Contribution of each caffeoyl residue of the pigment molecule of gentiodelphin to blue color development

Article abstract of DOI:10.1016/S0031-9422(00)00049-2

To clarify the function of each caffeoyl residue in the diacylated anthocyanin gentiodelphin, a pigment from the blue flower of Gentiana makinoi, two mono-deacyl derivatives were compared for both color development and stability. In neutral solution, 3,5-di-O-β-D-glucopyranosyl-3'-O-(6-O- caffeoyl-β-D-glucopyranosyl)delphinidin was both bluer and more stable than 3,3'-di-O-β-D-glucopyranosyl-5-O-(6-O-caffeoyl-β-D- glucopyranosyl)delphinidin. Conformational analysis of each derivative under acidic conditions revealed only the 3'-O-caffeoylglucopyranosyl derivative to demonstrate intramolecular stacking. Additionally, the acyl residue in the B- ring contributed more to blue color development than that in the A-ring. (C) 2000 Elsevier Science Ltd.

Full text of DOI:10.1016/S0031-9422(00)00049-2

Phytochemistry 54 (2000) 85±92
www.elsevier.com/locate/phytochem
Contribution of each ca€eoyl residue of the pigment molecule of
gentiodelphin to blue color development
Kumi Yoshidaa,*, Yuki Toyamab, Kiyoshi Kamedaa, Tadao Kondoc
aSchool of Life Studies, Sugiyama Jogakuen University, Hoshigaoka, Chikusa, Nagoya 464-8662, Japan
bGraduate School of Human Informatics, Nagoya University, Chikusa, Nagoya 464-8601, Japan
cChemical Instrument Center, Nagoya University, Chikusa, Nagoya 464-8602, Japan
Received 19 August 1999; received in revised form 3 January 2000
Abstract
To clarify the function of each ca€eoyl residue in the diacylated anthocyanin gentiodelphin, a pigment from the blue ¯ower of
Gentiana makinoi, two mono-deacyl derivatives were compared for both color development and stability. In neutral solution,
3,5-di-O-b-D-glucopyranosyl-3'-O-(6-O-ca€eoyl-b-D-glucopyranosyl)delphinidin was both bluer and more stable than 3,3'-di-O-b-
D-glucopyranosyl-5-O-(6-O-ca€eoyl-b-D-glucopyranosyl)delphinidin. Conformational analysis of each derivative under acidic
conditions revealed only the 3'-O-ca€eoylglucopyranosyl derivative to demonstrate intramolecular stacking. Additionally, the
acyl residue in the B-ring contributed more to blue color development than that in the A-ring. 7 2000 Elsevier Science Ltd. All
rights reserved.
Keywords: Gentiana makinoi; Gentiodelphin; Acylated anthocyanin; Intramolecular stacking; Flower color development; Conformational analysis
1. Introduction
portant role in both color development and stability
(Brouillard, 1981, 1982; Goto et al., 1982; Goto and
Kondo, 1991; Figueiredo et al., 1996). It has been
suggested that the latter is due to intramolecular stack-
ing of the aromatic acid to the anthocyanidin nucleus
by hydrophobic interactions (Hoshino et al., 1981;
Goto et al., 1982; Hoshino et al., 1982; Brouillard,
1982; Goto, 1987; Goto and Kondo, 1991; Yoshida et
al., 1991).
Polyacylated anthocyanins, which contain two or
more aromatic acid residues, are a unique group of
¯ower pigments (Goto et al., 1982; Goto et al., 1986;
Goto, 1987; Brouillard, 1988; Goto and Kondo, 1991;
Brouillard and Dangles, 1994). They are very stable in
neutral or weakly acidic aqueous solutions (Saito et
al., 1971; Yoshitama and Hayashi, 1974; Goto et al.,
1982; Goto and Kondo, 1991), whereas simple antho-
cyanins are quickly decolorized by hydration at the 2-
position of the anthocyanidin nucleus (Brouillard and
Dubois, 1977; Brouillard and Delaporte, 1977; Brouil-
lard, 1982; Mazza and Brouillard, 1987). Since polya-
cylated anthocyanins become unstable on removal of
acyl residues, these aromatic acids must play an im-
We have proposed a folding conformation for poly-
acylated anthocyanins and provided evidence of an
intramolecular stacking conformation for gentiodel-
phin (1), a dica€eoyl anthocyanin (Goto et al. 1982;
Yoshida et al., 1992). In acidic methanol, only the caf-
feic acid attached to the glucose of B-ring is stacked to
the anthocyanidin nucleus with both aromatic rings in
parallel, strongly suggesting that the roles of each acyl
residue di€er in intramolecular stacking conformation
and ¯ower color development. However, only kinetic
and equilibrium studies of structurally diverse monoa-
cyl anthocyanins from Pharbitis nil have thus far been
* Corresponding author. Tel.: +81-052-781-1186; fax: +81-052-
782-7265. Present address: Graduate School of Human Informatics,
Nagoya University, Nagoya 464-8601, Japan.
E-mail address: yoshida@cic.nagoya-u.ac.jp (K. Yoshida).
0031-9422/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved.
PII: S0031-9422(00)00049-2

86
K. Yoshida et al. / Phytochemistry 54 (2000) 85±92

K. Yoshida et al. / Phytochemistry 54 (2000) 85±92
87
reported, and no variation in the contribution to color
development was established (Dangles et al., 1993).
We therefore prepared two kinds of mono-deacyl de-
rivatives from 1, in order to permit a direct compari-
son of their respective contributions to blue color
development and stability, and also undertook various
NMR spectroscopic experiments in order to assess
their respective conformations and degree of intramo-
lecular stacking.
2. Results and discussion
Gentiodelphin (1) was puri®ed from blue petals of
Gentiana makinoi according to the procedure described
previously (Yoshida et al., 1992) with slight modi®-
cation. By controlled alkaline hydrolysis, 1 (52.9 mg)
gave a reaction mixture consisting of three products
(2±4) and unreacted 1. Using preparative HPLC, the
reaction mixture was separated to give 2 (4.3 mg), 3
(7.5 mg), 4 (10.0 mg) and recovered gentiodelphin (1,
4.0 mg). Compound 4 was deduced to be a fully deacy-
lated anthocyanin by co-chromatography with an
authentic sample (Yoshida et al., 1992).
5
Fig. 1. UV±Vis spectra of 1±4 in aqueous solutions (5 Â 10 M in
0.1 M phosphate bu€er at pH 6.5, path length; 10 mm). ÐÐ: 1, ± Á
± Á ±: 2, - - - - : 3, Á Á Á Á 4.
radiation at each anomeric proton (Table 1). In 3 the
6-position of glucose Q was esteri®ed, because of the
lower shift of the methylene protons (H-6a and 6b)
of glucose Q. Thus, the structure of 3 was deduced to
be 3,5-di-O-b-D-glucopyranosyl-3'-O-(6-O-ca€eoyl-b-D-
glucopyranosyl)delphinidin.
High resolution FABMS of 2 and 3 gave the same
molecular formula of C42H47O25, corresponding to loss
of one ca€eoyl residue from gentiodelphin. By com-
parison of its NMR spectroscopic data with those in
literature, 2 was identi®ed to be albireodelphin A, pre-
viously isolated from blue ¯owers of G. albireo (Hoso-
kawa et al., 1997; Table 1). To unambiguously
determine the position of the ca€eoyl residue, all the
1H-NMR spectral signals of 3 were assigned by 1D-
HOHAHA and NOE di€erence spectra with ir-
Compounds 1±4 were individually dissolved in aqu-
5
eous bu€er solution (5  10 M, pH 6.5 in 0.1 M
phosphate bu€er) and then the UV±Vis spectra of
each were recorded (Fig. 1). Compound 3 showed a
similar blue color as that of 1, whereas the color of 2
and 4 in the bu€er solution (pH 6.5) was purplish
blue. The spectrum of 3 but not 2 was similar to that
of 1. Stability of the color in aqueous solution was
recorded by monitoring the decrease in absorbance at
5
lmax (520±530 nm) at 5 Â 10
M (Fig. 2). Gentiodel-
phin (1) was stable while the bis-deca€eoyl pigment (4)
Fig. 2. Stability of 1 (*), 2 (Q), 3 (R) and 4 (^) in aqueous sol-
5
utions at 258C (5 Â 10
M in 0.1 M phosphate bu€er at pH 6.5,
path length: 10 mm). After half month, 1 and 3 retained color, but 2
an 4 were decolorized.

88
K. Yoshida et al. / Phytochemistry 54 (2000) 85±92
Table 2
Ratios of gg, gt an tg rotamers at the exo-cyclic C5±C6 bond of
each glucoside in 1±4a
residue on the 5-O-glucoside. Di€erences in stacking
were concluded to be responsible for the variation in
both the color and the stability of 2 and 3.
To clarify the conformation of the aromatic acid
residue and the anthocyanidin nucleus in 2 and 3, a
1H-NMR spectroscopic study was carried out. Since
compounds 1±4 were slightly soluble in the bu€er (pH
Compound
Glucoside
% of each rotamer
gg
gt
tg
1
2
3
4
R
*
Q
R
*
Q
R
*
Q
R
*
Q
45
49
22
49
45
54
46
55
24
45
58
58
55
51
82
51
55
46
54
45
75
55
42
42
0
0
4
0
0
0
0
0
1
0
0
0
1
6.5), H-NMR spectra were measured in 10% TFAd-
CD3OD as described in a previous report (Yoshida et
al., 1992). Shift of proton signals toward higher ®eld
under these conditions indicates the existence of face-
to-face stacking with two aromatic rings in parallel
(Goto et al., 1982, 1986; Goto, 1987). The signals of
the delphinidin nucleus of 3 were shifted up®eld 0.02±
0.3 ppm more than that of 4, while those of 2 were
una€ected (Table 1). The signals of the ca€eic acid
residue of 3 were also shifted 0.35±0.74 ppm up®eld as
compared with those of methyl E-ca€eate as a result
of the large ring current (chemical shift of methyl E-
ca€eate in TFAd-CD3OD: H-a 6.28 ppm, H-b 7.56
ppm, H-2; 7.07 ppm, H-5; 6.78 ppm, H-6; 6.95 ppm),
while those of 2 were not shifted. Therefore, the caf-
feoyl residue of 3 should stack face-to-face to the
anthocyanidin nucleus, while that of 2 should not. The
conformation was also analyzed by 1H-NOE. In all
pigments, 1±4, strong NOEs were observed between
a NMR spectral data were measured in acidic methanol solutions
and the ratios were calculated by the modi®ed Karplus equation: J5,
6proS= 1.3 gg + 2.7 gt + 11.7 tg, J5, 6proR = 1.33 gg + 11.5 gt +
5.8 tg, gg + gt + tg = 1.
was quickly decolorized. Interestingly, the stabilities of
the monoca€eoyl derivatives, 2 and 3, were di€erent.
Compound 3 with a ca€eoyl residue on the 3-O-gluco-
side was more stable than compound 2 with a ca€eoyl
Fig. 3. Torsional bonds of 2 and 3 during molecular modeling.

K. Yoshida et al. / Phytochemistry 54 (2000) 85±92
89
Fig. 4. Stereo-views (parallel view) of the optimized conformers in acidic methanol, obtained by molecular modeling. Upper: optimized confor-
mer of 2. Lower: optimize conformer of 3.
R-1 and H-4, *-1 and H-6, and Q-1 and H-2', re-
spectively, but no NOE was detected between R-1 and
H-2', 6', or *-1 and H-4, indicating that the anomeric
linkage conformation of each glucoside was ®xed.
Since weak NOEs give distance information useful for
analyzing molecular stacking interactions, the NOE
di€erence spectra of 2 and 3 were measured at 108C
and the long distance NOEs between the signals of
chromophore and ca€eic acid residue were analyzed
(Table 1). NOEs of negative 1±2% were observed for
compound 3 between the signals of the nucleus, H-4,
H-8, H-2' and H-6', and the signals of the ca€eic acid
residue, indicating an intramolecular stacking confor-
mation, but no long distance NOE was observed in 2.
From the coupling constant between H-5 and H-6 of
glucoside, J5, 6a and J5, 6b, the ratio of each rotamer of
the exo-cyclic C5±C6 bonds was calculated (Table 2,
Bruyn and Anteunis, 1976; Nishida et al., 1988). The
gt rotamer was predominant in the Q glucose of 3,
similar to that in 1. However, the other glucosides, R,
* and Q in 2 and R and * in 3, showed equal rota-
mer ratios ( gg:gt = ca. 1:1), therefore, acylation of 6-
OH of the glucoside Q forms the C5±C6 bond gt con-
formation. These results are consistent with those
reported for gentiodelphin (1) (Yoshida et al., 1992).
Thus, the ca€eic acid residue in the B-ring has the
greater contribution to formation of face-to-face intra-
molecular stacking.
To further investigate the conformations of 2 and
3, computer-assisted molecular modeling was per-
formed using QUANTA97/CHARMm software with
various constraints deduced from NMR spectro-
scopic analysis. Number of 5158 initial conformers
that were generated by grid rotation of the bonds
(Fig. 3), and then each conformer was optimized by
conjugate gradient energy minimization processes

90
K. Yoshida et al. / Phytochemistry 54 (2000) 85±92
using various constraints. Bond angle constraints
were adopted at the exo-cyclic C5±C6 bonds of all
the glucoside residues following the rotamer ratios
shown in Table 2. Since the dihedral angle between
the A- and B-ring, and the glucosidic bonds of R-
glucose were not so diverse by analysis of the
NMR spectroscopic data, those dihedral angles were
®xed at the same conformation as that obtained
from the X-ray data of commelinin (Kondo et al.,
1992). Between those protons for which a strong
NOE was observed, a 2.520.5 A distance constraint
was adopted, and between the protons for which a
weak NOE was observed, a 4.0 21.0 A distance con-
straint was adopted. From all the generated and mini-
mized conformers, 1000 numbers were selected, then
the Adopted-Basis Newton±Raphson method was used
to obtain an optimized conformation. As shown in
Fig. 4, the ca€eoyl residue of 3 stacked to the chromo-
phore at a distance approaching the van der Waal's
distance. On the other hand, the ca€eic acid residue in
2 had no interaction with the chromophore nucleus.
To determine whether the ca€eoyl residue of 2 stacks
to the nucleus in water, we carried out molecular mod-
eling with the dielectric constant e ˆ 70 for water,
using the same procedure described above. Under
these conditions the acyl moiety in 2 also stacked to
the nucleus with two aromatic rings in parallel from
the front side (Fig. 5).
that both the acyl moieties of 1 stack to the nucleus,
but the stacking manner of the 3'-acyl residue was
much closer to the nucleus than that of the 5-acyl resi-
due. In gentiodelphin 1, the contribution of each acyl
moiety to blue color development and stability is
di€erent and must be estimated by linear calculation.
3. Experimental
3.1. General procedures
UV±Vis spectra were recorded on a JASCO Ubest-
55 spectrometer. NMR spectra were obtained with a
JEOL GX500 spectrometer (1H: 500 MHz, 13C: 125
MHz) in a 5 mm f tube at variable temperature, using
10% TFAd±CD3OD as solvent. Chemical shifts were
recorded as parts per million (ppm) with the CD2HOD
resonance set at 3.325 ppm as a standard. FABMS
data were recorded on a JEOL MStation using gly-
cerol±HCl as the matrix, and ESIMS data were
recorded on a JEOL JMS-700 using the ¯ow injection
mode. Analytical and preparative HPLC were carried
out using an ODS-column (Develosil ODS-HG5 4.6
mm f  250 mm, 20 mm f  250 mm and 50 mm f Â
250 mm, Nomura Chemical). HPLC was performed at
408C with linear gradient elution from 10 to 30% aq.
acetonitrile (CH3CN) solution containing 0.5% TFA
in 30 min, or isocratic elution with various concen-
trations of aq. CH3CN solution containing 0.5% TFA.
As shown by the UV±Vis spectra and stability in
aqueous solution (pH 6.5), the contribution of each
ca€eoyl residue to blue color development is di€erent.
The mono-O-ca€eoyl anthocyanin, with its acyl moiety
on the glucoside in the B-ring, showed a blue color
and was relatively stable under neutral conditions.
Although the evidence for stacking conformation was
obtained in acidic methanol only, molecular modeling
studies in neutral aqueous and acidic methanol con-
ditions suggest the same result. The di€erence of stab-
ility among 1±3 in neutral aqueous solution suggests
3.2. Isolation of pigments
From fresh blue petals of G. makinoi, crude pig-
ments were obtained using the procedure described
previously (Yoshida et al., 1992). Crude pigment (9.6
g) was dissolved in 1.5 l of water containing 0.5%
TFA, then poured onto an Amberlite XAD-7 column
(50 mm f  300 mm). The column was stepwise eluted
Fig. 5. Stereo-view (parallel view) of the optimized conformers of 2 obtained by molecular modeling in water.

K. Yoshida et al. / Phytochemistry 54 (2000) 85±92
91
from 20 to 60% aq. MeOH containing 0.5% TFA.
The 30±50% MeOH fraction contained the major pig-
ments. The fraction was evaporated in vacuo, dis-
solved into 0.5% aq. TFA solution, and then the
solution was subjected to chromatography using a pre-
parative ODS±HPLC which was eluted with 5±50%
aq. CH3CN solution containing 0.5% TFA. Gentiodel-
phin (1) was eluted in the 19% aq. CH3CN fraction.
The eluent was dried in vacuo, yielding 1 (319 mg) as
a dark-red amorphous TFA salt.
O2/R10000 workstation (Silicon Graphics) using the
molecular modeling package QUANTA97/CHARMm
23.2, with dielectric constants, e ˆ 32:6 in methanol
and e ˆ 70:0 in water. The starting conformation was
constructed using the X-ray data of malonylawobanin
in commelinin (Kondo et al., 1992; Kondo et al.,
unpublished results). The conformation search for 2
was carried out using a grid scan routine with rotation
of the torsional angles, C0, F0, T1 and T2 by 608
increments, and rotation of T3 and T4 by 1808. All the
generated conformers were minimized by conjugate
gradient energy minimization until energy gradient tol-
erance was equal to 0.05. During minimization, dihe-
dral constraints gg at C5±C6 bond of glucose *, and
NOE constraints 2.5 2 0.5 A between the anomeric
protons and the nucleus protons were adopted.
Among the minimized conformers, 1000 from the low-
est energy level were selected and then the energy mini-
mization for each was achieved using Adopted-Basis
Newton±Raphson methods without any constraints
but T9 to obtain the optimized conformer. The confor-
mation search for 3 was performed with rotation of
the torsional angles, C00, F00, T5 and T6 by 608 incre-
ments, and rotate T7 and T8 by 1808. During minimiz-
ation, dihedral constraints gt at the C5±C6 bond of
glucose Q, and NOE constraints 2.5 2 0.5 A between
the anomeric protons and the nucleus protons and 4.0
2 1.0 A between the protons for which weak NOEs
were observed, were adopted using the same pro-
cedure.
3.3. Alkaline hydrolysis
Compound 1 (52.9 mg, 0.043 mmol) was dissolved
in MeOH (1.25 ml) at room temperature, bubbled with
argon gas (Ar), then cooled to 08C. To this solution,
1.0 N NaOH (1.25 ml) was added, and the mixture
was allowed to stand at 08C under Ar atmosphere for
30 min. To the reaction mixture, 6 N aq. HCl was
added at 08C, then the solution was warmed to room
temperature. The mixture was evaporated in vacuo to
1/10 volume and diluted ®ve fold with water, then sub-
jected to preparative ODS±HPLC (20 mm f  250
mm). The column was eluted from 10 to 50% aq.
CH3CN solution containing 0.5% TFA. Compounds
4, 3, 2 and 1 were eluted in 10, 12, 15 and 19%
CH3CN fractions, respectively. Fractions were evapor-
ated in vacuo to give 1 (4.0 mg, 7.6%), 2 (4.3 mg,
9.4%), 3 (7.5 mg, 16.4%) and 4 (10.0 mg, 25.8%).
3.3.1. 3,3'-Di-O-b-D-glucopyranosyl-5-O-(6-O-ca€eoyl-
b-D-glucopyranosyl)delphinidin (albireodelphin A, 2)
UV±Vis (0.01% HCl±MeOH) nm (e: absorption
coecient): 279 (14,300), 296 (12,700), 331 (12,200),
529 (21,700); HR-FABMS calcd. for C42H47O25 m/z
951.2406; found m/z: 951.2421 (+1.5 mmu).
Acknowledgements
The authors thank Mr. U. Maeda, Chemical Instru-
ment Center, Nagoya University for help with molecu-
lar modeling. This study was ®nancially supported by
the Ministry of Education, Science, Sports and Culture
of Japan with a Grant-in-Aid for Scienti®c Research
on Priority Areas No. 08219237, Scienti®c Research
(C) No. 09660147, COE Research No. 07CE2004, and
a Grant-in-Aid from the Ministry of Agriculture, For-
estry and Fisheries of Japan.
3.3.2. 3,5-Di-O-b-D-glucopyranosyl-3'-O-(6-O-ca€eoyl-
b-D-glucopyranosyl)delphinidin (3)
UV±Vis (0.01% HCl±MeOH) nm (e): 279 (18,600),
297 (15,800), 332 (15,200), 537 (30,100); HR-ESIMS
calcd. for C42H47O25 m/z 951.2406; found m/z:
951.2399 ( 0.7 mmu).
3.4. Absorption spectra
References
Each isolated anthocyanin (TFA salt) was dissolved
2
in water at a concentration of 5 Â 10 M. Individual
Brouillard, R., 1981. Phytochemistry 20, 143.
Brouillard, R., 1982. Anthocyanins as food colors, P.1. Academic
Press, New York.
5
solutions were diluted to 5 Â 10 M with 0.1 M phos-
phate bu€er (pH 6.5), with UV±Vis spectra sub-
sequently measured in a quartz cell (d = 10 mm) at
258C.
Brouillard, R., 1988. The ¯avonoids, advances in research since
1980. Chapman & Hall, London, p. 525.
Brouillard, R., Dangles, O., 1994. The ¯avonoids, advances in
research since 1986. Chapman & Hall, London, p. 565.
Brouillard, R., Delaporte, B., 1977. J. Am. Chem. Soc. 99, 8461.
Brouillard, R., Dubois, J.-E., 1977. J. Am. Chem. Soc. 99, 1359.
Bruyn, A., Anteunis, M., 1976. Carbohydrate Res. 47, 311.
3.5. Molecular modeling
Conformational analysis was performed on an IRIS

92
K. Yoshida et al. / Phytochemistry 54 (2000) 85±92
Hoshino, T., Matsumoto, U., Goto, T., Harada, N., 1982.
Tetrahedron Letters 23, 433.
Dangles, O., Saito, N., Brouillard, R., 1993. J. Am. Chem. Soc. 115,
3125.
Hoshino, T., Matsumoto, U., Harada, N., Goto, T., 1981.
Tetrahedron Letters 22, 3621.
Figueiredo, P., Elhabiri, M., Toki, K., Saito, N., Dangles, O.,
Brouillard, R., 1996. Phytochemistry 41, 301.
Kondo, T., Yoshida, K., Nakagawa, A., Tamura, H., Kawai, T.,
Goto, T., 1992. Nature 358, 515.
Goto, T., 1987. Structure. Progress in the Chemistry of Organic
Natural Products 52, 113.
Mazza, G., Brouillard, R., 1987. J. Agric. Food Chem 35, 422.
Nishida, N., Hori, Y., Ohrui, H., Meguro, H., 1988. J. Carbohydrate
Chem. 7, 239.
Goto, T., Kondo, T., 1991. Angew. Chem. Int. Ed. Engl 30, 17.
Goto, T., Kondo, T., Tamura, H., Imagawa, H., Iino, A., Takeda,
K., 1982. Tetrahedron Letters 23, 3695.
Saito, N., Osawa, Y., Hayashi, K., 1971. Phytochemistry 10, 445.
Yoshitama, K., Hayashi, K., 1974. Bot. Mag. Tokyo 87, 33.
Yoshida, K., Kondo, T., Goto, T., 1991. Tetrahedron Letters 32,
5579.
Goto, T., Tamura, H., Kawai, T., Hoshino, T., Harada, N., Kondo,
T., 1986. Ann. N.Y. Acad. Sci. 71, 155.
Hosokawa, K., Fukushi, E., Kawabata, J., Fujii, C., Ito, T.,
Yamamura, S., 1997. Phytochemistry 45, 167.
Yoshida, K., Kondo, T., Goto, T., 1992. Tetrahedron 48, 4313.