Food Chemistry 446 (2024) 138884

Contents lists available at ScienceDirect

Food Chemistry
journal homepage: www.elsevier.com/locate/foodchem

Identification of mozambioside roasting products and their bitter taste

receptor activation
Coline Czech a, b, Tatjana Lang b, Angelika Graßl b, Alexandra Steuer a, b, Antonella Di Pizio b, c,
Maik Behrens b, Roman Lang b, *
a
TUM Graduate School, TUM School of Life Sciences Weihenstephan, Technical University of Munich, Alte Akademie 8, 85354 Freising, Germany
b
Leibniz Institute for Food Systems Biology at the Technical University of Munich, Lise-Meitner-Str. 34, 85354 Freising, Germany
c
Chemoinformatics and Protein Modelling, School of Life Science, Technical University of Munich, Lise-Meitner-Str. 34, 85354 Freising, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: Arabica coffee contains the bitter-tasting diterpene glycoside mozambioside, which degrades during coffee
Bitter taste receptors roasting, leading to yet unknown structurally related degradation products with possibly similar bitter-receptor-
Coffee activating properties.
Molecular docking
The study aimed at the generation, isolation, and structure elucidation of individual pyrolysis products of
Structure elucidation
mozambioside and characterization of bitter receptor activation by in vitro analysis in HEK 293T-Gα16gust44
Mozambioside pyrolysis products
cells.
The new compounds 17-O-β-D-glucosyl-11-hydroxycafestol-2-on, 11-O-β-D-glucosyl-16-desoxycafestol-2-on,
11-O-β-D-glucosyl-(S)-16-desoxy-17-oxocafestol-2-on, 11-O-β-D-glucosyl-15,16-dehydrocafestol-2-on, and 11-O-
β-D-glucosyl-(R)-16-desoxy-17-oxocafestol-2-on were isolated from pyrolyzed mozambioside by HPLC and
identified by NMR and UHPLC-ToF-MS. Roasting products 11-O-β-D-glucosyl-(S)-16-desoxy-17-oxocafestol-2-on,
11-O-β-D-glucosyl-15,16-dehydrocafestol-2-on, and 11-O-β-D-glucosyl-(R)-16-desoxy-17-oxocafestol-2-on had
lower bitter receptor activation thresholds compared to mozambioside.
Molecular docking simulations revealed the binding modes of the compounds 11-O-β-D-glucosyl-15,16-dehy­
drocafestol-2-on and 11-O-β-D-glucosyl-(R)-16-desoxy-17-oxocafestol-2-on and their aglycone 11-hydroxycafes­
tol-2-on in the two cognate receptors TAS2R43 and TAS2R46.
The newly discovered roasting products 17-O-β-D-glucosyl-11-hydroxycafestol-2-on, 11-O-β-D-glucosyl-(S)-16-
desoxy-17-oxocafestol-2-on, 11-O-β-D-glucosyl-15,16-dehydrocafestol-2-on, and 11-O-β-D-glucosyl-(R)-16-
desoxy-17-oxocafestol-2-on were detected in authentic roast coffee brew by UHPLC-ToF-MS and could contribute
to coffee’s bitter taste impression.

1. Introduction (United States Department of Agriculture [USDA], 06/2023). Europe

(24.4 %), the United States (15.7 %), and Brazil (13.4 %) are the most
Coffee is a crop of considerable global economic importance (Food significant markets. Combined, they consumed ~53.5 % of the world­
and Agricultural Organization of the United Nations, FAO). In 2020, the wide production (USDA, 12/2023).
largest producers of the two commercially relevant varieties Coffea Raw coffee is processed into roasted coffee by dry heating, typically
Arabica L. and Coffea canephora (ratio ~53:47) were Brazil (36.3 %), at temperatures around 220 ◦ C and above (Illy & Viani, 2005, Chapter
Vietnam (17.3 %), Indonesia (7.4 %), and Colombia (8.1 %) (FAO, 03/ 4). The roasting is considered the critical step in the production process.
2023). Roughly 10.2 × 106 t of raw coffee were harvested in 2022/23 It changes the flavor from pea-like into alluring toasty notes and turns

Abbreviations: CGA, chlorogenic acid; CQL, caffeoyl quinide; DKP, diketopiperazine; HPLC, high-performance liquid chromatography; TAS2R, Taste receptor type
2; NMR, Nuclear magnetic resonance; ToF-MS, Time-of-flight mass spectrometry; UHPLC, ultra-high-performance liquid chromatography; PDA, photo-diode array
detection; SD, standard deviation; ECL, extracellular loop.
* Corresponding author.
E-mail addresses: c.czech.leibniz-lsb@tum.de (C. Czech), t.lang.leibniz-lsb@tum.de (T. Lang), a.grassl.leibniz-lsb@tum.de (A. Graßl), a.steuer.leibniz-lsb@tum.de
(A. Steuer), a.dipizio.leibniz-lsb@tum.de (A. Di Pizio), m.behrens.leibniz-lsb@tum.de (M. Behrens), r.lang.leibniz-lsb@tum.de (R. Lang).

https://doi.org/10.1016/j.foodchem.2024.138884
Received 23 November 2023; Received in revised form 7 February 2024; Accepted 25 February 2024
Available online 1 March 2024
0308-8146/© 2024 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
C. Czech et al. Food Chemistry 446 (2024) 138884

the beans brittle and dark brown (Clarke & Mcrae, 1987, Chapter 4; analyses of the elucidated structures; 5) detection of the elucidated
Parliament, 2000, Chapter 20). pyrolysis products of 1 in authentic coffee brew.
Most of the roasting takes place in Europe. In 2022, Italy (295 × 103
t), Germany (251 × 103 t), and Switzerland (107 × 103 t) were among 2. Materials and methods
the leading producers in the world in terms of exported volume of
roasted coffee (FAO, 12/2023). Lipofectamine 2000 and Fluo4-AM were from Thermo Fisher Sci­
Coffee brew, in principle, is the water extract of ground and roasted entific (Darmstadt, Germany), somatostatin 14 was from Bachem
coffee seeds (Parliament, 2000, Chapter 20). The beverage is consumed (Bubendorf, Switzerland). All other chemicals were purchased from
worldwide and offered in many variations, e.g., cold brew, filter coffee, Sigma Aldrich (Taufkirchen, Germany) and used as received. Solvents
or espresso. Part of this popularity is linked to the pharmacologic for UHPLC-ToF-MS analysis were from J.T. Baker (Deventer, The
properties of the contained alkaloid caffeine. Immediate benefits, e.g., Netherlands). Millipore water was from an Advantage A 10 System
increased alertness and focus, improved mood, reduced fatigue, and (Millipore, Molsheim, France). Mozambioside (1) was isolated from
long-term effects, e.g., prevention of cognitive decline, have been re­ coffee (100 % Arabica, Colombia), as reported (Lang et al., 2015). The
ported (Nehlig, 2018). purity of 1 was verified by NMR (cf. Supporting Figs. 7, 8, and 11), ToF-
However, the hedonic value is the central driver of coffee con­ MS (Supporting Table 1), and analytical UHPLC-PDA (cf. 2.2 and the
sumption, with toasty notes, acidity, and bitterness being crucial char­ Section 1. “Isolation of mozambioside (1)” in the Supporting Informa­
acteristics of the flavor profile (Batali et al., 2022). Caffeine is the best- tion). Purity was >99 %.
known bitter-tasting coffee ingredient. It activates human bitter re­
ceptors TAS2R7, -R10, -R14, -R43, and -R46 (Meyerhof et al., 2010), and 2.1. Pyrolysis of mozambioside
brews from commercial roasted coffees contain roughly 600–700 mg/l
(Weiss et al., 2010). However, coffee with reduced caffeine content has Mozambioside (1, 80 mg) was dissolved in millipore water (~10 ml)
been judged equally bitter (Šeremet et al., 2022), underscoring the and air-dried overnight in a crystallizing dish. The dry material was then
presence of additional bitter compounds. exposed to heat on a heating plate (220 ◦ C, 4.5 min) and cooled to room
The roasting process is known to generate several new classes of temperature. The material was taken up with water/methanol (50/50,
bitter compounds from precursors, contributing to the brew’s overall ~10 ml), ultrasonicated (30 min), and membrane filtered (0.45 µm, Ø
taste. Chlorogenic acids (CGAs) are esters of quinic acid and hydrox­ 25 mm, Rotalibo®, Carl Roth, Karlsruhe, Germany). The clear filtrate
ycinnamates such as ferulic acid, p-coumaric acid, and caffeic acid. With containing the roasting products was analyzed by UHPLC-PDA (2.2),
6–10 % of the dry matter, CGAs are highly abundant in raw coffee fractionated via preparative HPLC (2.3), and individual compounds
(Clifford, 1999). Applying the senso-analytical concept of “taste-dilution were purified by semi-preparative HPLC (2.4).
analysis” Frank et al. (2006, 2008) identified caffeoyl-quinides
(“chlorogenic acid lactones”, CQLs) as essential bitter compounds in 2.2. Ultra-high-performance liquid chromatography-photo-diode array
roasted coffee. CQLs are formed from CGAs by heat-induced lactoniza­ detection (UHPLC-PDA)
tion of the quinic acid moiety, with the 3-caffeoylquinic-1,5-lactone and
4-caffeoylquinic-1,5-lactone as the dominant derivatives (Farah et al., The UHPLC-system (Shimadzu, Duisburg, Germany) was a Nexera XS
2005). Pyrolysis of caffeic acid leads to the formation of tetra- with two LC-40D XS pumps, autosampler (SIL-40C-XS), column oven
oxygenated phenylindanes (Guillot et al., 1996), which Frank et al. (CTO-40S) and PDA detector (SPD-M40) set to record the spectra be­
(2007) described as “harshly bitter”. These bitter compounds origi­ tween 190 and 500 nm. Samples (5 µl) were injected onto a C18 column
nating from CGAs are soluble in hot water and, therefore, are extracted (Kinetex C18, 100 × 2.1 mm, 1.7 μm, Phenomenex, Aschaffenburg,
into the coffee brew (Blumberg et al., 2010). As another bitter-tasting Germany). The flow rate was 400 µl/min. Eluent A was 0.1 % formic
compound class, Ginz and Engelhardt (2000, 2001) reported that dike­ acid in water, and eluent B was 0.1 % formic acid in acetonitrile. The
topiperazines (DKPs) formed by roasting coffee’s protein fraction are binary gradient started with 5 % B for 2 min, increased to 30 % in 12
extracted into the coffee brew, contributing to the taste. Data from Stark min, to 100 % in 1 min (1.5 min isocratic), then lowered to 5 % in 0.5
and Hofmann (2005) on the bitterness of roasted cocoa nibs support the min and held for 3 min. The total analysis time was 20 min.
assumption that DKPs could contribute to coffee bitterness.
Mozambioside (1 in Fig. 1, D) is a diterpene glycoside in Arabica 2.3. Fractionation of roasting products by preparative high-performance
coffee (Richter & Spiteller, 1979) that elicits a bitter taste impression of liquid chromatography (HPLC)
roughly twice that of caffeine (Prewo et al., 1990). Since 1 was found to
be particularly abundant in caffeine-free coffee species, e.g., Coffea Roasting products were fractionated by preparative HPLC using a
pseudozanguebariae Bridsen, it has been hypothesized that 1 served as a Büchi PurePrep system (Büchi, Flawil, Switzerland) with ELSD and UV
substitute for caffeine as a plant defense (Prewo et al., 1990). The detection (280 nm) on a PFP column (Luna PFP(2), 21.2 mm × 250 mm,
compound activates bitter taste receptors TAS2R43 and TAS2R46 (Lang 5µ, Phenomenex, Aschaffenburg, Germany). A binary gradient was run
et al., 2020) and has a reported human bitter threshold of 60 ± 10 µM with millipore water (0.1 % formic acid, eluent A) and acetonitrile (0.1
(Lang et al., 2015). Quantitative studies revealed that roughly 80 % of 1 % formic acid, eluent B). Gradient elution started with a 20 ml/min flow
is pyrolyzed during coffee roasting, leading to ~10 µM residual at 100 % A (isocratic for 2 min). Subsequently, B was increased to 10 %
mozambioside in the coffee brew (Lang et al., 2015). within 10 min, to 27 % within 25 min, to 80 % within 2 min (isocratic for
The coffee diterpene glycoside mozambioside (1) is hypothesized as 1 min), and returned to starting conditions within 1.5 min (isocratic for
a precursor for a group of yet unknown structurally related pyrolysis 6 min). Peaks were collected in a fraction collector with 2 ml per glass
products that could contribute to coffee bitter taste. tube and analyzed by UHPLC-PDA and UHPLC-ToF-MS.
Therefore, the objectives of the study were: 1) model pyrolysis of 1
and isolation of non-volatile pyrolysis products; 2) structure elucidation 2.4. Purification of roasting products by semi-preparative high-
of non-volatile pyrolysis products using Nuclear magnetic resonance performance liquid chromatography (HPLC)
spectroscopy (NMR) and Ultra-high-performance liquid
chromatography-Time-of-Flight-mass spectrometry (UHPLC-ToF-MS); Fractions containing individual roasting products were purified by
3) screening for activation of 26 human bitter taste receptors, acquisi­ semi-preparative HPLC on a Jasco system (Pfungstadt, Germany)
tion of activation threshold and dose–response data; 4) investigation of coupled to a UV/VIS detector (280 nm). The samples (100 µl) were
possible compound-receptor binding modes by molecular docking injected manually onto a Hyperclone C18 (250 mm × 10 mm, 5 µm,

2
C. Czech et al. Food Chemistry 446 (2024) 138884

Fig. 1. (A) The mixture of roasting products activated the human bitter receptors TAS2R43 (black) and TAS2R46 (blue). The asterisk indicates the threshold
concentration, the lowest concentration at which receptor-transfected cells show a significantly higher signal (Student’s t-test, P < 0.05) than empty vectors. (B)
UHPLC-PDA (281 nm) analysis revealed that pyrolysis of mozambioside (1) gave rise to degradation products. (C) Shows excerpts of HSQC spectra with key signals
for structure elucidation (cf. Table 2 and Table 3, and Supplemental Information for full spectra). (D) Structures of new roasting products 2–6 formed by pyrolysis of
1, roasting product 7, and aglycone 8 (obtained by enzymatic cleavage of 2); Glc: β-D-glucose. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

3
C. Czech et al. Food Chemistry 446 (2024) 138884

Phenomenex, Aschaffenburg, Germany) column. Eluents were 0.1 % achieved on a reversed-phase column with guard column of the same
formic acid in millipore water (eluent A) and acetonitrile (eluent B); the type (Kinetex C18, 100 × 2.1 mm, 1.7 µm, Phenomenex, Aschaffenburg,
flow rate was 4.7 ml/min. The binary gradient started with 5 % B Germany) and gradient elution with 0.1 % formic acid (mobile phase A)
(isocratic for 5 min), was increased to 30 % within 20 min (isocratic for and 0.1 % formic acid in acetonitrile (mobile phase B) at a flow rate of
4 min), increased to 100 % within 3 min (isocratic for 3 min), and 0.3 ml/min with a column oven temperature of 40 ◦ C. The injected
subsequently lowered to 5 % within 5 min (isocratic for 5 min). Peaks sample volume was 5 µl. The LC gradient started at 5 % B for 3 min, was
were collected manually and dried by lyophilization. increased in 10 min to 30 % B, in 4 min to 100 % B (2 min), and 1 min to
17-O-β-D-glucosyl-11-hydroxycafestol-2-on (2). Retention time in the starting conditions at 5 % B (5 min).
UHPLC-ToF-MS (cf. 2.6): 10.67 min, ToF-MS (ESI+) found m/z 531.2215
(C26H36O10 + Na, [M+Na]+]); cf. Table 2 (13C data) and Table 3 (1H
data). 2.8. Functional calcium mobilization assay
11-O-β-D-glucosyl-16-desoxycafestol-2-on (3). Retention time in
UHPLC-ToF-MS (cf. 2.6): 13.33 min, ToF-MS (ESI+) found m/z 515.2258 Functional screening. The functional screening experiment was per­
(C26H36O9 + Na, [M+Na]+]); cf. Table 2 (13C data) and Table 3 (1H formed as in previous publications (Lang et al., 2020; Ziegler & Behrens,
data). 2021). Briefly, HEK 293T-Gα16gust44 cells were grown on 96-well
11-O-β-D-glucosyl-(S)-16-desoxy-17-oxocafestol-2-on (4). Retention plates under regular conditions (DMEM, 10 % FCS, 1 % penicillin/
time in UHPLC-ToF-MS (cf. 2.6): 14.00 min, ToF-MS (ESI+) found m/z streptomycin, 1 % glutamine; 37 ◦ C, 5 % CO2, saturated air humidity)
513.2103 (C26H34O9 + Na, [M+Na]+]); cf. Table 2 (13C data) and and transiently transfected with cDNA constructs coding for 26 func­
Table 3 (1H data). tional human TAS2Rs (Meyerhof et al., 2010; Lang et al., 2023) using
11-O-β-D-glucosyl-15,16-dehydrocafestol-2-on (5). Retention time in lipofectamine 2000. The empty vector was transfected as a negative
UHPLC-ToF-MS (cf. 2.6): 14.17 min, ToF-MS (ESI+) found m/z 513.2105 control (mock). After 24 h, cells were treated for 1 h with Fluo4-AM dye
(C26H34O9 + Na, [M+Na]+]); cf. Table 2 (13C data) and Table 3 (1H in the presence of probenecid (2.5 mM), then washed with C1 buffer
data). (130 mM NaCl, 5 mM KCl, 2 mM CaCl2, 10 mM glucose, 10 mM HEPES;
11-O-β-D-glucosyl-(R)-16-desoxy-17-oxocafestol-2-on (6). Retention pH 7.4) and placed in a fluorometric imaging plate reader (FLIPRTetra,
time in UHPLC-ToF-MS (cf. 2.6): 14.42 min, ToF-MS (ESI+) found m/z Molecular Devices, San Jose, USA). The roasted mozambioside was
513.2105 (C26H34O9 + Na, [M+Na]+]); cf. Table 2 (13C data) and dissolved in C1 buffer and further diluted 1 to 10 with C1 buffer. After
Table 3 (1H data). application of test substances to the cells, changes in fluorescence (at
510 nm excitation and 488 nm emission) were measured. Cell viability
2.5. Formation of 11-hydroxycafestol-2-on (8) by enzymatic hydrolysis was tested by subsequently adding somatostatin 14 (100 nM) to the
of 17-O-β-D-glucosyl-11-hydroxycafestol-2-on (2) cells.
Receptor activation threshold and dose response. HEK 293T-
17-O-β-D-glucosyl-11-hydroxycafestol-2-on (2, 12 mg, 23.6 µmol) Gα16gust44 cells were transfected with TAS2R43, TAS2R46, and an
was dissolved in sodium acetate buffer (5 g/100 ml, pH 4.8, 5 ml) and empty vector as for the screening experiments. The test substances (2–6,
stirred (12 h, 37 ◦ C) with β-glucuronidase (Helix pomatia, 500 µl). After 8) were diluted with C1 buffer and measured analogously to the
the addition of acetonitrile (10 ml), the precipitate was removed by screening experiment. Results were based on three independent (n = 3)
centrifugation (4 ◦ C, 5 min, 4000 rpm), and the supernatant was evap­ measurements, each performed in duplicate. When calculating the
orated in a stream of air. The residue was taken up in water (3 ml) and compound-specific changes in fluorescence (Δ F/F), data from identi­
purified by semi-preparative HPLC (cf. 2.4). 11-hydroxycafestol-2-on cally treated mock controls were subtracted and normalized based on
was obtained as an off-white foam. background fluorescence. Dose-response curves were established by
11-hydroxycafestol-2-on (8). Retention time in UHPLC-ToF-MS (cf. plotting signal amplitudes versus log compounds concentration. The half
2.6): 11.18 min, ToF-MS (ESI+) found m/z 369.1675 (C20H26O5 + Na, maximal effective concentration (EC50) was identified by nonlinear
[M+Na]+]); cf. Table 2 (13C data) and Table 3 (1H data). regression using the equation y = (max − min)/[1 + (x/EC50)-Hillslope]
+ min, where x is the substance concentration. Plots were generated
2.6. Nuclear magnetic resonance spectroscopy (NMR) using SigmaPlot 14.0.
Statistics. Threshold concentrations were defined as the lowest con­
1D and 2D nuclear magnetic resonance (NMR) data were recorded centration with bitter receptor responses significantly larger than the
on an AV500 NMR spectrometer (500 MHz, Bruker Avance III, Bruker, control (Student’s t-test, P < 0.05, Table 1). EC50 values of mozambio­
Rheinstetten, Germany). ROESY spectra were recorded on an AV600 side before and after roasting were compared by Student’s t-test (un­
NMR spectrometer. Data was processed using the Topspin software paired, two-tailed, Table 1). Differences between efficacies (Emax, ΔF/F,
(version 4.1.3, Bruker, Rheinstetten, Germany) and MestReNova soft­
ware (version 14.2.1; Mestrelab Research S.L., Santiago de Compostella, Table 1
Spain). Proton signals were deduced from 1H and 1H,13C-HSQC spectra Threshold concentration and EC50 values of mozambioside before and after
due to the substantial overlap of signals. Annotated spectra are compiled pyrolysis.
in the Supporting Information (Supporting Figs. 7–42). Compound Max. conc. Threshold (µM)b EC50 (µM)
(µM)a
TAS2R43 TAS2R46 TAS2R43 TAS2R46c
2.7. Ultra-high-performance liquid chromatography-time-of-flight-mass-
d
spectrometry (UHPLC-ToF-MS) 1 1000 30 300 50.6 ± –
13.8
e
1, roasted 300 3 300 26.6 ± –
Exact mass data acquisition for compound characterization was 13.1
achieved on a TripleToF 6600 mass spectrometer (Sciex, Darmstadt, a
Maximum applied concentration.
Germany) coupled to an ExionLC UHPLC system (Sciex, Darmstadt, b
Receptor activation threshold is the lowest concentration at which receptor-
Germany). Ionization was positive electrospray (ESI+). MS source pa­ transfected cells show a significantly higher signal (Student’s t-test, P < 0.05)
rameters were as follows: ion spray voltage 5.5 kV, source temperature than empty vector-transfected cells.
450 ◦ C, heater gas 65 psi, nebulizer gas 55 psi, curtain gas (nitrogen) 35 c
No saturation was reached, therefore no EC50 could be calculated.
psi. Survey scan (information-dependent acquisition, IDA), CE 35 V, d
Literature data from Lang et al., 2020.
product ion DP 80 V, Spread 15 V. Chromatographic separation was e
Mixture of compounds obtained by pyrolysis at 220 ◦ C for 4.5 min (cf. 2.1).

4
C. Czech et al. Food Chemistry 446 (2024) 138884

Table 2
13
C NMR data of compounds 1–8.
11-O-β-D-glucosyl- 17-O-β-D- 11-O-β-D-glucosyl- 11-O-β-D-glucosyl- 11-O-β-D-glucosyl- 11-O-β-D-glucosyl- Bengalensol 11-
cafestol-2-on glucosyl- 16-desoxycafestol- (S)-16-desoxy-17- 15,16- (R)-16-desoxy-17- (7)***,b hydroxycafestol-
(mozambioside, 1)*,a cafestol-2- 2-on (3)* oxocafestol-2-on dehydrocafestol-2-on oxocafestol-2-on 2-on (8)*
on (2)* (4)** (5)* (6)**
13
#C C δ (ppm)a
1 54.3, CH2 52.9, CH2 52.9, CH2 52.4, CH2 53.8, CH2 52.3, CH2 54.4, CH2 54.3, CH2
2 187.6, C 186.2, C 186.2, C 184.5, C 187.5, C 184.3, C 185.6, C 187.7, C
3 148.0, C 146.6, C 146.5, C 146.1, C 148.1, C 146.3, C 147.3, C 148.0, C
4 144.6, C 143.4, C 143.4, C 141.6, C 144.6, C 141.5, C 142.5, C 144.8, C
5 45.9, CH 45.1, CH 46.1, CH 43.8, CH 45.8, CH 43.8, CH 46.0, CH 46.2, CH
6 23.4, CH2 22.0, CH2 22.3, CH2 22.2, CH2 22.5, CH2 22.0, CH2 22.7, CH2 23.4, CH2
7 41.4, CH2 39.9, CH2 37.3, CH2 39.0, CH2 38.5, CH2 39.0, CH2 36.9, CH2 41.4, CH2
8 43.9, C 42.4, C 42.9, C 43.1, C 48.3, C 42.2, C 45.0, C 43.8, C
9 59.3, CH 61.0, CH 56.2, CH 54.2, CH 49.8, CH 54.1, CH 56.5, CH 62.7, CH
10 45.0, C 43.8, C 43.6, C 42.9, C 45.3, C 43.0, C 43.6, C 45.3, C
11 74.4, CH 64.8, CH 72.1, CH 70.0, CH 72.9, CH 68.5, CH 77.4, CH 65.9, CH
12 35.3, CH2 35.1, CH2 38.2, CH2 37.4, CH2 35.7, CH2 33.3, CH2 40.9, CH2 36.8, CH2
13 45.9, CH 44.5, CH 41.6, CH 35.2, CH 41.0, CH 35.5, CH 42.8, CH 45.9, CH
14 38.1, CH2 36.1, CH2 44.7, CH2 38.0, CH2 44.8, CH2 40.3, CH2 44.0, CH2 37.8, CH2
15 51.5, CH2 50.7, CH2 39.7, CH2 39.0, CH2 136.2, CH 39.4, CH2 53.3, CH2 52.1, CH2
16 83.5, C 80.9, C 37.3, CH 52.4, CH 150.2, C 51.6, CH 90.0, C 83.1, C
17 67.6, CH2 74.0, CH2 66.6, CH2 204.4, CH 61.7, CH2 207.4, CH 65.9, CH2 67.3, CH2
18 111.4, CH 110.0, CH 110, CH 110.5, CH 111.5, CH 110.5, CH 111.2, CH 111.4, CH
19 150.4, CH 149.0, CH 148.9, CH 148.3, CH 150.4, CH 148.9, CH 149.2, CH 150.4, CH
20 15.9, CH3 14.3, CH3 14.1, CH3 15.0, CH3 15.7, CH3 15.1, CH3 16.8, CH3 15.7, CH3
1′ 103.1, CH 103.5, CH 100.3, CH 99.5, CH 100.6, CH 98.3, CH – –
2′ 75.3, CH 74.0, CH 74.0, CH, CH 73.7, CH 75.5, CH 73.5, CH – –
3′ 78.4, CH 76.5, CH 77.0, CH 76.9, CH 77.9, CH 77.0, CH – –
4′ 71.7, CH 70.5, CH 70.3, CH 70.0, CH 71.7, CH 69.9, CH – –
5′ 77.8, CH 76.5, CH 76.4, CH 76.6, CH 77.8, CH 76.6, CH – –
6′ 62.8,CH2 61.5,CH2 61.3,CH2 60.7,CH2 62.6,CH2 60.7, CH2 – –

NMR spectra were recorded in *CD3-OD, **d6-DMSO;*** CD3-CN.

a
NMR data were in accordance with published data from Richter & Spiteller, 1979 and Lang et al., 2015.
b
NMR data were in accordance with published data from Hasan et al., 1994 and Lang et al., 2020.

Table 4) of roasting products were analyzed using Brown-Forsythe and Release 2023-2: Maestro, Schrödinger, LLC, New York, NY, 2023). The
Welch ANOVA (no equal SDs) with correction for False Discovery Rate model of TAS2R43 and docking poses of 1, 2, 3, 4, and 7, and refined
in multiple comparisons (two-step-up method of Benjamini, Krieger, and poses of 5, 6, and 8 are available at https://github.com/dipizio/TAS
Yekutieli). Statistics were done in GraphPad Prism 9.3.0 for Windows 2R-models.
(GraphPad Software, San Diego, California, USA, https://www.graphpa
d.com). 3. Results & discussion

3.1. Bitter receptor screening of the mixture of mozambioside roasting

2.9. Molecular modeling and in silico calculations
products

The structures of compounds 1–8 were manually built using the 2D

The concentration of the bitter-tasting Arabica coffee compound
Sketcher and the BUILD tool available in Maestro (Schrödinger Release
mozambioside (11-O-β-D-glucosyl-cafestol-2-on, 1) drops during coffee
2023-2: Maestro, Schrödinger, LLC, New York, NY, 2023). The 3D
roasting from initially ~1 µmol/g to ~0.2 µmol/g, possibly generating
structure and protonation state at pH 7 ± 1 were generated with LigPrep
new products (Lang et al., 2015). Pyrolysis of mozambioside under
(Schrödinger Release 2023-2: Maestro, Schrödinger, LLC, New York, NY,
model conditions comparable to authentic coffee roasting gave rise to
2023).
several new structurally related degradation products. To investigate
The receptor structure of TAS2R46 (PDB ID: 7XP6) was used as a
whether these new compounds also activated bitter taste receptors, the
template for modeling the structure of TAS2R43 using Prime
mixture of degradation products was subjected to a functional bitter
(Schrödinger Release 2023-2: Maestro, Schrödinger, LLC, New York, NY,
taste receptor screening using 26 different human TAS2Rs expressed in
2023). The sequence identity between TAS2R43 and the template for the
HEK 293T-Gα16gust44 cells (Meyerhof et al., 2010; Lang et al., 2023).
region modeled is 88 %. Following previous works, the structure was
The screening revealed the absence of responses in the majority of the
selected without ECL2, and TAS2R43 was modeled without ECL2 (Zie­
TAS2Rs (Supporting Fig. 1); however, the two mozambioside-responsive
gler et al., 2023). Docking simulations of all ligands were performed on
receptors, TAS2R43 and TAS2R46 (Lang et al., 2020), showed signifi­
the CryoEM structure of TAS2R46 and the generated model of TAS2R43.
cant activation by the compound mixture (Fig. 1A, Table 1).
The receptor binding sites were prepared using the “Receptor Grid
Determination of the dose–response relationships of TAS2R43 and
Generation” tool within the centroid of the ligand SY9 given by the
TAS2R46 expressing HEK 293T-Gα16gust44 cells with increasing con­
experimental structure PDB ID: 7XP6. Glide Standard Precision
centrations of mixture of mozambioside roasting products revealed
(Schrödinger Release 2023-2: Maestro, Schrödinger, LLC, New York, NY,
considerable differences in sensitivities among the two responsive re­
2023) was used for the docking studies. Twenty poses per ligand were
ceptors. Whereas the TAS2R43 exhibited a threshold concentration of 3
saved. The docking poses of compounds 5, 6, and 8 with the lowest Glide
µM (based on the mozambioside concentration initially subjected to
scores were used as input for a ligand-receptor refinement with Prime
roasting), TAS2R46 expressing cells required 300 µM to reach statisti­
(Schrödinger Release 2023-2: Prime, Schrödinger, LLC, New York, NY,
cally significant activation relative to control experiments (Student’s t-
2023). The 2D and 3D representations of compounds 5, 6, and 8 within
test, P < 0.05, Table 1). Comparing these concentrations with previously
TAS2R43 and TAS2R46 were generated with Maestro (Schrödinger

5
C. Czech et al. Food Chemistry 446 (2024) 138884

Table 3
1
H NMR data of compounds 1–8.
11-O-β-D-glucosyl- 17-O-β-D- 11-O-β-D-glucosyl- 11-O-β-D-glucosyl- 11-O-β-D-glucosyl- 11-O-β-D-glucosyl- Bengalensol 11-
cafestol-2-on glucosyl- 16-desoxycafestol- (S)-16-desoxy-17- 15,16- (R)-16-desoxy-17- (7)***,c hydroxycafestol-
(mozambioside, 1)*,b cafestol-2- 2-on (3)* oxocafestol-2-on dehydrocafestol-2-on oxocafestol-2-on 2-on (8)*
on (2)* (4)** (5)* (6)**
1
#C H δ (ppm) (J, Hz)
1 2.46 d (16.0) 2.44 2.48 d (14.5) 2.45 d (16.0) 2.45 d (16.0) 2.45 d (16.0) 2.45 d (16.0) 2.42 d (16.3)
2.78 d (16.0) d (16.5) 2.82 d (16.1) 2.63 d (15.6) 2.80 d (16.0) 2.58 d (16.0) 2.66 d (16.0) 2.76 d (16.3)
2.77
d (16.5)
2 – – – – – – – –
3 – – – – – – – –
4 – – – – – – – –
5 2.91 dd (2.6, 12.6) 2.89 dd (2.3, 2.94 dd (2.7, 10.3) 2.93 dd (2.3, 12.6) 2.95 dd (2.6, 12.4) 2.95 dd (2.5, 12.5) 2.86 dd (2.7, 2.87 dd (2.2, 12.5)
12.4) 12.5)
6 1.65 dd (3.2, 13.5) 1.66 m 1.63 m 1.50 dd (2.5, 11.0) 1.64 dd (4.0, 10.1) 1.50 dd (1.7, 12.0) 1.63 m 1.65 d (3.3)
2.06 dd (3.2, 13.5) 2.04 m 2.05 m 2.02 dd (2.5, 12.0) 2.08 m 2.03 m 2.07 m 2.04 d (2.6)
7 1.82 m 1.82 m 1.15 dd (3.2, 8.1) 1.50 m 1.76 t (2.4) 1.58 t (7.2) 1.59 m 1.80 m
1.86 m 1.77 m 1.97 m 1.77 td (3.3, 13.3) 1.83 m
8 – – – – – – – –
9 1.89 s 1.66 s 1.95 s 1.90 s 1.83 s 1.85 s 2.08 s 1.67 m
10 – – – – – – – –
11 3.93 d (7.4) 3.91 dd (2.0, 4.04 d (5.7) 3.95 d (5.6) 4.07 d (5.8) 3.91 d (5.7) 4.42 t (3.0) 3.88 d (7.0)
7.5)
12 1.83 m 1.83 m 1.86 m 1.79; m 1.90 dd (2.4, 6.0) 1.67 m 1.80 m 1.80 m
2.35 d (16.0) 2.20 m 2.10 m 1.97 s 2.10 m 1.98 m 2.20a 2.04 t (6.5)
13 2.09a 2.17 a 2.59a 2.50a 2.46 a 2.56a 2.48 t (6.5) 2.09 dd (2.3, 2.4)
14 1.78 m 1.83 m 1.92 m 1.74 m 1.54 dd (5.0, 10.2) 1.12 m 1.32a 1.79 m
1.85 m 2.19 dd (2.3, 6.7) 0.80 dd (3.9, 7.5) 2.16 d (10.6) 1.92 dd (5.5, 8.7) 2.20a
15 1.35 dd (1.2, 15.0) 1.43 1.63 m 1.56 m 5.36 d (0.9) 1.54 s 1.54 d (3.5) 1.40 s
2.26 dd (1.2, 15.0) d (14.0) 1.86 m 2.07 dd (1.5, 13.8) 1.60 d (3.3) 2.20 dd (0.5, 14.6)
2.17
d (14.0)
16 – – 2.16 dd (1.9, 6.2) 3.17 m – 2.68 m – –
17 3.78 d (12.0) 3.59 d (9.8) 3.30a 9.59 d (1.1) 4.12 dd (1.3, 12.0) 9.80 s 3.54 d (11.6) 3.77 d (11.4), 4.18
4.28 d (12.0) 4.80 d (9.5) 4.23 dd (1.7, 12.1) 3.64 d (12.0) d (11.4)
18 6.61 d (1.7) 6.63 d (1.6) 6.62 d (1.7) 6.70 d (1.7) 6.62 d (1.7) 6.70 d (1.7) 6.57 d (1.7) 6.61 d (1.6)
19 7.79 d (1.7) 7.81 d (1.7) 7.81 d (1.7) 7.97 d (1.7) 7.80 d (1.8) 7.97 d (1.7) 7.73 d (1.8) 7.79 d (1.5)
20 0.88 s 1.31 s 0.90 s 0.74 s 0.94 s 0.75 s 1.03 s 0.87 s
1′ 4.34 d (8.0) 4.30 d (8.2) 4.36 d (8.0) 4.24 d (7.7) 4.30 d (7.8) 4.24 d (7.7) – –
2′ 3.17 dd (7.6, 9.0) 3.23 t (9.2) 3.21 m 2.95 m 3.13 dd (7.6, 2.2) 2.92 m – –
3′ 3.37 t (8.9) 3.38 t (9.0) 3.44 m 3.18 m 3.37 t (9.0) 3.18 m – –
4′ 3.26 m 3.27 m 3.39 m 3.10 m 3.31a 3.09 m – –
5′ 3.26 m 3.28 m 3.39 m 3.08 m 3.26 m 3.06 m – –
6′ 3.67 dd (1.7, 10.0) 3.67 dd (1.9, 3.72 dd (5.3, 6.5) 3.50 dd (5.3, 11.0) 3.69 dd (5.3, 12.1) 3.49 m – –
3.92 dd (1.7, 10.1) 10.1) 3.95 dd (3.4, 5.7) 3.68 ddd (1.5, 8.0, 3.93 dd (2.2, 12.0) 3.66 ddd (1.8, 8.0,
3.89 dd (1.9, 9.0) 12.0)
11.1)

NMR spectra were recorded in *CD3-OD, **d6-DMSO;*** CD3-CN.

a
Assigned from 2D spectra (HSQC) due to interference with other signals in 1H.
b
NMR data were in accordance with published data from Richter & Spiteller, 1979 and Lang et al., 2015.
c
NMR data were in accordance with published data from Hasan et al., 1994 and Lang et al., 2020.

published data obtained for unroasted mozambioside (1), roasted 1 was 3.2. Model roasting of isolated mozambioside to generate degradation
10 times more potent (3 µM vs. 30 µM [Lang et al., 2020]) at the products
TAS2R43 and similarly potent at the TAS2R46 (300 µM).
The EC50-concentration determined for TAS2R43 was 26.6 ± 13.1 Mozambioside (1) was roasted at a temperature of 220 ◦ C for 4.5 min
µM, whereas 1 exhibited 50.6 ± 13.8 µM (Table 1, Student’s t-test, P = to generate degradation products. Based on the peak area in UHPLC-
0.09). Similar to previous results obtained with 1, TAS2R46 expressing PDA analysis (281 nm), roughly 70 % of 1 was converted into degra­
cells did not show signal saturation, and hence, no EC50-concentration dation products. These were isolated by preparative HPLC on PFP ma­
could be calculated. The receptor data show that pyrolysis of 1 resulted terial (2.3) and subsequent semi-preparative purification on C18 (2.4,
in a significant increase in potency, in terms of activation threshold, for see Fig. 1B). The pure isolates were investigated by UHPLC-PDA (2.2),
TAS2R43 activation (Table 1, P < 0.05). This observation suggested the NMR spectroscopy (2.6), and UHPLC-ToF-MS (2.7) for structure eluci­
generation of new and more potent compounds. dation. Comparison of the respective data of 1 with those of the isolates
The results of these initial experiments suggested that the roasting clarified the new structures 2–6. As detailed in Sections 3.3–3.6 (see
process turned mozambioside (1) into more potent bitter receptor- below), five new compounds were identified, namely 17-O-β-D-glucosyl-
activating products. To clarify the structures and characterize their 11-hydroxycafestol-2-on (2), 11-O-β-D-glucosyl-16-desoxycafestol-2-on
bitter activation properties, 1 was pyrolyzed on a preparative scale. (3), 11-O-β-D-glucosyl-(S)-16-desoxy-17-oxocafestol-2-on (4), 11-O-β-D-
Individual roasting products were isolated for structure elucidation via glucosyl-15,16-dehydrocafestol-2-on (5), and 11-O-β-D-glucosyl-(R)-16-
UHPLC-ToF-MS and NMR spectroscopy and characterized for bitter taste desoxy-17-oxocafestol-2-on (6). In addition, the known roasting product
receptor activation on TAS2R43 and TAS2R46. bengalensol (7, Hasan et al., 1994; Lang et al., 2020) was identified and
purified (3.7). Further, the aglycone of compounds 1–6 (11-

6
C. Czech et al. Food Chemistry 446 (2024) 138884

Table 4 This observation suggested an intramolecular migration of the hexose

Maximal tested concentrations (µM) and activation threshold concentrations of from the hydroxyl group at C11 in 1 to that of C17 in 2. This assumption
the new compounds 2–6 and 8, with human bitter receptors TAS2R43 and was confirmed in the HMBC spectrum, where the glycosidic proton H-1′
TAS2R46. at 4.36 ppm (d, J = 7.9 Hz) correlated with the carbon C17 (74.0 ppm)
Compound Max. Threshold (µM)2 Efficacy (Emax, ΔF/F)3 (cf. Supporting Figs. 12–15). Roasting product 2 was therefore identified
conc.
TAS2R43 TAS2R46 TAS2R43 TAS2R46 as 17-O-β-D-glucosyl-11-hydroxycafestol-2-on (2 in Fig. 1D). The 13C
(µM)1
and 1H NMR data are tabled in Table 2 and Table 3.
2 300 – 300 – 0.020 ±
0.002a 3.4. Structure elucidation of 11-O-β-D-glucosyl-16-desoxycafestol-2-on
3 30 – – – –
4 300 100 30 0.102 ± 0.098 ±
(3)
0.005a,† 0.006b,†
5 300 30 100 0.129 ± 0.043 ± Roasting product 3 eluted after 12.8 min in UHPLC-PDA analysis and
0.006b,† 0.006c,‡ showed UV maxima at 221 and 280 nm, similar to 1, thus suggesting an
6 300 10 100 0.303 ± 0.072 ±
unchanged chromophore. In the UHPLC-ToF-MS (ESI+), the pseudo
0.008c,† 0.006d,‡
8 300 10 30 0.171 ± 0.098 ± molecular ion was m/z 493.2461 ([M+H]+), which corresponds to a
0.011d,† 0.005b,‡ sum formula C26H36O9 (Δ+5.9 ppm), indicating the loss of one oxygen.
The daughter ion with m/z 313.1822 (C20H24O3, [M-hexose-H2O + H]+)
Data are means ± SD of n = 3, each performed in duplicates. 1 maximum applied
concentration; 2 threshold concentration is the lowest concentration at which
showed the cleavage of an intact hexose (–C6H12O6, –180 Da) and
receptor-transfected cells show a significantly higher signal (Student’s t-test, P subsequently indicated structural changes within the aglycone. We,
< 0.05) than empty vector controls. 3 One-way Brown-Forsythe and Welch therefore, assumed 3 to be a deoxygenation product of 1. The NMR
ANOVA comparison between efficacies of roasting products with correction for spectra showed that the signals of roasting product 3 were nearly su­
False Discovery Rate (two-stage step-up method of Benjamini, Krieger, and perimposable with 1, with some exceptions. Compared to 1, not six but
Yekutieli) in multiple comparisons. Different letters (a-d) in the same column only five quaternary carbons and one additional CH group were detec­
indicate significant differences between compounds (P < 0.05), and different ted, indicating one quaternary carbon was reduced into a methine
symbols (†,‡) in the same line indicate significant differences between TAS2R43 group. Indeed, the hydroxylated carbon C16 signal at 83.5 ppm in 1 was
and TAS2R46 (P < 0.05). replaced by a new signal at 37.3 ppm with a proton resonating at 2.16
ppm in 3. In the HMBC spectrum, this new CH signal showed correla­
hydroxycafestol-2-on) was obtained by enzymatic cleavage of 2 (3.8) tions to C11, C13, and C17, confirming that deoxygenated C16 triggered
and inspected for bitter receptor activation. the new CH signal. The remaining eleven CH groups (C5, C9, C11, C13,
Fig. 1, C shows significant correlations and changes in the 2D NMR C18, C19, C1′, C2′, C3′, C4′, C5′), eight CH2 groups (C1, C6, C7, 12, C14,
spectra (HSQC) relative to the parent compound mozambioside (1), C15, C17, C6′), and one methyl group (C20) were similar to those of 1,
which allowed unequivocal determination of the roasting products’ thus confirming no further structural changes (cf. Supporting
structures. The NMR data revealed that pyrolysis of 1 primarily initiated Figs. 16–19). In summary, roasting product 3 was formed upon deoxy­
dehydration processes of the hydroxylated five-membered ring genation of the hydroxyl group at C16, leading to 11-O-β-D-glucosyl-16-
C13–C16. As detailed in Sections 3.3–3.6 (see below), five new com­ desoxycafestol-2-on (3 in Fig. 1D). The 13C and 1H NMR data are tabled
pounds 2–6 were identified. The loss of one molecule of water from in Table 2 and Table 3.
carbon C16 gave rise to the quantitatively dominating three new com­
pounds 11-O-β-D-glucosyl-(S)-16-desoxy-17-oxocafestol-2-on (4), 11-O- 3.5. Structure elucidation of 11-O-β-D-glucosyl-(S)-16-desoxy-17-
β-D-glucosyl-15,16-dehydrocafestol-2-on (5), and 11-O-β-D-glucosyl-(R)- oxocafestol-2-on (4) and 11-O-β-D-glucosyl-(R)-16-desoxy-17-
16-desoxy-17-oxocafestol-2-on (6). 17-O-β-D-glucosyl-11-hydrox­ oxocafestol-2-on (6)
ycafestol-2-on (2) was isolated as a minor product, putatively formed by
intramolecular migration of the glucosyl moiety from C11 to C17. A In UHPLC-PDA analysis, roasting products 4 and 6 eluted after 13.3
second minor product was 11-O-β-D-glucosyl-16-desoxycafestol-2-on min and 13.7 min, respectively. The UV spectra had maxima at 219 and
(3), in which the tertiary hydroxyl of C16 was removed. A complete loss 280 nm (4) and 196 and 280 nm (6), respectively. In the UHPLC-ToF-MS
of the glucose and intramolecular etherification accompanied by the (ESI+), the pseudo molecular ions were m/z 491.2311 (4, [M+H])+ and
elimination of water formed compound 7, also known as bengalensol, m/z 491.2303 (6, [M+H]+), each corresponding to a sum formula
first reported by Hasan et al. (1994). The new structures are given in C26H34O9 (4: Δ + 7.2 ppm, 6: Δ + 5.6 ppm). The calculated sum formula
Fig. 1D. indicated the loss of one molecule of water (-18 Da) compared to educt
1. The product ions of 4 and 6 were similar with m/z 311.1658 (4,
3.3. Structure elucidation of 17-O-β-D-glucosyl-11-hydroxycafestol-2-on C20H22O3, [M-hexose-H2O + H]+, Δ + 5.2 ppm) and m/z 311.1654 (6,
(2) C20H22O3, [M-hexose-H2O + H]+, Δ + 3.9 ppm). In both roasting
products, the daughter ions pointed to the cleavage of an intact hexose.
Roasting product 2 eluted after 10.2 min in the UHPLC-PDA with UV Therefore we expected the loss of one molecule of water from the
maxima at 219 and 281 nm. In the UHPLC-ToF-MS analysis, the pre­ aglycone.
cursor ion in ESI+ mode was m/z 531.2204 with the predicted sum In the NMR, the signals of 4 and 6 were nearly superimposable. Both
formula C26H36O10Na ([M+Na]+). A daughter ion with m/z 329.1757 spectra differed slightly from the reference signals of 1, with three
(C20H24O4, [M-hexose-H2O + H]+, Δ-2.5 ppm) indicated the cleavage of important exceptions. Most strikingly, in both roasting products, the
an intact hexose moiety (–C6H12O6, − 180 Da). The NMR showed in hydroxylated quaternary carbon C16 signal was replaced by a CH group
total six quaternary carbons (C2, C3, C4, C8, C10, C16), eleven CH (4: 52.4 ppm/3.17 ppm, 6: 51.6 ppm/2.68 ppm). Further, in both 4 and
groups (C5, C9, C11, C13, C18, C19, C1′, C2′, C3′, C4′, C5′), eight CH2 6, the signal of a new carbonyl (aldehyde) group appeared at 204.4
groups (C1, C6, C7, C12, C14, C15, C17, C6′) and one CH3 group (C20). ppm/9.59 ppm (4) and 207.4 ppm/9.80 ppm (6), respectively. In 4, the
When comparing the NMR spectra of 2 with educt 1, the signals and aldehyde proton H17 was a doublet (J = 1.1 Hz), while in 6, a singlet
correlations were superimposable, except for the CH-11 and CH2-17 appeared. In 4 and 6, the new carbonyl replaced the hydroxylated
shifts, which were interchanged. In roasting product 2, CH-11 and CH2- methylene C17 since the CH2-17 signals had vanished. Finally, in both
17 signals resonated at 64.8 ppm and 74.0 ppm, respectively, while in roasting products, the methylene C15 was shifted upfield from ~52 ppm
compound 1, CH-11 and CH2-17 resonated at 74.4 ppm and 67.6 ppm. in 1 to 39.0 ppm (4) and 39.4 ppm (6), respectively. For both 4 and 6,

7
C. Czech et al. Food Chemistry 446 (2024) 138884

five quaternary carbons (C2, C3, C4, C8, C10), 13 CH groups (C5, C9, pointing to equal chromophore structures. In the UHPLC-ToF-MS (ESI+)
C11, C13, C16, C17, C18, C19, C1′, C2′, C3′, C4′, C5′), seven CH2 groups 7 showed a pseudo molecular ion with m/z 329.1759 ([M+H]+) corre­
(C1, C6, C7, C12, C14, C15, C6′), and one CH3 group (C20) appeared in sponding to a sum formula of C20H24O4 (Δ + 3.5 ppm). Therefore, the
the spectrum and, apart from the deviations mentioned, were similar to molecule was expected to be a dehydrated derivative of the aglycone 11-
those of 1 (cf. Fig. 1C, D and Supporting Figs. 20–24 and 29–33). Given hydroxycafestol-2-on. This assumption was confirmed by the NMR data,
the difference in retention time, the similarity of MS indicating water which agreed with the data reported earlier (Hasan et al., 1994; Lang
loss from the aglycone, and the emerging aldehyde signals in the NMR, et al., 2020) for the diterpenoid bengalensol (7 in Fig. 1, D). The 13C and
1
we expected the two roasting products 4 and 6 to be formed by the H NMR data are tabled in Table 2 and Table 3.
elimination of water from hydroxylated carbon C16 leading to the enol,
which subsequently tautomerized into (R)- and (S)-aldehydes 4 and 6 3.8. Generation of 11-hydroxycafestol-2-on (8) for structure–activity
(Supporting Fig. 4). relationships
Since a new chiral center at C16 was formed, the roasting products 4
and 6 were subjected to ROESY experiments. We focused on the chiral The glycoside 1 cannot be cleaved by treatment with acid or hy­
carbons C16 and C11 and the aldehyde group C17 to analyze the ste­ drolytic enzymes (Richter & Spiteller, 1979). However, isolated roasting
reochemistry. For roasting product 4, the ROESY did not show corre­ product 2 (17-O-β-D-glucosyl-cafestol-2-on) was susceptible to enzy­
lations between H11 and H16 or H17, but clear correlation signals of matic cleavage with β-glucuronidase from Helix pomatia and afforded
H16 and H17 (Supporting Fig. 24) appeared. Another outstanding cor­ the aglycone 11-hydroxycafestol-2-on (8). The compound eluted at 10.4
relation was found between H16 and H15. When considering other key min in the UHPLC-PDA run and showed a UV maximum at 220 and 281
correlations, such as between H11 and H20 or between H5 and H9, and nm, indicating no changes in the chromophore. UHPLC-ToF-MS analysis
missing correlations between H11 and H5 or H9, we resolved the spatial (ESI+) showed a pseudo molecular ion with m/z 347.1889 ([M+H]+)
arrangement for 4 as an (S)-isomer, as shown in Supporting Fig. 42B. with the predicted sum formula C20H26O5 (Δ-3.5 ppm). In total signals
In 6, the critical correlations with CH-5, CH-9, CH-11, and CH3-20 of six quaternary carbons (C2, C3, C4, C8, C10, C16), six CH groups (C5,
were consistent with those of 4, but the correlations of H16 with H12 C9, C11, C13, C18, C19), seven CH2 groups (C1, C6, C7, C12, C14, C15,
and with H13 were different (Supporting Fig. 33). When considering the C17), and one methylene group (C20) were detected. The shifts of these
aldehyde group CHO-17, the difference to 4 was a correlation signal signals were in good agreement with those of the aglycone of 1 and,
with H12, indicating an (R)-configuration at C16, as shown in Sup­ therefore, were unambiguously assigned. Compound 8 was not formed
porting Fig. 42C. The structures of 4 and 6 are given in Fig. 1D. The 13C in the model roasting experiments in isolatable amounts but was enzy­
and 1H NMR data are tabled in Table 2 and Table 3. matically prepared for structure-receptor-activity investigations. The
structure is given in Fig. 1D. The 13C and 1H NMR data are tabled in
3.6. Structure elucidation of 11-O-β-D-glucosyl-15,16-dehydrocafestol-2- Table 2 and Table 3.
on (5)
3.9. Detection of mozambioside and its roasting products in roasted
In UHPLC-PDA analysis, roasting product 5 eluted after 13.5 min and Arabica coffee
showed the same UV spectrum as 1 with maxima at 219 and 280 nm,
suggesting an unchanged chromophore. In the UHPLC-ToF-MS (ESI+) Model roasting of pure mozambioside resulted in the formation of
the pseudo molecular ion was m/z 491.2250 ([M+H]+). This observa­ the new compounds 2–6 and bengalensol (7). Their occurrence was
tion pointed towards the sum formula of C26H34O9 (Δ-5.2 ppm) and investigated in an authentic sample of roasted coffee brew. An aliquot
indicated the loss of one water molecule (− 18 Da) compared to educt 1. was diluted and injected into the UHPLC-ToF-MS system. A solution of
A prominent daughter ion appeared in the fragment spectrum with m/z the isolates served as a reference standard for comparison of retention
311.1630 (C20H22O3, [M-hexose-H2O + H]+, Δ-3.8 ppm) and confirmed times and exact mass data (Fig. 2, A). We identified 1, 2, and 4–7 in the
the cleavage of an intact hexose molecule (-C6H12O6, − 180 Da) from the roasted coffee brew. Mozambioside (Fig. 2, B) formed abundant pro­
aglycone. The NMR showed signals of six quaternary carbons (C2, C3, tonated molecular ions ([M+H]+), while the isobaric derivative 2
C4, C8, C10, C16), twelve CH groups (C5, C9, C11, C13, C15, C18, C19, exclusively formed the sodiated species ([M+Na]+, Fig. 2, C). The
C1′, C2′, C3′, C4′, C5′), seven CH2 groups (C1, C6, C7, C12, C14, C17, C6′) roasting product 3 and the aglycone 11-hydoxycafestol-2-on (8) were
and one methyl group (C20). The NMR data of 5 were nearly superim­ not found. Roasting products 4–6 were abundant but not baseline-
posable to those of educt 1, except that in 5, the CH2-15 signal dis­ separated under the chosen conditions (Fig. 2, D). Finally, bengalensol
appeared, and a new CH group emerged (5.36 ppm/136.2 ppm). Since (7) was found in the brew, eluting late at 15.3 min (Fig. 2, E). The
the ToF-MS data suggested the loss of one water molecule from the presence of these bitter receptor activators in the authentic coffee
aglycone, and the hydroxylated methylene at C17 (61.7 ppm) was intact sample suggested a possible contribution to the taste profile. However,
(cf. roasting products 4 and 6), we expected the formation of a double quantitative data are needed for the clarification of this hypothesis.
bond upon water elimination. A prominent new CH group resonated as a
doublet (J = 0.9 Hz) at 5.36 ppm in the 1H spectrum and 136.2 ppm in 3.10. Characterization of bitter taste receptor activation
the 13C spectrum, assigned to the newly formed CH group at C15.
Further, the signal of quaternary carbon C16 was shifted from 83.5 ppm The new structures 2–6 formed by thermal treatment of mozam­
in 1 far into the low field to 150.2 ppm, suggesting a quaternary double- bioside (1) and their aglycone 11-hydroxycafestol-2-on (8) underwent
bond carbon. These were clear hints for water elimination between functional studies to identify the taste receptors responsible for medi­
carbons C15 and C16 and the formation of a double bond (Supporting ating their bitter taste. After confirming that the roasted mozambioside
Fig. 5). Correlations between the proton H15 and the carbons C7 and C9 mixture elicited responses of the same two receptors, TAS2R43 and
in the HMBC spectrum confirmed the structure as 11-O-β-D-glucosyl- TAS2R46 as their precursor mozambioside (1) with a significantly
15,16-dehydrocafestol-2-on (5 in Fig. 1D, cf. Supporting Figs. 25–28). increased potency, in terms of receptor activation threshold, at the re­
The 13C and 1H NMR data are tabled in Table 2 and Table 3. ceptor TAS2R43 (P < 0.05, Table 1), the individual purified products
were investigated for their activation properties.
3.7. Identification of bengalensol (7) The roasting products 2–6 and 11-hydroxycafestol-2-on (8) were
individually investigated for activation of TAS2R43 and TAS2R46 in
In UHPLC-PDA analysis, roasting product 7 eluted after 14.4 min and concentrations ranging from 0.3 to 300 µM (Fig. 3). Note that roasting
showed comparable UV spectra as 1 with maxima at 211 and 220 nm, product 3 was only investigated between 0.3 and 30 µM because of the

8
C. Czech et al. Food Chemistry 446 (2024) 138884

Fig. 2. (A) Overlaid Total Ion Chromatograms (TICs) of standard solutions of 1–8 with retention times. B–E Extracted ion chromatograms (XICs) from the analysis of
roasted coffee brew: (B) mozambioside (1, Rt. 10.20 min, m/z 509.237 [M+H]+), (C) 1 (Rt. 10.19 min, [M+Na]+) and 17-O-β-D-glucosyl-11-hydroxycafestol-2-on (2,
Rt. 10.67 min, m/z 531.218 [M+Na]+), (D) isobaric roasting products 4–6 (m/z 491.226, [M+H]+): 11-O-β-D-glucosyl-(S)-16-desoxy-17-oxocafestol-2-on (4, Rt.
14.01 min), 11-O-β-D-glucosyl-15,16-dehydrocafestol-2-on (5, Rt. 14.12 min), 11-O-β-D-glucosyl-(R)-16-desoxy-17-oxocafestol-2-on (6, Rt. 14.25 min); (E) benga­
lensol (7, Rt. 15.36 min, m/z 329.17 ([M+H]+).

minimal yield during pyrolysis (Fig. 1B). products comprise the presence of yet unidentified highly potent acti­
As shown in Fig. 3, the roasting products 2–6 and the aglycone 8 vators in the mix or synergistic effects of the identified compounds.
exhibited a broad range of activities in the functional receptor assays. 2 Synergism of bitter compounds, however, has not been reported so far.
showed small activation of exclusively TAS2R46. Compared to control When testing the mixture of roasting products, the threshold con­
experiments, roasting product 3 did not elicit any responses of TAS2R43 centration for activation of TAS2R46 was 300 µM (Table 1). The lower
and TAS2R46. However, the remaining four compounds, 4–6 and 8, potencies of 2 and 3 apparently outweighed those of roasting products
resulted in heterogeneous activation profiles (cf. Table 2). 2 and 4 were 4–6 and 7, resulting in no observable change in the activation threshold
the only roasting products with lower activation thresholds for TAS2R46 for this bitter receptor (Table 2). Although lower in relative concentra­
than for TAS2R43. The efficacies of 4 were similar between the re­ tion in terms of peak area in Fig. 1B, compounds 2 and 3 present in the
ceptors. In contrast, 5, 6, and 8 were significantly more potent on mixture possibly bind to TAS2R46 but fail to induce pronounced acti­
TAS2R43 than on TAS2R46. The most pronounced difference between vation. Such competitive effects were witnessed before, during the co-
the two bitter taste receptors was obtained for compound 6, with a 10- application of the potent agonist mozambioside with the partial
fold higher sensitivity of TAS2R43 compared to TAS2R46. The efficacy agonist kahweol onto TAS2R43 expressing cells (Lang et al., 2020).
of the substances deviated amongst each other and with respect to the
activated receptors. 6 resulted in the most substantial increase in signal
amplitude with an efficacy roughly four times higher on TAS2R43 than 3.11. Investigation of putative binding modes
TAS2R46 (Table 2). EC50 concentrations were not calculated, as the
applied substance concentrations did not reach signal saturation. The functional assays revealed that 5, 6, and 8 were the most potent
Compared to their precursor 1, roasting product 4 was a better compounds towards TAS2R43, but they also activated TAS2R46.
agonist for TAS2R46, while 5 and 6 were better activators of TAS2R43. Therefore, we used docking simulations to investigate the binding
The individual threshold concentrations for TAS2R43 were higher (30 modes of these compounds within the orthosteric binding sites of
and 10 µM, respectively, Table 2) than that of the roasting mixture (3 TAS2R43 and TAS2R46 structural models.
µM, Table 1). Possibly, the minor product bengalensol (7), which has a The structure of TAS2R46 was recently solved (Xu et al., 2022).
reported threshold concentration of 3 µM (Lang et al., 2020), contrib­ Bitter taste receptors belong to the superfamily of G protein-coupled
uted substantially to the mixture’s superior potency. Other speculative receptors (GPCRs). They have a similar architecture to class A GPCRs,
possibilities for the increased potency of the mixture of roasting including seven transmembrane (TM) α-helices with an intracellular
carboxyl tail and an extracellular amino terminus. TM helices are

9
C. Czech et al. Food Chemistry 446 (2024) 138884

Fig. 3. Concentration-dependent receptor activation of TAS2R43 (blue) and TAS2R46 (red) of individual roasting products formed from 1 (cf. Fig. 1, B). The as­
terisks indicate the threshold concentrations, the lowest concentration at which receptor-transfected cells show a significantly higher signal (Student’s t-test, P <
0.05) than identically treated empty vector controls. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

connected by three intracellular (ICL1, ICL2, and ICL3) and three the position of the compounds within the receptor. Bengalensol (7)
extracellular loops (ECL1, ECL2, and ECL3) (see Supporting Fig. 43 for showed a similar binding mode as compound 8, whereas compounds
the 3D representation of TAS2R46). Using this structure as a template, 1–4 could not adapt the poses described above for 5 and 6 (all docking
we modeled the receptor structure of TAS2R43. The structure of poses are available at: https://github.com/dipizio/TAS2R-models/
TAS2R46 in complex with strychnine (PDB ID: 7XP6) was solved tree/main/roasting-products.
without the extracellular loop 2 (ECL2). Only the inactive receptor
model (PDB ID: 7XP5) has the ECL2 solved, but in this conformation, the 4. Conclusions
loop would overlap with the position of potential ligands within the
orthosteric binding site. As previous investigations pointed out, in cases In conclusion, five new compounds 2–6 and the known bengalensol
of high flexibility and modeling uncertainty of ECL2, molecular docking (7) were isolated from the model pyrolysis of coffee compound
simulations on GPCRs without the ECL2 might reach better performance mozambioside. Their structures were elucidated by instrumental tech­
(de Graaf et al., 2008; Born et al., 2013; Jaiteh et al., 2020). Therefore, niques. The four new roasting products 2 and 4–6 and their aglycone 8
we used the models without ECL2 for our investigations. activated human bitter receptors TAS2R43 and TAS3R46. 1 and the
The focus was on characterizing the binding mode of compound 8, as generated new compounds 2 and 4–7 were detected and unambiguously
it represents the core aglycone moiety of the series of structurally related identified in roasted coffee brew by UHPLC-ToF-MS. The new com­
roasting products of mozambioside. The finding was that a consistent pounds 5 and 6 significantly activated bitter receptors at concentrations
binding mode was obtained in both receptors: the ligand anchors to TM5 10 times lower than their parent compound 1. Docking simulations
(H-bond with T180) and TM7 (H-bond with R268) in TAS2R43, and revealed that skeleton 8 could bind with a similar orientation in both
TM5 (H-bond with T180) and TM6 (S248) in TAS2R46 (Fig. 4A and 4B). TAS2R43 and TAS2R46 binding sites. However, the different composi­
Compound 8 showed the same orientation in both receptors, with the tion and volume of the binding sites affect the poses of the glucosylated
hydroxymethyl group C17-OH of the aglycone moiety towards TM5 and derivatives.
the heterocyclic furane group towards TM3. These new compounds could contribute to either coffee taste given
With respect to 8, compound 5 has a double bond in C15, while the necessary concentration in the brew or, provided oral bioavail­
compound 6 has an aldehyde in C17 instead of the hydroxymethyl group ability, interact with extraorally expressed TAS2R43 and 46.
C17-OH. These differences did not affect the interactions with the re­ The astonishing complexity of bitter-tasting molecules arising from
ceptors. With the addition of the glucosyl moiety in compounds 5 and 6, the pyrolysis of a single substance may provide only a glimpse of the
the position of the aglycone moiety was maintained in both receptors, plethora of bitter substances and their bitter receptor-activating prop­
but the glucosyl substituent was accommodated in different subpockets. erties in coffee brews. Future experiments will add numerous further
Within TAS2R43, the compounds 5 and 6 reached the polar side chains coffee ingredients capable of activating bitter taste receptors. While such
of the residues R81 and Y85, K156, and N176 by building hydrogen research efforts should result in an ever-growing knowledge about the
bonds with the glucose moiety in the subpocket located between TM3, composition of coffee’s bitter compounds contributing to its hedonic
TM4, and TM5. Since these interactions pulled the compounds higher value, the real challenge will be the assessment of the physiological
compared to 8, the compounds lost the hydrogen bond with R268 impacts of individual compounds after consumption.
(Fig. 4A) but interacted with S248 (Fig. 4C).
Within TAS2R46 (Fig. 4D), the glucosyl moiety pointed to a sub­ CRediT authorship contribution statement
pocket at the core of the receptor, stabilized by hydrogen bonds with
residues W88 and N92 in TM3 and with E265 in TM7. Further, the Coline Czech: Writing – review & editing, Writing – original draft,
residues L58, N62, and F269 built a hydrophobic subpocket, stabilizing Visualization, Methodology, Investigation, Formal analysis. Tatjana

10
C. Czech et al. Food Chemistry 446 (2024) 138884

Fig. 4. Predicted binding modes of compounds 5, 6, and 8 within TAS2R43 and -46. (A) 8/TAS2R43: Compound 8 and interacting residues are represented as red as
a ball and stick, the receptor is shown as a grey cartoon, and hydrogen bonds are dashed lines. (B) 8/TAS2R46: Compound 8 and interacting residues are represented
as red as a ball and stick, the receptor is shown as a light green cartoon, and hydrogen bonds are dashed lines. (C) 5 and 6/TAS2R43: Compounds 5 and 6 and
interacting residues are represented as beige and cyan ball and stick, the glucose moiety is highlighted in pink; the receptor is shown as grey cartoon; hydrogen bonds
are dashed lines. (D) 5 and 6/TAS2R46: Compounds 5 and 6 and interacting residues are represented as beige and cyan ball and stick; the glucose moiety is
highlighted in pink; the receptor is shown as a light green cartoon; hydrogen bonds are dashed lines. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

Lang: Writing – original draft, Visualization, Investigation, Formal Data availability

analysis. Angelika Graßl: Investigation. Alexandra Steuer: Writing –
original draft, Visualization, Software, Investigation, Formal analysis. All data are in the main manuscript and the supplemental informa­
Antonella Di Pizio: Writing – review & editing, Writing – original draft, tion file.
Supervision, Software, Resources, Methodology, Investigation. Maik
Behrens: Writing – review & editing, Writing – original draft, Super­ Acknowledgment
vision, Resources, Methodology, Investigation. Roman Lang: Writing –
review & editing, Writing – original draft, Supervision, Resources, We thank V. Schlagbauer for contributions in an earlier phase of the
Methodology, Investigation, Conceptualization. project and M. Riedmaier, A. Beusch, and L. Reim for technical assis­
tance. We also thank O. Frank and J. K. Kreißl for their support in
Declaration of competing interest recording NMR spectra. A.D.P’s research is supported by the Leibniz
Programme for Women Professors (grant: P116/2020).
The authors declare the following financial interests/personal re­
lationships which may be considered as potential competing interests: Appendix A. Supplementary material
Professor Dr. Antonella Di Pizio reports financial support was provided
by Leibniz Association. If there are other authors, they declare that they Supplementary data (Supporting Information - Identification of
have no known competing financial interests or personal relationships mozambioside roasting products and their bitter taste receptor activa­
that could have appeared to influence the work reported in this paper. tion) to this article can be found online at https://doi.org/10.1016/j.
foodchem.2024.138884.

11
C. Czech et al. Food Chemistry 446 (2024) 138884

References Jaiteh, M., Rodriguez-Espigares, I., Selent, J., & Carlsson, J. (2020). Performance of
virtual screening against GPCR homology models: Impact of template selection and
treatment of binding site plasticity. PLoS Computational Biology, 16, Article
Batali, M. E., Lim, L. X., Liang, J., Yeager, S. E., Thompson, A. N., Han, J.,
e1007680. https://doi.org/10.1371/journal.pcbi.1007680
Ristenpart, W. D., & Guinard, J.-X. (2022). Sensory analysis of full immersion coffee:
Lang, R., Klade, S., Beusch, A., Dunkel, A., & Hofmann, T. (2015). Mozambioside is an
Cold brew is more floral, and less bitter, sour, and rubbery than hot brew. Foods, 11
arabica-specific bitter-tasting furokaurane glucoside in coffee beans. Journal of
(16), Article 2440. https://doi.org/10.3390/foods11162440
Agricultural and Food Chemistry, 63(48), 10492–10499. https://doi.org/10.1021/acs.
Blumberg, S., Frank, O., & Hofmann, T. (2010). Quantitative studies on the influence of
jafc.5b04847
the bean roasting parameters and hot water percolation on the concentrations of
Lang, T., Di Pizio, A., Risso, D., Drayna, D., & Behrens, M. (2023). Activation profile of
bitter compounds in coffee brew. Journal of Agricultural and Food Chemistry, 58,
TAS2R2, the 26th human bitter taste receptor. Molecular Nutrition & Food Research. ,
3720–3728. https://doi.org/10.1021/jf9044606
Article 2200775. https://doi.org/10.1002/mnfr.202200775
Born, S., Levit, A., Niv, M. Y., Meyerhof, W., & Behrens, M. (2013). The human bitter
Lang, T., Lang, R., Di Pizio, A., Mittermeier, V. K., Schlagbauer, V., Hofmann, T., &
taste receptor TAS2R10 is tailored to accommodate numerous diverse ligands. The
Behrens, M. (2020). Numerous compounds orchestrate coffee’s bitterness. Journal of
Journal of Neuroscience, 33(1), 201–213. https://doi.org/10.1523/
Agricultural and Food Chemistry, 68(24), 6692–6700. https://doi.org/10.1021/acs.
JNEUROSCI.3248-12.2013
jafc.0c01373
Clarke, R. J., & Macrae, R. (Eds.) (1987). Coffee – Volume 2: Technology. Dordrecht:
Meyerhof, W., Batram, C., Kuhn, C., Brockhoff, A., Chuboda, E., Bufe, B., Appendino, G.,
Springer. doi: 10.1007/978-94-009-3417-7.
& Behrens, M. (2010). The molecular receptive ranges of human TAS2R bitter taste
Clifford, M. N. (1999). Chlorogenic acids and other cinnamates – nature, occurrence and
receptors. Chemical Senses, 35, 157–170. https://doi.org/10.1093/chemse/bjp09
dietary burden. Journal of the Science of Food and Agriculture, 79(3), 362–372.
Nehlig, A. (2018). Interindividual differences in caffeine metabolism and factors driving
https://doi.org/10.1002/(SICI)1097-0010(19990301)79:3<362::AID-
caffeine consumption. Pharmacological Reviews, 70, 384–411. https://doi.org/
JSFA256>3.0.CO;2-D
10.1124/pr.117.014407
De Graaf, C., Foata, N., Engkvist, O., & Rognan, D. (2008). Molecular modeling of the
Parliament, T. H. (2000). An overview on coffee roasting. In: T. H. Parliament, C.-T. Ho,
second extracellular loop of G-protein coupled receptors and its implication on
and P. Schieberle (Eds.), Caffeinated beverages – Health benefits, physiological effects,
structure-based virtual screening. Proteins, 71, 599–620. https://doi.org/10.1002/
and chemistry. ACS Symposium Series, ISBN 0-8412-3654-2.
prot.21724
Prewo, R., Guggisberg, A., Lorenzi-Riatsch, A., Baumann, T. W., & Wettstein-Bättig, M.
Illy, A., & Viani, R. (Eds.) (2005). Espresso coffee – The science of quality (2nd ed.).
(1990). Crystal structure of mozambioside, a diterpene glycoside of coffea
Elsevier Academic Press, ISBN: 0-12-370371-9.
pseudozanguerbariae. Phytochemistry, 3, 990–992. https://doi.org/10.1016/0031-
Food and Agriculture Organization of the United Nations, FAO (2024). https://www.fao.
9422(90)80065-O
org/markets-and-trade/commodities/coffee/en/, accessed 26. January 2024.
Richter, H., & Spiteller, G. (1979). Über ein neues furokauranglycosid aus grünen
Food and Agriculture Organization of the United Nations, FAO (24. March 2023).
kaffeebohnen. Chemische Berichte, 112, 1088–1092. https://doi.org/10.1002/
Erntemenge der führenden Anbauländer von Kaffee weltweit in den Jahren 2020
cber.19791120403
und 2021 (in 1.000 Tonnen) [Graph]. In: Statista, accessed 29. January 2024, http
Šeremet, D., Fabečić, P., Vojvodić Cebin, A., Mandura Jarić, A., Pudić, R., & Komes, D.
s://de.statista.com/statistik/daten/studie/216872/umfrage/kaffeeproduktion-fuehr
(2022). Antioxidant and sensory assessment of innovative coffee blends of reduced
ender-laender/.
caffeine content. Molecules, 27(2), 448. https://doi.org/10.3390/
FAO. (21. December 2023). Exportmenge der führenden Exportländer von geröstetem
molecules27020448
Kaffee weltweit im Jahr 2022 (in Tonnen) [Graph]. In: Statista, accessed 29. January
Stark, T., & Hofmann, T. (2005). Structures, sensory activity, and dose/response
2024, https://de.statista.com/statistik/daten/studie/224989/umfrage/fuehrende-e
functions of 2,5-diketopiperazines in roasted cocoa nibs (Theobroma cacao). Journal
xportlaender-von-geroestetem-kaffee/.
of Agricultural and Food Chemistry, 53(18), 7222–7231. https://doi.org/10.1021/
Farah, A., De Paulis, T., Trugo, L. C., & Martin, P. R. (2005). Effect of roasting on the
jf051313m
formation of chlorogenic acid lactones in coffee. Journal of Agricultural and Food
USDA Foreign Agricultural Service. (22. June, 2023). Erntemenge von Kaffee weltweit in
Chemistry, 53, 1505–1513. https://doi.org/10.1021/jf048701t
den Jahren 2011/12 bis 2023/24* (in 1.000 Säcken à 60 Kilogramm) [Graph]. In:
Frank, O., Blumberg, S., Krümpel, G., & Hofmann, T. (2008). Structure determination of
Statista, accessed 29. January 2024, https://de.statista.com/statistik/daten/studie/
3-O-caffeoyl-epi-γ-quinide, an orphan bitter lactone in roasted coffee. Journal of
319733/umfrage/erntemenge-von-kaffee-weltweit/.
Agricultural and Food Chemistry, 56, 9581–9585. https://doi.org/10.1021/jf802210a
USDA Foreign Agricultural Service. (22. Juni, 2023). Konsum von Kaffee in den
Frank, O., Blumberg, S., Kunert, C., Zehentbauer, G., & Hofmann, T. (2007). Structure
wichtigsten Absatzmärkten weltweit in den Jahren 2021/2022 bis 2023/2024* (in
determination and sensory analysis of bitter-tasting 4-vinylcatechol oligomers and
1.000 Säcken à 60 Kilogramm) [Graph]. In Statista (accessed 07/19/2023) https://
their identification in roasted coffee by means of LC-MS/MS. Journal of Agricultural
de.statista.com/statistik/daten/studie/216868/umfrage/kaffeeverbrauch-fuehre
and Food Chemistry, 55, 1945–1954. https://doi.org/10.1021/jf0632280
nder-laender/.
Frank, O., Zehentbauer, G., & Hofmann, T. (2006). Bioresponse-guided decomposition of
Weiss, C., Rubach, M., Lang, R., Seebach, E., Blumberg, S., Frank, O., Hofmann, T., &
roast coffee beverage and identification of key bitter taste compounds. European
Somoza, V. (2010). Measurement of the intracellular pH in human stomach cells: A
Food Research and Technology, 222, 492–508. https://doi.org/10.1007/s00217-005-
novel approach to evaluate the gastric acid secretory potential of coffee beverages.
0143-6
Journal of Agricultural and Food Chemistry, 58(3), 1976–1985. https://doi.org/
Ginz, M., & Engelhardt, U. H. (2000). Identification of proline-based diketopiperazines in
10.1021/jf903614d
roasted coffee. Journal of Agricultural and Food Chemistry, 48, 3528–3532. https://
Xu, W., Wu, L., Liu, S., Liu, X., Cao, X., Zhou, C., Zhang, J., Fu, Y., Guo, Y., Wu, Y.,
doi.org/10.1021/jf991256v
Tan, Q., Wang, L., Liu, J., Jiang, L., Fan, Z., Pei, Y., Yu, J., Cheng, Y., Zhao, S.,
Ginz, M., & Engelhardt, U. H. (2001). Identification of new diketopiperazines in roasted
Hao, X., Liu, Z.-J., & Hua, T. (2022). Structural basis for strychnine activation of
coffee. European Food Research and Technology, 213, 8–11. https://doi.org/10.1007/
human bitter taste receptor TAS2R46. Science, 377(6612), 1298–1304. 10.1126/sc
s002170100322
ience.abo1633.
Guillot, F. L., Malnoë, A., & Stadler, R. H. (1996). Antioxidant properties of novel
Ziegler, F., & Behrens, M. (2021). Bitter taste receptors of the common vampire bat are
tetraoxygenated phenylindan isomers formed during thermal decomposition of
functional and show conserved responses to metal ions in vitro. Proceedings of the
caffeic acid. Journal of Agricultural and Food Chemistry, 44, 2503–2510. https://doi.
Royal Society B: Biological Sciences, 288, Article 20210418. https://doi.org/10.1098/
org/10.1021/jf9508155
rspb.2021.0418
Hasan, C. M., Huda, Q., Lavaud, C., Connolly, J. D., & Haque, M. E. (1994). Bengalensol,
Ziegler, F., Steuer, A., Di Pizio, A., & Behrens, M. (2023). Physiological activation of
a new 16-epicafestol derivative from the leaves of Coffea bengalensis. Natural
human and mouse bitter taste receptors by bile acids. Communications Biology, 6(1),
Product Letters, 5, 55–60. https://doi.org/10.1080/10575639408043935
612. https://doi.org/10.1038/s42003-023-04971-3

12