2-O-β-D-Glucopyranosyl-carboxyatractyligenin from Coffea L. inhibits adenine nucleotide translocase in isolated mitochondria but is quantitatively degraded during coffee roasting

Article abstract of DOI:10.1016/j.phytochem.2013.03.022

Atractyloside (1) and carboxyatractyloside (2) are well-known inhibitors of the adenine nucleotide translocase (ANT) in mitochondria, thus effectively blocking oxidative phosphorylation. Structurally related derivatives atractyligenin (3), 2-O-b-D-glucopyranosyl-atractyligenin (4), 3′-O-β-D-glucopyranosyl-2′-Oisovaleryl-2β-(2-desoxy- atractyligenin)-β-D-glucopyranoside (5), and 2-O-β-D-glucopyranosyl- carboxyatractyligenin (6) were isolated from raw beans of Coffea L. and the impact of 1-6 on ANT activity was evaluated in isolated mitochondria. Among the coffee components, 6 significantly inhibited ANT activity leading to reduced respiration. Quantitative analysis in commercial coffees, experimental roastings of coffee, and model experiments using purified compound 6 consistently revealed a complete degradation during thermal treatment. In comparison, raw coffee extracts were found to contain high levels of 6, which are therefore expected to be present in food products enriched with raw coffee extracts. This implies the necessity of analytically controlling the levels of 6 in raw coffee extracts when used as additives for food products.

Full text of DOI:10.1016/j.phytochem.2013.03.022

Phytochemistry 93 (2013) 124–135
Contents lists available at SciVerse ScienceDirect
Phytochemistry
journal homepage: www.elsevier.com/locate/phytochem
2-O-b-D-Glucopyranosyl-carboxyatractyligenin from Coffea L. inhibits adenine
nucleotide translocase in isolated mitochondria but is quantitatively degraded
during coffee roasting
Roman Lang a, Tobias Fromme c, Anja Beusch a, Anika Wahl a, Martin Klingenspor c, Thomas Hofmann a,b,
⇑
a Chair of Food Science and Molecular Sensory Science, Technische Universität München, Lise-Meitner-Str. 34, 85354 Freising, Germany
b ZIEL Bioanalytics Unit, Technische Universität München, Alte Akademie 10, 85354 Freising, Germany
c ZIEL Unit Molecular Nutritional Medicine, Technische Universität München, Gregor-Mendel-Str. 2, 85350 Freising, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
Atractyloside (1) and carboxyatractyloside (2) are well-known inhibitors of the adenine nucleotide trans-
Received 6 January 2013
Received in revised form 19 March 2013
Available online 2 May 2013
locase (ANT) in mitochondria, thus effectively blocking oxidative phosphorylation. Structurally related
derivatives atractyligenin (3), 2-O-b-
isovaleryl-2b-(2-desoxy-atractyligenin)-b-
D D
-glucopyranosyl-atractyligenin (4), 30-O-b- -glucopyranosyl-20-O-
D-glucopyranoside (5), and 2-O-b-D-glucopyranosyl-carboxya-
tractyligenin (6) were isolated from raw beans of Coffea L. and the impact of 1–6 on ANT activity was
evaluated in isolated mitochondria. Among the coffee components, 6 significantly inhibited ANT activity
leading to reduced respiration. Quantitative analysis in commercial coffees, experimental roastings of
coffee, and model experiments using purified compound 6 consistently revealed a complete degradation
during thermal treatment. In comparison, raw coffee extracts were found to contain high levels of 6,
which are therefore expected to be present in food products enriched with raw coffee extracts. This
implies the necessity of analytically controlling the levels of 6 in raw coffee extracts when used as addi-
tives for food products.
Keywords:
Coffea L.
Rubiaceae
Coffee
Adenine nucleotide translocator
UPLC-Time-of-Flight-MS quantification
2-O-b-D-Glucopyranosyl-
carboxyatractyligenin
Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction
Santi and Luciani, 1978). Compared to 1, its carboxylated deriva-
tive 2 has been reported to show a 10-fold increased toxicity due
Atractyloside (1) and carboxyatractyloside (2) are toxic sulfo-
glucosides of the norditerpenoid atractyligenin (3), present all
around the globe in the plant kingdom as compiled by Obatomi
and Bach (1998) (Fig. 1). 1 and 2 have first been isolated from
the rhizome of Atractylis gummifera (L.), a thistle growing in the
mediterrean region, but further sources comprise other Asteraceae
e.g. Callilepsis laureola (L.), Xanthium strumarium (L.) and Weidelia
glauca (Ort.). Ingestion of these plants by cattle, or application of
ethnomedicinal ointments and remedies (Daniele et al., 2005) pre-
pared from such material have been found responsible for fatal
intoxications through kidney or liver necrosis (Stewart and Steenk-
amp, 2000; Stuart et al., 1981; Zaim et al., 2008). 1 has been re-
ported as a competitive inhibitor of the adenine nucleotide
translocase (ANT) disrupting the exchange of ADP and ATP across
the mitochondrial inner membrane and thus effectively blocking
oxidative phosphorylation. LD50 values were found to range from
143 mg/kg (rat, i.p.) to 15 mg/kg (dogs, i.v.) (Luciani et al., 1978;
to its irreversible interaction with ANT, whereas the aglycon 3
was found to be at least 150 times less active (Vignais et al., 1978).
Roasted coffee, enjoyed by millions of people every day for its
alluring flavor and arousing properties, has been reported to con-
tain atractyligenin (Aeschbach et al., 1982; Ludwig et al., 1974)
as well as its water-soluble glycosides 2-O-b-
tractyligen (4) and 30-O-b- -glucopyranosyl-20-O-isovaleryl-2b-(2-
desoxy-atractyligenin)-b- -glucopyranoside (5) (Obermann and
D-glucopyranosyla-
D
D
Spiteller, 1976). The huge differences in absolute content of atrac-
tyligenin glucosides in the seeds of Coffea arabica and Coffea cane-
phora var. robusta (Maier and Wewetzer, 1978; Maetzel and Maier,
1983) led to the proposal of using these phytochemicals for
authenticity control (Aeschbach et al., 1982). Animal trials with
rodents showed that compounds 4 and 5 apparently did not
induce any kidney damage (Hopps et al., 1997). Besides 4 and 5,
the carboxylated derivatives 2-O-b-
tractyligenin (6) and 30-O-b- -glucopyranosyl-20-O-isovaleryl-2b-
(2-desoxy-carboxyatractyligenin)-b- -glucopyranoside (7) were
D-glucopyranosyl-carboxya-
D
D
proposed to be present in raw coffee on the basis of mass spectro-
metric studies (Bradbury and Balzer, 1999).
The high content of natural antioxidants and putative chemo-
protective compounds pushed forward recent developments con-
⇑
Corresponding author at: Chair of Food Science and Molecular Sensory Science,
Technische Universität München, Lise-Meitner-Str. 34, 85354 Freising, Germany.
Tel.: +49 (0)8161 712902; fax: +49 (0)8161 712949.
E-mail address: thomas.hofmann@tum.de (T. Hofmann).
0031-9422/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.phytochem.2013.03.022

R. Lang et al. / Phytochemistry 93 (2013) 124–135
125
sidering raw coffee extracts as neutraceuticals and health-benefi-
cial food additives, respectively (Hoelzl et al., 2010; Volz et al.,
2012). Human exposure through functional foods containing in-
creased amounts of atractyligenin- and carboxyatractyligenin-glu-
cosides might yet be negligible. However, for a reliable risk
assessment, data are urgently needed on the amounts of atractyli-
genin- and carboxyatractyligenin-glucosides in raw and roasted
coffee beans as well as their activity in blocking oxidative phos-
phorylation by means of adenine nucleotide translocase inhibition.
The aims of the present investigation were, therefore, to isolate
atractyligenin- and carboxyatractyligenin-glucosides from raw cof-
fee beans, to compare their inhibitory activity on adenine nucleo-
tide translocase in mitochondria with those of the well-known
inhibitors atractyloside (1) and carboxyatractyloside (2), respec-
tively, and to quantitate their abundance in coffee beans before
and after roasting.
et al., 2012), the glycosidic proton H–C10 was found at 4.44 ppm
as a dublett showing a coupling constant of 7.9 Hz, thus indicating
a b-linkage. 1H,13C-heteronuclear couplings of H–C10 to carbon C2,
observed in the HMBC spectrum, confirmed the substitution of the
aglycon at position 2. Signal assignment based on the homonuclear
couplings detected in the 1H–1H-correlation spectrum and the
heteronuclear couplings in the HSQC spectrum confirmed 2-O-b-
D-glucopyranosylatractyligenin (4) as the constituent in HPLC
fraction 7. To the best of our knowledge, the NMR data of the intact
glucoside have not been reported yet.
Rechromatography of fraction 11 gave compound 5 showing
the ToF-MS spectrum displayed in Fig. 2. NMR spectroscopy
revealed besides the glycosidic proton H–C10 at 4.64 ppm
(J = 8.6 Hz), an additional glycosidic proton signal resonate at
4.34 ppm as a dublett (J = 7.9 Hz), thus indicating a b-linked dig-
lucoside (cf. Table 1 and exact mass data in 4. Experimental). Final
signal assignment by means of 1H–1H-correlation spectra and the
1H–13C heteronuclear correlation spectroscopy (HSQC, HMBC) led
to the unequivocal identification of compound 5 as 30-O-b-
pyranosyl-20-O-isovaleryl-2b-(2-desoxy-atractyligenin)-b-
D
-gluco-
2. Results and discussion
D
-gluco-
pyranoside. To the best of our knowledge, the NMR data of the
intact glucoside have not been reported yet.
2.1. Isolation of atractyligenin- and carboxyatractyligenin glucosides
from coffee beans
UPLC-ToF-MS analysis of HPLC fraction 6 revealed the pseudo-
molecular ion m/z 481.2443 similar to that found for compound
4. However, an additional ion with m/z 525.2339 was observed
with about 30% intensity, implying an additional carboxy function
in the molecule and matching well with a calculated sum formula
of 6 (C26H37O11, [MꢀH]ꢀ, cf. Fig. 2). 13C NMR spectroscopy
(d4-methanol) showed two carbonyl carbons resonating at 175.6
and 175.9 ppm, both showing heteronuclear couplings with the
adjacent protons H–C3 and H–C5. Carbon C4 was shifted downfield
to 59.3 ppm and no connected proton was observed in the HSQC
spectrum. The 13C chemical shifts of the remaining carbon atoms
were almost identical to those observed for 4. Taking all spectro-
scopic data into account, the compound isolated from fraction 6
was identified as 2-O-b-glucopyranosyl-carboxyatractyligenin (6).
To the best of our knowledge, the NMR data of this glucoside have
not been published yet.
Following a procedure reported earlier (Ludwig et al., 1974),
atractyligenin (3) was isolated from roasted coffee brew after enzy-
matic liberation from its glucosides by means of liquid–liquid
extraction, followed by silica gel fractionation and subsequent
purification by preparative RP-HPLC. The NMR data were in good
accordance with those reported recently (Brucoli et al., 2012). For
the isolation of atractyligenin glycosides, a lyophilized hot water
extract prepared from powdered raw coffee was extracted with
methanol, the methanol extractables were separated by adsorption
chromatography on XAD-2 resin followed by preparative RP-HPLC
to afford 15 subfractions which were screened for compounds 4–6
by means of UPLC-ToF-MS.
The ToF-MS spectrum of the compound eluting in HPLC fraction
no. 7 indicated the presence of 2-O-b-D-glucopyranosylatractylige-
nin (4) (Fig. 2) and was confirmed by means of 1D/2D NMR spec-
troscopy. While the 1H and 13C NMR signals of the aglycon were
assigned based on published data (Hanson et al., 1976; Brucoli
In contrast to compounds 4–6, target compound 7 detected in
HPLC-fraction 13 co-eluted with another compound showing a
Fig. 1. Structures of atractyloside (1), carboxyatractyloside (2), atractyligenin (3), 2-O-b-glucopyranosyl-atractyligenin (4), 30-O-b-
desoxy-atractyligenin)-b-
-glucopyranoside (5), 2-O-b-glucopyranosylcarboxyatractyligenin (6), 30-O-b- -glucopyranosyl-20-O-isovaleryl-2b-(2-desoxy-carboxyatractylige-
nin)-b- -glucopyranoside (7) and 2,15-didehydroatractyligenin (8, IS).
D
-glucopyranosyl-20-O-isovaleryl-2b-(2-
D
D
D

126
R. Lang et al. / Phytochemistry 93 (2013) 124–135
Fig. 2. Exact mass spectra of compounds 3–7 acquired in negative electrospray (50–1000 Da).
pseudomolecular ion of m/z 515.1191 (calculated m/z 515.1180 for
25H23O12 -1.0 mDa), most likely a di-caffeoyl quinic acid. Unfor-
are well known to feed electrons into the respiratory chain via
complex II which drives proton export from the mitochondrial ma-
trix into the inter-membrane space by complexes III and IV (Brand
and Nicholls, 2011). In this setting, oxygen consumption is a mea-
sure of electron flux and, in consequence, of proton export. In the
absence of ADP, protons re-enter the matrix by proton leak across
the mitochondrial inner membrane. This proton leak is constantly
compensated by the respiratory chain (state 4 respiration) in order
to keep up the membrane potential. In the presence of ADP, this
adenine nucleotide is readily taken up into the matrix via the
ANT and subsequently phosphorylated to ATP at complex V driven
by proton motive force, thus resulting in a significantly higher res-
piration rate (state 3 respiration). The increased oxygen consump-
tion of state 3 above state 4 respiration is ANT dependent and, in
consequence, only state 3 respiration is sensitive to ANT inhibition.
In the presence of ADP, an ANT inhibitor therefore reduces state 3
respiration until it reaches the lower state 4 respiration rate upon
complete ANT blockage (Brand and Nicholls, 2011).
Using this experimental set-up, the coffee-derived components
3–6 as well as atractyloside (1) and carboxyatractyloside (2), both
well-known ANT inhibitors, were tested for their inhibitory effect
on oxidative phosphorylation by respirometry of isolated mito-
chondria. In the presence of carboxyatractyloside (2), we observed
a diminished state 3 oxygen consumption in a dose dependent
manner representing ANT inhibition with a calculated IC50 of
369 ꢁ 10ꢀ9 M (Fig. 3A). The basal level (state 4) was reached with
compound 2 at a concentration of ꢂ10ꢀ5 M, indicating complete
blockage of adenine nucleotide transporters. As expected, the
IC50 value of the less potent atractyloside (1) was roughly 270
times higher (98 ꢁ 10ꢀ6 M), the basal state 4 respiration level
was reached at a concentration of ꢂ10ꢀ3 M. The carboxylated
derivative 6 isolated from raw coffee diminished oxygen consump-
C
, D
tunately, any attempt to further purify compound 7 by means of
re-chromatography or crystallization failed due to its enormous
instability. Therefore, compound 7 was characterized by exact
mass data of the pseudomolecular ion and its fragments generated
by application of low collision energy. The pseudomolecular ion m/
z 771.3441 ([MꢀH]ꢀ) observed for 7 in negative electrospray was
well in line with the theoretical monoisotopic mass of 30-O-b-
D-
glucopyranosyl-20-O-isovaleryl-2b-(2-desoxy-carboxyatractylige-
nin)-b- -glucopyranoside. Collision energy induced decarboxyl-
D
ation of the pseudomolecular ion m/z 771.3441 and led to the
fragment ion m/z 727.3539 matching well the [MꢀH]ꢀ of com-
pound 5 (Fig. 2). In addition, the fragment ion m/z 643.2963 indi-
cated the loss of the esterified isovaleric acid moiety.
Furthermore, the ions m/z 481.2434 and 319.1909 reflected the
pseudomolecular ion of compound 4 and its aglycon atractyligenin
(3), formed by a cleavage of one and two hexose moieties, respec-
tively. On the basis of its ToF-MS data of the mother ion and the
fragment ions the structure of compound 7 was proposed as 30-
O-b-
tyligenin)-b-
D
-glucopyranosyl-20-O-isovaleryl-2b-(2-desoxy-carboxyatrac-
-glucopyranoside.
D
2.2. Activity of atractyligenin- and carboxyatractyligeninglucoside on
adenine nucleotide translocator in isolated mitochondria
As atractyloside (1) and carboxyatractyloside (2) are well-
known inhibitors of the adenine nucleotide translocator (ANT) in
mitochondria, the following experiments aimed at evaluating the
inhibitory effect of the coffee-derived components 3–6 on oxida-
tive phosphorylation by respirometry of isolated mitochondria. In
the presence of the substrate succinate, such isolated mitochondria

R. Lang et al. / Phytochemistry 93 (2013) 124–135
127
Table 1
13C and 1H NMR data of compounds 4–6 and 8.
Position Compound 4
Compound 5
Compound 6
Compound 8
dH
dC
dH
dC
dH
dC
dH
dC
1
2
3
4
5
6
7
0.83 t, J = 11.8 Hz, 1H; 2.31 dd, 48.6,
0.78 t, J = 12.1 Hz,
1H; 2.31 m, 1H
4.27 m, 1H
48.8,
CH2
74.6,
CHOH
37.1,
CH2
44.6,
CH
50.7,
CH
0.90 t, J = 12.4 Hz, 1H; 2.35
dd, J = 12.2, 3.6 Hz, 1H
4.24 m, 1H
48.7,
CH2
73.1,
CHOH
41.0,
CH2
2.52 dd, J = 6.9, 6.9 Hz,
1H; 1.83–2.14 m, 5H
56.4,
CH2
211.9,
C@O
44.3,
CH2
46.6,
CH
48.5,
CH
J = 13.3, 4.4 Hz, 1H
CH2
4.31 m, 1H
73.8,
CHOH
37.4,
CH2
45.2,
CH
1.29 m, 1H; 2.51 m, 1H
2.65 m, 1H
1.29–1.55 m, 6H;
1.89 m, 1H
2.65 m, 1H
1.35–1.56 m, 5H; 2.74 m, 1H
2.53 m, 2H
59.3, C
3.07 m, 1H
1.36–1.54 m, 5H
50.0,
CH
1.29–1.55 m, 6H
1.56–1.77 m, 4H
1.29–1.55 m, 6H
1.88–1.90 m, 3H
52.3,
CH
24.3,
CH2
36.4,
CH2
2.15 m, 1H
1.58–1.72 m, 4H; 1.95 d,
J = 13.2 Hz, 1H
26.8,
CH2
26.5,
CH2
35.7,
CH2
1.58–1.73 m, 4H; 1.88–
1.90 m, 3H
1.35–1.56 m, 5H; 1.58–
1.73 m, 4H
1.83–2.14 m, 5H
25.6,
CH2
34.3,
CH2
1.36–1.54 m, 5H; 1.58–1.72 m,
4H
36.4,
CH2
1.83–2.14 m, 5H; 1.37–
1.50 m, 5H
8
9
49.8, C
54.7,
CH
49.9, C
49.2, C
55.2,
CH
53.4, C
51.9,
CH
1.09 d, J = 7.7 Hz, 1H
1.09 d, J = 7.9 Hz, 1H 54.6,
1.15 d, J = 7.4 Hz
1.43 d, H = 8.5 Hz, 1H
CH
10
11
41.9, C
19.4,
CH2
33.7,
CH2
43.8,
CH
35.7,
CH2
83.8,
CHOH
160.6,
C
109.2,
CH2
41.9, C
19.4,
CH2
33.7,
CH2
43.7,
CH
36.3,
CH2
83.7,
CHOH
160.5,
C
40.8, C
19.4,
CH2
33.7,
CH2
43.8,
CH
37.5,
CH2
83.6,
CHOH
160.4,
C
109.2,
CH2
175.6,
C@O
175.9,
C@O
18.1,
CH3
44.7, C
19.4,
CH2
33.0,
CH2
39.5,
CH
37.2,
CH2
212.1,
C@O
151.2,
C
115.6,
CH2
1.36–1.54 m, 5H; 1.58–1.72 m,
4H
1.36–1.54 m, 5H; 1.58–1.72 m,
4H
1.29–1.55 m, 6H;
1.56–1.77 m, 4H
1.29–1.55 m, 6H;
1.56–1.77 m, 4H
2.70 m, 1H
1.35–1.56 m, 5H; 1.58–
1.73 m, 4H
1.35–1.56 m, 5H; 1.58–
1.73 m, 4H
1.37–1.50 m, 5H; 1.68 m,
1H
1.70 m, 1H; 1.83–2.14 m,
5H
12
13
14
15
16
17
18
19
20
10
2.71 m, 1H
2.72 m, 1H
3.08 m, 1H
1.36–1.54 m, 5H; 1.89 d,
J = 10.4, 1H
3.76 s, 1H
1.56–1.77 m, 4H
3.78 s, 1H
1.35–1.56 m, 5H; 1.88–
1.90 m, 3H
3.78 s, 1H
2.39 d, J = 11.7 Hz, 1H;
1.37–1.50 m, 5H
5.08 s, 1H; 5.18 s, 1H
5.07 s, 1H; 5.18 s,
1H
109.3,
CH2
5.08 s, 1H; 5.19 s, 1H
5.91 s, 1H; 5.34 s, 1H
–
179.5,
C@O
17.4,
CH3
102.9,
CH
75.3,
CHOH
78.3,
CHOH
71.8,
CHOH
77.9,
CHOH
62.9,
CH2OH
179.8,
C@O
17.3,
CH3
177.9,
C@O
17.6,
CH3
1.02 s, 3H
1.00 s, 3H
1.03 s, 3H
1.06 s, 3H
4.44 d, J = 7.9 Hz, 1H
3.13 dd, J = 8.9, 8.9 Hz, 1H
3.37 m, 1H
4.64 d, J = 8.6 Hz, 1H 101.2,
4.43 d, J = 7.8 Hz, 1H
3.12 dd, J = 8.2, 8.2 Hz, 1H
102.9,
CH
CH
73.9,
CHOH
78.1,
CHOH
71.4,
CH2OH
77.9,
CHOH
62.7,
20
4.87 overlapped by
solvent
3.33 overlapped by
solvent
3.29 overlapped by
solvent
75.2,
CHOH
78.1,
CHOH
71.7,
CHOH
77.9,
CHOH
62.8,
CH2OH
30
3.30 m, overlapped by
solvent
3.37 dd, J = 8.2, 8.2 Hz, 1H
40
3.27 m, 1H
50
3.29 m, 1H
3.39 overlapped by
solvent
3.29 m, overlapped by
solvent
3.68 m, 1H; 3.85, m, 1H
60
3.68 dd, J = 12.0, 5.0 Hz, 1H;
3.86 dd, 12.2, 2.0 Hz, 1H
3.64–3.96 m, 4Ha
CH2OH
100
200
300
400
500
4.34 d, J = 7.9 Hz, 1H 105.5,
CH
3.19 dd,
74.9,
J = 8.4,8.4 Hz, 1H
3.76 dd, J = 8.4,
8.4 Hz, 1H
CHOH
85.2,
CHOH
70.2,
3.51 m, 1H
CHOH
78.3,
3.28–3.40
overlapped by
solvent
CHOH
600
1000
2000
3000
4000
4000
3.64–3.96 m, 4Ha
62.7,
CH2OH
174.0,
C@O
2.30 d, J = 7.7 Hz, 2H 44.2,
CH2
26.7,
CH
23.2,
CH3
23.2,
CH3
2.11 m, 1H
1.02 s, 3H
1.01 s, 3H
Compounds were measured in d4-methanol.
a
These signals overlapped with each other.

128
R. Lang et al. / Phytochemistry 93 (2013) 124–135
tration of ꢂ10ꢀ3 to 10ꢀ2 M. In comparison, compounds 1, 2, and 6
showed no effect on state 4 respiration, thus indicating their spec-
ificity for ANT inhibition (Fig. 3C). On the other hand, atractyligenin
(3), and glucosides 4 and 5 isolated from coffee showed no signif-
icant inhibitory effect (Fig. 3B). Comparing the chemical structure
of the non-active compound 4 and the active derivative 6 indicated
the additional carboxy-group bound to C4 to be required for the
ANT inhibitory activity of 2-O-b-D-glucopyranosyl-carboxyatrac-
tyligenin (6).
As a constituent of coffee, compound 6 may render health risks
to the consumer as it acts in a similar way to 1 and 2. Therefore
accurate data on their quantity in raw and roasted coffee are
needed for a further risk assessment of these atractyligenin- and
carboxyatractyligenin glucosides in human diet.
2.3. Quantitative analysis of atractyligenin- and carboxyatractyligenin
glucosides in coffee
2.3.1. Method development
For the quantitation of atractyligenin and its glucosides in cof-
fee samples by means of LC–MS, 2,15-didehydroatractyligenin (8)
was synthesized as a structurally related internal standard by oxi-
dizing the hydroxyl functions of isolated atractyligenin (3) to car-
bonyl groups by using Dess–Martin-periodinan. After purification
of the target compound 8 (46% yield), the structure of 2,15-dide-
hydroatractyligenin was confirmed by NMR and ToF-MS experi-
ments. The exact mass of the pseudo molecular ion m/z 315.1595
([MꢀH]ꢀ, cf. Fig. 2) observed in negative electrospray ionization
propagated the sum formula C19H23O4 ([MꢀH]ꢀ) and corresponded
well with the calculated mass of 8. As expected, two new carbon
signals resonating at 211.9 and 212.1 ppm were observed and
the carbons HO–C2 at 65.2 ppm and HO–C15 at 83.7 ppm were
lacking, thus indicating the successful conversion of atractyligenin
(3) into 2,15-didehydroatractyligenin (8) (Fig. 4).
Fig. 3. (A) Dose-dependent inhibition of state
atractyloside (1, ꢁ), carboxyatractyloside (2, d), and 2-O-b-
4
mitochondrial respiration by
-glucopyranosyl-
carboxyatractyligenin from raw coffee (6, j); B) effects of the coffee compounds
atractyligenin (3, ꢁ), 2-O-b-
-glucopyranosyl-atractyligenin (4, j) and 30-O-b-
glucopyranosyl-20-O-isovaleryl-2b-(2-desoxy-carboxyatractyligenin)-b-
-glucopy-
D
D
D-
D
ranoside (5, d); C) effects of the inhibiting compounds 1 (ꢁ), 2 (d) and 6 (j) on
state 4 respiration. The concentration (M) is given on the x-axis (logarithmic scale).
tion through ANT blockage similarly to 1 and 2. The IC50 value of
318 ꢁ 10ꢀ6 M was about three times higher than that of 1. A com-
plete ANT blockage (state 4) was reached at an estimated concen-
Fig. 4. 13C NMR spectra of hemisynthetic 2,15-didehydroatractyligenin (upper panel) and atractyligenin (lower panel) measured in d4-methanol (100 MHz).

R. Lang et al. / Phytochemistry 93 (2013) 124–135
129
Table 2
Data on quantifier mass traces, linear range, Limit of Quantification, precision and accuracy of the calibration curves of compound 3–6.
Compounda
Quantifier Ion (m/z)b
Retention time (min)
Linear range (
lM)
LoQ (pmol)c
Precision (RSD, %)
Accuracy (%)
R2
3 [MꢀH]ꢀ, 3b [MꢀH]ꢀ
4 [MꢀH]ꢀ, 4b [MꢀH]ꢀ
5 [MꢀH]ꢀ
319.19 0.02
481.24 0.02
727.35 0.02
525.23 0.02
315.15 0.02
2.16, 2.29d
1.96, 2.04e
2.29
1.91
2.50
2 – 60
2 – 60
2 – 30
2 – 30
–
0.2
0.2
0.2
0.5
–
0.9 – 6.3
2.1 – 8.5
1.5 – 6.0
2.7 – 9.0
–
94 – 106
93 – 105
94 – 102
91 – 101
–
0.997
0.994
0.998
0.991
–
6 [MꢀH]ꢀ
8 [MꢀH]ꢀ, IS
a
b
c
Compound 7 (Rt. 2.14 min, m/z 771.34 0.02, [MꢀH]ꢀ) was quantified using the calibration of structurally related compound 6.
Mass window of 0.02 Da was used.
Limit of Quantification defined as signal/noise ratio >10, absolute amount of analyte injected on column.
A second signal with the identical exact mass (cf. Fig. 8) observed in the degradation model experiments was tentatively termed iso-atractyligenin.
a second signal with the identical exact mass (cf. Fig. 8) observed in the degradation model experiments was tentatively termed 2-O-b-
d
e
D-glucopyranosyl-iso-
atractyligenin.
Table 3
Quantitative data for compounds 3–7 in raw and roasted coffees.
3 (nmol/g)
4 (nmol/g)
5 (nmol/g)
6 (nmol/g)
7 (nmol/g)a
Raw coffee Arabicab
–
–
–
341 17 (4.8%)
–
–
2185 82 (3.8%)
2816 64 (2.3%)
1741 118 (6.8%)
2612 139 (5.3%)
115 11 (9.5%)
–
–
686 34 (5.0%)
1156 10 (0.9%)
600 67 (11.1%)
1223 75 (6.2%)
2378 256 (10.8%)
272 32 (11.8%)
50 3 (5.3%)
–
–
2104 147 (7.0%)
460 15 (3.4%)
178 9 (4.9%)
–
Raw coffee Robusta Ic
Raw coffee Robusta IIc
Commercial coffee Ib
Commercial coffee IIc
Commercial Espressoc
Commercial instant coffeec,d
237 16 (6.7%)
225 6 (2.9%)
279 13 (4.5%)
285 18 (6.5%)
–
–
–
55 7 (13.5%)e
28 8 (17.0%)e
a
b
c
This compound was quantified using the calibration curve of the structurally related compound 6.
Data are means standard deviation (RSD, %) of n = 5.
Data are means standard deviation (RSD, %) of n = 3.
Spray-dried instant coffee contained 30% raw coffee extract as claimed by the manufacturer.
Signal was detected but below LoQ.
d
e
After preparing methanolic stock solutions of the individual
mercially available medium roasted coffee (commercial coffee I)
and a powdered raw coffee (raw Arabica). As given in Table 3, pre-
cision was 11% RSD or better for all the analytes.
UPLC-ToF-MS analysis of a raw coffee sample did not show any
trace amounts of atractyligenin and only low amounts of the com-
pounds 4 (341 nmol/g) and 5 (115 nmol/g) compared to large
amounts of the carboxylated derivatives 6 (2378 nmol/g) and 7
(2104 nmol/g) (Fig. 5A). In contrast, the carboxy-derivatives 6
compounds 3–6 and the internal standard 8, followed by appropri-
ate dilution, seven calibration standards were prepared with con-
centration ratios analyte/internal standard ranging from 0.25 to
12.5 and a fixed concentration of the internal standard. The stan-
dards were analyzed in triplicate and the obtained area ratios plot-
ted versus the concentration ratios followed by linear regression.
Linear range as well as precision and accuracy of the back-calcu-
lated standards are given in Table 2. Precision, expressed as rela-
tive standard deviation, was between 0.9% and 9.0%, accuracy
was between 91 and 106%. The Limit of Quantification (LoQ), deter-
mined by analysis of serial dilutions of a standard solution (signal/
noise ratio 10/1), ranged between 0.2 and 0.5 pmol as the absolute
amount of analyte injected onto the column (Table 2). The calibra-
tion of the carboxy-derivative 6 was used for quantitation of com-
pound 7. For quantitation of the analytes in coffee, the powdered
sample was placed in a glass vial and stirred after addition of the
internal standard in 50%-aqueous methanol. To estimate an appro-
priate equilibration time and ensure the sufficient stability of the
internal standard, a coffee sample was stirred after addition of 8
in aqueous methanol and aliquots of the supernatant were re-
moved after several time points (3–45 min), diluted, membrane-
filtered and analyzed by means of UPLC-ToF-MS. As the ratio of
analyte/internal standard reached a steady state after 15 minutes,
we considered the mixture equilibrated and, therefore, used 15
min as a suitable equilibration time for quantitative analysis of
the coffee samples. During the experiment the peak area of the
internal standard 8 showed a variance of 5.1%. Replicate analysis
of a single prepared coffee sample stored in the autosampler
(10 °C) and analyzed on two consecutive days (n = 6 on day 1,
n = 6 on day 2, after 24 h) showed 179565 6086 counts (3.4%)
and 1736355 9591 counts (5.5%), respectively, indicating suffi-
cient stability of 8 during quantitative analysis of coffee samples.
Precision of the method, expressed as relative standard deviation,
was evaluated by replicate analysis (n = 5 per sample) of a com-
and
7
could not be detected in the roasted coffee sample
was the major compound with levels of
(Fig. 5B). Here,
4
2185 nmol/g, followed by 5 (686 nmol/g) and the aglycone atrac-
tyligenin (1, 237 nmol/g).
2.3.2. Analysis of raw and roasted coffees
Analysis of two raw coffees (Robusta), a commercial medium
roast coffee and an espresso-type coffee revealed comparable data
to those obtained for the coffees used for method development.
However, in the raw Robusta coffees, neither atractyligenin, nor
its glucosides were detected, and the concentrations of the carbox-
ylated derivatives 6 (272 nmol/g and 50 nmol/g) and 7 (460 nmol/g
and 178 nmol/g) were lower compared to the values in the raw
Arabica (Table 3). On the other hand, the roasted samples did
not show any traces of 6 and 7, but contained 3 in levels ranging
from 225 to 279 nmol/g. 2-O-b-D-glucopyranosylatractyligenin 4
was the dominating derivative with levels of up to 2816 nmol/g,
followed by compound 5 (600–1156 nmol/g). In the dark roasted
espresso sample (100% Arabica beans), the abundance of com-
pound 4 was ꢂ40% lower compared to the medium roasted com-
mercial coffee, thus suggesting severe losses during intense
roasting (Table 3).
Analysis of an instant coffee, claimed to contain 30% raw coffee
extract, showed target compounds 3–5 in quantities comparable to
those found in regular roasted coffees (3: 285 nmol/g, 4:
2612 nmol/g, 5: 1223 nmol/g). However, the concentrations of 6
and 7 were by far lower than expected for a coffee sample contain-

130
R. Lang et al. / Phytochemistry 93 (2013) 124–135
Fig. 5. Quantitative UPLC-ToF MS run (section from 0.0 to 3.75 min) of compounds 3–5 in roasted coffee (A) and 6 and 7 in raw coffee (B).
ing raw coffee extract. Quantitative analysis of compounds 6 and 7
in an aqueous raw coffee extract before and after freeze-drying
showed a high recovery of 93 1% (6) and 94 8% (7), thus demon-
strating the stability of these glucosides upon lyophilization and
suggesting their severe degradation upon spray-drying as part of
the instant coffee manufacturing process.
Taking all these data into consideration, it might be concluded
that 6 and 7 were initially present in raw coffee, but degraded to
generate compounds 3–5 by decarboxylation during thermal treat-
ment in coffee processing such as, e.g. bean roasting or spray-dry-
ing of instant coffee.
anosylatractyligenin 4 as well as glucoside 5 degraded substan-
tially during roasting. Atractyligenin (3) was increasingly formed
within the first 4 min of roasting, then reached a steady state
around 6 min with ꢂ350 nmol/g prior to degradation upon further
roasting. In the sample roasted for 10 min, roughly 25% and 17% of
compounds 4 (989 103 nmol/g) and 5 (282 26 nmol/g) were left
compared to the values of sample roasted for 1 min (4:
3466 271 nmol/g, 5: 2421 20 nmol/g). Since the concentra-
tion/time data pairs of 4 and 5 each fell on a straight line in the
semi-logarithmic plot (Fig. 6B), we concluded that their degrada-
tion during coffee roasting followed first order kinetics with calcu-
lated half-life times of 4.12 (4) and 2.50 min (5) at a roasting
temperature of 255 °C. The observation that even light roasting
apparently led to complete degradation of 6 and 7 was well in line
with the data found for commercial roast coffees (Table 3).
2.3.3. Influence of coffee roasting on compounds 4–7
In an explorative experiment, we roasted raw coffee beans at
255 °C for 1, 2, 3, 4, 6, and 10 min to afford light to very dark
roasted coffee samples. Each sample was frozen in liquid nitrogen,
ground, spiked with the internal standard solution, equilibrated,
and finally extracted and analyzed by means of UPLC-ToF-MS. Each
sample was studied in triplicates. We were unable to detect any
trace of the carboxylated derivatives 6 and 7, and only compounds
2.3.4. Model roasting experiments with compound 6
In
a final experiment, we heated purified compound 6
(35,000 pmol per experiment) at 180, 240, and 260 °C, respectively,
for up to 10 min and determined the residual amount as well as its
degradation products formed (Fig. 7). Heating at 180, 240 and
3–5 were abundant (Fig. 6A). The dominating 2-O-b-D-glucopyr-

R. Lang et al. / Phytochemistry 93 (2013) 124–135
131
5000
4000
3000
2000
1000
0
10000
A
B
Cpd 3
Cpd 4
Cpd 5
1000
100
0
10
0
10
1
2
3
4
6
1
2
3
4
6
time (min)
time (min)
Fig. 6. Concentration of compound 3 (dots), 4 (crosses) and 5 (circles) plotted against roasting time (255 °C). Data represent mean standard deviation of n = 3 independent
extractions, each analyzed in triplicates. Note that compounds 6 and 7 could not be detected in the samples. A) linear scale, B) semi-logarithmic scale.
A
B
40000
30000
20000
10000
0
Cpd6 180 °C
Cpd6 240 °C
Cpd6 260°C
Cpd4 180 °C
Cpd4 240 °C
Cpd4 260°C
30000
20000
10000
0
0
10
1
2
3
4
6
0
10
1
2
3
4
t (min)
6
t (min)
Cpd4b 180 °C
Cpd4b 240 °C
Cpd4b 260°C
C
D
Cpd3b 240°C
Cpd3b 260°C
Cpd3 240°C
Cpd3 260°C
10000
8000
6000
4000
2000
0
4000
3000
2000
1000
0
0
10
1
2
3
4
6
0
10
1
2
3
4
6
t (min)
t (min)
Fig. 7. Concentration time profiles of compound 6 and its degradation products during roasting at 180, 240 °C and 260 °C. (A) 2-O-b-glucopyranosyl-carboxyatractyligenin
(6), (B) 2-O-b-glucopyranosyl-atractyligenin (4), (C) putative isomer 2-O-b-glucopyranosyl-iso-atractyligenin (4b), (D) atractyligenin (3) and putitative isomer iso-
atractyligenin (3b). Note, that compounds 3 and 3b were formed only at 240 and 260 °C. Error bars are standard deviation of triplicate analysis.
260 °C led to a loss of ꢂ35%, 67%, and 96%, respectively, of the ini-
tially present amount of 6 already after 1 min. During further
roasting, a maximum degradation of ꢂ60% (13,863 547 pmol,
180 °C), 90% (3951 273, 240 °C) and > 99% (538 67 nmol,
260 °C) was found after 10 min (Fig. 7A). The observation that 6 de-
graded virtually completely was well in line with the data obtained
for the authentic coffee samples, in which 6 could not be detected.
Compound 4 was formed by decarboxylation of 6, particularly dur-
ing the first minute (Fig. 7B). At 180 °C, a steady state was reached
after 3 min with a maximum yield of ꢂ36% (12,488 928 pmol).
The yield was ꢂ52% (18,495 394 pmol) at 240 °C after 2 min,
however, further roasting led to degradation. Roasting at 260 °C in-
stantly gave 4 in a yield of ꢂ79% after 15 s (27,526 1892 pmol),
but increasing the roasting time above 1 min led to heavy losses.
Degradation of compound 4 followed first order kinetics (one
phase decay), with an estimated half-life time of ꢂ1 min. In the
ToF-MS quantitation trace of compound 4 (m/z 481.2444) a second
peak appeared at a retention time of 2.05 min (Fig. 8). This finding
gave first evidence for the formation of 2-O-b-D-glucopyranosyl-
iso-atractyligenin (4b), possibly formed by removal of the carboxyl
group C19 instead of C18. The peak area found for 4b increased
similar to compound 4 with roughly 20% of its intensity (Fig. 7C).
The highest yields were obtained by roasting at 240 °C for 2 min
(9271 404 pmol), thereafter the compound degraded again. At
260 °C, the yields of compound 4b were lower and degradation
was more evident (Fig. 7C). Atractyligenin (Fig. 7D) was only
formed when heating was done at 240 °C and 260° reached a max-
imum after 10 min and a yield of ꢂ3% (240 °C) based on the ini-
tially present amount of compound 6. At 260 °C, compound 3
was generated in a yield of up to 8% (2703 216 pmol) after
4 min, thereafter this compound degraded again. Similar to 4 and
4b, a second compound with the same exact mass m/z 319.1910
was observed in the chromatogram after ꢂ1 min of roasting
(Fig. 8). The amount of this degradation product, tentatively
termed iso-atractyligenin (3b) and probably arising from decar-
boxylation of C19 and removal of the glucose moiety, showed a

132
R. Lang et al. / Phytochemistry 93 (2013) 124–135
Fig. 8. Excerpt of an UPLC-ToF-MS analysis (0.9 to 2.55 min) of the model roasting of 6 at 260 °C for 360 s (left panel) and corresponding extracted exact mass spectra (left
panel). 4 (Rt. 1.96 min), 4b (Rt. 2.04), 3 (Rt. 2.16), 3b (Rt. 2.29).
steady increase during the roasting experiment. While 4b was not
detectable at 180 °C, maximum amounts of ꢂ0.9% and 2% based on
compound 6 were found when thermal treatment was performed
at 240 and 260 °C, respectively. Fig. 8 shows a UPLC-ToF-MS anal-
ysis of a model roasting of compound 6 (260 °C, 360 s). The ex-
tracted exact mass spectra of 4 and 4b, and 3 and 3b were
super-imposable, substantiating the assumption of the presence
of isomers.
4. Experimental
4.1. Chemicals and materials
Nuclear Magnetic Resonance (NMR) spectra were recorded on a
Bruker 400 MHz DRX spectrometer. Chemicals were obtained from
Sigma-Aldrich (Steinheim, Germany) unless otherwise stated. Sol-
vents for isolation procedures, MS-grade methanol and acetoni-
trile, analytical silica thin layer chromatography sheets
(0.25 mm) and silica gel for column chromatography were pur-
chased from Merck (Darmstadt, Germany). Water for chromato-
graphic separations was obtained from a Milli Q Gradient A10
system (Millipore, Schwalbach, Germany). Roast coffee samples
for analysis and method development was purchased from local
markets, raw coffee was purchased via the internet from Coffee-
shop24 (Rellingen, Germany). Methanol-d4 for NMR experiments
was from Eurisotop (Gif sur Ivette, France).
Taking all these data into consideration, it can be concluded
that 2-O-b-
rapid decarboxylation to give 2-O-b-
genin (4) and, to a minor extent, its isomer 2-O-b-
D
-glucopyranosyl-carboxyatractyligenin (6) undergoes
-glucopyranosyl-atractyli-
-glucopyrano-
D
D
syl-iso-atractyligenin (4b), particularly at higher temperatures.
With more severe roasting, compound 4 and 4b decompose to
give small amounts of the aglycon atractyligenin (3) and the
corresponding isomer 3b. Compared to these model experiments
performed with purified 2-O-b-D-glucopyranosyl-carboxyatractyli-
genin (6), the degradation of 6 during roasting of authentic coffee
beans appeared to be more exhaustive as not even traces of resid-
ual 6 could be detected in any sample (Fig. 6, Table 3). In addition,
the yields of putative isomers 3b and 4b accounted for <5% of the
4.2. Isolation of atractyligenin (3) from roast coffee
Compound 3 was isolated from roast coffee following a proce-
dure reported earlier (Ludwig et al., 1974). Briefly, roast coffee
brew (60 g/L) was adjusted to pH 4.7 by addition of solid sodium
acetate (82 g) and glacial acetic acid (60 mL) and incubated (48 h,
37 °C) with crude enzyme preparation (sulfatase/b-glucuronidase
from Helix pomatia, 150,000 U). The coffee solution was extracted
with ethylacetate (5 ꢁ 200 mL), the organic layers combined and
evaporated. The residue was chromatographed on silica gel
(350 ꢁ 20 mm, 5% water) using dichloromethane/methanol (85/
15, v/v) as eluent and collecting fractions (20 mL) which were
checked by TLC (silica gel, dichloromethane/methanol 85/15, van-
illin/sulfuric acid spraying reagent, Rf. 0.7, red spot).
determined amount of atractyligenin (3) and 2-O-b-D-glucopyr-
anosylatractyligenin (4) in coffee, respectively, suggesting that
carboxyl C19 is stabilized by or even bound to matrix within the
coffee seed, thus being protected from decarboxylation.
3. Concluding remarks
Summarizing the compiled data, 2-O-b-D-glucopyranosyl-carb-
oxyatractyligenin (6) isolated from raw coffee inhibited mitochon-
drial oxidative phosphorylation through ANT blockage. While the
activity was about three times lower than that of the well-known
toxic atractyloside (1), abundance of 6 in coffee brew could be a se-
vere risk to the consumer. However, quantitative studies of 6 in
commercial and experimental roast coffees as well as model stud-
ies by means of UPLC-ToF-MS demonstrated that this compound is
fully decomposed upon heat exposure. In comparison, raw coffee
extracts were found to contain high levels of the ANT-inhibiting
6 that, if not degraded by thermal drying procedures such as spray
drying, is expected to be present in food products enriched with
raw coffee extracts. This implies the necessity of analytically con-
trolling the levels of 6 in raw coffee extracts when used as addi-
tives for food products.
Atractyligenin was isolated from the combined silica gel frac-
tions 10–17 by means of preparative HPLC (RP18, Hypersil,
250 ꢁ 21.4 mm, 5
lm Thermo, Darmstadt) using a water/acetoni-
trile gradient by increasing the acetonitrile content from 2% to
100% within 40 min at a flow of 18 mL minꢀ1. Monitoring the efflu-
ent at 200 nm, the peak eluting at ꢂ16 min was collected, the sol-
vent was separated in vacuum to afford atractyligenin (3) after
freeze-drying as a white, amorphous powder (60 mg; purity
>98%). Atractyligenin (3): UPLC-ToF-MS (ESIꢀ) m/z 319.1914 calcu-
lated for C19H27O4 ([MꢀH]ꢀ), found m/z 319.1913 (
D
0.2 mDa). 1H
NMR (MeOD, 400 MHz) 0.70 (1H, t, J = 11.76 Hz, Ha/b-C1), 1.01
(3H, s, H–C20), 1.06 (1H, d, J = 7.65 Hz, H–C9), 1.26 (1H, dt,

R. Lang et al. / Phytochemistry 93 (2013) 124–135
133
J = 12.24, 5.81 Hz, Ha/b-C3), 1.39–1.49 (5H, m, H–C5, Ha/b-C7, Ha/
b-C11, Ha/b-C12, Ha/b-C14), 1.61–1.71 (4H, m, Ha/b-C6, Ha/b-C7,
Ha/b-C11, Ha/b-C12), 1.89 (1H, d, J = 10.40 Hz, Ha/b-C14), 1.95
(1H, d, J = 13.16 Hz, Ha/b-C6), 2.19 (1H, dd, J = 13.31, 4.40 Hz, Ha/
b-C1), 2.38 (1H, m, Ha/b-C3), 2.65 (1H, td, J = 5.35, 1.94 Hz, H–
C4), 2.71 (1H, m, H–C13), 3.77 (1H, s, H–C15), 4.17 (1H, m, H–
C2), 5.08 (1H, s, Ha–C17), 5.19 (1H, s, Hb-C17); 13C NMR (MeOD,
100 MHz) 17.41 (C20), 19.37 (C11), 26.80 (C6), 33.71 (C12), 36.35
(C7), 37.36 (C14), 38.48 (C3), 41.94 (C10), 43.82 (C13), 45.06
(C4), 49.06 (C8), 50.00 (C5), 50.55 (C1), 54.64 (C9), 65.21 (C2),
83.72 (C15), 109.21 (C17), 160.47 (C16), 178.96 (C19).
man Animal Welfare law. Liver was dissected, minced and homog-
enized in ice-cold STE buffer (250 mM sucrose, 5 mM tris, 2 mM
EGTA, pH 7.4, 4 °C). The homogenate was centrifuged (1000g,
3 min), the obtained supernatant decanted and centrifuged again
(11,800g, 10 min). The pelleted mitochondrial fraction was re-sus-
pended in STE buffer and an aliquot was treated with 2.5% sodium
desoxycholate to measure mitochondrial protein concentration by
the Biuret method. An aliquot of the mixture corresponding to
800 lg mitochondrial protein was dissolved to 1 mL KHE buffer
(120 mM KCl, 5 mM KH2PO4, 3 mM HEPES, 1 mM EGTA, 2 mM
MgCl2, 5 mg/mL fatty acid free BSA, pH 7.2, 37 °C) to measure oxy-
gen consumption by a Clark type electrode (Model 10, Rank Broth-
ers, Cambridge, UK). Signals were transduced by a PowerLab unit
and analyzed by the LabChart software (ADInstruments, Castle Hill,
Australia). Mitochondria were energized by addition of succinate
(aq. 2 mM) (state 4). State 3 respiration was induced by addition
of ADP (aq. 1 mM). Aqueous solutions of compounds 1–6 (10ꢀ3
4.3. Isolation of atractyligenin derivatives (4–6) from raw coffee
A lyophilized raw coffee extract was prepared following pre-
cisely the procedure reported by Müller and Hofmann (2005).
The lyophilisate (50 g) was extracted with methanol (2 ꢁ 200 mL)
under sonication (40 °C, 30 min). The glycosidic fraction of raw cof-
fee was obtained by adsorption on XAD-2 resin (200 g in water,
400 mm ꢁ 75 mm) and elution with methanol as reported by
Obermann and Spiteller (1976). The glycosidic fraction was sepa-
rated by preparative HPLC (RP18, Hyperclone, 250 ꢁ 21.4 mm,
to 10ꢀ9 M) were used for titration (1
lL).
4.5. Synthesis of 2,15-didehydroatractyligenin (8)
Dess–Martin-periodinan (25 mg) was added to a solution of 3
(11 mg, 34.3 lmol) in acetonitrile (1.5 mL) and the mixture was
5l, Phenomenex, Darmstadt, Germany). Eluents were 0.1% aque-
ous formic acid (A) and acetonitrile (B) delivered at a flow of
20 mL minꢀ1. The eluent composition was 5% B (5 min) followed
by a gradient from 5% to 45% B within 35 min. Monitoring the efflu-
ent at 200 nm, a total of 15 HPLC fractions were collected. Com-
pound 6 eluted after 20.5 min (fraction 6), compound 4 after
22 min (fraction 7), and compound 5 after 28 min in fraction 11.
The collected compounds were separated from solvent in vacuum,
freeze-dried, and purified by re-chromatography if necessary on
the same column. Purification of compound 7 was unsuccessful
as it co-eluted with a dicaffeoyl-quinic acid and decomposed read-
ily during purification by re-chromatography or crystallization.
Compound 7 was therefore characterized by exact mass data of
the intact molecule and the fragment ions (Fig. 2). The structures
of the purified compounds 4–6 were confirmed by means of ToF-
MS and NMR experiments. The complete NMR data of the gluco-
sides are summarized in Table 1.
stirred (room temperature, 5 min). When a white precipitate oc-
curred, the mixture was stirred for further 10 min, evaporated
and taken up in water (2 mL) and centrifuged. The supernatant
was separated by preparative HPLC (RP18, Hyperclone,
21.5 ꢁ 250 mm, 5
l, Phenomenex, Darmstadt). Using aqueous for-
mic acid (0.1%) as eluent A, gradient elution involved increasing
eluent B (acetonitrile) from 5% to 50% within 20 min. Monitoring
the effluent at 200 nm, the peak eluting after 19 min was collected,
freed from solvent in vacuum, and freeze-dried to yield compound
8 as a white crystalline powder (5.1 mg, 15.8 lmol, yield: 46%).
UPLC-ToF-MS (ESIꢀ) m/z 315.1602 calculated for C19H23O4
([MꢀH]ꢀ), found m/z 315.1595 (
D-0.7mDa), cf. Table 1 for NMR
data.
4.6. Ultra performance liquid chromatography/time of flight mass
spectrometry (UPLC/ToF-MS)
2-O-b-D-glucopyranosyl-atractyligeninglucoside (4): UPLC-
ToF-MS (ESIꢀ) m/z 481.2443 calculated for C25H37O9 ([MꢀH]ꢀ),
found m/z 481.2444 (
D
0.1 mDa), cf. Table 1 for NMR data.
The Acquity UPLC core system (Waters, Bedford, MA) consisted
of a binary solvent manager, sample manager, column oven, and
was hyphenated to a Waters Synapt G2S HDMS mass spectrometer
(Waters, Manchester, UK) operating in negative electrospray ioni-
30-O-b-
D
-glucopyranosyl-20-O-isovaleryl-2b-(2-desoxy-atrac-
tyligenin)-b-
D
-glucopyranoside (5): UPLC-ToF-MS (ESIꢀ) m/z
727.3546 calculated for C36H55O15 ([MꢀH]ꢀ), found m/z 727.3550
(D
0.4 mDa); cf. Table 1 for NMR data.
zation. Separation was achieved on
a BEH C18 column
2-O-b-
D
-glucopyranosyl-carboxyatractyligeninglucoside (6):
(50 mm ꢁ 2 mm, 1.7 m) with water and acetonitrile (0.1% formic
l
UPLC-ToF-MS (ESIꢀ) m/z 525.2341 calculated for C26H37O11
acid as modifier) at a flow rate of 0.6 mL minꢀ1. After isocratic elu-
tion with 95% water (1 min), acetonitrile was increased to 50%
within 2.5 min, then to 90% within 0.5 min followed by isocratic
elution (0.5 min) prior to re-establishment of the initial conditions.
Re-equilibration was done for 1.3 min. MS-data were acquired
using the MSe experiment (mass acquisition alternating between
no collision and low collision energy) in the mass range of 50–
1000 Da (centroid data, 10 Hz). Capillary voltage was 1 kV, sam-
pling cone voltage 20 V, source offset 50, source temperature
([MꢀH]ꢀ), found m/z 525.2339 (
D-0.2mDa); cf. Table 1 for NMR
data.
30-O-b-
D
-glucopyranosyl-20-O-isovaleryl-2b-(2-desoxy-carb-
-glucopyranoside (7): Retention time:
oxyatractyligenin)-b-
D
2.14 min (cf. UPLC-MS method for gradient details); UPLC-ToF-
MS (ESIꢀ): m/z 771.3444 calculated for C37H55O17, ([MꢀH]ꢀ), found
m/z 771.3441 (D-0.3 mDa); fragments: m/z 727.3539 (C36H55O15,
[MꢀCO2–H]ꢀ), m/z 643.2963 (C31H46O14
,
[Mꢀisovaleric acid–
CO2–H]ꢀ), m/z 481.2434 (C25H37O9, [liberated compound 4–H]ꢀ),
m/z 319.1906 (C19H27O4, [liberated compound 3–H]ꢀ). Cf. Fig. 2
for details.
120 °C, cone gas flow 50 L hꢀ1, desolvation gas flow 850 L hꢀ1
,
desolvation temperature was 500 °C. Masslynx 4.1 (Waters) was
used for operating the system. Quantitative data were acquired
with Quanlynx used for peak detection and integration. Mass data
were online-corrected using the pentapeptide leucine-enkephaline
(Tyr-Gly-Gly-Phe-Leu, m/z 554.2615, [MꢀH]ꢀ) dissolved in aque-
ous acetonitrile (1/1, v/v, 0.1% formic acid) at a final concentration
4.4. Respirometry of isolated mitochondria
Isolation of mitochondria and determination of ANT activity
was performed as previously described (Jastroch et al., 2012).
Briefly, liver mitochondria were isolated from a male mouse of
the 129 Sv/ec Tac strain bred in our own facility according to Ger-
of 1 ng/
l
L and infused into the ion source (5
l
L ꢁ minꢀ1) during
data acquisition. Scan time for the lock mass was 0.3 s (15 s inter-
val). Mass-calibration of the Synapt G2S in the mass range m/z

134
R. Lang et al. / Phytochemistry 93 (2013) 124–135
50–1200 was done with sodium formate (5 mM in 2-propanol/
water 9/1, v/v).
4.7. Model roasting of 2-O-b-D-glucopyranosyl-
carboxyatractyligeninglucoside (6)
For investigation of the degradation of compound 6 during
roasting, aliquots (20 lL) of a standard solution (1.7 mM in meth-
4.6.1. Standard solutions, Limit of Quantification, calibration curves
and linearity
anol) were pipetted into glass vials and evaporated in a stream of
nitrogen, resulting in ꢂ35 nmol solid per vial. For each tempera-
ture series (180, 240 and 260 °C), nine vials were prepared. The
vials were then inserted into a thermostated heating block for
roasting. After the respective heating time (15, 30, 45, 60, 120,
180, 270, 360, 600 s), the vial was removed from the block and
water (1 mL) was added to cool down the content and stop the
pyrolysis. Finally, the internal standard 8 (1.5 mL, 5 nmol/mL in
50% aqueous methanol) was added, the mixture vortexed, diluted
Stock solutions were prepared by individually dissolving com-
pounds 3–6 and 8 in pure methanol (5 mL) to obtain concentra-
tions of 1
were combined to yield
l
mol/mL. Appropriate aliquots of the analytes 3–6
mixed analytes standard with
a
100 nmol/mL in 10% aqueous methanol per compound. The inter-
nal standard (8) was diluted to 5 nmol/mL (50% aqueous metha-
nol). Seven calibration standards were prepared with a fixed
concentration of the IS (2.5 nmol/mL) and concentration ratios
analyte/IS from 0.25 to 12.5. Each standard was analysed in tripli-
cate by UPLC-ToF-MS and the area ratios of analyte and internal
standard (cf. Table 2 for the exact mass traces used) plotted against
the respective concentration ratios followed by linear regression.
Accuracy and precision were obtained by back-calculation of the
analyte concentrations using the respective calibration curve (cf.
Table 2). Compound 7 was determined using calibration of the
structurally related compound 6. The Limit of Quantification
(LoQ), defined as the concentration with a signal/noise ratio >10
and estimated by analysis of serial dilution of a standard solution,
is expressed as absolute amount of analyte injected onto the col-
umn (Table 2).
with water (1 + 1, v/v) and analyzed by UPLC-ToF-MS (1 lL).
4.8. Statistical analysis
Processing of raw data was done using Excel 2010 (Ó Microsoft
Corporation). Quantitative data are expressed as mean standard
deviation unless otherwise noted. To determine IC50 values we em-
ployed the Prism 5.02 software (GraphPad, La Jolla, USA). A non-
linear regression was calculated using drift corrected, mean respi-
ration values (y = bottom + (top ꢀ bottom)/(1 + 10^(logIC50 ꢀ x); x
is the log of the concentration, y starts at the bottom and goes to
the top with a sigmoidal shape). In the case of 6, the bottom value
was manually set to equal state 4 respiration.
4.6.2. Analysis of coffee samples
Quantitative analysis of coffee powder was performed by addi-
tion of the internal standard 8 (1 mL, 5 nmol/mL in 50% aqueous
methanol) to the sample (50 mg) in a glass vial. After equilibration
by light shaking (15 min), water (1 mL) was added, and an aliquot
Acknowledgements
Ms. Maren Ilse, Ms. Barbara Suess and Mr. Benedikt Kohles are
acknowledged for skillful acquisition of the NMR spectra.
of the membrane-filtrate (500
lL) was diluted with water (500
lL)
and, then, injected into the UPLC-ToF-MS system (1
lL). Precision
of the analysis was evaluated by replicate extraction and analysis
(n = 5) of individual samples of powdered raw coffee (Arabica)
and powdered roasted coffee (commercial coffee I), respectively,
and expressed as relative standard deviation (RSD, %, cf. Table 3).
References
Aeschbach, R., Kusy, A., Maier, H.G., 1982. Diterpenes in coffee I. Atractyligenin. Z.
Lebensm. Unters. Forsch. 175, 337–341.
Bradbury, A.G.W., Balzer, H.H.,1999. Carboxyatractyligenin and atractyligenin
glycosides in coffee. Colloque Scientifique International Sur Le Cafe, 18th, pp.
71–77.
4.6.3. Evaluation of equilibration time and stability of the internal
standard 8
Brand, M.D., Nicholls, D.G., 2011. Assessing mitochondrial dysfunction in cells.
Biochem. J. 435, 297–312.
A coffee sample (500 mg) was mixed with the IS working solu-
tion (5 mL, 5 nmol/mL in 50% aqueous methanol) and stirred at
Brucoli, F., Borrello, M.T., Stapleton, P., Parkinson, G.N., Gibbons, S., 2012. Structural
characterization and antimicrobial evaluation of atractyloside, atractyligenin,
and 15-didehydroatractyligenin methyl ester. J. Nat. Prod. 75, 1070–1075.
Daniele, C., Dahamna, S., Firuzi, O., Sekfali, N., Saso, L., Mazzanti, G., 2005. Atractylis
gummifera L. poisoning: an ethnopharmacological review. J. Ethnopharmacol.
97, 175–181.
room temperature. In intervals, an aliquot (50
lL) was removed,
diluted with water (150 L) and analyzed by UPLC-ToF-MS. The
l
area ratios of analytes and internal standards were plotted against
time. The equilibrium was considered reached when further
incubation failed to change the area ratio of analyte and internal
standard. Stability of the internal standard 2,15-didehydroatractyl-
igenin (8) during sample processing and storage in the autosam-
pler was verified by replicate analysis of a single prepared coffee
analysis on two consecutive days (n = 6 on day 1, n = 6 on day 2,
after 24 h).
Hanson, J.R., Siverns, M., Piozzi, F., Savona, G., 1976. The 13C nuclear magnetic
resonance spectra of kauranoid diterpenes. J. Chem. Soc. Perkin I 1, 114–117.
Hoelzl, C., Knasmüller, S., Wagner, K.-H., Elbling, L., Huber, W., Kager, N., Ferk, F.,
Ehrlich, V., Nersesyan, A., Neubauer, O., Desmarchelier, A., Marin-Kuan, M.,
Delatour, T., Verguet, C., Bezençon, C., Besson, A., Grathwohl, D., Simic, T., Kundi,
M., Schilter, B., Cavin, C., 2010. Instant coffee with high chlorogenic acid levels
protects humans against oxidative damage of macromolecules. Mol. Nutr. Food
Res. 54, 1722–1733.
Hopps, V., Vetri, P., Biondi, F., Fontana, G., Mantia, G., Venturella, F., Cascio, G., 1997.
Effects of atractyligenin and some of its natural derivatives on renal function.
Fitoterapia 68, 29–34.
Jastroch, M., Hirschberg, V., Klingenspor, M., 2012. Functional characterization of
UCP1 in mammalian HEK293 cells excludes mitochondrial uncoupling artefacts
and reveals no contribution to basal proton leak. Biochim. Biophys. Acta 1817,
1660–1670.
Luciani, S., Carpenedo, F., Tarjan, E.M., 1978. Effects of Atractyloside and
Carboxyatractyloside in the whole animal. In: Santi, R., Luciani, S. (Eds.),
Atractyloside: Chemistry, Biochemistry and Toxicology. Piccin Medical Books,
Padova, Italy, pp. 109–124.
4.6.4. Evaluation of the lyophilization procedure on compounds 6
and 7
Powdered raw coffee (2.5 g) was suspended in warm water
(35 °C, 100 mL) and incubated (5 min) prior to filtration. Aliquots
of the filtrate (1 mL, n = 3) were spiked with the internal standard
working solution (5 nmol/mL, 1 mL), vortexed and the membrane
filtrate (0.45
UPLC-ToF-MS (1
zen and lyophilized (48 h, 20 °C, 120
up in water (1 mL), the internal standard working solution
(5 nmol/mL, 1 mL) added and the membrane-filtrate (0.45 m,
500 L) diluted with water (500 L) prior to UPLC-ToF-MS (1 L).
l
m, 500
L). Aliquots of the filtrate (1 mL, n = 3) were fro-
bar). The residue was taken
lL) diluted with water (500 lL) prior to
l
Ludwig, H., Obermann, H., Spiteller, G., 1974. Atractyligenin
component of roast coffee. Chem. Ber., 2409–2411.
– an essential
l
Maetzel, U., Maier, H.G., 1983. Diterpenoids in coffee II. Glycosides of atractyligenin.
Z. Lebensm. Unters. Forsch. 176, 281–284.
Maier, H.G., Wewetzer, H., 1978. Determination of diterpene glycosides in coffee. Z.
Lebensm. Unters. Forsch. 167, 105–107.
l
l
l
l

R. Lang et al. / Phytochemistry 93 (2013) 124–135
135
Müller, C., Hofmann, T., 2005. Screening of raw coffee for thiol binding site
precursors using ‘‘in bean’’ model roasting experiments. J. Agric. Food Chem. 53,
2623–2629.
Obatomi, D.K., Bach, P.H., 1998. Biochemistry and toxicology of the diterpenoid
glycoside atractyloside. Food Chem. Toxicol. 36, 335–346.
Obermann, H., Spiteller, G., 1976. The structures of the ‘‘coffee-atractylosides’’.
Chem. Ber. 109, 3450–3461.
Santi, R., Luciani, S., 1978. Atractyloside: Chemistry, Biochemistry and Toxicology.
Piccin Medical Books, Padova, Italy.
Stewart, M.J., Steenkamp, V., 2000. The biochemistry and toxicity of atractyloside: a
review. Ther. Drug. Monitor. 22, 641–649.
Vignais, P.M., Vignais, P.V., Defaye, G., 1978. Stucture–activity relationship of
atractyloside and diterpenoid derivatives on the oxidative phosphorylation and
adenine nucleotide translocation in mitochondria. In: Santi, R., Luciani, S. (Eds.),
Atractyloside: Chemistry, Biochemistry and Toxicology. Piccin Medical Books,
Padova, Italy, pp. 39–68.
Volz, N., Boettler, U., Winkler, S., Teller, N., Schwarz, C., Bakuradze, T., Eisenbrand, G.,
Haupt, L., Griffiths, L.R., Stiebitz, H., Bytof, G., Lantz, I., Lang, R., Hofmann, T.,
Somoza, V., Marko, D., 2012. Effect of coffee combining green coffee bean
constituents with typical roasting products on the Nrf2/ARE pathway in vitro
and in vivo. J. Agric. Food Chem. 60, 9631–9641.
Zaim, N., Guemouri, L., Lamnaouer, D., Benjouad, A., 2008. Study of four cases of
poisoning by Atractylis gummifera L. in Marocco. Thérapie 63, 49–54.
Stuart, B.P., Cole, R.J., Gosser, H.S., 1981. Cocklebur (Xanthium strumarium, L. var.
strumarium) intoxication in swine. Review and redefinition of the toxic
principle. Vet. Pathol. 18, 368–383.