Identification of (furan-2-yl)methylated benzene diols and triols as a novel class of bitter compounds in roasted coffee

Article abstract of DOI:10.1016/j.foodchem.2010.11.008

Preliminary studies demonstrated that the identification of unknown bitter taste compounds in roasted coffee, by means of an analytical fractionation approach, is hampered by their limited oxidative, as well as chemical stability. A synthetic-constructive

Full text of DOI:10.1016/j.foodchem.2010.11.008

Food Chemistry 126 (2011) 441–449
Contents lists available at ScienceDirect
Food Chemistry
journal homepage: www.elsevier.com/locate/foodchem
Identification of (furan-2-yl)methylated benzene diols and triols as a novel class
of bitter compounds in roasted coffee
⇑
Stefanie Kreppenhofer, Oliver Frank, Thomas Hofmann
Chair of Food Chemistry and Molecular Sensory Science, Technische Universität München, Liese–Meitner-Strasse 34, 85354 Freising, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 30 June 2010
Received in revised form 24 August 2010
Accepted 1 November 2010
Preliminary studies demonstrated that the identification of unknown bitter taste compounds in roasted
coffee, by means of an analytical fractionation approach, is hampered by their limited oxidative, as well
as chemical stability. A synthetic-constructive strategy was followed in the present investigation by per-
forming targeted reactions of putative coffee-related precursors to give candidate bitter taste molecules.
Binary mixtures of a di and trihydroxybenzene, namely pyrogallol, hydroxyhydroquinone, catechol, or 3-
and 4-methylcatechole, and a furan derivative, namely furfuryl alcohol, furan-2-aldehyde, or 5-(hydroxy-
methyl)furan-2-aldehyde, all of which are known to be present in roasted coffee, were thermally treated.
The reaction products were identified as (furan-2-yl)methylated benzene diols and triols, by means of
LC–MS and NMR experiments, and their bitter taste thresholds determined by means of sensory analysis.
Finally, LC–MS/MS studies verified the natural occurence of 4-(furan-2-ylmethyl)benzene-1,2-diol,
4-(furan-2-ylmethyl)benzene-1,2,3-triol, 4-(furan-2-ylmethyl)-5-methylbenzene-1,2-diol, and 3-(furan-
2-ylmethyl)-6-methylbenzene-1,2-diol as a novel class of bitter taste compounds in roasted coffee.
Depending on their chemical structure, the bitter taste recognition threshold of these compounds ranged
Keywords:
Coffee
Bitter taste
Furfuryl alcohol
Furan-2-aldehyde
5-(Hydroxymethyl) furan-2-aldehyde
Catechol
Pyrogallol
3-Methylcatechol
4-Methylcatechol
Hydroxyhydroquinone
between 100 and 537 lmol/l.
Ó 2010 Elsevier Ltd. All rights reserved.
1. Introduction
dicaffeoylquinic acids giving rise to a series of intensely bitter tast-
ing caffeoyl quinides (Blumberg et al., 2010; Clifford, 1979; Frank
et al., 2006, 2008). Moreover, the caffeoylquinic acids and, once
formed, their corresponding lactones, have been recently reported
to be degraded to 4-vinylcatechol, which oligomerises to produce a
family of polyhydroxylated phenylindans exhibiting a harsh and
lingering bitter taste profile (Blumberg et al., 2010; Frank et al.,
2007).
Although O-caffeoylquinic acids are long-known to be decom-
posed into a series of di and trihydroxybenzenes such as pyrogallol,
hydroxyhydroquinone, catechol, 3- and 4-methylcatechol, and 4-
ethylcatechol (Clifford, 1979; Haffenden & Yaylayan, 2005; Lang,
Müller, & Hofmann, 2006; Tressl, Bahri, Koeppler, & Jensen,
1978), their role as transient intermediates in the generation of
candidate bitter tastants upon coffee roasting is unknown. Simi-
larly, furan derivatives such as furfuryl alcohol (Shibamoto et al.,
1981) and 5-(hydroxymethyl)furan-2-aldehyde (Belitz, 1977;
Moon & Shibamoto, 2009), the later of which originates from Mail-
lard-type and caramelisation reactions of carbohydrates (Antal,
Mok, & Richards, 1990; Lewkowski, 2001; Richards, 1956), respec-
tively, have been suggested to contribute to the development of
the coffee’s bitterness. Any information, however, on the mecha-
nisms involved in transforming such reactive furans into candidate
bitter taste molecules upon coffee roasting is lacking.
Throughout history, from 14th century Arabian merchants, to
the famous coffee houses of Vienna, to the comfort of your own
home, a freshly brewed coffee has been associated with pleasure,
reward, and enjoyment by stimulating our senses from the alluring
smell of the composition of aroma-active volatiles to the first sip
inducing a well-balanced taste impression centering around bitter-
ness and sourness.
Although it is the roasting procedure which is reported to gen-
erate an orchestra of volatile key odourants (Mayer, Czerny, &
Grosch, 2000; Semmelroch & Grosch, 1995, 1996) and non-volatile
bitter taste compounds (Blumberg, Frank, & Hofmann, 2010; Frank,
Blumberg, Krümpel, & Hofmann, 2008; Frank, Blumberg, Kunert,
Zehentbauer, & Hofmann, 2007; Frank, Zehentbauer, & Hofmann,
2006), the knowledge available on the later group of molecules is
far from comprehensive. By application of a sensomics approach
to coffee beverages and coffee-related model systems, followed
by LC–MS/MS and 1D/2D NMR studies, coffee roasting was found
to induce the transesterification, epimerization, and lactonisation
of non-bitter 3-O-, 4-O-, and 5-O-caffeoylquinic acids, as well as
⇑
Corresponding author. Tel.: +49 8161/71 2902; fax: +49 8161/71 2949.
E-mail address: thomas.hofmann@wzw.tum.de (T. Hofmann).
0308-8146/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodchem.2010.11.008

442
S. Kreppenhofer et al. / Food Chemistry 126 (2011) 441–449
As the high reactivity and chemical instability of many bitter
or minus one dilution step; that is, a threshold value of 134
for 4-(furan-2-ylmethyl)benzene-1,2-diol represents a range from
67 to 268 mol/l for the astringent impression.
lmol/l
compounds have limited their unequivocal identification in
roasted coffee, by means of iterative fractionation experiments in
previous studies, the objective of the present investigation was to
thermally treat binary mixtures of di/trihydroxybenzenes and
reactive furan derivatives identified in coffee, namely furfuryl alco-
hol, furan-2-aldehyde, and 5-(hydroxymethyl)furan-2-aldehyde,
respectively. The reaction products formed should be identified
in their chemical structure by means of LC–MS/MS and 1D/2D
NMR experiments, their bitter taste threshold concentrations
determined by means of human sensory studies, and finally, their
natural occurrence in freshly prepared coffee brew verified by
means of HPLC–MS/MS.
l
2.3. Roasting experiments with a binary mixture of catechol and
furfuryl alcohol
A binary mixture of catechol (0.04 mmol) and furfuryl alcohol
(0.2 mmol), was intimately mixed with silica gel (900 mg) and ace-
tic acid (500 ll; 1% in water). After drying this mixture for 120 min
at 50 °C in a lab oven, the temperature was raised to 180 °C and kept
for 10 min. After cooling, the reaction mixture was extracted twice
with acetone/water (7/3, v/v; 5 ml), the organic solvent was sepa-
rated in vacuum and then analysed by means of RP-HPLC/DAD.
2. Materials and methods
2.4. Model reactions with mixtures of di/trihydroxybenzenes and furan
derivatives in solution
2.1. Chemicals and materials
2.4.1. Analytical scale
The following compounds were obtained commercially: pyro-
gallol, hydroxyhydroquinone, catechol, 4-methylcatechol, 3-meth-
ylcatechol, resorcinol, caffeic acid, chlorogenic acid, furfuryl
alcohol, furan-2-aldehyde, 5-(hydroxymethyl)furan-2-aldehyde
(Sigma–Aldrich, Steinheim, Germany), ethanol, methanol, formic
acid, acetic acid (Merck KGA, Darmstadt, Germany). Solvents were
of HPLC grade (J.T. Barker, Deventer, Netherlands). Deionized water
used for the chromatography was purified by means of Milli-Q-
Gradient A10 system (Millipore, Billerica, USA). Deuterated sol-
vents were supplied by Euriso-Top (Gif-Sur-Yvette, France). For
gradient flash chromatography RP-18-Bulk material (LiChroprepÒ
A binary solution of 5-O-caffeoylquinic acid, caffeic acid, cate-
chol, 3-methylcatechol, 4-methylcatechol, resorcinol, pyrogallol,
or hydroxyhydroquinone (0.04 mmol each), respectively, with fur-
furyl alcohol, furan-2-aldehyde, or 5-(hydroxymethyl)furan-2-
aldehyde (0.04–0.4 mmol each), respectively, in aqueous acetic
acid (1 ml, 1% in water) was thermally treated in a closed Pyrex
glass vial (10 ml) at 100 °C using a heated alumina block (Table
1). After various time intervals (10 min, and 1, 2, 4, and 8 h), the
reaction mixture was cooled to room temperature, membrane
filtered, and aliquots (30 ll) were directly monitored for reaction
products by means of RP-HPLC/DAD.
RP-18, 25–40 lm, Merck KGaA, Darmstadt, Germany) was used.
Sensory analyses were performed with bottled water (Evian, Da-
none, Wiesbaden, Germany). Coffee beans (Arabica Brazil, Santos),
roasted for 240 s at 230 °C, were obtained from the food industry.
Reference compounds of 3-((2-furylmethyl)sulfanyl)benzene-
1,2-diol and 3-((2-furylmethyl)sufanyl)benzene-1,2,4-triol were
synthesised and purified as reported previously (Müller,
Hemmersbach, Van’t Slot, & Hofmann, 2006).
Table 1
Reaction conditions used and reaction products formed in binary model systems of di/
trihydroxybenzenes and furfuryl alcohol, furan-2-aldehyde, and 5-(hydroxy-
methyl)furan-2-aldehyde, respectively.
Di/
Time [min]b at
100 °C
Phenol/furan
ratioc
Reaction
productd
trihydroxybenzenea
Furfuryl alcohol
Catechol
Resorcinol
Pyrogallol
120
120
60
1/5
1/5
1/5
1/5
1/1
1/1
1/1–1/10
1/1–1/10
1
2
3
4
5
6
2.2. Sensory analyses
Hydroxyhydroquinone 10
2.2.1. Panel training
4-Methylcatechol
3-Methylcatechol
Caffeic acid
120
120
n.r.
n.r.
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 with no history of known taste disor-
ders, participated for at least two years in sensory training sessions
with purified reference compounds by using the sip-and-spit
method as earlier reported in detail (Brock and Hofmann, 2008;
Frank et al., 2006; Frank et al., 2007; Frank et al., 2008; Haseleu,
Intelmann, & Hofmann, 2009; Scharbert, Holzmann, & Hofmann,
2004; Stark, Bareuther, & Hofmann, 2005).
Chlorogenic acid
Furan-2-aldehyde
Catechol
Resorcinol
Pyrogallol
Hydroxyhydroquinone n.r.
4-Methylcatechol
3-Methylcatechol
Caffeic acid
n.r.
120
60
1/1–1/10
1/5
1/10
1/1–1/10
1/1–1/10
1/1–1/10
1/1–1/10
1/1–1/10
7
8
n.r.
n.r.
n.r.
n.r.
Chlorogenic acid
2.2.2. Bitter taste recognition threshold concentrations
5-(Hydroxymethyl)furan-2-aldehyde
Catechol
Resorcinol
Pyrogallol
n.r.
120
120
1/1–1/10
1/5
1/5
1/1–1/10
1/1–1/10
1/1–1/10
1/1–1/10
1/1–1/10
To overcome carry-over effects of astringent and bitter com-
pounds, threshold concentrations of test compounds were deter-
mined in bottled water by means of the recently developed half-
tongue test (Brock and Hofmann, 2008; Frank et al., 2007; Frank
et al., 2008; Hufnagel and Hofmann, 2008; Scharbert et al., 2004;
Stark et al., 2005, 2010). Serial 1:1 dilutions of the samples (1 ml)
were presented in order of increasing concentrations to a trained
panel of twelve persons in three different sessions, using the sip-
and-spit method. The geometric mean of the last and second to last
concentrations was calculated and taken as the individual recogni-
tion threshold. At the start of the session and before each trial, the
subject rinsed with water and expectorated. Values between indi-
viduals and three separate sessions differed by not more than plus
9
10
Hydroxyhydroquinone n.r.
4-Methylcatechol
3-Methylcatechol
Caffeic acid
n.r.
n.r.
n.r.
n.r.
Chlorogenic acid
a
Binary mixtures of di/trihydroxybenzenes (0.04 mmol) and furan derivatives
(0.04–0.4 mmol) in aqueous acetic acid (1 ml, 1% in water) were thermally treated
at 100 °C for different time intervals (10–480 min).
b
Optimum heating time was determined in analytical scale model systems.
Optimum ratio of reactants was determined in analytical scale model systems.
Structures of reaction products are given in Figs. 1–3. n.r.: no reaction product
c
d
detected.

S. Kreppenhofer et al. / Food Chemistry 126 (2011) 441–449
443
2.4.2. Preparative scale
2.4.4. 4-(Furan-2-ylmethyl)benzene-1,3-diol, 2 (Fig. 1)
A binary solution of catechol, 3-methylcatechol, 4-methylcate-
chol, resorcinol, pyrogallol, or hydroxyhydroquinone (2.0 mmol
each), respectively, and furfuryl alcohol (2–10 mmol), furan-2-
aldehyde (10–20 mmol), or 5-(hydroxymethyl)furan-2-aldehyde
(10 mmol), respectively, in aqueous acetic acid (15 ml, 1% in
water) was thermally treated in a Pyrex glass vial (60 ml) at
100 °C for the times given in Table 1. After cooling, the reaction
mixtures were diluted with methanol (10 ml) and separated by
means of gradient flash chromatography using a Buechi sepacore
system (Flawil, Switzerland) equipped with a 150 ꢀ 40 mm i.d.
UV/Vis (MeOH): kmax = 212 nm, 272 nm; LC–TOF–MS: m/z
189.055545 (measured for [C11H9O3]ꢁ), 189.055718 (calculated
for [C11H9O3]ꢁ); MS–ESIꢁ: m/z 189 (100, [MꢁH]ꢁ); MS/MS
(ꢁ54 V): m/z 161 (80), 67 (100), 49 (55), 41 (40); 1H NMR
(400 MHz; d3-MeOD, COSY): d [ppm] 3.80 [s, 2H, HꢁC(7)], 5.80
[dd, 1H, J = 0.9, 3.0 Hz, HꢁC(9)], 6.22 [dd, 1H, J = 2.4, 8.0 Hz,
HꢁC(4)], 6.25 [dd, 1H, J = 1.8, 3.0 Hz, HꢁC(10)], 6.30 [d, 1H, J =
2.4 Hz, HꢁC(2)], 6.80 [d, 1H, J = 8.0 Hz, HꢁC(5)], 7.32 [s, 1H, J =
1.8 Hz HꢁC(11)]; 13C NMR (100 MHz, d3-MeOD, 135-DEPT, HMQC,
HMBC): d [ppm] 28.6 [CH2, C(7)], 103.5 [CH, C(2)], 106.7 [CH, C(9)],
107.6 [CH, C(4)], 111.3 [CH, C(10)], 117.4 [C, C(6)], 131.8 [CH, C(5)],
142.1 [CH, C(11)], 156.9 [C, C(8)], 157.1 [C, C(1)], 158.1 [C, C(3)].
column filled a slurry of RP-18 material (LiChroprep, 25–40
Merck KGaA Darmstadt, Germany). Monitoring the effluent at
220 nm, chromatography was performed at flow rate of
lm,
a
40 ml/min, starting with a mixture (100/0, v/v) of aqueous formic
acid (0.1% in water; solvent A) and methanol (solvent B) for
8 min, then increasing B to 80% within 34 min, followed by an in-
crease of B to 100% within 2 min, and finally, maintaining B at
100% for 10 min. Using an automated fraction collector, crude
fractions showing UV absorbance were collected individually,
freed from organic solvents in vacuum and then further purified
by means of preparative RP-HPLC using a 250 ꢀ 21.2 mm i.d.,
2.4.5. 4-(Furan-2-ylmethyl)benzene-1,2,3-triol, 3 (Fig. 1)
UV/Vis (MeOH): kmax = 220 nm, 280 nm; LC–TOF–MS: m/z
205.051234 (measured for [C11H9O4]ꢁ), 205.050632 (calculated
for [C11H9O4]ꢁ); MS–ESIꢁ: m/z 205 (100, [MꢁH]ꢁ); MS/MS
(ꢁ54 V): m/z 161 (100), 67 (65), 49 (35), 41 (20); 1H NMR
(400 MHz; d3-MeOD, COSY): d [ppm] 3.83 [s, 2H, HꢁC(7)], 5.90
[dd, 1H, J = 0.9, 3.2 Hz, HꢁC(9)], 6.24 [dd, 1H, J = 1.8, 3.1 Hz,
HꢁC(10)], 6.27 [d, 1H, J = 8.3 Hz, HꢁC(4)], 6.37 [d, 1H, J = 8.3 Hz,
HꢁC(5)], 7.31 [dd, 1H, J = 0.9, 1.8 Hz HꢁC(11)]; 13C NMR
(100 MHz, d3-MeOD, 135-DEPT, HMQC, HMBC): d [ppm] 28.8
[CH2, C(7)], 106.6 [CH, C(9)], 107.7 [CH, C(4)], 111.1 [CH, C(10)],
118.2 [C, C(6)], 121.1 [CH, C(5)], 134.2 [C, C(2)], 142.0 [CH,
C(11)], 145.3 [C, C(1)], 145.7 [C, (C3)], 156.7 [C, C(8)].
5 lm, Microsorb 100–5 C18 column (Varian, Darmstadt,
Germany). Monitoring the effluent at 220 and 280 nm, chroma-
tography was performed at a flow rate of 18 ml/min with a mix-
ture (25/75, v/v) of aqueous formic acid (0.1% in water; solvent A)
and methanol (solvent B) for 5 min, increasing the methanol con-
tent to 60% within 20 min, then to 100% within 5 min, and finally,
maintaining solvent B at 100% for 5 min. After separation of the
solvent in vacuum, the residue was suspended in water (10 ml),
freeze-dried twice, then analysed by means of LC–MS/MS and
1D/2D NMR experiments.
2.4.6. 5-(Furan-2-ylmethyl)benzene-1,2,4-triol, 4 (Fig. 1)
UV/Vis (MeOH): kmax = 204 nm, 296 nm; LC–TOF–MS: m/z
205.051234 (measured for [C11H9O4]ꢁ), 205.050632 (calculated
for [C11H9O4]ꢁ); MS–ESIꢁ: m/z 205 (100, [MꢁH]ꢁ); MS/MS
(ꢁ54 V): m/z 137 (100), 124 (50), 67 (15), 41 (20); 1H NMR
(400 MHz; d3-MeOD, COSY): d [ppm] 3.77 [s, 2H, HꢁC(7)], 5.93
[d, 1H, J = 3.1 Hz, HꢁC(9)], 6.26 [dd, 1H, J = 1.8, 3.1 Hz,
HꢁC(10)], 6.34 [s, 1H, HꢁC(2)], 6.47 [s, 1H, HꢁC(5)], 7.32 [d, 1H,
J = 1.8 Hz, HꢁC(11)]; 13C NMR (100 MHz, d3-MeOD, 135-DEPT,
HMQC, HMBC): d [ppm] 27.1 [CH2, C(7)], 103.1 [CH, C(2)], 105.3
[CH, C(9)], 109.8 [CH, C(10)], 115.3 [C, C(6)], 116.6 [CH, C(5)],
137.5 [C, C(1)], 140.7 [CH, C(11)], 143.9 [C, C(3)], 147.5 [C, (C4)],
155.2 [C, C(8)].
2.4.3. 4-(Furan-2-ylmethyl)benzene-1,2-diol, 1 (Fig. 1)
UV/Vis (MeOH): kmax = 220 nm, 284 nm; LC–TOF–MS: m/z
189.05457 (measured for [C11H9O3]ꢁ), 189.055718 (calculated for
[C11H9O3]ꢁ); MS–ESIꢁ: m/z 189 (100, [MꢁH]ꢁ); MS/MS (ꢁ54 V):
m/z 161 (100), 67 (65), 49 (35), 41 (20); 1H NMR (400 MHz; d3-
MeOD, COSY): d [ppm] 3.77 [s, 2H, HꢁC(7)], 5.96 [d, 1H, J =
3.1 Hz, HꢁC(9)], 6.27 [pt, 1H, J = 1.8 Hz, HꢁC(10)], 6.52 [dd, 1H,
J = 1.9, 8.0 Hz, HꢁC(5)], 6.64 [d, 1H, J = 1.9 Hz, HꢁC(1)], 6.68 [d,
1H, J = 8.0 Hz, HꢁC(4)], 7.33 [d, 1H, J = 1.8 Hz HꢁC(11)]; 13C
NMR (100 MHz, d3-MeOD, 135-DEPT, HMQC, HMBC): d [ppm]
34.6 [CH2, C(7)], 106.8 [CH, C(9)], 111.3 [CH, C(10)], 116.4 [CH,
C(4)], 117.0 [CH, C(1)], 121.1 [CH, C(5)], 131.4 (C, C(6)], 142.5
[CH, C(11)], 144.9 [C, C(2)], 146.3 [C, (C3)], 156.8 [C, C(8)].
2.4.7. 4-(Furan-2-ylmethyl)-5-methylbenzene-1,2-diol, 5 (Fig. 1)
UV/Vis (MeOH): kmax = 220 nm, 282 nm; LC–TOF–MS: m/z
203.07151 (measured for [C12H11O3]ꢁ), 203.071368 (calculated
for [C12H11O3]ꢁ); MS–ESIꢁ: m/z 203 (100, [MꢁH]ꢁ); MS/MS
Fig. 1. Chemical structures of reaction products 1–6 formed upon reaction of di/trihydroxybenzenes with furfuryl alcohol.

444
S. Kreppenhofer et al. / Food Chemistry 126 (2011) 441–449
(ꢁ64 V): m/z 159 (70), 67 (100), 49 (50), 41 (40); 1H NMR
(400 MHz; d3-MeOD, COSY): d [ppm] 2.11 [s, 3H, HꢁC(12)], 3.76
[s, 2H, CꢁH(7)], 5.85 [d, 1H, J = 3.0 Hz, HꢁC(9)], 6.26 [t, 1H, J =
2.0, 3.0 Hz, HꢁC(10)], 6.55 [s, 1H, HꢁC(5)], 6.57 [s, 1H, HꢁC(2)],
7,33 [s, 1H, J = 2.0 Hz, HꢁC(11)]]; 13C NMR (100 MHz, d3-MeOD,
135-DEPT, HMQC, HMBC): d [ppm] 18.7 [CH3, C(12)], 32.3 [CH2,
C(7)], 106.7 [CH, C(9)], 111.2 [CH, C(10)], 118.0 [CH, C(5)], 118.3
[CH, C(2)], 128.5 [C, C(6)], 128.8 [C, C(1)], 142.3 [CH, C(11)],
144.1 [C, C(3)], 144.7 [C, (C4)], 156.4 [C, C(8)].
2.4.8. 3-(Furan-2-ylmethyl)-6-methylbenzene-1,2-diol, 6 (Fig. 1)
UV/Vis (MeOH): kmax = 220 nm, 280 nm; LC–TOF–MS: m/z
203.070836 (measured for [C12H11O3]ꢁ), 203.071368 (calculated
for [C12H11O3]ꢁ); MS–ESIꢁ: m/z 203 (100, [MꢁH]ꢁ); MS/MS
(ꢁ56 V): m/z 175 (90), 67 (100), 49 (45), 41 (25); 1H NMR
(400 MHz; d3-MeOD, COSY): d [ppm] 2.11 [s, 3H, HꢁC(12)], 3.82
[s, 2H, HꢁC(7)], 5.80 [dd, 1H, J = 0.8, 3.1 Hz, HꢁC(9)], 6.25 [dd,
1H, J = 1.8, 3.1 Hz, HꢁC(10)], 6.48 [d, 1H, J = 8.0 Hz, HꢁC(5)],
6.57 [d, 1H, J = 8.0 Hz, HꢁC(4)], 7.32 [d, 1H, J = 1.8 Hz,
HꢁC(11)]; 13C NMR (100 MHz, d3-MeOD, 135-DEPT, HMQC,
HMBC): d [ppm] 12.0 [CH3, C(12)], 33.0 [CH2, C(7)], 106.7 [CH,
C(9)], 111.3 [CH, C(10)], 113.1 [CH, C(4)], 121.8 [CH, C(5)], 124.7
[C, C(6)], 129.7 [C, C(3)], 142.2 [CH, C(11)], 144.7 [C, C(1)], 144.8
[C, (C2)], 156.8 [C, C(8)].
Fig. 2. Chemical structures of reaction products 7 and 8 formed upon reaction of
resorcinol and pyrogallol, respectively, with furan-2-aldehyde.
HMQC, HMBC): d [ppm] 28.6 [CH2, C(7)], 37.6 [CH, C(12)], 103.4
[CH, C(2)], 103.5 [CH, C(15, 150)], 107.0 [CH, C(9)], 107.1 [CH,
C(17, 170)], 107.5 [CH, C(4)], 109.0 [CH, C(10)], 117.7 [C, C(6)],
121.9 [C, C(13, 130)], 131.1 [CH, C(18, 180)], 131.7 [CH, C(5)],
155.1 [C, C(1)], 156.6 [C, C(14, 140)], 156.9 [C, C(16, 160)], 157.6
[C, C(8)], 157.6 [C, C(11)], 157.8 [C, C(3)].
2.4.9. 4,40-(Furan-2-ylmethanediyl)dibenzene-1,3-diol, 7 (Fig. 2)
UV/Vis (MeOH): kmax = 224 nm, 280 nm; LC–TOF–MS: m/z
297.077153 (measured for [C17H13O5]ꢁ), 297.076847 (calculated
for [C17H13O5]ꢁ); MSꢁESIꢁ: m/z 297 (100, [MꢁH]ꢁ); MS/MS
(ꢁ56 V): m/z 253 (55), 187 (60), 109 (45), 41 (20); 1H NMR
(400 MHz; d3-MeOD, COSY): d [ppm] 5.70 [d, 1H, J = 3.1 Hz,
HꢁC(9)], 5.84 [s, 1H, HꢁC(7)], 6.18 [dd, 2H, J = 2.4, 8.0 Hz,
HꢁC(4, 40)], 6.25 [dd, 1H, J = 1.8, 3.1 Hz, HꢁC(10)], 6.28 [d, 2H,
J = 2.4 Hz, HꢁC(2, 20)], 6.60 [d, 2H, J = 8.0 Hz, HꢁC(5, 50)], 7.34
[d, 1H, J = 1.8 Hz, HꢁC(11)]; 13C NMR (100 MHz, d3-MeOD, 135-
DEPT, HMQC, HMBC): d [ppm] 36.1 [CH, C(7)], 102.1 [CH, C(4,
40)], 105.6 [CH, C(5, 50)], 106.8 [CH, C(9)], 109.4 [CH, C(10)], 120.1
[C, C(6, 60)], 129.1 [CH, C(2, 20)], 140.7 [CH, C(11)], 155.3 [C, C(1,
10)], 156.3 [C, C(3, 30)], 158.2 [C, C(8)].
2.4.12. 2-(Bis-(2,3,4-trihydroxyphenyl)methyl)-5-(2,3,4-trihydroxy-
benzyl))furan 10 (Fig. 3)
UV/Vis (MeOH): kmax = 216 nm, 272 nm; LC–TOF–MS: m/z
467.0997 (measured for [C24H19O10]ꢁ), 467.098360 (calculated
for [C24H19O10]ꢁ); MS–ESIꢁ: m/z 419 (100, [MꢁH]ꢁ); MS/MS
(ꢁ66 V): m/z 341 (100), 175 (65), 125 (40), 109 (15), 41 (5); 1H
NMR (400 MHz; d3-MeOD, COSY): d [ppm] 3.77 [s, 2H, HꢁC(7)],
5.59 [dd, 1H, J = 0.9, 3.1 Hz, HꢁC(9)], 5.73 [d, 1H, J = 3.1 Hz,
HꢁC(10)], 5.82 [s, 1H, HꢁC(12)], 6.21 [d, 2H, J = 8.5 Hz, HꢁC(17,
170)], 6.24 [d, 2H, J = 8.0 Hz, HꢁC(18, 180)], 6.36 [d, 1H, J =
8.3 Hz, HꢁC(4)], 6.39 [d, 1H, J = 8.3 Hz; HꢁC(5)]; 13C NMR
(100 MHz, d3-MeOD, 135-DEPT, HMQC, HMBC): d [ppm] 28.9
[CH2, C(7)], 38.9 [CH, C(12)], 107.0 [CH, C(17, 170)], 107.4 [CH,
C(9)], 107.8 [CH, C(4)], 109.2 [CH, C(10)], 118.5 [C, C(6)], 120.7
[CH, C(18, 180)], 121.2 [CH, C(5)], 122.6 [C, C(13, 130)], 134.2 [C,
C(15, 150)], 134.3 [C, C(2)], 144.8 [CH, C(14, 140)], 145.1 [C, C(1)],
145.5 [C, C(16, 160)], 145.6 [C, C(3)], 155.1 [C, C(8)], 157.3 [C,
(C(11)].
2.4.10. 4,40-(Furan-2-ylmethanediyl)dibenzene-1,2,3-triol, 8 (Fig. 2)
UV/Vis (MeOH): kmax = 220 nm, 272 nm; LC–TOF–MS: m/z
329.067259 (measured for [C17H13O7]ꢁ), 329.066676 (calculated
for [C17H13O7]ꢁ); MS–ESIꢁ: m/z 329 (100, [MꢁH]ꢁ); MS/MS
(ꢁ64 V): m/z 283 (25), 203 (35), 125 (100), 41 (10); 1H NMR
(400 MHz; d3-MeOD, COSY): d [ppm] 5.72 [d, 1H, J = 3.5 Hz,
HꢁC(9)], 5.88 [s, 1H, HꢁC(7)], 6.17 [d, 2H, J = 8.5 Hz, HꢁC(5, 50)],
6.21ꢁ6.26 [m, 3H, HꢁC(6, 60, 10)], 7.33 [d, 1H, J = 1.8 Hz,
HꢁC(11)]; 13C NMR (100 MHz, d3-MeOD, 135-DEPT, HMQC,
HMBC): d [ppm] 38.2 [CH, C(7)], 107.4 [CH, C(4, 40)], 108.4 [CH,
C(9)], 110.8 [CH, C(10)], 120.6 [CH, C(5,50)], 122.3 [C, C(6, 60)],
134.2 [C, C(2, 20)], 142.1 [CH, C(11)], 144.9 [C, C(1, 10)], 145.6 [C,
C(3, 30)], 159.3 [C, C(8)].
2.5. Preparation of the coffee beverage
After grinding the coffee beans by means of a batch mill (IKA,
Staufen, Germany), a portion (54 g) of coffee powder, placed in a
coffee filter (No. 4, Melitta, Germany), was percolated with boiling
water until the filtrate reached a volume of 1.0 l. The coffee bever-
age obtained was immediately cooled to room temperature in an
ice-bath prior to HPLC–MS/MS analysis.
2.4.11. 2-(Bis-(2,4-dihydroxyphenyl)methyl)-5-(2,4-dihydroxy-
benzyl))furan 9 (Fig. 3)
2.6. High-performance liquid chromatography (HPLC)
UV/Vis (MeOH): kmax = 216 nm, 280 nm; LC–TOF–MS: m/z
419.116258 (measured for [C24H23O7]ꢁ), 419.113627 (calculated
for [C24H23O7]ꢁ); MS–ESIꢁ: m/z 419 (100, [MꢁH]ꢁ); MS/MS
(ꢁ54 V): m/z 309 (100), 253 (25), 159 (75), 149 (40), 109 (50), 41
(20); 1H NMR (400 MHz; d3-MeOD, COSY): d [ppm] 3.75 [s, 2H,
HꢁC(7)], 5.57 [d, 1H, J = 3.1 Hz, HꢁC(10)], 5.74 [d, 1H, J = 3.1 Hz,
HꢁC(9)], 5.80 [s, 1H, HꢁC(12)], 6.19 [dd, 2H, J = 2.3, 8.2 Hz,
HꢁC(17, 170)], 6.20 [dd, 1H, J = 2.3, 8.0 Hz, HꢁC(4)], 6.28 [m, 3H,
HꢁC(2, 15, 150)], 6.63 [d, 2H, J = 8.2 Hz, HꢁC(18, 180)], 6.81 [d,
1H; J = 8.2 Hz, HꢁC(5)]; 13C NMR (100 MHz, d3-MeOD, 135-DEPT,
The analytical HPLC apparatus (Kontron, Eching, Germany) con-
sisted of a low-pressure gradient system 525 HPLC pump, an M800
gradient mixer,
540 + diode array detector. Chromatography was performed on
m, Luna Phenyl-Hexyl column (Phenomenex,
a type 560 autosampler, and a DAD type
250 ꢀ 4.6 mm i.d., 5
l
Aschaffenburg, Germany), operated with a flow rate of 0.8 ml/
min. Semi-preparative chromatography was done using a HPLC
system consisting of two Sykam S1122 high pressure pumps
(Eresing, Germany), an Sunchrom Spectraflow 600 DAD detector
(Friedrichsdorf, Germany), a Rheodyne injector with a 2 ml loop

S. Kreppenhofer et al. / Food Chemistry 126 (2011) 441–449
445
Fig. 3. Chemical structures of reaction products 9 and 10 formed upon reaction of resorcinol and pyrogallol, respectively, with 5-(hydroxymethyl)furan-2-aldehyde.
(Alsbach a.d. Bergstrasse, Germany), and a 250 ꢀ 21.2 mm i.d.,
m, Microsorb 100–5 C18 column (Varian, Darmstadt, Germany)
operated with a flow rate of 18 ml/min.
compound 9, and m/z 467 ? 341 and 467 ? 174 for compound
10. Nitrogen served as nebulizer gas (65 psi) and as turbo gas
(350 °C) for solvent drying (55 psi). After injection of the sample
5
l
(5 ll), chromatography was performed with a flow rate of 250 ll
using a solvent gradient starting with aqueous formic acid (0.1%
in water) for 3 min, then increasing the methanol content to 40%
within 10 min, then to 60% within 5 min, to 70% within 12 min,
and finally, to 100% within 2 min, thereafter maintaining the meth-
anol content for an additional 3 min.
2.7. Gradient flash chromatography
The gradient flash chromatography apparatus (Buechi, Flawil,
Switzerland) consisted of two C-605 pumps with a C-615 pump
manager, a C-635 UV/Vis detector, and a C-660 fraction collector.
Chromatography was performed with a flow rate of 40 ml/min
on
a
self-packed 150 ꢀ 40 mm i.d. polypropylene cartridge
2.10. Nuclear magnetic resonance (NMR) spectroscopy
filled with LiChroprep, 25–40
Darmstadt, Germany).
lm, RP-18 Material (Merck KGaA,
The 1H and 13C NMR experiments were conducted on a DRX 400
spectrometer (Bruker, Rheinstetten, Germany). The samples were
dissolved in d3-MeOD (Euriso-Top, Giv-Sur-Yvette, France) and
then analysed in 178 ꢀ 5 mm NMR tubes (Norell, Sp5000 Landis-
ville, USA) at 298 K. Chemical shifts were determined using tetra-
methylsilane (TMS) as the internal standard in the proton
dimension and from the carbon signal of d3-MeOD (49.3 ppm) in
the carbon dimension. While data processing was performed using
Topspin Version 1.3 (Bruker, Rheinstetten), the individual data
interpretation was done by means of MestReNovaÒ 5.1.0–2940
(Mestrelab Research S.L., Santiago de Compostela, Spain).
2.8. LC/time-of-flight mass spectrometry (LC/TOF–MS)
High resolution mass spectra were measured on a Bruker Mi-
cro–TOF mass spectrometer (Bruker Daltronics, Bremen, Germany)
and referenced with sodium formate.
2.9. High-performance liquid chromatography / tandem mass
spectrometry (HPLC–MS/MS)
The Agilent 1200 Series HPLC-system consisted of a pump, a
degasser and an autosampler (Agilent, Waldbronn, Germany) and
was connected to a 4000 Q Trap triple quadrupole/linear ion trap
mass spectrometer (Applied Biosystems/MDS Sciex, Darmstadt,
Germany) with an electrospray ionisation (ESI) device running in
negative ionisation mode, with a spray voltage of ꢁ4500 V. The
quadruples operated at unit mass resolution. Nitrogen served as
a curtain gas (20 psi) with the declustering potential was set at
ꢁ10 to ꢁ80 V in the ESIꢁ mode. The mass spectrometer operated
in the full scan mode monitoring negative ions. Fragmentation of
[MꢁH]ꢁ molecular ions into specific product ions was induced by
collision with nitrogen (5 ꢀ 10ꢁ5 Torr) and a collision energy of
ꢁ10 to ꢁ70 V. For instrumentation control and data acquisition,
the Sciex Analyst software (v1.5) was used.
3. Results and discussion
3.1. Preliminary studies
Preliminary studies demonstrated that the identification of pre-
viously unknown bitter taste compounds in roasted coffee, by
means of an analytical fractionation approach is hampered by their
limited oxidative, as well as chemical stability. In consequence, a
synthetic-constructive strategy based on the targeted reaction of
putative taste precursors to give candidate bitter taste molecules,
followed by the verification of their natural occurrence in roasted
coffee, is believed to be a more promising approach. As di and tri-
hydroxybenzenes such as pyrogallol, hydroxyhydroquinone, cate-
chol, 3- and 4-methylcatechole, and 4-ethylcatechol (Clifford,
1979; Haffenden & Yaylayan, 2005; Lang et al., 2006; Tressl et al.,
1978) as well as furan derivatives such as furfuryl alcohol, furan-
2-aldehyde, 5-(hydroxymethyl)-furan-2-aldehyde (Belitz, 1977;
Moon & Shibamoto, 2009; Shibamoto et al., 1981) have been iden-
tified as rather reactive molecules in roasted coffee samples, the
screening of thermally treated binary mixtures of di/trihydroxy-
benzenes and furan derivatives was supposed to be a fruitful strat-
egy for the discovery of previously undetermined, candidate bitter
tasting reaction products in roasted coffee.
For HPLC–MS/MS analysis of the furan-2-ylmethylated benzene
diols and triols in a coffee beverage, an analytical 150 ꢀ 2.1 mm
i.d., 5 lm, Luna Phenyl–Hexyl column (Phenomenex, Aschaffen-
burg, Germany) was connected to the mass spectrometer operating
in the multiple reaction monitoring (MRM), detecting negative
ions. For a duration of 50 ms, the mass transitions m/z 189 ? 161
and 189 ? 67 were used for the analysis of
1 and 2, m/z
205 ? 67 and 205 ? 41 for compound 3, m/z 205 ? 67 and
205 ? 43 for compound 4, m/z 203 ? 67 and 203 ? 159 for com-
pound 5, m/z 203 ? 67 and 203 ? 175 for compound 6, m/z
297 ? 187 and 297 ? 253 for compound 7, m/z 329 ? 125 and
329 ? 203 for compound 8, m/z 419 ? 309 and 419 ? 159 for

446
S. Kreppenhofer et al. / Food Chemistry 126 (2011) 441–449
3.2. Reaction of catechol and furfuryl alcohol under low-moisture and
aqueous conditions, respectively
respectively. Interestingly, reaction products were generated
exclusively in the model systems containing the di and trihydroxy-
benzenes, whereas caffeic acid and 5-O-caffeoyl quinic acid did not
show any reactivity under the conditions applied in the model
experiments. The optimised reaction conditions, which afforded
the reaction products from di/trihydroxybenzenes and furfuryl
alcohol in the highest yields, are summarised in Table 1.
For structure determination of the respective target com-
pounds, the individual model reactions were carried out in a pre-
parative scale, the crude reaction mixtures were separated by
means of gradient flash chromatography, followed by semi-pre-
parative RP-HPLC to give the target molecules 2–6 (Fig. 1). The
structure of this reaction product was unequivocally determined
by means of LC–MS/MS, LC–TOF–MS, and 1D/2D NMR spectro-
scopic experiments.
As preliminary quantitative studies on roasted coffee revealed
catechol and furfuryl alcohol as the quantitatively predominating
dihydroxybenzene and furan derivative, respectively, a binary mix-
ture of these compounds was thermally treated at 180 °C for
10 min on silica gel, before the presence of catalytic amounts of
acetic acid. After cooling, the reaction mixture was extracted with
acetone/water to separate the reaction products from the silica gel
and the extract obtained was analysed by means of RP-HPLC con-
nected to a diode array detector or a mass spectrometer, respec-
tively. HPLC analysis revealed one main reaction product (1),
eluting after 21 min and exhibiting UV absorption maxima at 220
and 284 nm, besides the educts furfuryl alcohol and catechol. LC–
MS analysis (ESIꢁ) revealed m/z 189 as the pseudomolecular ion
([MꢁH]ꢁ) and high resolution mass spectrometry indicated an ele-
mental composition of C11H10O3 for the target molecule.
As the amounts of 1 were too low for an unequivocal structure
elucidation by means of NMR spectroscopy, a binary mixture of
catechol and furfuryl alcohol was heated at 100 °C for 120 min in
aqueous solution containing small amounts of acetic acid. Compar-
ison of the HPLC chromatograms recorded for the dry-heated and
the aqueous model system revealed the same peak pattern for both
reaction systems. As the up-scaling of the model reaction was
rather simple for the aqueous system, this later reaction was per-
formed in preparative scale. After cooling, the reaction mixture
was separated by means of gradient flash chromatography on
RP-18 material, followed by final purification of reaction product
1 using preparative RP-18 HPLC, and subsequent lyophilisation to
give the target compound, showing the pseudomolecular ion m/z
189 ([MꢁH]ꢁ) as a pale, amorphous powder in a purity of more
than 98% (HPLC, NMR).
Reaction of resorcinol and furfuryl alcohol revealed compound
2 (Fig. 1) exhibited a pseudmolecular ion of m/z 189 ([MꢁH]ꢁ), fit-
ting well with that of compound 1. Differing from the 1H NMR
spectrum of 1, the signal of the proton HꢁC(4) was high-field
shifted to 6.22 ppm in compound 2 and also the homonuclear cou-
3
4
plings J4,5 (8.0 Hz) and J2,4 (2.4 Hz) confirmed the linkage of the
(furan-2-yl)methyl moiety to C(6) of the resorcinol moiety.
Strengthened by the heteronuclear connectivities (HMBC) ob-
served between C(7), HꢁC(5), as well as HꢁC(9), compound 2
was identified as 4-(furan-2-ylmethyl)benzene-1,3-diol (Fig. 1),
thus confirming data reported earlier (Kenzo, 1961).
LC–MS and LC–TOF–MS analysis of compounds 3 and 4 (Fig. 1),
isolated from the reaction mixture of furfuryl alcohol with pyrogal-
lol and hydroxyhydroquinone, respectively, showed a pseudomo-
lecular ion of m/z 205 ([MꢁH]ꢁ) and an elemental composition of
C
11H10O4 for both isomers. The 1H NMR spectrum of 3 and 4,
respectively, showed the proton signals expected for the (furan-
2-yl)methyl moiety, two coupling adjacent arene protons
(3J = 8.3 Hz) of the pyrogallol moiety of compound 3, as well as
two non-coupling, isolated aromatic protons in the hydroxyhydro-
quinone moiety of compound 4. Taking all spectroscopic data into
account, these compounds were identified as the previously not re-
ported 4-(furan-2-ylmethyl)benzene-1,2,3-triol (3) and 5-(furan-
2-ylmethyl)benzene-1,2,4-triol (4), respectively (Fig. 1).
Compounds 5 and 6 (Fig. 1), isolated from the reaction mixture
of furfuryl alcohol with 4-methylcatechol and 3-methylcatechol,
respectively, both showed a molecular weight of 204 Da. The 1H
NMR spectra of 5 and 6 exhibited the proton signals assigned for
the (furan-2-yl)methyl moiety and the proton resonances expected
for 4-methylcatechol and 3-methylcatechol, respectively, lacking
one arene proton each. Signal assignment by means of an HMBC
experiment led to the unequivocal identification of the previously
not reported 4-(furan-2-ylmethyl)-5-methylbenzene-1,2-diol (5)
and 3-(furan-2-ylmethyl)-6-methylbenzene-1,2-diol (6), respec-
tively (Fig. 1).
The 1H NMR spectrum recorded for 1 showed seven resonance
signals integrating for eight protons. The resonance signals ob-
served at 3.77, 5.96, 6.26, and 7.31 ppm were assigned as the pro-
tons of a (furan-2-yl)methyl moiety, whereas the three signals
detected at 6.68, 6.52, and 6.64 ppm were proposed to belong to
the aromatic protons HꢁC(1), HꢁC(4), and HꢁC(5) of catechol.
The large coupling constant of 8 Hz found for the protons HꢁC(4)
and HꢁC(5) demonstrated the vicinal position of these two protons
in the catechol moiety and indicated the linkage of the (furan-2-
yl)methyl moiety to C(6) of the catechol ring. This was further
strengthened by the heteronuclear connectivity found in the HMBC
spectrum between the carbon atom C(7) of the (furan-2-yl)methyl
moiety and the protons HꢁC(1) and HꢁC(5) of the catechol moiety.
Taking all the spectroscopic data into consideration, compound 1
was unequivocally identified as 4-(furan-2-ylmethyl)benzene-
1,2-diol (Fig. 1). Although the structure of compound 1 was pro-
posed already in 1960 on the basis of IR experiments (Kenzo,
1960; Kenzo, 1961), a report on its final confirmation by means
of LC–MS and NMR spectroscopy is still lacking to the best of our
knowledge.
3.4. Reaction of di/trihydroxybenzenes and furan-2-aldehyde
To compare the compounds formed from coffee-related di and
trihydroxybenzenes in the reaction with furfuryl alcohol, with
those formed in the presence of furan-2-aldehyde, analytical scale
reactions were performed with binary mixtures of catechol, resor-
cinol, pyrogallol, hydroxyhydroquinone, 3-methylcatechol, 4-
methylcatechol, caffeic acid, or 5-O-caffeoyl quinic acid, respec-
tively, and furan-2-aldehyde in 1% aqueous acetic acid heated at
100 °C varying in the phenol/furan ratio (1:1, 1:5, 1:10) and the
heating time (10, 60, 120, 240, 480 min). HPLC analysis revealed
that reaction products were generated exclusively in the model
systems containing resorcin and pyrogallol, respectively, the aro-
matic ring of both molecules is substituted with hydroxyl groups
in the meta-position (Table 1).
3.3. Reaction of di/trihydroxybenzenes and furfuryl alcohol
In order to investigate the reaction products formed from other
coffee-related di and trihydroxybenzenes and furfuryl alcohol, first
comparative analytical scale experiments were performed with
binary mixtures of catechol, resorcin, pyrogallol, hydroxyhydroqui-
none, 3-methylcatechol, 4-methylcatechol, caffeic acid, or 5-O-caf-
feoyl quinic acid, respectively, combined with furfuryl alcohol in
1% aqueous acetic acid heated at 100 °C, varying in the phenol/fur-
an ratio (1:1, 1:5, 1:10) and the heating time (10, 60, 120, 240,
480 min). Each reaction mixture was analysed by means of RP-
HPLC coupled to a diode array detector and a mass spectrometer,

S. Kreppenhofer et al. / Food Chemistry 126 (2011) 441–449
447
For structure determination, the individual model reactions
were carried out in a preparative scale, the crude reaction mixtures
were separated by means of gradient flash chromatography, fol-
lowed by semi-preparative RP-HPLC to afford the purified target
molecules 7 and 8 (Fig. 2), which were identified by means of
LC–MS/MS, LC–TOF–MS, and 1D/2D NMR spectroscopy.
HꢁC(5) of compound 9. By means of a heteronuclear multiple-bond
coherence experiment (HMBC), heteronuclear correlations were ob-
served between carbon C(7) and proton HꢁC(5), as well as between
the methin carbon atom C(12) and the arene protons HꢁC(18) and
´
HꢁC(18), respectively, thus demonstrating the structure of com-
pound 9 as the previously not reported 2-(bis-(2,4-dihydroxy-
phenyl)methyl)-5-(2,4-dihydroxybenzyl) furan (Fig. 3).
LC–MS and LC–TOF–MS analysis of compound 7, isolated from
the model reaction of resorcinol and furan-2-aldehyde, revealed a
pseudomolecular ion of m/z 297 ([MꢁH]ꢁ) and an elemental com-
position of C17H14O5, thus indicating a reaction product formed
from one molecule of furan-2-aldehyde and two molecules of res-
orcin. The 1H NMR spectrum of 7 exhibited seven resonance signals
integrating for nine protons. The proton signals of the (furan-2-yl)-
methyl moiety showed integrals of one, whereas the resonance sig-
nals observed at 6.18, 6.28, and 6.58 ppm integrated for two
protons each, thus confirming one (furan-2-yl)methyl moiety and
two resorcinol moieties in the target molecule. In addition, the
Compound 10, isolated from the reaction mixture of pyrogallol
and 5-(hydroxymethyl)furan-2-aldehyde, showed a pseudomolec-
ular ion of m/z 419 ([MꢁH]ꢁ) in the LC–MS spectrum and an ele-
mental composition of
C24H20O10 by means of LC–TOF–MS
analysis. The 1H NMR data of compound 10 were rather similar
to those found for compound 9, with the exception of the lack of
one arene proton for each benzene moiety. Taking all the spectro-
scopic data into account, the structure of compound 10 was clearly
determined as the previously unknown 2-(bis-(2,3,4-trihydroxy-
phenyl)methyl)-5-(2,3,4-trihydroxybenzyl)furan (Fig. 3).
3
3
homonuclear coupling constants J4,5 = 8.0 Hz and J2,4 = 1.5 Hz, as
well as the heteronuclear couplings observed between C(7) and
HꢁC(5)/HꢁC(50), in the HMBC spectrum indicated that the (fur-
an-2-yl)methyl moiety is bridging two resorcin molecules at posi-
tion C(6) and led to the unequivocal identification of molecule 7 as
4,40-(furan-2-ylmethanediyl) dibenzene-1,3-diol (Fig. 2).
LC–MS analyses of compound 8, generated from pyrogallol and
furan-2-aldehyde, revealed a molecular weight of 330 Da and indi-
cated the presence of two pyrogallol moieties and one furan-2-
aldehyde moiety in the target molecule. Differing from the 1H
NMR spectrum of compound 7, the spectrum of 8 was lacking
the arene protons HꢁC(2)/HꢁC(20), thus leading to the identifica-
tion of the target compound as the previously unknown 4,40-(fur-
an-2-ylmethanediyl)dibenzene-1,2,3-triol (8, Fig. 2).
3.6. Taste recognition threshold concentrations
Prior to sensory analysis, the purity of the reaction products 1–
10 as well as their parent di/trihydroxybenzenes, were checked by
1H NMR spectroscopy as well as HPLC–MS. To overcome carry-over
effects of astringent and bitter compounds, taste threshold concen-
trations of the individual compounds were determined in bottled
water (pH 6.0) by means of the half-tongue test using the sip-
and-spit method (Brock and Hofmann, 2008; Frank et al., 2007;
Frank et al., 2008; Scharbert et al., 2004; Stark et al., 2005). The
trained panel of twelve sensory subjects recognised an astringent
mouthfeel as well as a clear bitter taste for each of the compounds
evaluated. Depending on their chemical structure, the recognition
3.5. Reaction of di/trihydroxybenzenes and 5-(hydroxymethyl)furan-
2-aldehyde
thresholds ranged between 16 and 900
lmol/L for astringency
and between 100 and 1800 mol/l for bitter taste (Table 2). For
l
each class of compounds, the modification of the di and trihydr-
oxybenzenes with the different furan moieties induced a signifi-
cant change in their taste threshold concentrations. With the
As 5-(hydroxymethyl)furan-2-aldehyde is combining the reac-
tivity of furfuryl alcohol and furan-2-aldehyde in the same mole-
cule, the question arose as to which reaction products were
formed in the reaction with the coffee-related di and trihydroxy-
benzenes. Among all the analytical model reactions performed (Ta-
ble 1), resorcinol and pyrogallol each exclusively gave a specific
reaction product when reacted with 5-(hydroxymethyl)furan-2-
aldehyde. After semi-preparative up-scaling of these reaction sys-
tems, the reaction products 9 and 10 were isolated by means of
gradient flash chromatography, followed by RP-HPLC, before their
chemical structures were identified by means of LC–MS/MS, LC–
TOF–MS and 1D/2D NMR spectroscopic experiments.
Table 2
Human taste threshold concentrations of di/trihydroxybenzenes and selected reac-
tion products.
Compounda
Threshold
concentration
[l
mol/l]b
Bitter Astringent
Benzene-1,2-diol (catechol)
4-(Furan-2-ylmethyl)benzene-1,2-diol (1)
3-((2-Furylmethyl)sulfanyl)benzene-1,2-diol (11)
1313
537
350
671
134
n.d.
LC–MS and LC–TOF–MS analysis of compound 9, generated from
resorcin and 5-(hydroxymethyl)furan-2-aldehyde, revealed the
pseudomolecular ion m/z 419 ([MꢁH]ꢁ) and the elemental composi-
tion C24H24O7, thus indicating that three molecules of resorcin re-
acted with one molecule 5-(hydroxymethyl)furan-2-aldehyde. The
1H NMR spectrum exhibited 10 resonance signals integrating for a
total of 14 carbon-bound protons. The resonance signal observed
at 3.74 ppm integrated for two protons and was assigned as the
methylene group HꢁC(7), whereas the three protons resonating at
5.57, 5.74, and 5.80 ppm were assigned as the protons HꢁC(9),
HꢁC(10), and HꢁC(12) of the former 5-(hydroxymethyl)furan-2-
aldehyde moiety. The remaining proton signals observed in the 1H
NMR spectrum corresponded to the arene protons of three resorcin
moieties. The signals detected at 6.19, 6.28, and 6.63 ppm integrated
Benzene-1,3-diol (resorcinol)
1800
812
177
340
900
333
66
4-(Furan-2-ylmethyl)benzene-1,3-diol (2)
4,40-(Furan-2-ylmethanediyl)dibenzene-1,3-diol (7)
2-(Bis-(2,4-dihydroxyphenyl)methyl)-5-(2,4-
dihydroxybenzyl) furan (9)
110
Benzene-1,2,3-triol (pyrogallol)
297
328
457
455
78
62
42
16
4-(Furan-2-ylmethyl)benzene-1,2,3-triol (3)
4,40-(Furan-2-ylmethanediyl)dibenzene-1,2,3-triol (8)
2-(Bis-(2,3,4-trihydroxyphenyl)methyl)-5-(2,3,4-
trihydroxybenzyl) furan (10)
Benzene-1,2,4-triol (hydroxyhydroquinone)
5-(Furan-2-ylmethyl)benzene-1,2,4-triol (4)
3-((2-Furylmethyl)sufanyl)benzene-1,2,4-triol (12)
1336
317
135
74
20
n.d.
4-Methylbenzene-1,2-diol (4-methylcatechol)
4-(Furan-2-ylmethyl)-5-methylbenzene-1,2-diol (5)
540
279
363
60
3
for two protons each, showed the coupling constants J17/18 = 8.2,
3-Methylbenzene-1,2-diol (3-methylcatechol)
3-(Furan-2-ylmethyl)-6-methylbenzene-1,2-diol (6)
371
100
131
79
3
3
3
0
0
0
0
J17 /18 = 8.2, J17/15 = 2.3, J17 /15 = 2.3, and were assigned as
HꢁC(17)/HꢁC(170), HꢁC(15)/HꢁC(150), and HꢁC(18)/HꢁC(180).
The third resorcin moiety showed three resonances at 6.20, 6.28,
and 6.81 ppm corresponding to the protons HꢁC(4), HꢁC(2) and
a
Compound numbering refers to the chemical structures given in Figs. 1–4.
Taste threshold concentrations were determined by means of half-tongue test
in bottled water (pH 6.0), n.d.: not determined.
b

448
S. Kreppenhofer et al. / Food Chemistry 126 (2011) 441–449
exception of the pyrogallol derivatives (3, 8, 10), the various reac-
tion products exhibited lower bitter taste threshold concentrations
than the parent di and trihydroxybenzenes, e.g. the bitter taste
threshold of resorcin decreased by a factor of 2.2 (2), 10.2 (7),
and 5.3 (9) after reaction with furfuryl alcohol, furan-2-aldehyde,
and 5-(hydroxymethyl)furan-2-aldehyde, respectively (Table 2).
Interestingly, the chemical structures of compound 1 and 4
shared the identical carbo skeleton with that of 3-((2-furylmeth-
yl)sulfanyl)benzene-1,2-diol (11) and 3-((2-furylmethyl)sufa-
nyl)benzene-1,2,4-triol (12), which were recently identified in
coffee beverages and found to be formed upon oxidative coupling
of catechol and hydroxyhydroquinone, respectively, with the
odour-active 2-furfurylthiol as shown in Fig. 4 (Müller, Hemmers-
bach, Van’t Slot, & Hofmann, 2006; Müller and Hofmann, 2007).
This structural similarity prompted us to investigate the taste
activity of the thio ethers 11 and 12 (Table 2). Compounds 11
and 12 showed bitter taste recognition thresholds of 350 and
135
lmol/l being rather close to the data found for 1 (537
lmol/
l) and 4 (317
lmol/l), thus demonstrating that the additional sul-
Fig. 4. Chemical structures and formation of the thio ethers 11 and 12, previously
identified in roasted coffee (17, 18).
phur atom in 11 and 12 does not significantly influence the bitter
taste activity of that class of compounds.
Fig. 5. HPLC–MS/MS chromatogram (MRM mode) of the analysis of (furan-2-yl)methylated benzene diols and triols 1, 3, 5, and 6 in coffee brew.
Fig. 6. Reaction mechanism leading to the formation of (furan-2-yl)methylated benzene diols and triols 1, 3, 5, 6 from furfuryl alcohol and 5-O-chlorogenic acid upon coffee
roasting.

S. Kreppenhofer et al. / Food Chemistry 126 (2011) 441–449
449
Belitz, H.D. (1977). Taste-active substances in coffee. In: Proceedings of the 7th
colloquium of the international coffee science association; ASIC, (pp. 3–252).
Bremen, Hamburg, 24.
Blumberg, S., Frank, O., & Hofmann, T. (2010). Quantitative studies on the influence
of bean roasting parameters and hot water percolation on the concentrations of
bitter compounds in coffee brew. Journal of Agricultural and Food Chemistry, 58,
3720–3728.
Brock, A., & Hofmann, T. (2008). Identification of the key astringent compounds in
spinach (Spinacia oleracea) by means of the taste dilution analysis. Chemical
Precipitation, 4, 268–281.
3.7. Identification of bitter compounds in coffee brew
In order to verify the occurrence of the bitter tasting reaction
products formed from di/trihydroxybenzenes and furan derivatives
in coffee beverages, a beverage freshly prepared from medium
roasted coffee beans was analysed for compounds 1–10 by means
of HPLC–MS/MS (ESIꢁ) operating in the multiple reaction monitor-
ing (MRM) mode. Prior to analysis, characteristic mass transitions
and optimised instrument settings were selected for each individ-
ual compound in tuning runs.
Comparison of the retention time as well as the characteristic
mass transitions of each target compound detected in coffee with
those of the corresponding reference compound led to the
unequivocal identification of the bitter tastants 1, 3, 5, and 6 in
the coffee sample (Fig. 5). In addition, the identity of these com-
pounds in coffee was confirmed by means of co-chromatography
with the corresponding reference compound isolated from the
model reaction systems. Interestingly, all the compounds (1, 3, 5,
6) detected in roasted coffee are generated upon reaction of di
and trihydroxybenzenes with furfuryl alcohol, whereas the com-
pounds 2, 4, and 7–10, which are generated from furan-2-aldehyde
and 5-(hydroxymethyl)furan-2-aldehyde, respectively, could not
be detected in the coffee brew.
Clifford, M. N. (1979). Chlorogenic acids
– Their complex nature and routine
determinations in coffee beans. Food Chemistry, 4, 63–71.
Frank, O., Blumberg, S., Krümpel, G., & Hofmann, T. (2008). Structure determination
of 3-O-caffeoyl-epi-quinide, an orphan bitter lactone in roasted coffee. Journal of
Agricultural and Food Chemistry, 56, 9581–9585.
Frank, O., Blumberg, S., Kunert, C., Zehentbauer, G., & Hofmann, T. (2007). Structure
determination and sensory analysis of bitter-tasting 4-vinylcatechol oligomers
and their identification in roasted coffee by means of LC-MS/MS. Journal of
Agricultural and Food Chemistry, 55, 1945–1954.
Frank, O., Zehentbauer, G.,
&
Hofmann, T. (2006). Bioresponse-guided
decomposition of roasted coffee beverage and identification of key bitter taste
compounds. European Food Research and Technology, 222, 492–508.
Haffenden, L. J. W., & Yaylayan, V. A. (2005). Mechanism of formation of redox-
active hydroxylated benzenes and pyrazine in 13C-labeled glycin/D-glucose
model systems. Journal of Agricultural and Food Chemistry, 53, 9742–9746.
Haseleu, G., Intelmann, I., & Hofmann, T. (2009). Structure determination and
sensory evaluation of novel bitter compounds formed from b-acids of hops
(Humulus lupulus L.) upon wort boiling. Food Chemistry, 116, 71–81.
Hufnagel, J. C., & Hofmann, T. (2008). Quantitative reconstruction of the nonvolatile
sensometabolome of a red wine. Journal of Agricultural and Food Chemistry, 56,
9190–9199.
Kenzo, T. (1960). Condensation products of furfuryl alcohol. V. Condensation
products of furfuryl alcohol with polyhydric phenols. Nippon Kagaku Zasshi, 81,
336–337.
4. Conclusions
Kenzo, T. (1961). Condensation products of furfuryl alcohol. VII. Some characteristic
infrared absorption bands of furan derivatives. Nippon Kagaku Zasshi, 82,
373–376.
Lang, R., Müller, C., & Hofmann, T. (2006). Development of a stable isotope dilution
analysis with liquid chromatography–tandem mass spectrometry detection for
the quantitative analysis of di and trihydroxybenzenes in foods and model
systems. Journal of Agricultural and Food Chemistry, 54, 5755–5762.
Lewkowski, J. (2001). Synthesis, chemistry and application of 5-hydroxymethyl-
furfural and its derivatives. ARKIVOC, 17, 54.
Mayer, F., Czerny, M., & Grosch, W. (2000). Sensory study of the character impact
aroma compounds of a coffee beverage. European Food Research and Technology,
211, 272–276.
Moon, J.-K., & Shibamoto, T. (2009). Role of roasting conditions in the profile of
volatile flavor chemicals formed from coffee beans. Journal of Agricultural and
Food Chemistry, 57, 5823–5831.
Müller, C., Hemmersbach, S., Van’t Slot, G., & Hofmann, T. (2006). Synthesis and
structure determination of covalent conjugates formed from the sulfury-roasty
smelling 2-furfurylthiol and di- or trihydroxybenzenes and their identification
in coffee brew. Journal of Agricultural and Food Chemistry, 54, 10076–10085.
Müller, C., & Hofmann, T. (2007). Quantitative studies on the formation of phenol/2-
furfurylthiol conjugates in coffee beverages toward the understanding of the
molecular mechanisms of coffee aroma staling. Journal of Agricultural and Food
Chemistry, 55, 4095–4102.
Richards, E. L. (1956). Nonenzymatic browning: The reaction between glucose and
glycerine the dry state. Biochemical Journal, 64, 639–644.
Scharbert, S., Holzmann, N., & Hofmann, T. (2004). Identification of the astringent
taste compounds in black tea infusions by combining instrumental analysis and
human bioresponse. Journal of Agricultural and Food Chemistry, 52, 3498–3508.
Semmelroch, P., & Grosch, W. (1995). Analysis of roasted coffee powders and brews
The data obtained in this investigation clearly demonstrate that,
once formed upon coffee roasting by degradation of O-caffeoylqui-
nic acids, the di and trihydroxybenzenes catechol, pyrogallol,
3-methylcatechol, and 4-methylcatechol do react with furfuryl
alcohol to give (furan-2-yl)methylated benzene diols and triols as
a novel class of bitter taste compounds in roasted coffee. On the
basis of the model reactions performed, it can be concluded that
acid-catalysed dehydration of furfuryl alcohol gives the furfuryl
cation as a transient intermediate, which reacts via an electrophilic
substitution with di and trihydroxybenzenes to yield the bitter
tasting (furan-2-yl)methylated benzenes 1, 3, 5, and 6 (Fig. 6). In
contrast, the sulphur-containing bitter compounds 11 and 12,
earlier identified in roasted coffee (Müller, Hemmersbach, Van’t
Slot, & Hofmann, 2006; Müller and Hofmann, 2007) are generated
upon oxidative coupling of the odour-active 2-furfurylthiol to the
di and trihydroxybenzene catechol and hydrohydroquinone,
respectively (Fig. 4).
To demonstrate the percent abundance of these four novel bit-
ter taste compounds to the bitterness of coffee beverages, the
development of a stable isotope dilution analysis for their accurate
quantification is currently in progress.
by gas chromatography–olfactometry of headspace samples. Lebensmittel
Wissenschaft –und Technologie, 28, 310–313.
Semmelroch, P., & Grosch, W. (1996). Studies on character impact odorants of coffee
brews. Journal of Agricultural and Food Chemistry, 44, 537–543.
-
Acknowledgement
Shibamoto, T., Harada, K., Mihara, J., Nishimura, K., Yamaguchi, A., Aitoku, A., et al.
(1981) Application of HPLC for evaluation of coffee flavor quality. In: G.
charambous, G. Inglett (Eds.), The quality of foods and beverages (vol. 2) (p. 311),
New York: Academic Press.
Stark, T., Bareuther, S., & Hofmann, T. (2005). Sensory-guided decomposition of
roasted cocoa nibs (Theobroma cacao) and structure determination of taste-
active polyphenols. Journal of Agricultural and Food Chemistry, 53, 5407–5418.
Stark, T., Wollmann, N., Wenker, K., Lösch, S., Glabasnia, A., & Hofmann, T. (2010).
Matrix-calibrated LC–MS/MS quantitation and sensory evaluation of oak
ellagitannins and their transformation products in red wines. Journal of
Agricultural and Food Chemistry, 58, 6360–6369.
This research project was supported by grants from FEI (For-
schungskreis der Ernährungsindustrie e.V., Bonn, Germany), the
AiF (Arbeitsgemeinschaft industrieller Forschung), and the German
Ministry of Economics (Project No. 15752N). In addition, we thank
Kraft Foods for financial support.
References
Tressl, R., Bahri, D., Koeppler, H., & Jensen, A. (1978). Diphenols and caramel
compounds in roasted coffees of different varieties. II. Zeitschrift fur
Lebensmittel-untersuchung und -forschung, 167, 111–114.
Antal, M. J., Mok, W. S. L., & Richards, G. N. (1990). Mechanism of formation of 5-
hydroxymethyl-2-furaldehyde from
D-fructose and sucrose. Carbohydrate
Research, 199, 91–109.