55 to 6.2 µmol/L.
Sequential Solvent Extraction of Cocoa Samples. Nonalkalized
roasted cocoa powder (100 g) and alkalized roasted cocoa powder (100
g), respectively, were extracted with n-pentane (5 × 300 mL) at room
temperature for 30 min. After centrifugation, the organic layers were
combined and freed from solvent in vacuum to give the n-pentane
solubles (fraction I). The residual cocoa material was then extracted
five times with a mixture (7:3, v/v; 300 mL each) of acetone and water
for 45 min at room temperature with stirring. The acetone/water extract
was freed from organic solvent under reduced pressure at 30 °C, the
aqueous solution obtained was extracted with dichloromethane (4 ×
200 mL), and the combined organic layers were freed from solvent in
a vacuum to give the dichloromethane extractables (fraction II). The
remaining aqueous layer was adjusted to pH 5.0 with aqueous
hydrochloric acid (1 mol/L) and extracted with ethyl acetate (5 × 200
mL), the organic phase was freed from solvent in a vacuum to give
the ethyl acetate extractables (fraction III), and the aqueous phase was
freeze-dried to give the water solubles (fraction IV). In addition, the
insoluble residue of the powdered cocoa was freeze-dried to give
fraction V. The individual fractions were freeze-dried two times to
remove trace amounts of solvents, their yields were determined by
weight, and their taste profiles were evaluated in aqueous solutions as
given in Table 1.
Gel Permeation Chromatography (GPC). Acetone/water extracts
(fractions III/IV; 1 g each) obtained from alkalized (AC) and nonal-
kalized cocoa (NC) powder, respectively, were dissolved in a mixture
(50:50, v/v; pH 3.5; 10 mL) of methanol and water, membrane filtered
(45 µm), and then applied onto the top of a water-cooled 100 × 5 cm
glass column XK50 (Amersham Pharmacia Biotech, Uppsala, Sweden)
filled with a slurry of Sephadex LH 20 material (Amersham Pharmacia
Biotech) in the same solvent mixture. Using a flow rate of 2.3 mL/

9512 J. Agric. Food Chem., Vol. 54, No. 25, 2006
Stark and Hofmann
Figure 1. GPC/TDA chromatograms of the solvent fraction III/IV isolated from (A) nonalkalized and (B) alkalized cocoa powder.
-
Figure 2. LC-MS/MS (ESI ) spectrum of a velvety astringent compound isolated from GPC fraction 10 of alkalized cocoa.
min, chromatography was performed with methanol/water (50:50, v/v;
pH 3.5; 0.8 L), followed by methanol/water (70:30, v/v; pH 3.5; 1.5
L), methanol/water (90:10, v/v; pH 3.5; 1 L), and, finally, methanol (2
L). Monitoring the effluent by means of a UV-1575 type UV-vis
detector (Jasco, Groâ-Umstadt, Germany) operating at 270 nm, 16
fractions were collected from the AC and NC powders, respectively,
by a fraction collector, and the individual fractions were freed from
solvent in a vacuum at 30 °C and were then freeze-dried twice. The
residue of each GPC fraction was used for the TDA as well as for
chemical analysis.
TDA. Aliquots of the GPC fractions and the HPLC fractions,
respectively, were dissolved in “natural” ratios in exactly 5.0 mL of
bottled water (pH 6.0) and, then, sequentially diluted 1+1 with bottled
water. The serial dilutions of each of these fractions were then presented
to the sensory panel in order of ascending concentrations, and each
dilution was evaluated for astringency by means of the half-mouth test.
At the start of the session and before each trial, the subject rinsed with
water and expectorated. 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 the blank as a proof for the correctness of the data. The geometric
mean of the last and the second to last concentrations was calculated
and taken as the dilution at which a sensory difference between the
diluted extract and the blank sample could just be detected. This dilution
was defined as the taste dilution (TD) factor (8). The TD factors
evaluated by four different assessors in three different sessions were
averaged. The TD factors between individuals and separate sessions
did not differ by more than one dilution step. On the basis of the TD
factors obtained, the individual GPC fractions were rated in their taste
impact (Figure 1).
HPLC Analysis of GPC Fraction 10. GPC fraction 10 isolated
from AC powder was solubilized in a mixture (30:70, v/v, 1 mL) of
methanol and formic acid (0.1%) and then analyzed by means of HPLC-
DAD as well as LC-MS. A series of compounds were detectable
exhibiting UV-vis absorption maxima at 218, 231, and 280 nm and
showing the typical mass spectrum exemplified in Figure 2.
Preparation of Flavan-3-ol-C-glycosides (1-11, Figure 3) Using
a Model Alkalization Process. Analytical Scale. (-)-Epicatechin (0.04
mmol) was mixed with D-glucose (0.04, 0.08, 0.2, 0.4, 0.8, or 4 mmol,
respectively), K2CO3 (0.21 mmol), and water (1 g). After homogeniza-
tion, the alkalization process was performed by heating at 80 °C with
stirring. After 10, 20, 40, and 60 min, the alkalization process was
stopped by adjusting the pH to 5.0 using aqueous hydrochloric acid (1
mol/L) and, after membrane filtration, aliquots (10-50 µL) of the
mixture were analyzed by means of RP-HPLC-DAD.
PreparatiVe Scale. (-)-Epicatechin (2 mmol) was mixed with
D-glucose or D-galactose (20 mmol, each), K2CO3 (10.42 mmol), and
water (25 g). After homogenization, the alkalization process was

Structure Determination of Flavan-3-ol-C-hexosides in Cocoa
J. Agric. Food Chem., Vol. 54, No. 25, 2006 9513
8.0 Hz, H-C(5′)], 6.85 [d, 1H, J ) 2.0 Hz, H-C(2′)]; 13C NMR (100
MHz, MeOD, HMQC, HMBC) δ 27.8 [C-4], 62.8 [C-6′′], 68.5 [C-3],
71.7 [C-4′′], 73.5 [C-2′′], 76.5 [C-1′′], 79.4 [C-3′′], 82.3 [C-5′′], 82.4
[C-2], 97.1 [C-6], 101.0 [C-4a], 104.2 [C-8], 114.7 [C-2′], 116.2 [C-5′],
119.5 [C-6′], 132.4 [C-1′], 146.1 [C-3′], 146.3 [C-4′], 155.1 [C-8a],
157.1 [C-7], 157.5 [C-5].
(-)-Catechin-6-C-â-D,8-C-â-D-diglucopyranoside, 2 (Figure 3): CD
(MeOH) λmax(∆ꢀ) ) 239 (-1.79), 286 (+0.46), 301 (-0.39); UV-vis
(MeOH/0.1% HCOOH, pH 2.5; 2:8, v/v): λmax ) 218, 231, 280 nm;
1
MS (ESI-), m/z 613 (100%, [M - 1]-), 493 (13%, [M - 121]-); H
NMR (400 MHz, MeOD, COSY) δ 2.45 [dd, 1H, J ) 7.4, 16.4 Hz,
H-C(4a)], 2.73 [dd, 1H, J ) 5.2, 16.4 Hz, H-C(4b)], 3.24-3.38 [m,
4H, H-C(4, 5′′, 5′′′, 3′′′)], 3.43 [m, 2H, H-C(4, 3′′)], 3.54 [dd, 1H, J
) 9.2, 9.6 Hz, H-C(2′′)], 3.62-3.77 [m, 5H, H-C(6′′′a,b, 2′′′, 6′′a,b)],
3.87 [ddd, 1H, J ) 5.2, 7.2 Hz H-C(3)], 4.61 [d, 1H, J ) 6.8 Hz,
H-C(2)], 4.77 [d, 1H, J ) 9.6 Hz, H-C(1′′′)], 4.80 [d, 1H, J ) 9.6
Hz, H-C(1′′)], 6.59 [dd, 1H, J ) 2.0, 8.0 Hz, H-C(6′)], 6.65 [d, 1H,
J ) 8.0 Hz, H-C(5′)], 6.71 [d, 1H, J ) 2.0 Hz, H-C(2′)]; 13C NMR
(100 MHz, MeOD, HMQC, HMBC) δ 26.7 [C-4], 60.6 [C-6′′′], 61.0
[C-6′′], 67.1 [C-3], 69.8 [C-4′′′/4′′], 70.1 [C-4′′/C-4′′′], 72.0 [C-2′′′],
73.4 [C-2′′], 75.5 [C-1′′′/C-1′′], 75.9 [C-1′′/C-1′′′], 77.8 [C-3′′/C-3′′′],
78.0 [C-3′′′/C-3′′], 81.0 [C-5′′/C-5′′′/C-2], 81.1 [C-5′′/C-5′′′/C-2], 81.2
[C-2/C-5′′′/C-5′′], 101.0 [C-4a], 103.2 [C-8], 104.6 [C-6], 113.4 [C-2′],
114.7 [C-5′], 118.2 [C-6′], 130.7 [C-1′], 144.8 [C-3′/C-4′], 144.9 [C-4′/
C-3′], 153.4 [C-8a], 154.4 [C-7, C-5].
(-)-Catechin-6-C-â-D-glucopyranoside, 3 (Figure 3): CD (MeOH)
λ
max(∆ꢀ) ) 240 (-5.01), 283 (+1.00), 301 (-0.20); UV-vis (MeOH/
0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231, 280 nm; MS (ESI-),
1
m/z 451 (100%, [M - 1]-), 331 (20%, [M - 121]-); H NMR (400
MHz, MeOD, COSY) δ 2.53 [dd, 1H, J ) 8.0, 16.0 Hz, H-C(4a)],
2.90 [dd, 1H, J ) 5.4, 16.0 Hz, H-C(4b)], 3.42 [m, 1H, J ) 2.4, 4.4
Hz, H-C(5′′)], 3.50 [m, 1H, J ) 8.8 Hz, H-C(3′′)], 3.54 [m, 1H,
H-C(4′′)], 3.69 [dd, 1H, J ) 8.8, 9.6 Hz, H-C(2′′)], 3.81 [dd, 1H, J
) 4.4, 12.0 Hz, H-C(6a′′)], 3.88 [dd, 1H, J ) 2.4, 12.0 Hz, H-C(6b′′)],
3.98 [ddd, 1H, J ) 5.4, 7.6, 8.0 Hz, H-C(3)], 4.59 [d, 1H, J ) 7.6
Hz, H-C(2)], 4.88 [d, 1H, J ) 9.6 Hz, H-C(1′′)], 6.02 [s, 1H,
H-C(8)], 6.72 [dd, 1H, J ) 2.0, 8.4 Hz, H-C(6′)], 6.77 [d, 1H, J )
8.4 Hz, H-C(5′)], 6.83 [d, 1H, J ) 2.0 Hz, H-C(2′)]; 13C NMR (100
MHz, MeOD, HMQC, HMBC) δ 28.5 [C-4], 62.2 [C-6′′], 68.7 [C-3],
71.2 [C-4′′], 74.6 [C-2′′], 77.5 [C-1′′], 79.6 [C-3′′], 82.5 [C-5′′], 82.9
[C-2], 96.3 [C-8], 102.2 [C-4a], 105.5 [C-6], 115.2 [C-2′], 116.1 [C-5′],
120.0 [C-6′], 132.1 [C-1′], 146.32 [C-3′/C-4′], 146.35 [C-4′/C-3′], 156.1
[C-8a], 156.4 [C-7/C-5], 156.5 [C-5/C-7].
Figure 3. Chemical structures of flavan-3-ol-C-glycosides 1−11.
performed for 10 min at 80 °C with stirring. After cooling, the
alkalization process was stopped by the addition of aqueous hydro-
chloric acid (1 mol/L) until a pH of 5.0 was reached. Thereafter, the
reaction mixture was concentrated under reduced pressure (40 °C) to
about 10 mL and was then applied onto the top of a water-cooled 140
× 40 mm RP-18 column, LiChroprep, 25-40 µm (Merck) conditioned
with aqueous formic acid (0.1% in water; pH 2.5). Chromatography
was performed using aqueous formic acid (0.1% in water; pH 2.5) as
the effluent, followed by aqueous formic acid (0.1% in water; pH 2.5)
containing increasing amounts of methanol. Monitoring the effluent at
270 nm, the fractions containing the title compounds were freed from
solvent in a vacuum to give 10 subfractions (f1-f10), freeze-dried,
and finally purified by means of HPLC (Jasco) consisting of a HPLC
pump system PU 2087, a high-pressure gradient unit, and a PU-2075
UV detector using a preparative RP-18 column, HyperClone micro ODS
(C18), 21.2 × 250 mm, 5 µm (Phenomenex, Aschaffenburg, Germany),
as the stationary phase. Subfractions f4-f10 were dissolved in a mixture
(30:70, v/v) of methanol and aqueous formic acid (0.1% in water; pH
2.5), and, after membrane filtration, aliquots (0.5-2.0 mL) were
fractionated by RP-HPLC. Monitoring the effluent at 280 nm, chro-
matography was performed with a mixture (5:95, v/v) of methanol and
aqueous formic acid (0.1% in water, pH 2.5) for 10 min, increasing
the methanol content to 40% over 30 min and then to 100% over 5
min, thereafter eluting with methanol for 10 min at a flow rate of 18.0
mL/min. After removal of solvents in a vacuum, the title compounds
were suspended in water (10 mL) and freeze-dried two times to afford
the corresponding flavan-3-ol-C-glycosides 1-11 (Figure 3) as white,
amorphous powders in high purities of >99%.
(-)-Epicatechin-8-C-â-D-glucopyranoside, 4 (Figure 3): CD (MeOH)
λ
max(∆ꢀ) ) 239 (-3.33), 282 (-1.16), 301 (-0.12); UV-vis (MeOH/
0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231, 280 nm; MS (ESI-),
1
m/z 451 (100%, [M - 1]-), 331 (20%, [M - 121]-); H NMR (400
MHz, MeOD, COSY) δ 2.87 [ddd, 2H, J ) 2.0, 4.0, 16.0 Hz,
H-C(4a,b)], 3.37 [m, 1H, H-C(5′′)], 3.48 [dd, 1H, J ) 8.4, 8.8 Hz,
H-C(3′′)], 3.50 [dd, 1H, J ) 8.8, 9.6 Hz, H-C(4′′)], 3.81 [dd, 1H, J
) 4.4, 12.0 Hz, H-C(6a′′)], 3.88 [dd, 1H, J ) 2.4, 12.0 Hz, H-C(6b′′)],
4.07 [m, 1H, H-C(2′′)], 4.16 [s, 1H, H-C(3)], 4.91 [s, 1H, H-C(2)],
4.91 [d, 1H, J ) 9.6 Hz, H-C(1′′)], 6.04 [s, 1H, H-C(6)], 6.78 [d,
1H, 8.4 Hz, H-C(5′)], 6.86 [dd, 1H, J ) 1.6, 8.4 Hz, H-C(6′)], 7.06
[d, 1H, J ) 1.6 Hz, H-C(2′)]; 13C NMR (100 MHz, MeOD, HMQC,
HMBC) δ 28.4 [C-4], 61.2 [C-6′′], 66.0 [C-3], 70.2 [C-4′′], 72.0 [C-2′′],
75.2 [C-1′′], 78.5 [C-2], 78.6 [C-3′′], 80.7 [C-5′′], 95.7 [C-6], 98.6
[C-4a], 102.9 [C-8], 113.7 [C-2′], 114.6 [C-5′], 117.5 [C-6′], 131.0
[C-1′], 144.2 [C-3′/C-4′], 144.6 [C-4′/C-3′], 154.1 [C-8a], 155.7 [C-7],
156.7 [C-5].
(-)-Catechin-8-C-â-D-glucopyranoside, 1 (Figure 3): CD (MeOH)
(-)-Epicatechin-6-C-â-D,8-C-â-D-diglucopyranoside, 5 (Figure 3):
UV-vis (MeOH/0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231,
280 nm; MS (ESI-), m/z 613 (100%, [M - 1]-), 493 (16%, [M -
121]-); 1H NMR (400 MHz, MeOD, COSY), 2.86 [ddd, 2H, J ) 2.2,
4.0, 17.2 Hz, H-C(4a,b)], 3.36-3.42 [m, 2H, H-C(5′′, 5′′′)], 3.46-
3.56 [m, 4H, H-C(3′′, 3′′′, 4′′, 4′′′)], 3.64 [dd, 1H, J ) 8.8, 9.2 Hz,
H-C(2′′)], 3.76-3.87 [m, 4H, H-C(6′′′a,b, 6′′a,b)], 3.92 [dd, 1H, J
) 9.6 Hz H-C(2′′′)], 4.17 [m, 1H, H-C(3)], 4.89 [s, 1H, H-C(2)],
4.91 [d, 1H, J ) 9.6 Hz, H-C(1′′)], 4.93 [d, 1H, J ) 9.6 Hz,
H-C(1′′′)], 6.77 [d, 1H, J ) 8.0 Hz, H-C(5′)], 6.83 [dd, 1H, J ) 1.6,
8.0 Hz, H-C(6′)], 7.03 [d, 1H, J ) 1.6 Hz, H-C(2′)]; 13C NMR (100
λ
max(∆ꢀ) ) 241 (+3.09), 284 (+1.32), 297 (-0.45); UV-vis (MeOH/
0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231, 280 nm; MS (ESI-),
1
m/z 451 (100%, [M - 1]-), 331 (41%, [M - 121]-); H NMR (400
MHz, MeOD, COSY) δ 2.57 [dd, 1H, J ) 7.0, 16.0 Hz, H-C(4a)],
2.78 [dd, 1H, J ) 5.2, 16.0 Hz, H-C(4b)], 3.36-3.42 [m, 3H, H-C(5′′,
3′′, 4′′)], 3.71 [dd, 1H, J ) 4.8, 12.0 Hz, H-C(6a′′)], 3.87 [dd, 1H, J
) 1.6, 12.0 Hz, H-C(6b′′)], 4.01 [ddd, 1H, J ) 5.2, 6.8, 7.0 Hz,
H-C(3)], 4.03 [dd, 1H, J ) 9.2 Hz, H-C(2′′)], 4.77 [d, 1H, J ) 6.8
Hz, H-C(2)], 4.85 [d, 1H, J ) 9.2 Hz, H-C(1′′)], 6.02 [s, 1H,
H-C(6)], 6.71 [dd, 1H, J ) 1.6, 8.0 Hz, H-C(6′)], 6.77 [d, 1H, J )

9514 J. Agric. Food Chem., Vol. 54, No. 25, 2006
Stark and Hofmann
MHz, MeOD, HMQC, HMBC) δ 28.2 [C-4], 60.6 [C-6′′′/C-6′′], 60.8
[C-6′′/C-6′′′], 66.0 [C-3], 69.7 [C-4′′′/4′′], 70.0 [C-4′′/C-4′′′], 72.1
[C-2′′′], 73.3 [C-2′′], 75.4 [C-1′′], 76.0 [C-1′′′], 78.08 [C-3′′/C-3′′′],
78.12 [C-3′′′/C-3′′], 78.6 [C-2], 80.8 [C-5′′/C-5′′′], 81.1 [C-5′′′/C-5′′],
99.9 [C-4a], 103.5 [C-8], 104.8 [C-6], 113.7 [C-2′], 114.6 [C-5′], 117.6
[C-6′], 130.8 [C-1′], 144.3 [C-3′/C-4′], 144.6 [C-4′/C-3′], 153.1 [C-8a],
154.1 [C-7], 154.6 [C-5].
Hz, H-C(6b′′)], 3.95-4.00 [m, 2H, H-C(3, 4′′)], 4.01 [dd, 1H, J )
9.6 Hz, H-C(2′′)], 4.57 [d, 1H, J ) 7.2 Hz, H-C(2)], 4.81 [d, 1H, J
) 10.0 Hz, H-C(1′′)], 5.94 [s, 1H, H-C(8)], 6.70 [dd, 1H, J ) 2.0,
8.4 Hz, H-C(6′′)], 6.76 [d, 1H, J ) 8.4 Hz, H-C(5′′)], 6.82 [d, 1H,
J ) 2.0 Hz, H-C(2′′)]; 13C NMR (100 MHz, MeOD, HMQC, HMBC)
δ 27.0 [C-4], 61.6 [C-6′′], 67.3 [C-3], 69.2 [C-4′′], 70.7 [C-2′′], 74.6
[C-3′′], 75.8 [C-1′′], 79.3 [C-5′′], 81.4 [C-2], 94.5 [C-8], 100.4 [C-4a],
104.6 [C-6], 113.8 [C-2′], 114.7 [C-5′], 118.6 [C-6′], 130.7 [C-1′],
144.85 [C-3′/C-4′], 144.86 [C-4′/C-3′], 154.5 [C-8a], 154.8 [C-7/C-5],
154.9 [C-5/C-7].
(-)-Epicatechin-6-C-â-D-glucopyranoside, 6 (Figure 3): UV-vis
(MeOH/0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231, 280 nm;
1
MS (ESI-), m/z 451 (100%, [M - 1]-), 331 (20%, [M - 121]-); H
(-)-Epicatechin-8-C-â-D-galactopyranoside, 10 (Figure 3): UV-
vis (MeOH/0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231, 280
nm; MS (ESI-), m/z 451 (100%, [M - 1]-), 331 (33%, [M - 121]-);
1H NMR (400 MHz, MeOD, COSY) δ 2.88 [ddd, 2H, J ) 2.0, 4.0,
16.4 Hz, H-C(4a,b)], 3.60 [ddd, 1H, J ) 1.6, 6.0 Hz, H-C(5′′)], 3.63
[ddd, 1H, Jae ) 3.0, 9.4 (aa) Hz, H-C(3′′)], 3.70 [dd, 1H, J ) 5.0, 11.4
Hz, H-C(6a′′)], 3.76 [dd, 1H, J ) 7.0, 11.4 Hz, H-C(6b′′)], 3.95 [d,
1H, J ) 2.8 Hz, H-C(4′′)], 4.15 [m, 1H, H-C(3)], 4.16 [dd, 1H, J )
8.8, Hz, H-C(2′′)], 4.89 [s, 1H, H-C(2)], 4.90 [d, 1H, J ) 8.8 Hz,
H-C(1′′)], 6.01 [s, 1H, H-C(6)], 6.78 [d, 1H, J ) 8.0 Hz, H-C(5′)],
6.80 [dd, 1H, J ) 2.0, 8.0 Hz, H-C(6′)], 7.03 [d, 1H, J ) 2.0 Hz,
H-C(2′)]; 13C NMR (100 MHz, MeOD, HMQC, HMBC) δ 28.3 [C-4],
61.6 [C-6′′], 66.0 [C-3], 69.1 [C-4′′], 70.9 [C-2′′′], 74.8 [C-3′′], 75.5
[C-1′′], 78.6 [C-2], 78.8 [C-5′′], 95.9 [C-6], 98.2 [C-4a], 103.5 [C-8],
113.7 [C-2′], 114.6 [C-5′], 117.5 [C-6′], 131.0 [C-1′], 144.3 [C-3′],
144.6 [C-4′], 153.2 [C-8a], 155.7 [C-7], 156.7 [C-5].
(-)-Epicatechin-6-C-â-D,8-C-â-D-digalactopyranoside, 11 (Figure
3): UV-vis (MeOH/0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218,
231, 280 nm; MS (ESI-), m/z 613 (100%, [M - 1]-), 493 (29%, [M
- 121]-), 331 (11%, [M - 283]-); 1H NMR (400 MHz, MeOD,
COSY) δ 2.77 [ddd, 2H, J ) 3.2, 4.4, 17.2 Hz, H-C(4a,b)], 3.51 [m,
4H, H-C(3′′, 3′′′, 5′′, 5′′′)], 3.63 [m, 4H, H-C(6′′a,b, 6′′′a,b)], 3.85
[d, 1H, J ) 2.8 Hz, H-C(4′′′)], 3.88 [dd, 1H, J ) 9.2, 9.6 Hz,
H-C(2′′)], 3.88 [d, 1H, J ) 2.8 Hz, H-C(4′′)], 4.06 [m, 1H, H-C(3)],
4.07 [dd, 1H, J ) 9.2 Hz, H-C(2′′′)], 4.76 [d, 1H, J ) 9.6 Hz,
H-C(1′′′)], 4.78 [s, 1H, H-C(2)], 4.78 [d, 1H, J ) 9.2 Hz, H-C(1′′′)],
6.67 [d, 1H, J ) 8.4 Hz, H-C(5′)], 6.70 [dd, 1H, J ) 1.6, 8.0 Hz,
H-C(6′)], 6.92 [d, 1H, J ) 1.6 Hz, H-C(2′)]; 13C NMR (100 MHz,
MeOD, HMQC, HMBC) δ 28.0 [C-4], 61.4 [C-6′′, C-6′′′], 65.9 [C-3],
69.0 [C-4′′/4′′′], 69.2 [C-4′′′/C-4′′], 70.6 [C-2′′′], 71.4 [C-2′′], 74.7
[C-3′′, C-3′′′], 75.5 [C-1′′, C-1′′′], 78.5 [C-2], 79.2 [C-5′′′, C-5′′], 99.2
[C-4a], 103.1 [C-8], 105.0 [C-6], 113.6 [C-2′], 114.6 [C-5′], 117.4
[C-6′], 130.8 [C-1′], 144.2 [C-3′/C-4′], 144.4 [C-4′/C-3′], 153.1 [C-8a],
153.8 [C-7, C-5].
(-)-Catechin: CD (MeOH) λmax(∆ꢀ) ) 232 (+1.91), 280 (+0.51);
UV-vis (MeOH/0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231,
280 nm.
(+)-Catechin: CD (MeOH) λmax(∆ꢀ) ) 230 (-0.85), 280 (-0.26);
UV-vis (MeOH/0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231,
280 nm.
(-)-Epicatechin: CD (MeOH) λmax(∆ꢀ) ) 239 (-1.41), 279 (-0.63);
UV-vis (MeOH/0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231,
280 nm.
Identification of Flavan-3-ol-C-glycosides 1-11 in Cocoa Powder.
Nonalkalized or alkalized roasted cocoa powder (5.0 g) was extracted
with n-pentane (5 × 30 mL) at room temperature for 30 min. The
residual cocoa material was then extracted five times with acetone/
water (70:30, v/v; 30 mL each) for 45 min at room temperature with
stirring. After centrifugation, the liquid layer was freed from acetone
under reduced pressure at 30 °C and then freeze-dried to give the
acetone/water extract. Aliquots (∼50 mg) of the acetone/water extract
were taken up in a methanol/water mixture (1:1, v/v; 10 mL), which
was acidified to pH 2.5 with formic acid (0.1% in water). After
membrane filtration, aliquots (5 µL) were analyzed by means of HPLC-
MS/MS, which was equipped with a 150 × 2 mm i.d., 5 µm, RP
phenylhexyl column (Phenomenex) operated with a flow rate of 0.2
mL/min. Chromatography was performed starting with aqueous formic
acid (0.1%, pH 2.5), held for 10 min, and the methanol content was
increased to 40% in 40 min, increased to 100% within 10 min, and,
finally, held at 100% for 10 min. By means of the multiple-reaction
monitoring (MRM) mode, the individual flavan-3-ol-C-glycosides 1,
NMR (400 MHz, MeOD, COSY) δ 2.76 [dd, 1H, J ) 3.2, 16.4 Hz,
H-C(4a)], 2.89 [dd, 1H, J ) 4.4, 16.4 Hz, H-C(4b)], 3.43 [m, 1H,
H-C(5′′)], 3.50 [dd, 1H, J ) 8.4, 8.8 Hz, H-C(3′′)], 3.54 [dd, 1H, J
) 8.0, 8.8 Hz, H-C(4′′)], 3.69 [dd, 1H, J ) 8.8, 9.6 Hz, H-C(2′′)],
3.81 [dd, 1H, J ) 4.6, 12.2 Hz, H-C(6a′′)], 3.88 [dd, 1H, J ) 2.2,
12.2 Hz, H-C(6b′′)], 4.21 [m, 1H, J ) 3.2, 4.4 Hz, H-C(3)], 4.86 [s,
1H, H-C(2)], 4.88 [d, 1H, J ) 9.6 Hz, H-C(1′′)], 6.01 [s, 1H,
H-C(8)], 6.77 [d, 1H, J ) 8.4 Hz, H-C(5′)], 6.81 [dd, 1H, J ) 1.6,
8.4 Hz, H-C(6′)], 6.98 [d, 1H, J ) 1.6 Hz, H-C(2′)]; 13C NMR (100
MHz, MeOD, HMQC, HMBC) δ 27.8 [C-4], 60.7 [C-6′′], 66.1 [C-3],
69.8 [C-4′′], 73.2 [C-2′′], 76.1 [C-1′′], 78.2 [C-3′′], 78.5 [C-2], 81.1
[C-5′′], 95.2 [C-8], 100.0 [C-4a], 103.8 [C-6], 114.0 [C-2′], 114.5 [C-5′],
118.0 [C-6′], 130.7 [C-1′], 144.5 [C-3′/C-4′], 144.6 [C-4′/C-3′], 154.5
[C-8a], 155.1 [C-7/C-5], 155.1 [C-5/C-7].
(-)-Catechin-8-C-â-D-galactopyranoside, 7 (Figure 3): CD (MeOH)
max(∆ꢀ) ) 239 (+1.94), 283 (+1.15), 301 (-0.30); UV-vis (MeOH/
λ
0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231, 280 nm; MS (ESI-),
m/z 451 (100%, [M - 1]-), 331 (35%, [M-121]-); 1H NMR (400 MHz,
MeOD, COSY) δ 2.56 [dd, 1H, J ) 6.6, 16.4 Hz, H-C(4a)], 2.70 [dd,
1H, J ) 5.2, 16.4 Hz, H-C(4b)], 3.51 [dd, 1H, Jae ) 3.0, 9.4 (aa) Hz,
H-C(3′′)], 3.58 [ddd, 1H, J ) 1.6, 5.2 Hz, H-C(5′′)], 3.68 [dd, 1H,
J ) 4.8, 11.2 Hz, H-C(6a′′)], 3.75 [dd, 1H, J ) 7.0, 11.2 Hz,
H-C(6b′′)], 3.90 [d, 1H, J ) 2.8 Hz, H-C(4′′)], 4.02 [ddd, 1H, J )
5.2, 6.4 Hz, H-C(3)], 4.10 [dd, 1H, J ) 9.2, 9.6, Hz, H-C(2′′)], 4.79
[d, 1H, J ) 6.0 Hz, H-C(2)], 4.82 [d, 1H, J ) 9.6 Hz, H-C(1′′)],
5.98 [s, 1H, H-C(6)], 6.67 [dd, 1H, J ) 2.0, 8.4 Hz, H-C(6′)], 6.74
[d, 1H, J ) 8.4 Hz, H-C(5′)], 6.80 [d, 1H, J ) 2.0 Hz, H-C(2′)]; 13
C
NMR (100 MHz, MeOD, HMQC, HMBC) δ 26.0 [C-4], 61.7 [C-6′′],
67.0 [C-3], 69.3 [C-4′′], 70.6 [C-2′′], 74.8 [C-3′′], 75.5 [C-1′′], 79.1
[C-5′′], 80.9 [C-2], 95.7 [C-6], 99.2 [C-4a], 103.4 [C-8], 113.3 [C-2′],
114.8 [C-5′], 118.0 [C-6′], 131.0 [C-1′], 144.7 [C-3′], 144.9 [C-4′],
152.9 [C-8a], 155.7 [C-7], 156.1 [C-5].
(-)-Catechin-6-C-â-D,8-C-â-D-digalactopyranoside, 8 (Figure 3):
CD (MeOH) λmax(∆ꢀ) ) 239 (-1.91), 286 (+0.61), 301 (-0.20); UV-
vis (MeOH/0.1% HCOOH, pH 2.5; 2:8, v/v): λmax ) 218, 231, 280
nm; MS (ESI-), m/z 613 (100%, [M - 1]-), 451 (35%, [M - 163]-),
493 (31%, [M - 121]-), 373 (15%, [M - 241]-), 331 (11%, [M -
283]-); 1H NMR (400 MHz, MeOD, COSY) δ 2.57 [dd, 1H, J ) 6.8,
16.4 Hz, H-C(4a)], 2.81 [dd, 1H, J ) 5.0, 16.4 Hz, H-C(4b)], 3.55
[dd, 1H, Jae ) 2.8, 9.6 (aa) Hz, H-C(3′′′)], 3.60-3,67 [m, 3H, H-C(3′′,
5′′, 5′′′)], 3.70-3,82 [m, 4H, H-C(6′′a,b, 6′′′a,b)], 3.94 [d, 1H, J )
2.8 Hz, H-C(4′′′)], 3.98-4.03 [m, 3H, H-C(3, 2′′, 4′′)], 4.09 [dd,
1H, J ) 9.6 Hz, H-C(2′′′)], 4.74 [d, 1H, J ) 6.4 Hz, H-C(2)], 4.84
[d, 1H, J ) 9.6 Hz, H-C(1′′′)], 4.87 [d, 1H, J ) 9.6 Hz, H-C(1′′)],
6.70 [dd, 1H, J ) 2.0, 8.0 Hz, H-C(6′)], 6.76 [d, 1H, J ) 8.0 Hz,
H-C(5′)], 6.82 [d, 1H, J ) 2.0 Hz, H-C(2′)]; 13C NMR (100 MHz,
MeOD, HMQC, HMBC) δ 26.4 [C-4], 61.6 [C-6′′/C-6′′′], 61.7 [C-6′′′/
C-6′′], 67.1 [C-3], 69.21 [C-4′′/4′′′], 69.25 [C-4′′′/C-4′′], 69.9 [C-2′′],
70.8 [C-2′′′], 74.4 [C-3′′′], 74.7 [C-3′′], 75.5 [C-1′′/C-1′′′], 75.6 [C-1′′′/
C-1′′], 79.3 [C-5′′/C-5′′′], 79.4 [C-5′′′/C-5′′], 81.2 [C-2], 100.2 [C-4a],
103.1 [C-8], 104.9 [C-6], 113.4 [C-2′], 114.7 [C-5′], 118.2 [C-6′], 130.8
[C-1′], 144.8 [C-3′/C-4′], 144.9 [C-4′/C-3′], 152.4 [C-8a], 153.4 [C-7/
C-5], 154.6 [C-5/C-7].
(-)-Catechin-6-C-â-D-galactopyranoside, 9 (Figure 3): CD (MeOH)
λ
max(∆ꢀ) ) 240 (-1.40), 283 (+0.33), 301 (-0.18); UV-vis (MeOH/
0.1% HCOOH, pH 2.5; 2:8, v/v) λmax ) 218, 231, 280 nm; MS (ESI-),
1
m/z 451 (100%, [M - 1]-), 331 (30%, [M - 121]-); H NMR (400
MHz, MeOD, COSY) δ 2.50 [dd, 1H, J ) 7.8, 16.2 Hz, H-C(4a)],
2.86 [dd, 1H, J ) 5.4, 16.2 Hz, H-C(4b)], 3.58 [dd, 1H, Jae ) 2.8, 9.6
(aa) Hz, H-C(3′′)], 3.64 [ddd, 1H, J ) 5.6, 6.4 Hz, H-C(5′′)], 3.72
[dd, 1H, J ) 4.8, 11.6 Hz, H-C(6a′′)], 3.79 [dd, 1H, J ) 7.0, 11.6

Structure Determination of Flavan-3-ol-C-hexosides in Cocoa
J. Agric. Food Chem., Vol. 54, No. 25, 2006 9515
3, 4, 6, 7, 9, and 11 (m/z 451.2 f 331.0) and 2, 5, 8, and 10 (m/z
613.1 f 373.0) were analyzed using the transition reactions monitored
for a duration of 150 ms given in brackets.
High-Performance Liquid Chromatography (HPLC). The HPLC
apparatus (Kontron, Eching, Germany) consisted of low-pressure
gradient system 525 HPLC pump, a M800 gradient mixer, a type 560
autosampler, and a DAD type 540+ diode array detector. Chromatog-
raphy was performed on 250 × 4.6 mm stainless steel columns packed
with RP phenylhexyl material, 5 µm (Phenomenex), operated with a
flow rate of 0.8 mL/min.
Liquid Chromatography-Mass Spectrometry (LC-MS/MS). LC-
MS/MS analysis was performed using an Agilent 1100 HPLC system
connected to the API 3200 LC-MS/MS (Applied Biosystems, Darm-
stadt, Germany) running in the negative electrospray ionization mode.
LC-Time-of-Flight Mass Spectrometry (LC-TOF-MS). High-
resolution mass spectra of compound 1 measured on a Bruker Micro-
TOF (Bruker Daltonics, Bremen, Germany) mass spectrometer with
flow injection and using sodium formate and PEG 600, respectively,
as the reference revealed the molecular formula C21H24O11 for the taste
compound.
Sensory evaluation of aqueous mixtures of the individual
solvent fractions isolated from NC and AC, respectively, by
means of taste profile analysis demonstrated the highest scores
for bitterness, astringency, and sourness in fraction IV, followed
by the ethyl acetate extractables (fraction III) evaluated with
somewhat lower taste intensities for bitter taste. Comparison
of the samples NC and AC clearly indicated that the alkalization
process induced a significant reduction of bitterness (4.0 f 2.5),
astringency (2.5 f 2.0), and sourness (1.5 f 1.0) of the cocoa
water solubles (fraction IV) (Table 1). Similarly to fraction IV,
a significant decrease of the intensities of the individual taste
modalities was observable for fraction III. Therefore, fractions
III and IV were recombined to fraction III/IV prior to further
analysis.
Sensory-Guided GPC Separation of Fraction III/IV. To
sort out the strongly taste-active compounds from the bulk of
less taste-active or tasteless substances and to visualize the
differences in the tastant signature of both the cocoa samples,
fraction III/IV isolated from NC and AC, respectively, was
separated by means of GPC. Using Sephadex LH-20 as the
stationary phase and methanol/water mixtures as the mobile
phase, the effluent was separated into 16 fractions, respectively,
which were collected separately and, then, investigated for the
taste impact. To achieve this, the 16 fractions obtained from
fraction III/IV of NC (Figure 1A) and AC (Figure 1B) were
freeze-dried, taken up in water, and then analyzed by means of
the TDA (8) using the recently developed half-mouth test
(1, 7).
Circular Dichroism (CD) Spectroscopy. For CD spectroscopy,
methanolic solutions of the samples were analyzed by means of a Jasco
J600 Spectro polarimeter (Hachioji, Japan).
1
Nuclear Magnetic Resonance Spectroscopy (NMR). H, COSY,
HMQC, HMBC, 13C, and DEPT-135 NMR measurements were
performed on an DMX 400 spectrometer (Bruker, Rheinstetten,
Germany). Evaluation of the experiments was carried out using 1D-
and 2D-WIN NMR (version 6.1) as well as MestReC software (version
4.4.6.0, Mestrelab Research, Santiago de Compostela, Spain). Tetra-
methylsilane was used as the internal standard.
In the NC sample, the highest TD factors of 512 and 256
were found for the puckering astringent taste of GP fractions
12-15, followed by fractions 4 and 16, which still showed
astringent taste after a dilution of 1+127 (Table 2). In contrast,
GPC fractions 8-11 exhibited a velvety, silky type of astringent
mouth-coating sensation evaluated with TD factors between 32
and 256, respectively. In the AC sample, the puckering
astringent taste of GPC fractions 4 and 14-16 were judged with
the highest TD factor of 128. In addition, the velvety astringent
fraction 10 was evaluated with a TD factor of 128 (Table 2).
RESULTS AND DISCUSSION
A freshly prepared aqueous suspension (pH 6.0) of nonal-
kalized, powdered cocoa (NC) imparted the typical complex
and attractive cocoa taste and was used as the reference for taste
profile analysis. In comparison, an aqueous suspension (pH 6.0)
of alkalized cocoa (AC) powder was sensorially evaluated. To
achieve this, a trained sensory panel was asked to rate the
intensity of the taste qualities bitter, sour, and astringent mouth-
coating on a scale from 0 (not detectable) to 5 (intensely
detectable). As expected, high scores were found for the intensity
of the bitter taste (4.7) and the astringent, mouth-coating taste
sensation (3.0) as well as the sourness (2.2) of the nonalkalized
cocoa sample (Table 1). In contrast, bitterness (3.1), astringency
(2.1), and sourness (1.1) were strongly decreased in the AC
powder. To gain a first insight into the influence of the
alkalization process on the cocoa taste compounds, both cocoa
powders were extracted sequentially with solvents of different
polarities.
Solvent Fractionation of AC and NC. The cocoa powders
were extracted with n-pentane to give the pentane solubles
(fraction I) after evaporation of the solvent in a vacuum. The
residual cocoa material was then extracted with aqueous acetone,
and, after removal of the acetone from the aqueous phase in
vacuum, the solution obtained was extracted with dichlo-
romethane. The combined organic layers were freed from
solvent in vacuum to give the dichloromethane extractables
(fraction II). The remaining aqueous layer was extracted with
ethyl acetate, the organic phase was freed from solvent to give
the ethyl acetate extractables (fraction III), and the aqueous
phase was freeze-dried to give the water solubles (fraction IV).
In addition, the insoluble residue of the cocoa samples was
freeze-dried to give fraction V. Besides the cocoa butter present
in fraction I and the nonsoluble residue, the solubles in fraction
IV were found with the highest yields of 16.7 and 14.2% in the
NC and AC samples, respectively (Table 1).
By comparison of the taste dilution chromatograms in Figure
1 obtained for the solvent fraction III/IV of the two cocoa
samples, the sensory data obtained for GPC fractions 11-16
of the AC sample showed clearly lower TD factors for their
puckering and rough astringent sensation. As these fractions
have been recently demonstrated to contain catechin, epicat-
echin, and the family of (4â f 8)-linked epicatechin oligomers
up to the decamer (1), these data show first evidence for the
degradation of these so-called procyanidins during the alkal-
ization process. In contrast, the puckering astringent fractions
4, 6, and 8 were evaluated with identical TD factors in both
cocoa samples (Table 2). As the tastes of these fractions have
been recently shown to be due to a family of N-phenylprope-
noyl-L-amino acids (1, 2), the present study demonstrates first
evidence that these cinnamoyl derivatives are relatively stable
throughout the alkalization process. In fractions 8-10, exhibiting
a velvety astringent taste, LC-MS/MS analysis confirmed our
previous findings (1) that quercetin, naringenin, luteolin, and
apigenin glycopyranosides are present in these fractions (Table
2).
The only fraction that showed a higher TD factor for the
velvety astringent taste after alkalization was GPC fraction 10.
An LC-MS full scan run revealed several compounds exhibiting
a molecular mass of 452 Da. Isolation of these peaks by means
of semipreparative RP-HPLC, followed by sensory analysis,
revealed that these compounds exhibited a smooth and velvety

9516 J. Agric. Food Chem., Vol. 54, No. 25, 2006
Stark and Hofmann
Table 2. Yield, Taste Quality, and Taste Dilution (TD) Factor of GPC Fractions Obtained from Solvent Fraction III/IV Isolated from Nonalkalized (NC)
and Alkalized Cocoa (AC) Powder
TD factor in
fractiona
taste qualityb
astringent
astringent
sour
NC
AC
bitter and astringent compounds identifiedc
1
2
3
8
4
32
<1
32
64
organic acids, sugars
sweet
4
4
5
puckering, astringent, sour
bitter
128
32
128
64
N-phenylpropenoyl amino acids, organic acids
caffeine, theobromine
6
7
8
9
10
11
12
13
14
15
16
puckering, astringent
puckering, astringent
velvety, astringent
velvety, astringent
velvety, astringent
puckering, astringent
puckering, astringent
puckering, astringent
puckering, astringent
puckering, astringent
puckering, astringent
64
32
32
32
64
32
64
16
128
16
64
N-phenylpropenoyl amino acids
N-phenylpropenoyl amino acids
apigenin, luteolin, kaempferol, naringenin, and quercetin glycosides
64
256
512
256
512
256
128
epicatechin, catechin
epicatechin, catechin
procyanidin B2
procyanidin C1
[epicatechin-(4âf8)]3-epicatechin
[epicatechin-(4âf8)]4-epicatechin, polymers
64
128
128
128
a Number of GPC fraction referring to Figure 1. b Taste quality and TD factor were determined by using a half-mouth test. c Compounds were identified by means of
HPLC-DAD and LC-MS/MS and were confirmed with reference compounds as reported recently (1).
astringent sensation. Electrospray ionization (ESI-) MS/MS
analysis of one representative compound revealed an [M - 1]-
ion with m/z 451 and a cleavage of 120 amu to give the fragment
ion m/z 331 as expected for a catechin C-glycoside (Figure 2).
Because the same compounds were found in significantly lower
concentrations in the NC sample, the question arose whether
these velvety mouth-coating compounds are formed nonenzy-
matically during the alkalization process.
the individual subfractions was performed by preparative RP-
HPLC, thus giving rise to highly pure compounds that were
analyzed by means of UV-vis, LC-MS, and NMR spectros-
copy.
The velvety astringent compound 1 isolated from fraction f4
was obtained as a white amorphous powder, showed the typical
absorption maxima expected for catechin and its epimeric
epicatechin, respectively, and showed an [M - 1]- ion with
m/z 451 as well as the fragment ion with m/z 331 by ESI- MS/
MS as shown in Figure 2. High-resolution LC-MS analysis
confirmed the target compound to have the molecular formula
Isolation and Structure Elucidation of Velvety Taste
Compounds in GPC Fraction 10. To gain more detailed insight
into the compounds imparting the velvety astringent sensation
in GPC fractions 10 of the AC sample, solutions of (-)-
epicatechin and glucose, known as the quantitatively predomi-
nant polyphenol and reducing carbohydrate, respectively, were
treated in a newly developed “model alkalization process”. To
achieve this, binary mixtures of (-)-epicatechin and D-glucose
in ratios of 1:1 to 1:10 were heated in an aqueous K2CO3
solution at 80 °C with stirring. After 10, 20, 40, and 60 min,
the alkalization process was stopped by acidification with
hydrochloric acid and, after membrane filtration, aliquots of the
mixture were analyzed by means of RP-HPLC coupled to a
diode array detector and a mass spectrometer, respectively. First,
the model alkalization experiments revealed a significant
epimerization of (-)-epicatechin into (-)-catechin, thus con-
firming earlier data reported in the literature (9). Second, a series
of compounds exhibiting a molecular mass of 452 Da was
detectable. The UV-vis and MS spectra as well as the retention
times of these compounds fitted well with those isolated from
GPC fraction 10 of the alkalized cocoa. Among the model
alkalization experiments, a molar ratio of 1:10 of (-)-epicatechin
to D-glucose and a heating time of 10 min at 80 °C were found
as the conditions showing the most efficient conversion of (-)-
epicatechin into these reaction products.
1
C21H24O11. The H NMR spectrum of compound 1 showed an
aromatic singlet for H-C(6) at 6.02 ppm and three aromatic
protons resonating at 6.71, 6.77, and 6.85 ppm showing an ABX
coupling system. In addition, the four aliphatic protons H-C(4a),
H-C(4b), H-C(3), and H-C(2) were observable at 2.57, 2.78,
4.01, and 4.77 ppm, coupling with each other and suggesting a
flavan-3-ol aglycon. Besides the signals of the flavan-3-ol
aglycone, the 1H NMR spectrum also exhibited seven aliphatic
protons resonating at 3.36-3.42 ppm [H-C(5′′, 3′′, 4′′)], 3.71
ppm [H-C(6a′′)], 3.87 ppm [H-C(6b′′)], 4.03 ppm [H-C(2′′)],
and 4.85 ppm [H-C(1′′)] as expected for a hexose unit.
Considering all of the coupling constants of the sugar moiety
in the molecule and, in particular, the coupling constant of J )
9.2 Hz observed for the protons H-C(1′′) and H-C(2′′) and
comparing these values with the H-C(1)/H-C(2) coupling
constants reported for â-D-glucopyranosides and R-D-glucopy-
ranosides (10, 11), the sugar moiety was with certainty identified
as the â-D-glucopyranose.
A comparison of the 13C NMR spectrum, in which 21 signals
appeared, with the results of the DEPT-135 experiment showing
13 signals, revealed 8 signals corresponding to quarternary
carbon atoms. Unequivocal assignment of these quarternary
carbon atoms and the hydrogen-substituted carbon atoms,
respectively, could be successfully achieved by means of
heteronuclear multiple bond correlation spectroscopy (HMBC)
To speed the isolation and full spectroscopic structure
determination of the key taste compounds in GPC fraction 10,
the alkali-catalyzed (-)-epicatechin/D-glucose reaction was
repeated on a preparative scale. The processed mixture was
concentrated under reduced pressure and then fractionated by
preparative column chromatography on RP-18 material. Moni-
toring the effluent at 270 nm, the effluents of the individual
peaks detected were collected and freed from solvent in a
vacuum to give 10 subfractions (f1-f10). Final purification of
2
3
optimized for JC,H and JC,H coupling constants and hetero-
nuclear multiple-quantum correlation spectroscopy (HMQC)
optimized for 1JC,H coupling constants, respectively. Addition-
ally, the HMBC experiment revealed a correlation between the
sugar proton H-C(1′′) resonating at 4.85 ppm and neighboring
carbon atoms C(7), C(8), and, in particular, C(8a), thus

Structure Determination of Flavan-3-ol-C-hexosides in Cocoa
J. Agric. Food Chem., Vol. 54, No. 25, 2006 9517
Figure 4. HMBC spectrum (400 MHz, d3-MeOD) of (−)-catechin-8-C-â-D-glucopyranoside (1).
demonstrating clearly the intramolecular 8-C-linkage of the â-D-
glucopyranose to its aglycone (Figure 4). The typical chemical
shift of the carbon atom C(1′) at 75.5 ppm confirmed the
C-linkage of the sugar part.
protons H-C(2), H-C(3), H-C(4a,b), H-C(2′), H-C(5′), and
H-C(6′), as well as the chemical shifts of their corresponding
C-atoms, indicated the 2(S),3(R)-configuration of the (-)-
catechin aglycone of compound 1 isolated from fraction f4.
Taking all of these spectroscopic data into consideration, the
structure of the velvety astringent compound 1 isolated from
fraction f4 could be unequivocally identified as the previously
not reported (-)-catechin-8-C-â-D-glucopyranoside (Figure 3).
By means of HPLC degustation, HPLC-MS/MS, and HPLC-
DAD, compound 1 was confirmed as an important taste
compound in GPC fraction 10, which was evaluated with a high
TD factor for its astringent and mouth-coating oral sensation
(Figure 1).
Using the same analytical strategy, LC-MS/MS studies and
1D- and 2D-NMR experiments led to the identification of
compound 2 detected in fraction f4 as (-)-catechin-6-C,8-C-
â-D-diglucopyranoside, compound 3 detected in fraction f5 as
(-)-catechin-6-C-â-D-glucopyranoside, and compounds 4-6
detected in fraction f9 as (-)-epicatechin-8-C-â-D-glucopyra-
noside, (-)-epicatechin-6-C,8-C-â-D-diglucopyranoside, and
(-)-epicatechin-6-C-â-D-glucopyranoside (Figure 3), respec-
tively.
Because LC-MS/MS experiments with cocoa sample AC
revealed, besides compounds 1-6, additional compounds with
typical mass transitions for epi- or catechin-C-hexopyranosides
in cocoa, additional alkalization model experiments were
performed with D-galactose as described above for D-glucose.
From this model experiment, the corresponding flavan-3-ol
galactopyranosides could be isolated and identified as (-)-
catechin-8-C-â-D-galactopyranoside (7), (-)-catechin-6-C,8-C-
â-D-digalactopyranoside (8), (-)-catechin-6-C-â-D-galactopy-
ranoside (9), (-)-epicatechin-6-C,8-C-â-D-digalactopyranoside
(10), and (-)-epicatechin-8-C-â-D-glucopyranoside (11) (Figure
3).
To clarify the configuration of the carbon atoms C(2) and
C(3) present in the aglycone of compound 1, in the following
(-)-epicatechin and (+)-catechin were investigated by NMR
spectroscopy and the unequivocal proton and carbon assign-
1
ments were successfully achieved by means of H, g-COSY,
13C, DEPT-135, HMQC, and HMBC. On comparison of the
chemical shifts and coupling constants of the protons H-C(4a,b),
H-C(3), and H-C(2), the diastereomers (-)-epicatechin and
(+)-catechin were clearly distinguishable; for example, the
chemical shift (2.75 ppm) and the vicinal coupling constant (2.8
Hz) of the proton H-C(4a) of (-)-epicatechin are significantly
different from those of the corresponding proton H-C(4a) of
(+)-catechin (2.52 ppm, 8.2 Hz). The chemical shifts of the
protons H-C(3) and H-C(2) of (-)-epicatechin at 4.20 and
4.82 were significantly different from those of the corresponding
protons H-C(3) and H-C(2) of (+)-catechin resonating at 4.00
and 4.58 ppm. It is interesting to note that the sequence of
increasing chemical shifts of the protons H-C(5′), H-C(6′),
and H-C(2′) of (-)-epicatechin is substantially different from
that of the (+)-catechin structure showing increasing chemical
shifts from H-C(6′) and H-C(5′) to H-C(2′). To complete
these investigations, (-)-catechin and (+)-epicatechin, isolated
from alkalization model experiments performed with (-)-
epicatechin and (+)-catechin, respectively, were analyzed by
means of NMR spectroscopy. The same chemical shifts and
coupling constants found for the diastereomers (-)-epicatechin
and (+)-catechin were found for the corresponding enantiomeric
(-)-catechin and (+)-epicatechin. By comparison of the chemi-
cal shifts and the coupling constants of the protons H-C(4a),
H-C(3), and H-C(2) and the sequence of increasing chemical
shifts of the protons H-C(5′), H-C(6′), and H-C(2′) of (-)-
epicatechin and (+)-catechin with the isolated C-glycoside 1,
its configuration at carbon atom C(2) and at C(3) could be
deduced. The coupling constants and the chemical shifts of the
As a final proof of the stereochemistry of all these taste
compounds, CD spectroscopic measurements were performed
with the commercially available reference isomers (-)-epicat-
echin, (-)-catechin, and (+)-catechin as well as the isolated

9518 J. Agric. Food Chem., Vol. 54, No. 25, 2006
Stark and Hofmann
nonenzymatic glycosylation chemistry, the corresponding bis-
glycosylated flavan-3-ols 2, 5, 8, and 10 are formed by an
additional glycosylation of a mono-C-hexoside involving glu-
cose or galactose.
To the best of our knowledge, compounds 1-3, 5, and 7-10
have not previously been reported in the literature. Although
compound 4 has been reported earlier in Pu-er tea (14) and,
together with compound 6, was isolated from cassia bark
(Cinnamomum cassia Blume) (15) and compound 11 has been
earlier reported in cocoa liquor (16), the velvety astringent taste
activity of (-)-epicatechin-C-glycopyranosides as well as their
potential formation by a nonenzymatic C-glycosylation of
polyphenols has not been previously described.
Identification of Flavan-3-ol-C-glycosides 1-11 in Non-
alkalized and Alkalized Cocoa Powder. To verify the occur-
rence of the flavan-3-ol-C-glycosides 1-11 in cocoa powder,
an aqueous acetone extract isolated from cocoa powder was
analyzed by RP-HPLC coupled to an LC-MS/MS instrument
running in the negative electrospray ionization mode as shown
in Figure 7. By means of the MRM mode, the mass transition
m/z 451.2 f 331.0 was recorded for the flavan-3-ol-C-
glycosides 1, 3, 4, 6, 7, 9, and 11 and the mass transition m/z
613.1 f 373.0 was monitored for the flavan-3-ol-C-glycosides
2, 5, 8, and 10.
LC-MS/MS analysis of the cocoa samples demonstrated for
the first time that besides the previously reported (-)-epicat-
echin-8-C-galactopyranoside (11) (16), the taste compounds
1-10 are present in cocoa. These data were further confirmed
by comparing the sensory quality, the mass spectra, and retention
times (RP-HPLC) with those obtained for the corresponding
reference compounds and, finally, by cochromatography. In
addition, recording the mass transition m/z 451.2 f 331.0 and
comparing the peak areas obtained for (-)-catechin- and (-)-
epicatechin-C-hexopyranosides in an aqueous acetone extract
isolated from nonalkalized (Figure 7A) and alkalized (Figure
7B) cocoa powder clearly demonstrated an increase of (-)-
catechin-C-glycopyranosides (1, 3, 7, and 9) upon alkalization.
In addition, alkalization induced a decrease of (-)-epicatechin-
C-glycopyranosides (4, 6, and 11), which can be explained by
an alkali-catalyzed epimerization of (-)-epicatechin-C-glyco-
sides into the corresponding (-)-catechin-C-glycosides during
the alkalization process as already observed for the alkali-treated
aglycones as described above. To investigate this observation
more precisely, the development of a stable isotope dilution
analysis for the accurate quantification of these taste compounds
is currently under progress.
Figure 5. (A) UV
catechin-6-C- -galactopyranoside (9; - - -) and (
-diglucopyranoside (2; - -); and (C) CD spectra of (
-glucopyranoside (4;- - -).
−
vis spectrum; (B) CD spectra of (
)-catechin-6-C,8-C-
)-epicatechin (s
−)-catechin (s), (−)-
â
-D
−
â
-
D
−
−
)
and (−)-epicatechin-8-C-â-D
(-)-catechin and the isolated C-glycosides 2, 4, and 9 (Figure
5). The CD spectra of (-)-epicatechin and (+)-catechin were
well in line with literature data (12). The data obtained clearly
demonstrated that the C-glycosides 2 and 9 isolated from
fractions f4 and f5 showed a positive CD band at about 280
nm, similar to that of (-)-catechin and C-glycoside 4 isolated
from fractions f9, which showed a negative CD band at about
280 nm similar to that of (-)-epicatechin.
Sensory Activity of Flavan-3-ol-C-glycosides. Prior to
Taking all of these data into consideration, the reaction
pathways leading to the formation of the flavan-3-ol-C-
glycosides by a previously not reported nonenzymatic C-
glycosylation of flavan-3-ols are proposed in Figure 6. Under
alkaline conditions, (-)-epicatechin (1) is epimerized to (-)-
catechin (2), thus confirming earlier reports (9). As exemplified
for the formation of the C-glycoside 6 in Figure 6, the carbonyl
atom of the open-chain form of the hexose (3) favorably formed
under alkaline conditions (13) is attacked by the A-ring of the
(-)-epicatechin at the 6-position to give the intermediate adduct
4. Upon water elimination, the unsaturated carbonyl 5 is formed,
which is cyclized to give the target glycoside 6 by means of a
Michael-type addition. Similarly, the reaction of the (-)-
epicatechin at the 8-position leads to the formation of glycoside
4, whereas the involvement of (-)-catechin instead of (-)-
epicatechin results in the corresponding diastereomeric glyco-
sides 1 and 3 (Figure 6). On the basis of this type of
sensory analysis, the purity of all compounds was checked by
HPLC-MS as well as H NMR spectroscopy. To study the
sensory activity of the flavan-3-ol-C-glycosides 1-11, the
human sensory recognition thresholds were determined in bottled
water (pH 6.0) using the half-mouth test (1, 7) (Table 3).
1
Compared to the flavan-3-ols (-)-epicatechin and (-)-
catechin exhibiting a puckering astringent as well as bitter taste
at threshold concentration levels between 600 and 1000 µmol/
L, the flavan-3-ol-C-glycosides were found to induce a smooth
astringent and velvety mouth-coating sensation at very low
threshold concentrations ranging from 1.1 to 99.5 µmol/L
without exhibiting any bitter taste (Table 3). In particular, the
threshold concentrations of 1.1 and 1.3 µmol/L determined for
(-)-catechin-6-C,8-C-â-D-digalactopyranoside and (-)-catechin-
6-C,8-C-â-D-diglucopyranoside were found to be extraordinarily
low, thus confirming our recent findings that the sugar moiety

Structure Determination of Flavan-3-ol-C-hexosides in Cocoa
J. Agric. Food Chem., Vol. 54, No. 25, 2006 9519
Figure 6. Proposed reaction pathway for the nonenzymatic glycosylation of (−)-epicatechin and (−)-catechin leading to the formation of (−)-epicatechin-
6-C-
â
-
D
-glucopyranoside (6), (
−)-epicatechin-8-C-
â
-
D
-glucopyranoside (4), (
−
)-catechin-6-C- -glucopyranoside (3), and (
â
-
D
−)-catechin-6-C-â-D
-gluco-
pyranoside (1).
Figure 7. LC-MS/MS analysis of (A) nonalkalized and (B) alkalized cocoa powder using the multiple reaction monitoring (MRM) mode.
of glycosylated phenols is a key driver for their sensory activity
glycosides showed somewhat higher threshold concentrations
(1, 7). In comparison, the monosubstituted flavan-3-ol-C-
ranging from 3.1 to 99.5 µmol/L.

9520 J. Agric. Food Chem., Vol. 54, No. 25, 2006
Stark and Hofmann
sponding 8-C-â-D-glucopyranoside (1) and 6-C-â-D-glucopy-
ranoside (3) of (-)-catechin showed that the aglycone strongly
influences the taste intensity of these compounds; for example,
(-)-catechin-8-C-â-D-glucopyranoside (1) and (-)-catechin-6-
C-â-D-glucopyranoside (3) possess threshold concentrations of
16.2 and 63.6 µmol/L, respectively, which are 5.3 and 6.4 times
above the threshold concentration determined for the corre-
sponding (-)-epicatechin derivative.
In addition to the sugar species and the aglycone, the position
of the individual monosaccharides at the aglycone also seems
to have an influence on the sensory activity of the flavan-3-ol-
C-glycosides; for example, all compounds bearing the glucose
or galactose linked at position 6 to the aglycone were evaluated
with relatively high thresholds of 99.5, 63.6, and 10.0 µmol/L,
respectively, whereas (-)-catechin-8-C-â-D-galactopyranoside
(7), (-)-catechin-8-C-â-D-glucopyranoside (1), and (-)-epicat-
echin-8-C-â-D-glucopyranoside (4), respectively, all bearing the
sugar moiety at position 8 next to the aglycone, showed
significantly lower thresholds of 49.8, 16.2, and 3.1 µmol/L,
respectively (Table 3).
Table 3. Human Recognition Taste Thresholds for the Oral Sensation
Induced by Flavan-3-ols and Flavan-3-ol-C-glycosides, Respectively
threshold concna
(µ
mol/L) for
compound
astringency bitterness
(−
(−
(−
(−
(−
(−
(−
(−
(−
(−
(−
(−
(−
)-epicatechin
800.0b
600.0b
99.5c
63.6c
49.8c
16.2c
15.2c
10.0c
5.1c
800.0
1000.0
ndd
nd
)-catechin
)-catechin-6-C-
)-catechin-6-C-
)-catechin-8-C-
)-catechin-8-C-
â
â
â
â
-
-
-
-
D
D
D
D
-galactopyranoside (9)
-glucopyranoside (3)
-galactopyranoside (7)
-glucopyranoside (1)
nd
nd
)-epicatechin-8-C-
â
â
-
-
D
-galactopyranoside (11)
-glucopyranoside (6)
nd
)-epicatechin-6-C-
D
nd
)-epicatechin-6-C,8-C-
â
-
D
-digalactopyranoside (10)
-glucopyranoside (4)
)-epicatechin-6-C,8-C- -diglucopyranoside (5)
nd
)-epicatechin-8-C-
â
-
D
3.1c
nd
â
-
D
2.5c
nd
)-catechin-6-C,8-C-
)-catechin-6-C,8-C-
â
-D
-digalactopyranoside (8)
-diglucopyranoside (2)
1.3c
nd
â
-D
1.1c
nd
a Taste threshold concentrations were determined by means of the half-tongue
test (7). b The astringent sensation was described as rough and extremely puckering.
c The astringent sensation was described as
mouthcoating. d Not detectable.
a very smooth and velvety
The data obtained for the flavan-3-ol-C-glycosides clearly
show that the sensory activity changes with variations in the
stereochemistry of the flavan-3-ol moiety and the position of
the linkage between aglycone and carbohydrate as well as in
the glycosylation pattern, thereby illustrating that oral thresholds
Besides the structure of the sugar, the aglycone moiety has a
significant influence on the perception of astringency. Com-
parison of (-)-epicatechin-8-C-â-D-glucopyranoside (4) and
(-)-epicatechin-6-C-â-D-glucopyranoside (6) with the corre-
Table 4. Influence of Flavan-3-ol-C-glycosides on the Taste Profile of a Cocoa Beveragea
additiveb
PDDc
taste profiled
typical cocoa-like bitterness
no additive (control)
(−
(−
(−
(−
(−
(−
)-catechin
)-catechin
)-catechin
)-catechin
)-catechin
)-catechin
−
−
−
−
−
−
8-C-
6-C-
8-C-
6-C-
6,8-C-
6,8-C-
â
â
â
â
-
-
-
-
â
â
D
D
D
D
-glcp (1)
-glcp (3)
-galp (7)
-galp (9)
6
6
6
6
8
8
softer, more pleasant, less bitter (++) than control
softer, more pleasant, less bitter (++) than control
softer, more pleasant, less bitter (++) than control
softer, more pleasant, less bitter (++) than control
very mild and pleasant, less bitter (+++) than control
very mild and pleasant, less bitter (+++) than control
-
-
D
-digalp (8)
-diglcp (2)
D
a The cocoa beverage was freshly prepared from cocoa powder (25 g/L) and bottled water stirred for 15 min at 50
mol/L) was used as additive. c Number of panelists detecting the difference from the control. Eight panelists were asked to compare the cocoa beverage containing no
additive (control) with the cocoa beverage spiked with an individual flavan-3-ol-C-glycoside by means of a triangle test. d If the sample was detected correctly (cf. footnote
c), the changes in taste and taste intensities were evaluated as being weak ( ), medium strong (++), and strong (+++).
°
C prior to use. b The individual C-glycoside (111
µ
+
Table 5. Influence of the Polyphenol Structure on the Bitterness of an Aqueous Theobromine Solution
additive
PDDa
taste (intensity)b
no additive (control)
bitter (5.0)
(
(
(
−
+
−
)-catechin-8-C-
2.2 mol/L
22.2 mol/L
222.2 mol/L
)-catechin
2.2 mol/L
22.2 mol/L
222.2 mol/L
)-epicatechin
2.2 mol/L
22.2 mol/L
222.2 mol/L
apigenin-8-C-
2.2 mol/L
22.2 mol/L
222.2 mol/L
apigenin 6-C-
2.2 mol/L
22.2 mol/L
222.2 mol/L
â-D-glcp (1)
µ
1
5
8
softer, more pleasant, and less intense bitterness (4.5)
softer, more pleasant, and less intense bitterness (3.5)
very mild, pleasant, and weakly bitter (1.0)
µ
µ
µ
µ
0
4
4
bitter (5.0)
more cocoa-like, but intensely bitter (5.0)
more cocoa-like, but rather bitter (4.0)
µ
µ
µ
0
4
4
bitter (5.0)
more cocoa-like, but intensely bitter (5.0)
more cocoa-like, but intensely bitter (4.5)
µ
â-D-glucopyranoside
µ
µ
0
0
4
bitter (5.0)
bitter (5.0)
bitter (4.5)
µ
−
µ
µ
â-D-glucopyranoside
0
0
4
bitter (5.0)
bitter (5.0)
bitter (4.5)
µ
a Number of panelists detecting the difference from the control. Eight panelists were asked to compare a solution of theobromine (3 mmol/L) with the a solution of
theobromine (3 mmol/L) spiked with different amounts of individual additives. b If the sample was detected correctly (cf. footnote a), the changes in taste and taste intensities
were evaluated in comparison to the control.

Structure Determination of Flavan-3-ol-C-hexosides in Cocoa
J. Agric. Food Chem., Vol. 54, No. 25, 2006 9521
determination of taste-active polyphenols. J. Agric. Food Chem.
2005, 53, 5407-5418.
(2) Stark, T.; Hofmann, T. Isolation, structure determination,
synthesis, and sensory activity of N-phenylpropenoyl-L-amino
acids from cocoa (Theobroma cacao). J. Agric. Food Chem.
2005, 53, 5419-5428.
(3) Stark, T.; Hofmann, T. Structures, sensory activity, and dose/
response functions of 2,5-diketopiperazines in roasted cocoa nibs
(Theobroma cacao). J. Agric. Food Chem. 2005, 53, 7222-7231.
(4) Stark, T.; Justus, H.; Hofmann, T. Quantitative analysis of
N-phenylpropenoyl-L-amino acids in roasted coffee and cocoa
powder by means of a stable isotope dilution assay. J. Agric.
Food Chem. 2006, 54, 2859-2867.
(5) Stark, T.; Bareuther, S.; Hofmann, T. Molecular definition of
the taste of roasted cocoa nibs (Theobroma cacao) by means of
quantitative studies and sensory experiments. J. Agric. Food
Chem. 2006, 54, 5530-5539.
(6) Wieser, H.; Belitz, H. D. Studies on the structure/activity
relationship of bitter tasting amino acids and peptides (in
German). Z. Lebensm. Unters. Forsch. 1975, 159, 65-72.
(7) Scharbert, S.; Holzmann, N.; Hofmann, T. Identification of the
astringent taste compounds in black tea infusions by combining
instrumental analysis and human bioresponse. J. Agric. Food
Chem. 2004, 52, 3498-3508.
(8) Frank, O.; Ottinger, H.; Hofmann, T. Characterization of an
intense bitter-tasting 1H,4H-quinolizinium-7-olate by application
of the taste dilution analysis, a novel bioassay for the screening
and identification of taste-active compounds in foods. J. Agric.
Food Chem. 2001, 49, 231-238.
(9) Courbat, P.; Weith, A.; Albert, A. Contribution a l’etude du
comportement de la catechin en milieu alcalin. HelV. Chim. Acta
1977, 60, 1665-1675.
(10) Altona, C.; Haasnoot, C. A. G. Prediction of anti and gauche
vicinal proton-proton coupling constants in carbohydrates: a
simple additivity rule for pyranose rings. Org. Magn. Reson.
1980, 13, 417-428.
of astringent compounds cannot be predicted from chemical
structures but have to be investigated on the basis of systematic
sensory studies with purified reference compounds.
Taste-Modifying Activity of Flavan-3-ol-C-glycosides. To
directly evaluate the influence of purified flavan-3-ol-C-
glycosides on the taste of cocoa powder, in the following, flavan-
3-ol-C-glycosides were added to a cocoa beverage and aqueous
theobromine solutions, respectively, and the taste profiles of
these solutions were compared to those of the corresponding
nonspiked samples as the control (Tables 4 and 5).
As summarized in Table 4, the addition of 111 µmol/L of
one of the mono-C-glycosides 1, 3, 7, or 9, respectively, to the
cocoa beverage was detected by six of eight panelists. Substitu-
tion of the mono-C-glycosides by the di-C-glycosides 2 and 8,
respectively, could be detected by all of the panelists. The
individual panelists reported that the bitterness of the cocoa
beverage was perceived as being significantly milder, softer,
and more pleasant in the presence of the C-glycosides.
To answer the question of whether the taste-modifying
activity of flavan-3-ol-C-glycoside is due to an interaction with
the theobromine perception, an aqueous solution of theobromine
(3 mmol/L) was spiked with increasing concentrations of (-)-
catechin-8-C-â-D-glucopyranoside (1) (Table 5). Spiking a
solution of theobromine (3 mmol/L) with an amount of 222
µmol of 1 was detected by all of the panelists and induced a
significant decrease of the bitterness (5.0 f 1.0) of theobromine,
thus indicating that the bitterness-mellowing effect of the
glycosides is due to a partial bitterness suppression of theobro-
mine. The addition of only 22 or 2.2 µmol of the glycoside
could be perceived by just five or one of eight panelists.
To investigate the influence of the polyphenol structure on
their modifying effect on theobromine bitter taste, the (-)-
catechin-8-C-â-D-glucopyranoside (1) was substituted by (+)-
catechin, (-)-epicatechin, apigenin-8-C-â-D-glucopyranoside,
and apigenin-6-C-â-D-glucopyranoside, respectively. As given
in Table 5, the addition of either (+)-catechin or (-)-epicatechin
to the theobromine at a concentration of 222.2 µmol/L was
detectable by just four of eight panelists, who described the
taste difference as a more “cocoa-like bitterness” with somewhat
lower bitter taste intensity (5.0 f 4.0). In comparison, apigenin-
8-C-â-D-glucopyranoside and apigenin-6-C-â-D-glucopyranoside
did not show any significant influence on the bitterness of the
theobromine.
(11) Markham, K. R.; Geiger, H. 1H nuclear magnetic resonance
spectroscopy of flavanoids and their glycosides in hexadeuteri-
odimethylsulfoxide. In The FlaVanoids; Harborne, J. B., Ed.;
Chapman and Hall: London, U.K., 1994; pp 441-497.
(12) Korver, O.; Wilkins, C. K. Circular dichroism spectra of
flavanols. Tetrahedron 1971, 27, 5459-5465.
(13) Isbell, H. S.; Frush, H. L.; Wade, C. W. R.; Hunter, C. E.
Transformations of sugars in alkaline solutions. Carbohydr. Res.
1968, 9, 163-175.
In summary, this is the first report demonstrating that
alkalization of cocoa is inducing the nonenzymatic C-glycosy-
lation of flavan-3-ols to give the newly identified flavan-3-ol-
C-glycosides. Human sensory experiments revealed that these
glycosides modify the bitter taste profile and decrease the bitter
taste intensity of cocoa beverages as well as theobromine
solutions, whereas the aglycones or other C-glycosides, such
as apigenin-8-C-â-D-glucopyranoside and apigenin-6-C-â-D-
glucopyranoside, did not show any significant activity. Aimed
at gaining a more comprehensive understanding of the sensory
contribution of the flavan-3-ol-C-glycosides to the typical
velvety astringent and mild bitter taste of alkalized cocoa
powder, the development of a stable isotope dilution analysis
is currently under progress.
(14) Zhou, Z. H.; Zhang, Y. J.; Xu, M.; Yang, C. R. Puerins A and
B, two new 8-C substituted flavan-3-ols from Pu-er tea. J. Agric.
Food Chem. 2005, 53, 8614-8617.
(15) Morimoto, S.; Nonaka, G. I.; Nishioka, I. Tannins and related
compounds. XXXVIII. Isolation and characterization of flavan-
3-ol glucosides and procyanidin oligomers from Cassia bark
(Cinnamomum cassia Blume). Chem. Pharm. Bull. 1986, 34,
633-642.
(16) Tsutomu, H.; Haruka, M.; Midori, N.; Naomi, O.; Toshio, T.;
Hideyuki, I.; Takashi, Y. Proanthocyanidin glycosides and related
polyphenols from cacao liquor and their antioxidant effect.
Phytochemistry 2002, 59, 749-758.
Received for review August 21, 2006. Revised manuscript received
October 6, 2006. Accepted October 6, 2006. This research project was
supported by Kraft Foods as well as by a grant from the Fonds der
Chemischen Industrie (FCI).
LITERATURE CITED
(1) Stark, T.; Bareuther, S.; Hofmann, T. Sensory-guided decom-
position of roasted cocoa nibs (Theobroma cacoa) and structure
JF062403+