3640 J. Agric. Food Chem., Vol. 56, No. 10, 2008
Limacher et al.
the sample via the septum, extraction from the headspace at 35 °C for
10 min, adsorption onto Carboxen fiber (polydimethylsiloxan, PDMS,
film thickness 75 µm, Supelco, Bellefonte, PA), and finally desorption
for 2 min. After GC separation, the data were acquired by MS. The
chromatographic conditions on a ZebronWAX capillary column (30
m × 0.25 mm, df 0.5 µm, Phenomenex, Torrance, CA) have recently
been described (9).
The aim of this work was to study the formation of furan
and its methyl derivative from sugars and specific amino acids
in food and model systems simulating food process conditions
such as roasting and pressure cooking. Reliable quantitative data
were obtained by solid phase microextraction (SPME) in
conjunction with gas chromatography/mass spectrometry (GC/
MS) using [2H4]-furan as the internal standard. Mechanistic
insight into the degradation steps was achieved by using 13C-
labeled precursors.
Quantitative analyses were performed in the selective ion monitoring
(SIM) mode and repeated in the full scan mode, whereas the
experiments with the labeled precursors were only analyzed in the full
scan mode. The parameters of the MS have recently been described
(9). Quantification was based on MS signals at m/z 68 for furan, m/z
72 for d4-furan, and m/z 82 for methylfuran. The following qualifiers
were used: m/z 39 and 69 for furan, m/z 42 for d4-furan, and m/z 53 for
methylfuran. The quantities of furan and 2-methylfuran were determined
relative to the internal standard (d4-furan). The response factor (RF)
was determined taking into consideration the differences in response
of the mass detector between analyte and standard (based on the slope
of the calibration curve). For methylfuran, the response factor addition-
ally was influenced by a different recovery rate of analyte and standard
during SPME adsorption. A stronger adsorption of the less polar
methylfuran explains the relatively low RF value of 0.36 as compared
to the RF value of 0.97 for furan.
Calculation of Labeling Percentage. The percent labeling distribu-
tion of furan and methylfuran was determined by subtracting the
naturally occurring percentage of 13C. An additional data treatment was
required for methylfuran (9), and in contrast to furan, methylfuran
generated a [M - 1]+ signal ([C5H5O]+, m/z 81) through the loss of
H· with a relative intensity of 63% as compared to the molecule ion
signal (m/z 82). Therefore, the [M - 1]+ signal of methylfuran has to
be taken into account when correcting the overall signal intensity to
calculate the percent labeling distributions in labeling experiments
studying methylfuran formation. The obtained percentages after cor-
rection lower than 1% were set to 0% by definition.
EXPERIMENTAL PROCEDURES
Materials. The following chemicals are commercially available:
L-alanine (ALA, >99.5%), D-arabinose (ARA, 99%), citric acid
(monohydrate), disodium hydrogenphosphate (dihydrate, 99%), D-
erythrose (ERY, g90%), D-fructose (FRU, 98%), furan (stabilized,
99%), d4-furan (stabilized, 98%), 2-furaldehyde (99%), D-glucose
anhydrous (GLU, 98%), 2-methylfuran (MF, 99%), D-phenylalanine
(PHE, 98%), L-threonine (THR, 98%), white quartz sand, and silicone
oil (oil bath, from -50 to 200 °C) (Sigma-Aldrich, Buchs, Switzerland);
L-serine (SER, 99%) and methanol (for analysis) (Merck, Darmstadt,
Germany); 3-methylfuran (97%) (Acros Organics, Geel, Belgium);
3-deoxyglucosone (3DG, 95%) (Toronto Research Chemicals, North
York, Canada); [U-13C6]-D-glucose, [1,2-13C2]-D-glucose, [U-13C6]-D-
fructose, [1-13C]-D-fructose, [U-13C3]-L-alanine, [3-13C]-L-alanine, and
[U-13C4]-L-threonine (Cambridge Isotope Laboratories, Andover, MA);
and [13C1]-D-arabinose (>98%, Omicron Biochemicals, South Bend,
IN).
Sample Preparation. Roasting Model Systems. Equimolar amounts
of precursors (0.1 mmol each) and sea sand (1 ( 0.05 g) were mixed
in a headspace vial (20 mL, clear glass, 75.5 mm × 22.5 mm, 20 mm
crimp, rounded bottom, Brechbu¨hler, Geneva, Switzerland) used as a
reaction vessel, which was sealed with a crimp cap. The samples were
heated at 200 °C for 10 min, simulating roasting conditions. Experi-
ments were carried out in duplicate by using two reaction vessels, both
simultaneously immersed into a silicon oil bath. A homogeneous
temperature distribution was achieved by magnetic stirring. More
experimental details are given in ref 9.
Aqueous Model Systems. Equimolar amounts of precursors (1 mmol
each) were placed in a volumetric flask (10 mL) and dissolved in the
buffer solution (pH 4 or pH 7) by stirring. Citric acid-phosphate buffer
solutions were used for buffering the aqueous model systems (10). An
aliquot of the homogeneous solution (1 mL ) 0.1 mmol of each
precursor) was transferred into the reaction vessel (20 mL headspace
vial) and then sealed with a crimp cap. The vials were then homogenized
using a Vortex shaker (Vortex Genie 2, Verrerie Carouge, Switzerland)
for 30 s. The aqueous samples were heated at 121 °C for 25 min,
simulating sterilization conditions. More experimental details are given
in ref 9.
Food Products. Vegetable puree was prepared from pumpkin bought
in a local grocery store (Migros) by using a plastic squeezer (Moulinex,
Switzerland). The pH values were immediately measured. For spiking
experiments, defined amounts of precursors were added into a flask
(50 mL) containing the vegetable puree (10 g) and stirred to obtain a
homogeneous sample. An aliquot (2 g) was then placed into the reaction
vessel (20 mL headspace vial) and sealed. The spiked samples were
heated at 123 °C for 22 min, simulating sterilization conditions.
Quantification of Furan and Methylfuran. Isotope Dilution
Assay. After heat treatment, the samples were cooled in a water bath
(∼10 °C) for 5 min. The internal standard (2H4-furan ) d4-furan) was
added through the septum with a gastight syringe (10 µL, Sigma-
Aldrich), and the vial was vortexed. The added volume of internal
standard (ca. 0.8 µg of 2H4-furan/mL of methanol) was adjusted (1-25
µL) in a way that analyte/standard ratios were within the linear range
of the calibration curve. The samples were left at room temperature
for at least 30 min to achieve equilibrium prior to analysis.
Carbon Module Labeling (CAMOLA). Model samples were
prepared by mixing equimolar amounts of fully labeled and unlabeled
precursors (for the CAMOLA approach, see ref 12) (i.e., [U-13C6]-
glucose and glucose), followed by heat treatment (roasting, aqueous
conditions), extraction of furan and methylfuran by SPME as described
previously, and GC/MS analysis.
RESULTS AND DISCUSSION
Quantification of Furan and 2-Methylfuran in Model
Systems. The formation of furan and 2-methylfuran was studied
in model systems simulating roasting and pressure cooking
(sterilization) conditions. This study focused on Maillard-type
reactions based on sugars, sugar derivatives, and amino acids
as putative furan precursors. The amounts of furan and 2-me-
thylfuran were quantified by SPME-GC/MS to evaluate the
efficiency of various precursors systems. The results are
expressed in micromol of furan or 2-methylfuran per mol of
precursor, and the relative standard deviation (RSD) is given
in percent. 3-Methylfuran also was detected in various model
systems (baseline separated from 2-methylfuran), but the
quantities were very low, and therefore, these data are not
reported in this paper.
Roasting Conditions. As shown in Table 1, the amounts of
furan generated from sugars (nos. 1-4) under roasting condi-
tions were in the range of about 70-340 µmol/mol. Arabinose
(no. 3) was the most efficient precursor of furan (335 µmol/
mol), whereas erythrose was the least efficient (no. 4, 75 µmol/
mol). Addition of phenylalanine to glucose resulted in an
increase of about 50% (no. 8, 124 µmol/mol), whereas the furan
amount decreased by 20% in the binary mixture of fructose and
phenylalanine (no. 9, 100 µmol/mol). In contrast, the presence
of the amino acids alanine, threonine, and serine (nos. 5-7) all
resulted in higher furan amounts, in particular, with glucose.
SPME-GC/MS. The analytical method for furan quantification
published by Goldmann et al. (11) was used after validation with some
modifications as recently reported (9). In short, sample preparation
involved the addition of a deuterated analogue of furan (d4-furan) to