No job name

J. Agric. Food Chem. 2002, 50, 2356−2364
Degradation of the Coffee Flavor Compound Furfuryl Mercaptan
in Model Fenton-type Reaction Systems
IMRE BLANK,*,† EDERLINDA C. PASCUAL,‡ STEÄPHANIE DEVAUD,† LAURENT B. FAY,† RICHARD H. STADLER,† CHAHAN YERETZIAN,† AND Nestec Ltd., Nestle´ Research Center, Vers-chez-les-Blanc, P.O. Box 44, CH-1000 Lausanne 26, Switzerland, and Scottish Crop Research Institute, Invergowrie, Dundee, DD2 5DA United Kingdom The stability of the coffee flavor compound furfuryl mercaptan has been investigated in aqueoussolutions under Fenton-type reaction conditions. The impact of hydrogen peroxide, iron, ascorbicacid, and ethylenediaminetetraacetic acid was studied in various combinations of reagents andtemperature. Furfuryl mercaptan reacts readily under Fenton-type reaction conditions, leading to upto 90% degradation within 1 h at 37 °C. The losses were lower when one or more of the reagentswas omitted or the temperature decreased to 22 °C. Volatile reaction products identified were mainlydimers of furfuryl mercaptan, difurfuryl disulfide being the major compound. In addition, a large numberof nonvolatile compounds was observed with molecular masses in the range of 92-510 Da. Theformation of hydroxyl and carbon-centered radicals was indicated by electron paramagnetic resonancespectra using R-(4-pyridyl-1-oxide)-N-tert-butylnitrone or 5-(diethoxyphosphoryl)-5-methyl-1-pyrroline-N-oxide as spin traps. Whereas •OH was generated by Fenton-type reactions, the C-centered radicalis probably a secondary product of the reaction of •OH with various organic molecules, the reactionwith furfuryl mercaptan appearing to be the most important. No evidence for S-centered radicals wasseen in the spin-trapping experiments, but a sulfur-containing radical was detected when measure-ments were made at 77 K in the absence of spin traps.
KEYWORDS:
Coffee; furfuryl mercaptan; thiols; Fenton chemistry; mass spectrometry; electron
paramagnetic resonance spectroscopy
INTRODUCTION
common constituent of liquid coffee (10, 11). In the presence Furfuryl mercaptan (2) has been reported to be a volatile
of transition metals in low oxidation states, H2O2 produces constituent of many foods and beverages (1), particularly when hydroxyl radicals (•OH) via the Fenton reaction (12): thermal treatment is involved in their production. In coffee, it was first described by Reichstein and Staudinger (2), and more recently 2 has been suggested to be an important odorant of
coffee (3). Its sensory relevance, evidenced by various groups
In coffee solutions, •OH is to some extent scavenged by (4, 5), is due to the “roasty, coffee-like aroma note” and low caffeine with the formation of 8-oxocaffeine (13). However, odor thresholds of 0.01 ng/L in air (6) and 0.01 µg/kg in water OH radicals are extremely reactive entities, able to extract hydrogen atoms from a wide range of organic molecules. They The concentration of 2 in roasted and ground coffee is
may, therefore, also attack odor-active thiols, such as furfuryl typically in the range of 1-2 mg/kg (8), but only about one- mercaptan, methyl mercaptan, 3-methyl-2-buten-1-thiol, 2-meth- third of this is detected in coffee brews (9). This might be due yl-3-furanthiol, and 3-methyl-3-mercaptobutyl formate (14), to low extractability during the preparation of coffee beverages which could potentially lead to alteration of the coffee aroma.
or the consequence of sensitivity to oxidative processes. In This study has investigated the stability of furfuryl mercaptan addition to atmospheric oxygen, oxidative processes in bever- in model systems under Fenton-type reaction conditions.
ages may be initiated by hydrogen peroxide (H2O2), which is a Experiments were conducted with the objectives of (i) determin-
ing the losses of 2, (ii) identifying the major volatile degradation
* Authors to whom correspondence should be addressed [I.B. (principal corresponding author) telephone +41 21 7858607, fax +41 21 7858554, products, (iii) characterizing the free radicals produced in these e-mail imre.blank@rdls.nestle.com; B.G. (EPR spectroscopy) telephone +44 reactions, and (iv) studying the mechanisms involved in the 1382 568532, fax +44 1382 562426, e-mail bgoodm@scri.sari.ac.uk].
oxidative/radical-induced degradation of 2. Chromatographic and
‡ Scottish Crop Research Institute.
mass spectrometric techniques were used to quantify the levels Degradation of Furfuryl Mercaptan by the Fenton Reaction J. Agric. Food Chem., Vol. 50, No. 8, 2002 Et2O (1 mL). The organic phase was centrifuged (5-15 min, 3500 Table 1. Experimental Design To Study the Effect of Fenton Reagents
rpm) and analyzed by GC coupled with a flame ionization detector on the Degradation of Furfuryl Mercaptan in Aqueous Model Systemsa (FID) and/or a mass spectrometer (MS). All reactions were performed at least in duplicate and samples injected twice into the GC.
(iii) Roles of Components of the Model Fenton-type Reaction System. The composition of the Fenton model systems described above (Table
1) was modified to study the roles of selected constituents. For example,
ascorbic acid was replaced by 4-hydroxy-2,5-dimethyl-3(2H)-furanone and FeCl3 by MnCl2 or CuCl. In some other model reactions, the concentrations of the reactants and reaction time, as well as the reaction temperature, were varied. All reactions, however, were performed at physiological (37 °C) and/or room (20-22 °C) temperature.
(iV) Volatile and NonVolatile Reaction Products. A simplified model a Values represent volumes expressed in µL. The total volume of each sample reaction was carried out for the characterization of volatile and nonvolatile compounds by mass spectrometric analysis. The model
reaction was composed of 2, H2O2, and FeCl3 in the ratio of 10:1:1 on
of 2 in model solutions under conditions of oxidative stress and
a molar basis. After 2 h at room temperature, neutral compounds were to identify products generated (15). Electron paramagnetic extracted with Et2O (2 × 50 mL). The combined etheral phases were resonance (EPR) spectroscopy was used to detect free radical dried over anhydrous Na2SO4 and concentrated to 20 mL. Volatile species, mainly by the chemical spin-trapping approach, in compounds were separated by distillation in high vacuum (3 × 10-5 which the reaction of spin traps (e.g., nitrones) with unstable mbar) and collected in two glass traps cooled with liquid nitrogen (17).
free radicals generates new more stable radicals (nitroxides), The distillate was concentrated to 1 mL using a Vigreux column (50× 1 cm) for MS analysis of volatile compounds. The residue was taken which can then be characterized (16).
for analysis of nonvolatile degradation products of 2.
(V) Kinetic Studies. Experiments were performed at 37 °C on samples MATERIALS AND METHODS
analogous to sample 1 in Table 1, except that 5 times higher volumes
Chemicals. The purity of all chemicals was at least of analytical
were used to obtain five data points from the same preparation. EtOH grade. Furfuryl mercaptan (2) was purchased from Aldrich (Buchs,
(0.1 mL) was added to 1 mL aliquots after 5, 10, 20, 40, or 120 min Switzerland) or Sigma (Dorset, U.K.). Benzyl mercaptan, cuprous to terminate the reaction. The cleanup procedure was as described in chloride (CuCl), difurfuryl monosulfide (5), difurfuryl disulfide (7),
section ii above. The concentrations of 2, 5, 7, and 8 were determined
4-hydroxy-2,5-dimethyl-3(2H)-furanone, and manganese(II) chloride (Vi) Spin Trapping of Unstable Free Radicals. Solutions for EPR 2) were from Aldrich. Ascorbic acid was from Aldrich or FSA Laboratory Supplies (Loughborough, U.K.). Tripotassium phosphate measurements at room temperature (20 ( 2 °C) and 37 °C were similar to samples 1, 4, and 5 (Table 1), except that a spin trap solution (100
3PO4), ferrous sulfate heptahydrate (FeSO4 7H2O), and R-(4-pyridyl- 1-oxide)-N-tert-butylnitrone (4-POBN) were from Sigma. Diethyl ether mM concentration in the final solution for 4-POBN or 5.3 mM for 2O), ethanol (EtOH), H2O2, ethylenediaminetetraacetic acid (EDTA, disodium salt), ferric chloride hexahydrate (FeCl ‚ Fenton-type reaction was initiated only after all of the other reagents were present. Samples were taken at intervals during the 37 °C 2HPO4), sodium dodecyl sulfate (SDS), and an- incubation and transferred rapidly to a quartz flat cell (Wilmad many) or Sigma. 5-(Diethoxyphosphoryl)-5-methyl-1-pyrroline-N-oxide WG812Q, Fluorochem, Old Glossop, Derbyshire, U.K.) for the EPR (DEPMPO) was from Calbiochem-Novabiochem (Beeston, Notting- measurements, which were made at room temperature. Spectra from incubations at room temperature were made on samples in flat cells, Sample Preparation. (i) Loss of Furfuryl Mercaptan (2) from
which remained in the EPR cavity for the duration of the experiment, Solutions at Room Temperature. Aqueous solutions of 2 (0.427 mmol/
as were additional measurements using DEPMPO as spin trap with solutions from which EDTA or ascorbic acid were omitted. Further mL) were mixed to obtain 10:1 ratios by volume of 2/FeCl
samples were also prepared with DEPMPO as spin trap and were 3 and 2/H2O2.
The samples (pH 5.1) were stored at room temperature (22 °C) for 8 analogous to samples 1 and 5, but without 2.
days, and the loss of 2 was monitored as a function of time. Aliquots
(Vii) Low-Temperature EPR Measurements. For low-temperature (77 (5 mL) were taken after 2, 5, 24, 48, 120, and 192 h. Benzyl mercaptan K) EPR measurements, a solution was prepared with just FeSO ‚ (0.5 mL) was added as internal standard and the pH adjusted to 4.0 (0.4 mM), H2O2 (38 mM), and 2 (87 mM). Samples were taken at
with aqueous HCl (0.1 M). Neutral compounds were extracted with intervals after mixing the reagents at room temperature, transferred to a 4 mm i.d. quartz tube, and frozen immediately in liquid nitrogen.
2O (5 mL), and 2 was quantified by gas chromatography (GC).
(ii) Loss of Furfuryl Mercaptan from Fenton-type Model Systems. For EPR measurements the tubes were transferred to a quartz “Finger The degradation of 2 was investigated in a systematic study for a series
Dewar” (Wilmad WG816B) filled with liquid nitrogen, which was of eight model Fenton-type reaction systems (Table 1) at pH 5.5, which
is close to the pH of coffee beverages. Sample 1 (complete Fenton (Viii) γ-Irradiation of Furfuryl Mercaptan. Samples of 2 and 10%
model) contained all of the reagents for a Fenton reaction. EDTA was 2 in water were placed in 3 mm i.d. quartz EPR tubes and frozen in
used to ensure complete solubilization of Fe(III), and ascorbic acid liquid nitrogen. They were maintained at that temperature during and was used for the reduction of Fe(III) to Fe(II), which initiated after irradiation with 0.6 MeV γ-rays (received dosage ∼ 100 Gy) from a 137Cs source of ∼63 TBq in a radiation facility in the School of 2O2 to yield OH radicals. In samples 2-5 one of the reagents was omitted, whereas samples 6 and 7 had more than one Biology at the University of St. Andrews.
component of the Fenton reaction system missing. The control sample Capillary GC. A Hewlett-Packard gas chromatograph (HP-5890)
8 was a buffered aqueous solution of 2.
equipped with an autosampler (HP-7673A) and cold on-column injector The aqueous solutions described in Table 1 were freshly prepared
was used. Samples were analyzed on an OV-1701 fused silica capillary before use, that is, ascorbic acid (20 mM), EDTA (25 mM), H2O2 column, 30 m × 0.25 mm i.d., film thickness 0.25 µm (J&W Scientific, 3 6H2O (10 mM), K2HPO4 (20 mM, pH 5.5), and 2 (33.3
Folsom, CA). The column pressure was 80 kPa using helium as carrier mM in 3.3% aqueous SDS), the latter being added as the last reagent gas. The effluent was split 1:1 to a flame ionization detector (FID) to the mixtures. Samples were stirred vigorously for 5 s and incubated and a flame photometric detector (FPD). For the oven temperature for 1 h at 37 °C, after which time EtOH (0.1 mL) was added to terminate the Fenton reaction. Chemical analyses were performed after the pH Quantification of Volatile Compounds. This was performed by
had been adjusted to pH 3.5 and neutral compounds extracted with GC-FID on a DB-5 capillary column (J&W Scientific) using benzyl J. Agric. Food Chem., Vol. 50, No. 8, 2002 Table 2. Degradation of Furfuryl Mercaptan (2) in Aqueous Model
Table 3. Effect of Fenton Reagents on the Decomposition of Furfuryl
Systems Containing Fenton Reagents as a Function of Timea Mercaptan (2) in Aqueous Model Systems
2 (control)
2/FeCl b
2/H2O2 (10:1)
concn of 2b (mg/50 mL)
loss of 2 (%)
a Amounts of 2 are related to its concentration at t ) 0 h (2.85 µmol/mL) and
expressed in percent. b See (i) under Sample Preparation for more details.
mercaptan as internal standard (49.4 mg/100 mL Et For more details see Table 1 and (ii) under Sample Preparation. b Mean values
added to the solutions after quenching with EtOH. The pH was adjusted with standard deviations were obtained using benzyl mercaptan as internal standard.
rapidly to 3.5, and the cleanup was performed as described in section The variation coefficients were 1−6% with 3.5% in average. c Initial concentration ii under Sample Preparation above. The response factors were of 2 was 21.6 mg/50 mL (3.8 mM). d Initial concentration of 2 was 18.2 mg/50 mL
determined in model solutions with known amounts of benzyl mercaptan and the compounds to be quantified, that is, 1.57 (2), 1.49 (5), and
1.66 (7). The response factor for 8 was set at 1.7.
Table 4. Effect of Selected Reagents on the Decomposition of Furfuryl
Mass Spectrometry. (i) Capillary GC-MS. Electron impact (EI) and
Mercaptan (2) in Aqueous Fenton Model Systems at 37 °C
positive chemical ionization (PCI) mass spectra were obtained on aFinnigan MAT 8430 mass spectrometer (Finnigan MAT, San Jose, CA) concn of 2a
at 70 and 150 eV, respectively. Ammonia was used as reagent gas for loss of 2 (%)
PCI. Volatile compounds were sampled via a cold on-column injector (HP-5890 GC) using the conditions described above. Relative abun- dances of the ions are given in percent.
(ii) Electrospray Ionization (ESI)-MS. Nonvolatile compounds were investigated by ESI-MS using a Finnigan TSQ 700 triple-quadrupole mass spectrometer equipped with an electrospray ionization source. Thisworked with a voltage of 4.5 kV and a transfer capillary heated at 150 a Initial concentration of 2 was 51.2 mg/25 mL (18 mM). Quantification is relative
°C. Argon was used as a sheath gas at a pressure of 40 psi. The samples to benzyl mercaptan as internal standard, added to the sample after quenching were introduced by continuous infusion at 5 µL/min using a Harvard with ethanol. For more details see (iii) under Sample Preparation. b FeCl3 was model 22 syringe pump. Data acquisition was performed on a DEC replaced by the same concentration of MnCl2 (10 mM). c FeCl3 was replaced by station 2100 running under Ultrix 4.2A (Digital Equipment) using the the same concentration of CuCl (10 mM). d Ascorbic acid was replaced by the Finnigan software package ICIS2, version 7.0. Mass spectra were same concentration of 4-hydroxy-2,5-dimethyl-3(2H)-furanone (20 mM).
acquired in positive mode by scanning from m/z 20 to 1000 in 1 s.
EPR Spectroscopy. EPR measurements were made at X-band
Loss of Furfuryl Mercaptan under Fenton-type Reaction
frequencies (∼9.5 GHz) using a Bruker ESP300E (Bruker U.K. Ltd., Conditions. Although 2 on its own is rather stable in aqueous
Banner Lane, Coventry, U.K.) computer-controlled spectrometer in- solutions, it is rapidly decomposed in the presence of H2O2 and corporating an ER4103TM cylindrical cavity. Microwave generation transition metals, as shown in Table 3. The concentration of 2
was by means of a klystron (ER041MR), and the frequency was remaining after a 1-h incubation at 37 °C was strongly dependent measured with a built-in frequency counter. All spectra were collected on the composition of the solution. Apart from the results for in 1024 data points using a modulation frequency of 100 kHz. A sample 2, these data are in good agreement with those published microwave power of 10 mW and modulation amplitude of 0.1 mT were in ref 15. The previously reported value of 45% for the loss of used for fluid solution measurements at room temperature. The 2 in sample 2 might be explained by trace impurities of transition
respective values for low temperature (77 K) spectra were 0.5 mWand 0.5 mT.
metals in the water used. These can promote the reaction of As is conventional in EPR spectroscopy, spectra were recorded as H2O2 in the presence of reducing agents and, thus, initiate first derivatives of the microwave absorption and displayed as functions decomposition of 2. Light also affects the stability of H2O2 and
of absorption versus magnetic field at a constant microwave frequency.
can lead to the generation of •OH.
In a small number of instances, second-derivative spectra were also The degradation of 2 is strongly temperature dependent
recorded to enhance resolution of the hyperfine structure from overlap- (Table 3). At room temperature, only ∼20% of 2 was
ping components. In most cases, spectral interpretations were confirmed decomposed in sample 1 after a reaction time of 1 h, in contrast by simulation using the Bruker Simfonia software package. However, to ∼90% at 37 °C. Successive repetitions of the experiment measurements of variations in intensity as a function of time are based led to similar results, that is, losses of 80-90% of 2 at 37 °C
on the maxima and minima of the first peak of first-derivative spectra and of 10-20% at room temperature. Negligible losses of 2
after smoothing by a double application of a 15-point polynomial were observed in samples 2 and 5-8, and only a small loss (5-10%) was observed in sample 3. In contrast, when EDTA
was omitted from the reaction mixture (sample 4), ∼70% of 2
RESULTS AND DISCUSSION
was lost (i.e., comparable to that observed at 37 °C).
Stability of Furfuryl Mercaptan in Aqueous Solutions. To
Replacing Fe(III) by either Mn(II) or Cu(I) in reactions at provide background information on the stability of 2, its
37 °C decreased the loss of 2 from 90 to 15-25% (Table 4).
concentration in aqueous solutions was monitored by GC-FID Changing the transition metal from Fe to Mn or Cu has also at room temperature over a time period of 8 days in the presence been reported to decrease the level of oxidation of caffeine to of either Fe(III) or H2O2. Compound 2 is relatively stable, and
8-oxocaffeine under Fenton reaction conditions (13). These only 12% was lost from an aqueous solution after 24 h (Table
results suggest, therefore, that Fe(III)/Fe(II) is more effective 2). The rate of loss was increased in the presence of either Fe-
than the other metal ions in cleaving H2O2, probably because (III) or H2O2, but >60% remained after 8 days of incubation.
of easier redox cycling between the two oxidation states.
Degradation of Furfuryl Mercaptan by the Fenton Reaction J. Agric. Food Chem., Vol. 50, No. 8, 2002 Table 5. Mass Spectrometric Data of Volatile Degradation Products of Furfuryl Mercaptan Found in Various Fenton Reaction Systemsa
mass spectrometric fragment ionsc (m/z, relative intensity) 96 (M+, 70), 96 (65), 67 (5), 42 (5), 40 (10), 39 (100), 38 (35), 37 (25), 29 (35) 114 (M+, 50), 81 (100), 53 (65), 52 (10), 51 (10), 50 (10), 45 (10), 39 (5), 27 (15) 162 (M+, 35), 81 (90), 53 (25), 51 (5), 39 (10), 28 (100), 27 (30) 194 (M+, 25), 113 (10), 85 (10), 81 (100), 53 (25), 45 (5), 43 (5), 27 (5) 194 (M+, 25), 126 (10), 113 (20), 85 (5), 81 (100), 53 (30), 51 (10), 45 (15), 27 (15) 212 (M+, 1), 194 (30), 161 (5), 113 (35), 100 (10), 81 (100), 53 (20), 45 (10), 43 (30), 27 (10) 226 (M+, 25), 161 (5), 85 (5), 81 (100), 53 (45), 51 (10), 45 (10), 27 (20) 258 (M+, 3), 193 (2), 161 (5), 113 (3), 85 (5), 81 (100), 53 (25), 51 (5), 45 (10), 27 (10) 384 (M+, 1), 225 (10), 193 (40), 161 (10), 62 (5), 81 (100), 53 (15), 45 (5), 27 (5) a Volatile compounds were isolated from the complete Fenton reaction model [sample 1, (ii) under Sample Preparation] except bifurfuryl (3), which was identified in the
simplified reaction sample as described under (iv) of Sample Preparation. b Linear retention indices were calculated on OV-1701 capillary columns. c Mass spectrometrywas performed using the electron impact (EI) and chemical ionization (CI, ammonia) technique. d Unknowns are sulfur-containing volatile compounds.
Figure 1. GC-MS identification of volatile compounds detected in a sample containing furfuryl mercaptan, hydrogen peroxide, and iron(III) ions in the
molar ratio of 10:1:1: (A) total ion current; (B) extract thereof showing only the trace of m/z 81. Numbering corresponds to that in Table 5.
9). 4-Hydroxy-2,5-dimethyl-3(2H)-furanone may, therefore, Table 6. Formation of Volatile Degradation Products of Furfuryl
Mercaptan under Fenton Conditionsa
represent an important component for the redox cycling of theFe during oxidative processes in liquid coffees.
Volatile Degradation Products of Furfuryl Mercaptan. As
reported previously (15), most of the volatile degradation products detected in sample 1 contain sulfur. Compound 7 was
the major degradation product of 2 followed by 5 and 8. Three
minor peaks detected by GC-FPD and GC-MS remain unknown (4, 6, and 9). Their structures are most likely related to 2,
because they all have a major fragment ion with m/z 81 (Table
5), which is characteristic of the furfuryl moiety.
The distribution of volatile degradation products of 2 is
a Concentrations are in µg/mL using benzyl mercaptan as internal standard.
influenced by the composition of the Fenton-type model The variation coefficients were <10%. b For experimental details see Table 1 and
systems. Compound 5 was found preferentially in sample 1,
(ii) under Sample Preparation. c Only trace amounts were detected (<1 µg/mL).
whereas 8 was detected at similar levels in samples 1, 4, and 5
(Table 6). Compound 7 was the most abundant volatile
Replacing ascorbic acid with the cyclic enoloxo compound degradation product of 2 in all samples, indicating that it is very
4-hydroxy-2,5-dimethyl-3(2H)-furanone in the Fenton reaction readily formed. The amounts varied from 3 to 80 µg/mL, the mixture led to ∼50-60% loss of 2 after a 1-h incubation at 37
highest amounts (50-80 µg/mL) being in samples 1, 4, and 5, °C (Table 4), a result which is comparable to that obtained with
for which the losses of 2 were greatest.
ascorbic acid. The amounts of 4-hydroxy-2,5-dimethyl-3(2H)- In a simplified model system, 2, H2O2, and FeCl3 in the molar
furanone in roasted and ground coffee are relatively high (∼100 ratio 10:1:1 was reacted for 2 h at room temperature to study mg/kg), and these are fully recovered in the coffee brews (8, the degradation products of 2. In the volatile fraction of this
J. Agric. Food Chem., Vol. 50, No. 8, 2002 Figure 2. ESI-MS of a sample containing furfuryl mercaptan, hydrogen peroxide, and FeCl3 in the molar ratio of 10:1:1.
presence of 2 and suggests that 2 can act as a reducing agent to
generate Fe(II).
Compounds 3, 5, and 7, which have been identified as
decomposition products of 2 in the present experiments, are
known volatile components of roasted coffee (18-20). The
aroma notes of 5 and 7 are described as burnt, sulfury, roasty,
and rubbery, but they lack the characteristic coffee-type aroma
of 2. When their low odor thresholds are considered, for
example, 0.0004 ng/L in air for 7 (21), it is likely that the loss
of 2 and concomitant formation of various difurfuryl sulfides
and other furfuryl-based moieties will lead to an imbalance in
coffee aroma during storage or under conditions of oxidative
Figure 3. Degradation of furfuryl mercaptan under Fenton-type reaction
conditions over a time period of 2 h.
Nonvolatile Degradation Products of Furfuryl Mercaptan.
Quantitation of the total concentrations of volatiles detected by
GC techniques showed that ∼40-50% of furfuryl mercaptan
equivalents were missing from samples 1, 4, and 5 at 37 °C,
for which losses of 2 were particularly high. Consequently, an
appreciable fraction of the products of 2 degradation in the
presence of Fenton reagents is thought to occur as nonvolatile
substances (15). These nonvolatile compounds might be ionic
low molecular weight compounds or polymeric materials that
cannot be analyzed by GC-hyphenated techniques. Therefore,
direct inlet MS was applied using electrospray ionization in an
attempt to characterize the residue of the simplified Fenton
reaction sample obtained by vacuum distillation. As shown in
Figure 2, a large number of compounds with mass range from
M ) 92 to 510 Da were found, with the majority being in the
range of M ) 124-262 Da. Unfortunately, these data do not
allow conclusions to be drawn about the chemical nature and
structure of the compounds, but they do indicate the chemical
Figure 4. General tendency in the formation of difurfuryl monosulfide
complexity of the nonvolatile fraction obtained from a simple Fenton reaction system. Some ions also suggest the presence O, 5), difurfuryl disulfide (0, 7), and difurfuryl trisulfide (∆, 8) over time
as degradation products of furfuryl mercaptan under Fenton conditions.
of volatile compounds, such as furfural (m/z 97), 2 (m/z 115),
5 (m/z 195), unknown 6 (m/z 213), and 7 (m/z 227), as minor
sample, the fragment ion at m/z 81 (Figure 1) revealed bifurfuryl
(3), which might be formed by dimerization of two furfuryl
Changes in Volatile Compounds over Time. To get an
radicals, in addition to 5, 7, and 8. An unknown compound (4)
insight into the dynamics of degradation of 2, a series of
at RI ) 1530, showing MS data similar to those of 5 (Table
measurements on sample 1 at 37 °C were performed to 5), was also detected. This experiment thus confirms that H2O2
determine the concentration of the principal volatile compounds, and FeCl3 are capable of initiating the Fenton reaction in the that is, 2, 5, 7, and 8, over a time period of 2 h. The initial
Degradation of Furfuryl Mercaptan by the Fenton Reaction J. Agric. Food Chem., Vol. 50, No. 8, 2002 Figure 5. Representative EPR spectra obtained during incubation of model solution 1 with the spin trap DEPMPO or 4-POBN at room temperature
(∼20 °C).
Figure 6. Variation of the intensities of the EPR signals from •OH ([)
and C-centered radical (9) adducts along with those of the unidentified
Figure 8. (A) EPR spectrum at 77 K from a rapidly frozen solution of
adduct with high- and low-field peaks separated by 11.6 mT (2) and furfuryl mercaptan in a simple model Fenton reaction system; (B) a 10.4 mT (×) as a function of incubation time with DEPMPO at 37 °C of simulation of spectrum (A) with g values of 2.002, 2.011, and 2.023 and (A) sample 1, (B) sample 1 minus furfuryl mercaptan, and (C) sample 4, a Gaussian line shape; (C) spectrum obtained at 77 K from furfuryl mercaptan that had received 100 Gy of γ-radiation from a 137Cs source.
The concomitant formation of 5, 7, and 8 over time is shown
in Figure 4. Compound 7 was detected as early as the 5 min
sample. It was the major product generated, the concentration
of which increased throughout the reaction period. In the 10-
min sample, 8 was found as a second degradation product. Its
concentration, however, increased less rapidly than that of 7,
and it reached a plateau after ∼40 min. Compound 5 was not
detected in the first 20 min, but its concentration increased
rapidly over the next 20 min. These data suggest a complex
series of reactions leading to the formation of these difurfuryl
derivatives and are consistent with the types of events expected
for free radical mediated processes.
Figure 7. Variation of the intensities of the EPR signals from the •OH
The data presented in this work clearly demonstrate the crucial adduct ([), along with that of the unidentified adduct with high- and low- role of H2O2 and iron in the degradation of 2. Reductive
field peaks separated by 12.2 mT (b), as a function of incubation time cleavage of the O-O bond in H2O2 is catalyzed by Fe(II) with with DEPMPO at 37 °C of (A) sample 5, which contained no ascorbic the generation of •OH, which is a powerful oxidant. •OH readily acid, and (B) sample 5, but without furfuryl mercaptan.
reacts with organic compounds, including thiols, and initiatesformation of volatile and nonvolatile degradation products with phase of furfuryl mercaptan decomposition was fast but reached 2. In addition to •OH, the formation of metal-based oxidizing
a plateau after ∼10-15 min of incubation when ∼25% of 2
species, such as Fe(II)OOH and iron(IV)-oxo complexes (13, was lost (Figure 3). Rapid decomposition set in again after ∼20
22), has also been reported to occur during the Fenton reaction.
min, and only ∼10% of 2 remained after a reaction period of
EPR Spectroscopy of Spin Trap Adducts. The products
determined by GC-MS analysis of sample 1 suggest that both J. Agric. Food Chem., Vol. 50, No. 8, 2002 Figure 9. Hypothetical mechanism showing the degradation of furfuryl mercaptan (2) under Fenton conditions and the formation of volatile reaction
products identified in this study, that is, bifurfuryl (3), difurfuryl monosulfide (5), difurfuryl disulfide (7), and difurfuryl trisulfide (8).
C- and S-centered radicals are involved as intermediates in the When 2 was excluded from the complete Fenton reaction
degradation of 2. Consequently, experiments using EPR spec-
system, the •OH adduct signal increased more rapidly than with troscopy have been performed in an attempt to provide the original solution. After 1 h of incubation in the presence of information on free radical species generated in the reaction DEPMPO, its intensity was approximately twice that which was observed when 2 was present (Figure 6A,B). This approxi-
When solutions equivalent to sample 1 were incubated with mately 2:1 ratio for the •OH adduct signal intensity for solutions DEPMPO at room temperature, the spectra obtained over a 20-h without and with 2 is in line with the 92:59 ratio that would be
period (Figure 5) were all dominated by the •OH adduct with
predicted on the basis of the combined molarities of 2,
hyperfine splittings a(31P) ) 4.72 mT, a(14N) ) 1.38 mT, and DEPMPO, EDTA, and ascorbic acid, if •OH showed no reaction a(1H) ) 1.40 mT (23). The ascorbate radical (a ) selectivity among them. The result does indicate, however, that (24) was seen in the early stages of incubation but decreased in the buffer is not reactive in these solutions. The intensity of intensity with time and was not detectable after ∼1 h. This the unknown radical adduct also increased more rapidly and its radical may, however, contribute to various redox processes in contribution to the overall EPR signal was appreciably higher the reaction mixtures, and the role of ascorbic acid may not be in the absence of 2, thus adding further support to it being
formed as a result of a reaction between DEPMPO and Fenton There were small amounts of a component from C-centered radical adducts [a(31P) ) 4.91 mT, a(14N) ) 1.47 mT, and a(1H) Compared to the results described above, removal of EDTA ) 2.13 mT] (23) in spectra from ∼6 h, but it was never more from the reaction mixture resulted in a major reduction in the than a minor fraction of the total signal. A further unidentified rate of generation of the •OH adduct signal in the presence of signal, with outermost peaks separated by 11.6 mT, is also DEPMPO (Figure 6C). Only very weak signals were observed
present in these spectra. It is related to the Fenton reaction, as from the C-centered radical adduct and the unidentified adduct it is also seen when DEPMPO is added to a simple Fenton with outer peak separations of 11.6 mT. This result is consistent system (unpublished results), and its spectrum can be analyzed with a reduction in the rate of generation of •OH. It illustrates in terms of the approximate hyperfine splittings, a(31P) ) 4.55 the important role played by EDTA in maintaining a sufficient mT, a(14N) ) 1.38 mT, and a(1H) ) 1.64 mT, along with either concentration of Fe(III) for effective redox cycling and compet- an additional doublet splitting with a ) 2.67 mT or a triplet ing with hydrolysis reactions, which lead to precipitation of with a ) 1.34 mT. If the latter assignment is correct, this could Fe(III) oxyhydroxide species (28). In addition to the components represent an adduct of an oxidized spin trap molecule, analogous described above, the EPR spectra contained a further weak signal to that seen with 4-POBN (25).
with outer peak separations of 10.4 mT. This adduct has not Results obtained at room temperature with 4-POBN as spin yet been identified, and its intensity showed little variation.
trap (Figure 5) were qualitatively similar to those described
The solution from which ascorbic acid was absent showed above for DEPMPO. In the early stages of incubation, the OH an initial rapid increase in •OH adduct signal, but this reached radical adduct [a(14N) ) 1.50 mT and a(1H) ) 0.16 mT] (26) a maximum after a few minutes (Figure 7A). This provides
was the dominant species, and small amounts of the ascorbate further evidence to support the suggestion that 2 can reduce
radical were seen, although it is partly obscured by the spectrum Fe(III) to Fe(II) for the Fenton reaction. This hypothesis is of the adduct. The signal from C-centered radical adducts [a(14N) supported by the observation that the rate of production of the ) 1.56 mT and a(1H) ) 0.26 mT] (26), however, grew EPR signal was much slower in a similar solution without 2
progressively with time and dominated the spectra after ∼15 (Figure 7B), although it did eventually achieve a similar
h. This result indicates that appreciable quantities of C-centered intensity. However, the relatively short duration of the •OH burst radical adducts were generated during longer incubation periods.
suggests that 2 oxidation products are not readily recycled back
Small amounts of a signal with a(14N) ) 1.42 mT and a(1H) ) to 2. An additional spectrum with outer peak separations of 12.2
1.46 mT were observed from ∼1 h. This reached a maximum mT from another unidentified adduct was observed with the around 3-4 h and then subsequently decreased in intensity. The solution containing 2. The intensity of this signal decreased
parameters are similar to those of the tert-butyl hydronitroxide radical (27), which is thought to be a breakdown product of In addition to reactions initiated by •OH from the Fenton reaction system, it is possible that radicals could be generated Degradation of Furfuryl Mercaptan by the Fenton Reaction J. Agric. Food Chem., Vol. 50, No. 8, 2002 from reactions between other components in the reaction derived from 2 (W. Andreoni, IBM Ru¨schlikon, personal
mixtures. For example, the reaction between 2 and Fe(III) could
communication), 2a and 2c, respectively, and it is conceiveable
produce Fe(II) and furfuryl mercaptan-derived radicals. These that the spin traps used in the present work show a preference radicals and the ascorbate radical, which is also formed as a result of reaction with Fe(III), may then participate in further The formation of 7 after the addition of 2 to the Fenton
reactions and could be the sources of one or more of the reaction system was very rapid, and this is the major furfuryl- unidentified radical adducts described above. Such reactions based moiety identified in the volatile fraction of the reaction could be more important in coffee brews than in the model mixture. Production of an S-centered free radical was also rapid.
system because of the greater abundance of potentially reactive The observation that dimerization of thiyl radicals (2a) is an
early event in the decomposition process of 2 (Figure 4) adds
Identification of the Presence of S-Containing Radicals.
support to the tentative identification of this radical as a dimeric There was no direct evidence for any S-centered radical species species, although it is also likely that sulfinyl radicals will be in the spin-trapping experiments, despite the GC-MS results present in this system. Disproportionation of 2a to the radicals
suggesting that such radicals are intermediates in the formation 2d and 2e and their subsequent reactions might represent a route
of some of the reaction products identified. However, when to the other volatiles, the formation of which is delayed relative higher concentrations of reagents were used and the reaction to the formation of 7. Other free radicals are also formed in the
mixture was frozen in liquid nitrogen to stabilize any radical(s) reaction medium (e.g., from ascorbate), and these may interfere that might be generated, an anisotropic radical signal was with the main reaction scheme. Such radical formation may observed at 77 K (Figure 8A) with g values of 2.002, 2.011,
explain the observation that the highest level of 7 was found in
and 2.023 (Figure 8B). The radical had moderate stability, with
sample 5, in which ascorbic acid was absent (Table 6).
a half-life of a few minutes, and is probably an intermediate in Furfurylsulfenic acid (MW ) 130 Da) was not detected in the decomposition of 2. It does not appear to be reactive with
the mass spectrometric measurements, although the identification either of the spin traps used in the present investigation.
of 7 and 8 as reaction products indicates that its formation is
There are several studies in the literature of S-containing probable. As shown in Figure 2, there is some evidence for
radicals with parameters similar to those observed here, but there the higher oxidation product furfurylsulfinic acid (MW ) 146 is no universal agreement on their identities. γ-Irradiation of Da), but not for furfurylsulfonic acid (MW ) 162 Da). In similar thioglycolic acid, methionine, or acetylmethionine yields a reaction systems, that is, ascorbate- and transition-metal- radical with g values virtually identical to those from furfuryl mediated oxidation of methanethiol, methanesulfenic acid (CH3- thiol in the present work. On the basis of ENDOR studies on SOH) has been proposed as an intermediate in the formation single crystals of thioglycolic acid, it has been argued that this of dimethyl disulfide and dimethyl trisulfide (36). However, its signal corresponds to radicals of the type (R2S)•+ (29). However, existence could not be substantiated, possibly because of the it has also been assigned to dimeric cationic radicals of the type high reactivity of CH3SOH (37). It is known that sulfenic acids (R2S-SR2)•+ (30, 31). The g values are also close to those of easily convert to thiosulfinate esters due to their dual electro- sulfinyl radicals (R2SO)•+ (32, 33) and the radical (RSSH)•- philic/nucleophilic character (38).
(34). A number of possible radicals can, however, be eliminated In conclusion, the processes by which furfuryl mercaptan is on the basis of EPR data, including general species such as degraded in Fenton-type reaction systems are complex. Evidence (RS)•, (RS)•+, (RSS)•, and RSSR•-.
has been found to support reaction pathways involving both C- When frozen 2 was exposed to γ-irradiation at 77 K, a
and S-centered free radical intermediates. These lead to the complex spectrum was obtained (Figure 8C). One of the
formation of a large number of volatile and nonvolatile products, components bears a close resemblance to that in Figure 8A,
among which are the odor-active compounds difurfuryl mono- suggesting that they may correspond to the same radical. The sulfide, difurfuryl disulfide, and difurfuryl trisulfide. However, fact that a completely different spectrum (not shown) was the differences in the nature of the aroma notes ascribed to 2
observed when a 10% solution of 2 was irradiated provides
and those of the volatile compounds described above indicate further evidence that a dimeric species derived from 2 has
that •OH-initiated degradation of 2 is likely to lead to a distortion
parameters similar to those in Figure 8. However, mercaptans
of the aroma in oxidized coffee brews. The existence of Fenton readily react with peroxyl radicals to generate sulfinyl radicals chemistry may, therefore, represent a significant factor that (33), which might also be generated in our model system.
affects the aroma composition of coffee beverages. The practical Indeed, sulfinyl radicals have been observed as a result of significance of other oxidation products has not been considered reaction of tert-butyl mercaptan in the Ti(III)/H2O2 Fenton reaction system (35). Unfortunately, the EPR spectra do notdiscriminate between the sulfinyl and dimeric cationic radicals, ACKNOWLEDGMENT
so it is not possible to identify their relative levels in the presentsystem.
We are grateful to S. Metairon for expert technical assistance.
Formation Mechanism. On the basis of the results obtained
in the present study, a hypothetical mechanism is proposed for LITERATURE CITED
the degradation of 2 under Fenton-type reaction conditions
explaining the formation of volatile degradation products
(1) Nijssen, L. M.; Visscher, C. A.; Maarse, H.; Willemsens, L. C.; (Figure 9). The initial phase involves attack by OH radicals
Boelens, M. H. Volatile Compounds in Foods. QualitatiVe and and the abstraction of an H atom. This leads to the generation QuantitatiVe Data, 7th ed.; TNO Nutrition and Food ResearchInstitute: Zeist, The Netherlands, 1996; pp 72.1-72.23.
of new free radical species, which have not yet been fully (2) Reichstein, T.; Saudinger, H. A new or improved method of characterized. Evidence has been found for the formation of producing artificial coffee oil. Patent Appl. UK 260,960, Feb C-centered radicals derived from 2 using EPR spectroscopy and
spin-trapping techniques, but the presence of S-centered radicals (3) Tressl, R. Formation of flavor components in roasted coffee. In was neither confirmed nor excluded by these measurements.
Thermal Generation of Aromas; Parliment, T. H., McGorrin, R.
Ab initio molecular calculations, however, indicate similar J., Ho, C. T., Eds.; ACS Symposium Series 409; American stability levels for the mononuclear S- and C-centered radicals Chemical Society: Washington, DC, 1989; pp 285-301.
J. Agric. Food Chem., Vol. 50, No. 8, 2002 (4) Holscher, W.; Vitzthum, O. G.; Steinhart, H. Identification and (23) Fre´javille, C.; Karoui, H.; Tuccio, B.; Le Moigne, F.; Culcasi, sensorial evaluation of aroma-impact compounds in roasted M.; Pietri, S.; Lauricella, R.; Tordo, P. 5-(Diethoxyphosphoryl)- Colombian coffee. Cafe Cacao The 1990, 34, 205-212.
5-methyl-1-pyrroline N-oxide: A new efficient phosphorylated (5) Blank, I.; Sen., A.; Grosch, W. Aroma impact compounds of nitrone for the in vitro and in vivo spin trapping of oxygen- Arabica and Robusta coffee. Qualitative and quantitative inves- centered radicals. J. Med. Chem. 1995, 38, 258-265.
tigations. In ASIC, Proceedings of the 14th International (24) Liu, Y.; Liu, Z.; Chen, P.; Wu, L. Generation of radical cationss Scientific Colloquium on Coffee, San Francisco, CA, July 14- A facile generation of radical cations via the action of an 17, 1991; ASIC: Paris, France, 1991; pp 117-129.
oxoammonium trifluoroacetate. Sci. Sinica (B) 1988, 31, 1062-
(6) Blank, I.; Sen, A.; Grosch, W. Potent odorants of roasted powder and brew of Arabica coffee. Z. Lebensm. Unters. Forsch. 1992,
(25) McCormick, M. L.; Buettner, G. R.; Britigan, B. E. The spin trap R-(4-pyridyl-1-oxide)-N-tert-butylnitrone stimulates peroxi- (7) Schieberle, P.; Hofmann, T. Investigation of the influence of dase-mediated oxidation of deferoxamine. J. Biol. Chem. 1995,
manufacturing parameters on the aroma performance of aromaticcompounds in cysteine/carbohydrate reaction mixtures. Lebens- mittelchemie 1996, 50, 105-108 (in German); Chem. Abstr.
(26) Buettner, G. R. Spin trapping: ESR parameters of spin adducts.
1996, 125, 326725v.
Free Radical Biol. Med. 1987, 3, 259-303.
(8) Semmelroch, P.; Laskawy, G.; Blank, I.; Grosch, W. Determi- (27) Hensley, K.; Aksenova, M.; Carney, J. M.; Harris, M.; Butter- nation of potent odorants in roasted coffee by stable isotope field, D. A. Amyloid -peptide spin trapping II: Evidence for dilution assay. FlaVour Fragrance J. 1995, 10, 1-7.
decomposition of the PBN spin adduct. NeuroReport 1995, 6,
(9) Semmelroch, P.; Grosch, W. Studies on character impact odorants of coffee brews. J. Agric. Food Chem. 1996, 44, 537-543.
(28) Stucki, J. W., Goodman, B. A., Schwertmann, U., Eds. Iron in (10) Nagao, M.; Fujita, Y.; Wakabayashi, H.; Nukaya, T.; Kosuge, Soils and Clay Minerals; D. Reidel Publishing: Dordrecht, The T.; Sugimura, T. Mutagens in coffee and other beverages.
EnViron. Health Perspect. 1986, 67, 89-91.
(29) Box, H. C.; Budzinski, E. E. Electron spin resonance detection (11) Stadler, R. H.; Turesky, R. J.; Mu¨ller, O.; Markovic, J.; Leong- Morgenthaler, P.-M. The inhibitory effects of coffee on radical- Chem. Soc., Perkin Trans. 2 1976, 553-555.
mediated oxidation and mutagenicity. Mutat. Res. 1994, 308,
(30) Gilbert, B. C.; Hodgeman, D. K. C.; Norman, R. O. C. Electron spin resonance studies. Part XXXVIII. Evidence for the formation (12) Korycka-Dahl, N. B.; Richardson, T. Activated oxygen species and oxidation of food constituents. CRC Crit. ReV. Food Sci. oxidation of sulphides. J. Chem. Soc., Perkin Trans. 2 1973,
Nutr. 1978, 10, 209-241.
(13) Stadler, R. H.; Fay, L. B. Antioxidative reactions of caffeine: Formation of 8-oxocaffeine (1,3,7-trimethyluric acid) in coffee (31) Petersen, R. L.; Nelson, D. J.; Symons, M. C. R. Unstable subjected to oxidative stress. J. Agric. Food Chem. 1995, 43,
Intermediates. Part 179. Electron spin resonance studies of radicals formed in irradiated organic sulphides and disulphides.
(14) Grosch, W. Flavour of coffee. A review. Food 1998, 42, 344-
J. Chem. Soc., Perkin Trans. 2 1978, 225-231.
(32) Symons, M. C. R.; Janes, R. Radical cations of trialkylphosphine (15) Blank, I.; Pascual, E. C.; Fay, L. B.; Stadler, R. H.; Goodman, oxides, trialkylphosphates, hexamethylphosphoramide, dimethyl B. A.; Yeretzian, C. Degradation of furfuryl mercaptan in Fenton- sulphoxide and various sulphones, sulphites and sulphates. J. type model systems. In Caffeinated BeVerages. Health Benefits, Chem. Soc., Faraday Trans. 1 1987, 83, 383-399.
Physiological Effects, and Chemistry; Parliment, T. H., Ho, C.- (33) Swarts, S. G.; Becker, D.; DeBolt, S.; Sevilla, M. D. Electron T., Schieberle, P., Eds.; ACS Symposium Series 754; American spin resonance investigation of the structure and formation of Chemical Society: Washington, DC, 2000; pp 230-240.
sulfinyl radicals; reaction of peroxyl radicals with thiols. J. Phys. (16) Pascual, E. C.; Blank, I.; Goodman, B. A.; Yeretzian, C. The Chem. 1989, 93, 155-161.
detection and characterization of free radicals generated during (34) Franzi, R.; Geoffroy, M.; Reddy, M. V. V. S.; Weber, J.
the decomposition of solutions of the coffee flavour compound Theoretical and single-crystal ESR study of a (RSSH)•- species.
furfuryl mercaptan. In ASIC, Proceedings of the 18th Interna- J. Phys. Chem. 1987, 91, 3187-3190.
tional Scientific Colloquium on Coffee, Helsinki, Aug 2-6, 1999; (35) Gilbert, B. C.; Laue, H. A. H.; Norman, R. O. C.; Sealy, R. C.
ASIC: Paris, France, 1999; pp 50-57.
Electron spin resonance studies. Part XLVI. Oxidation of thiols (17) Sen, A.; Laskawy, G.; Schieberle, P.; Grosch, W. Quantitative and disulphides in aqueous solution: formation of RS•, RSO•, determination of -damascenone in foods using a stable isotope dilution assay. J. Agric. Food Chem. 1991, 39, 757-759.
RSO2 , RSSR-•, and carbon radicals. J. Chem. Soc., Perkin Trans. (18) Gautschi, F.; Winter, M.; Flament, Y.; Willhalm, B.; Stoll, M.
2 1974, 892-900.
New developments in coffee aroma research. J. Agric. Food (36) Chin, H.-W.; Lindsay, R. C. Ascorbate and transition-metal Chem. 1967, 15, 15-23.
mediation of methanethiol oxidation to dimethyl disulfide and (19) Stoll, M.; Winter, M.; Gautschi, F.; Flament, I.; Willhalm, B.
dimethyl trisulfide. Food Chem. 1994, 49, 387-392.
Research on aromas. XIII. Coffee aroma. HelV. Chim. Acta 1967,
(37) Penn, R. E.; Block, E.; Revelle, L. K. Methanesulfenic acid. J. 50, 628-694 (in French); Chem. Abstr. 1967, 66, 94855v.
Am. Chem. Soc. 1978, 100, 3622-3623.
(20) Tressl, R.; Silwar, R. Investigation of sulfur-containing compo- (38) Block, E.; O’Connor, J. The chemistry of alkyl thiosulfinate nents in roasted coffee. J. Agric. Food Chem. 1981, 29, 1078-
esters. VII. Mechanistic studies and synthetic applications. J. Am. Chem. Soc. 1974, 96, 3929-3944.
(21) Gasser, U.; Grosch, W. Aroma extract dilution analysis of commercial meat flavorings. Z. Lebensm. Unters. Forsch. 1990,
190, 511 (in German); Chem. Abstr. 1990, 113, 229885s.
Received for review October 5, 2001. Revised manuscript received
(22) Wink, D. A.; Nims, R. W.; Saavedra, J. E.; Utermahlen, W. E.; January 11, 2002. Accepted January 14, 2002. We are grateful to the
Ford, P. C. The Fenton oxidation mechanism: Reactivities of Scottish Executive Rural Affairs Department for funding for B.A.G.
biologically relevant substrates with two oxidizing intermediates and the EPR facilities.
differ from those predicted for the hydroxyl radical. Proc. Natl.
Acad. Sci. U.S.A. 1994, 91, 6604-6608.

Source: http://www.imreblank.ch/JAFC_2002_50_2356.pdf