J. Agric. Food Chem.

2001, 49, 5425−5432 5425

2-Methyl-3-furanthiol and Methional Are Possible Off-Flavors in

Stored Orange Juice: Aroma-Similarity, NIF/SNIF GC-O, and GC
Analyses
Yair Bezman,† Russell L. Rouseff,‡ and Michael Naim*,†
Institute of Biochemistry, Food Science and Nutrition, Faculty of Agricultural, Food and Environmental
Quality Sciences, The Hebrew University of Jerusalem, P.O. Box 12, Rehovot 76-100, Israel and Citrus
Research and Educational Center, University of Florida, Lake Alfred, Florida 33850

The occurrence of methional in fresh orange juice, and possible occurrence of â-damascenone in
heated orange juice, has been previously suggested. Here we report on the occurrence of 2-methyl-
3-furanthiol in the headspace, collected by solid-phase micro-extraction, of fresh, pasteurized, and
stored orange juice. The contents of 2-methyl-3-furanthiol and methional were quantified, and the
relative level of â-damascenone was estimated, in the headspace of fresh, pasteurized, and stored
orange juices using the nasal impact frequency (NIF) and surface of NIF (SNIF) GC-Olfactometry
procedure. 2-Methyl-3-furanthiol concentrations were 2 ng/L in fresh and pasteurized Shamuti orange
juice, and 270 ng/L in stored juice of the same variety. Methional concentrations were 550, 830,
and 11550 ng/L in fresh, pasteurized, and stored pasteurized juices, respectively. â-Damascenone
content appeared to have increased during pasteurization and storage. Aroma-similarity experiments
strongly suggest that 2-methyl-3-furanthiol and methional, at the levels found in stored orange
juice (21 days at 35 °C), contribute to stored orange juice off-flavor.

Keywords: 2-Methyl-3-furanthiol; methional; â-damascenone; orange juice; NIF; SNIF; GC-O;

off-flavor; Citrus sinensis

INTRODUCTION Preliminary GC-O time-intensity experiments and

The instability of fresh orange juice aroma during previous observations (7) suggested the presence of
processing (e.g., heat treatment) and subsequent storage unreported aroma active compounds that were formed
has been extensively studied (1-3). Decreased levels of during thermal processing and/or subsequent storage.
characteristic fresh orange juice aroma compounds on Because these compounds might contribute to the
one hand, and off-flavor formation on the other, lead to objectionable odor of stored orange juice, it would be
the distinct aroma differences between fresh and pro- useful to characterize, identify, and evaluate their flavor
cessed juice (4). 4-Hydroxy-2,5-dimethyl-3(2H)-furanone contribution. Advanced GC-MS, sulfur detectors, and
(Furaneol), 4-vinylguaiacol, and R-terpineol have been various GC-O techniques (Charm, ADEA, and Osme)
previously reported as storage-related off-flavors that can be used to identify aroma impact compounds in
may contribute to the typical overall aroma of stored various food systems (9-11). However, due to the
commercial orange juice (1, 3, 5, 6). Marin et al. (4) used extremely low concentrations of potent odorants, direct
a gas chromatography-olfactometry (GC-O) aroma identification in the headspace by means of physical
dilution technique and Charm analysis for studying methods such as GC-MS is sometimes difficult (8). The
aroma activity in citrus juices. They noticed that newly introduced nasal impact frequency (NIF) and
processing altered the aroma profile of the juices, as the surface of NIF (SNIF) GC-O procedures (12, 13) enable,
four most odor-intense peaks in the unprocessed juice under certain circumstances, use of the high sensitivity
were greatly diminished during processing. Rouseff and of the human nose for the detection and even quanti-
co-workers (7) used the same GC-O technique to fication of ppt levels of certain volatiles. NIF results are
identify volatiles in thermally abused orange juice. The expressed as percentage detection frequency of an odor
aromagram of juice stored at 40 °C for 4 months by panelists, and SNIF expresses the integrated time
revealed five intense aroma peaks characterized as course of the NIF response. In the current study, we
bread, sulfur, rotten, and floral-like (two peaks), none used the NIF/SNIF procedure to monitor the presence
of which were observed in similar aromagrams of juice of three potent odor compounds in the headspace of
stored at 4 °C for 2 weeks. Recently, Hinterholzer and fresh hand-squeezed, pasteurized, and stored Shamuti
Schieberle (8) analyzed fresh hand-squeezed Valencia and Valencia orange juices.
orange juice using the GC-O aroma extraction dilution
analysis (AEDA) procedure, and identified 42 volatiles MATERIALS AND METHODS
with dilution values greater than 4. Materials. Fresh oranges (Citrus sinensis (L.) Osbeck),
cultivars Valencia and Shamuti, from trees grown on the
* To whom correspondence should be addressed. Tel: 972- campus of The Hebrew University, in Rehovot, Israel, were
8-9489276. Fax: 972-8-9476189. E-mail: [email protected]. used. Fruits were picked in March 2000 (Valencia) and in
† The Hebrew University of Jerusalem. January 2001 (Shamuti). Fruits were then immediately pro-
‡ University of Florida. cessed as described below.

10.1021/jf010724+ CCC: $20.00 © 2001 American Chemical Society

Published on Web 10/31/2001
5426 J. Agric. Food Chem., Vol. 49, No. 11, 2001 Bezman et al.

Chemicals. 2-Methyl-3-furanthiol (95%) and 3-(methyl- when a marker is available, they can serve as an important
thio)propionaldehyde (methional; 96%) were obtained from indicator for identification of an unidentified volatile by
Aldrich (Milwaukee, WI), and â-damascenone (98%) was a gift matching the Kovats retention indices (RI). Following SPME,
of Dr. Gary Takeoka of the USDA (WRRC, Albany, CA). Also volatiles were desorbed in the injector (230 °C) for 3 min, in
purchased were HPLC-grade methanol (J. T. Baker, Deventer, the splitless mode. Operating conditions were as follows:
Holland), butylated hydroxytoluene (BHT) and gallic acid column held at 33 °C for 5 min (DB-5MS) or 30 °C for 4 min
(Sigma, St. Louis, MO), and citric acid and potassium citrate (DB-WAX) and then increased at 8 °C/min to 230 °C or 6 °C/
(Frutarom Co., Haifa, Israel). min to 220 °C, respectively, then held for 8 or 10 min,
External Standard Solutions. Stock solutions (10000 respectively. Helium was used as the carrier gas with a linear
ppm) were prepared as follows: 20 µL of methional or velocity of 31 cm/s. A 0.75-mm i.d. SPME injector liner was
â-damascenone was dissolved in 1980 µL of methanol (0.05% used. At the end of the capillary column, the effluent was split
w/w BHT); and 20 µL of 2-methyl-3-furanthiol was dissolved (1:10 by volume) between the FID and a sniffing port (model
in 1980 µL of acetone (0.05% w/w BHT). These solutions were ODP-2, Gerstel GmbH, Germany). The FID and the sniffing
then stored at -20 °C until further use. Standard mixtures port were held at 250 and 270 °C, respectively. Linear Kovats
for calibration were obtained by serially diluting the stocks to RI of the compounds were calculated using a series of n-
the final concentrations. Calibration mixtures and dilutions alkanes.
were performed with citrate buffer (0.062 M citric acid, 0.018 High-Resolution Gas Chromatography-Mass Spec-
M potassium citrate, and 3.0 mM gallic acid as antioxidant, trometry. 2-Methyl-3-furanthiol was identified in a Saturn
pH 3.5). The final calibration solutions (50 mL each) were 2000 mass spectrometer (Varian, Walnut Creek, CA) connected
divided into aliquots of 1.5 mL and placed in 2-mL vials to a Varian 3800 gas chromatograph equipped with a DB-5
(Kimble Glass, Inc., Vineland, NJ), then stored immediately capillary column (30 m × 0.25 mm i.d., 0.25 µm; J&W
at -80 °C until analysis. Concentrations (ng/L) of volatiles in Scientific). Following the aforementioned SPME procedure,
each calibration mixture were 0.5, 5, 50, and 250 (2-methyl- samples were inserted into the injector (held at 250 °C) for 3
3-furanthiol); 100, 2000, 8000, and 10000 (methional); 10, 100, min, in splitless mode. The oven program was 30 °C for 5 min,
2000, and 15000 (â-damascenone) for mixtures 1, 2, 3, and 4, increasing to 160 °C at 5 °C/min and holding for 10 min, then
respectively. increasing to 280 °C at 40 °C/min and holding for 5 min. The
transfer line was held at 300 °C and the source was held at
Juice Treatments for GC-O Analysis. Chilled oranges
170 °C. Mass spectra in the electron impact (EI) mode were
(4 °C) were washed and halved. Juice was then hand-squeezed
generated at 70 eV. Chromatograms and spectra were recorded
with a kitchen juicer. A portion of the combined juice was
with Saturn GC/MS Workstation software, version 5.41 (Vari-
divided into 1.5-mL aliquots, sealed in 2-mL vials, and
an).
immediately stored at -80 °C (designated fresh juice). The
For comparison with the SPME results, volatile compounds
remaining juice was subjected to pasteurization with a micro-
were analyzed with a purge and trap (Tekmar 3000, Tekmar-
wave (500 W) for 4 min. Preliminary trials indicated that a
Dohrmann Co., Cincinnati, OH) coupled with the above GC
temperature of 100 °C (from an initial 20 °C) was reached after
and mass spectrometer. Twenty mL of orange juice (Shamuti,
heating 200 mL of juice for 3 min. After immediate cooling on
stored for 21 days at 35 °C) was placed in a 20-mL tube.
ice, juice was supplemented with 500 ppm sodium benzoate.
Volatile compounds were purged for 15 min (ambient temper-
This laboratory procedure was aimed at simulating industrial
ature) and trapped on Tenax Trap A, following by 2 min of
conditions common in Israel (98 °C for 1 min). A portion of
desorption at 200 °C, then 1 min of cryofocusing at -145 °C
that juice (designated pasteurized) was divided into aliquots
(CRYOFocusing Module, Tekmar). Volatiles were then des-
as described for fresh juice and stored at -80 °C. The
orbed at 200 °C for 1 min and introduced into a GC column,
remaining pasteurized juice was sealed in a 120-mL amber
SPB-624, 1.4-mm film thickness, 60 m × 0.25 mm (Supelco,
glass bottle and stored for 21 days at 35 °C. This juice was
Supelco Park, Bellefonte, PA). The column was programmed
then divided into aliquots as described for fresh and pasteur-
at 40 °C (1 min), heated at 4 °C/min to 150 °C, then 10 °C/min
ized juices and stored at -80 °C.
to 200 °C (held for 10 min), and then 20 °C/min to 240 °C (held
Solid-Phase Micro-Extraction (SPME) of Juice Vola- for 10 min). Mass spectra were conducted as described above
tiles. Following sample thawing, fresh, pasteurized or stored- with the transfer line held at 200 °C and the source held at
pasteurized orange juice (1 mL) was mixed with 1 mL of 170 °C.
saturated CaCl2 solution in a 4-mL amber vial (15 × 45 mm) Sniffing Procedure. On the basis of the approach of
and sealed with a screw-cap Teflon septum (National Scientific Pollien et al. (12), 10 panelists (7 females and 3 males) for the
Company). The saturated CaCl2 solution was used to deacti- Valencia experiment and 12 panelists (10 females and 2 males)
vate enzymes (14) in fresh juice and also as a means of salting for the Shamuti experiment, all students 20-25 years of age,
out volatiles. Each vial was then placed in a water bath set to were randomly selected. Panelists were experienced with odor
36 ( 0.3 °C and gently mixed. Samples were allowed to recognition of various volatiles and with the sniffing of the
equilibrate for 5 min prior to SPME, and they were maintained GC-O apparatus. Sniffing of samples derived from fresh,
at 36 °C throughout the 20 min assay. SPME of juice volatiles pasteurized, or stored juices was recorded at ranges of Kovats
was conducted by inserting a 2-cm stable flex fiber coated with RI values from 755 to 969 and from 1211 to 1599. Preliminary
50/30 µm of DVB/Carboxen/PDMS (Supelco, Bellefonte, PA). GC-O screenings verified that methional and â-damascenone
Preliminary trials at different SPME time periods (not shown) are eluted during these time ranges. Elution of each aroma
indicated that a 15-min SPME at 36 °C gave optimal extraction compound through the sniffing port was recorded (HP-3398A
conditions for obtaining NIF/SNIF results of the orange juice GC software) by pressing a button during the entire sensation
headspace. Shorter time periods resulted in lower SNIF values, of an odorant, resulting in a square signal. Panelists were
whereas times longer than 15 min showed lower SNIF values requested to assign odor properties to each volatile detected.
for 2-methyl-3-furanthiol and methional and no change in The 10/12 individual chromatograms for each juice treatment
â-damascenone. The full length of the coated fiber was were summed, yielding the final aromagram (detection fre-
therefore exposed to the orange juice headspace for 15 min. quency versus RI). Peak heights were related to their percent-
The fiber was then removed from the headspace and im- age detection frequencies (NIF) and peak areas (expressing
mediately inserted into the GC injector. the detection frequencies and the detection time duration) were
High-Resolution GC-O. GC-O analyses were performed the SNIF.
with an HP-4890 (Hewlett-Packard) gas chromatograph In addition, GC-O time-intensity experiments were per-
equipped with FID and with the following fused silica col- formed with three trained panelists in the indicated RI ranges
umns: a 5% phenyl-coated DB-5MS capillary column (30 m for the aforementioned orange juice treatments and authentic
× 0.32 mm i.d., 0.5 µm) and a DB-WAX capillary column (60 markers.
m × 0.32 mm i.d., 0.25 µm) (J&W Scientific, Folsom, CA). The Quantification of 2-Methyl-3-furanthiol, Methional
properties of these columns are very different and, therefore, and â-Damascenone in Shamuti Orange Juice Head-
Analyses of Off-Flavors in Stored Orange Juice J. Agric. Food Chem., Vol. 49, No. 11, 2001 5427

Table 1. Kovats RI Values Resulting from FID Chromatograms and GC-O Aromagrams of Authentic Markers of
2-Methyl-3-furanthiol, Methional, and â-Damascenone, and of Valencia Orange Juice, Using DB-WAX and DB-5 Capillary
Columns
DB-WAX DB-5
markers juice markers juice odor
compound FID GC-Oa GC-Oa FID GC-Oa GC-Ob description
2-methyl-3-furanthiol 1358 1357 1357 875 875 877 meaty/vitamin B
methional 1431 1431 1430 917 917 918 cooked potato
â-damascenone 1887 1886 1886 1365 1368 1366 rose/tobacco
a GC-O responses by three trained panelists. b Average NIF GC-O responses of orange juice data presented in Figure 1 using 10

panelists.

Table 2. Odor Threshold and Content (ng/L) and log OAV Values of 2-Methyl-3-furanthiol and Methional in Fresh,
Pasteurized, and Stored Shamuti Oranges Juice
odorant thresholda fresh pasteurized stored
2-methyl-3-furanthiol 5 1.8(-0.45)b 1.9 (-0.44) 270 (1.73)
methional 50 550 (1.04) 830 (1.22) 11550 (2.36)
a Threshold values are retronasal and were taken from refs 19, 29, and 30. b Numbers in parentheses are log OAV values.

space. Calibration curves were obtained (13) employing the Inc., Cary, NC). Juice aromagrams were analyzed for signifi-
described SPME procedure. Each of the calibration mixtures cance between treatments using one- and two-way analysis
(1 mL) was applied using the same GC system and operating of variance (ANOVA). Because the smell data were derived
conditions. The same 12-member panel that took part in the from 12 panelists, Friedman and Wilcoxon nonparametric
Shamuti orange juice trial was used to determine the calibra- statistical tests were also performed (17) to further explore
tion curves. Panelists were asked to record their odor sensation statistical differences. Significance was set at least at the level
from RI 755 to 969 and from 1211 to 1599 in the very same of p < 0.05. Odor activity values (OAV) (odor concentration/
manner as for the orange juice. Each panelist sniffed each of odor threshold) were expressed after logarithmic transforma-
the calibration mixtures once (resulting in four aromagrams tion.
per panelist) containing the concentrations indicated above. A data matrix representing the aroma similarity results was
Data recorded from all 12 panelists were summed. obtained, in which each cell in the matrix represents the mean
Sensory Similarity Experiments. Shamuti orange juice
similarity for all panelists for the corresponding comparison.
samples were treated and extracted as detailed above. The
This proximity matrix was then analyzed by the clustering
resulting juice was then pasteurized in a microwave oven (1000
program ADDTREE (16), yielding a tree structure of branches
W) for 4 min (1 min at 100 °C from an initial 20 °C, 500-mL
and subdivisions with aromas located at the ends of the
aliquots) and immediately cooled on ice. The combined pas-
branches.
teurized juice (5200 mL) was supplemented with 500 ppm
sodium benzoate. A 1000-mL portion of this juice was placed
in two 550-mL amber glass bottles, sealed, and stored for 21 RESULTS AND DISCUSSION
days at 35 °C. The remaining pasteurized juice was im-
mediately transferred to -80 °C storage. Identification of 2-Methyl-3-furanthiol, Methion-
Aliquots (25 mL) of juice from stored and pasteurized (after al, and â-Damascenone in Orange Juice. The pres-
thawing) samples were placed in 120-mL amber glass bottles
with Teflon-lined caps (Wheaton, Millville, NJ.) Aroma- ence of methional (8) in fresh orange juice, and possible
similarity, cluster hierarchical tree structure (HTS) (15, 16) presence of â-damascenone (18) in heated orange juice,
analyses were then performed. Ten students, 20 to 25 years has been previously proposed. Because of the extremely
of age, were trained to evaluate the aroma similarity of five low concentrations of potent odorants, their direct
orange juice treatments. Bottles containing the various juice identification in the headspace by means of physical
samples were brought to room temperature prior to each methods such as GC-MS is sometimes difficult (8).
aroma test. During a 2-h aroma session, four 15-min aroma After SPME, FID peaks for 2-methyl-3-furanthiol, me-
similarity tests were conducted. In each test, on a verbal signal
thional, and â-damascenone could not be detected (e.g.,
from the experimenter, a panelist opened the bottles (after
shaking) of two samples at 15-s intervals, smelled them, and at the ppt/ppb levels). Thus, FID RI values of authentic
rated the similarity level of their aromas on a scale of 1 to 20 markers of the three compounds, together with RI
(1 for no similarity, 20 for identical, and numbers between for values obtained from the time-intensity GC-O analy-
other ratings). Seven sample pairs were presented in coded, ses including the NIF/SNIF procedure, were then used
randomized order per test, resulting in 28 pairs tested during to identify selected potent odorants in fresh, pasteur-
each aroma session. ized, and stored orange juice. Time-intensity GC-O
Data Processing. Summing NIF (%) per RI of all 12 experiments suggested putative 2-methyl-3-furanthiol,
individual aromagrams (EXCEL 97, Microsoft Co., Seattle,
methional, and â-damascenone peaks in the Kovats RI
WA) produced the combined aromagram for each juice treat-
ment. Probit values were calculated using the following ranges indicated in Material and Methods. The obtained
EXCEL functions: FID RI values for authentic markers in two different
capillary GC columns (DB-5 and DB-WAX) were actu-
probits of NIFs ) NORMSNIV (NIF %) + 5 ally identical with those of GC-O using trained and
untrained panelists, resulting in identical odor descrip-
probits of SNIFs ) ∑ [probit(NIF) dt] t
tions (Table 1). The level of 2-methyl-3-furanthiol found
in stored juice was marginal but sufficient (about 270
where dt ) 1 RI unit, from the start to the end time of the ppt, Table 2) to enable its identification, although not
GC-O peak, and (NIF)t ) the instant NIF value (%) at time enough for quantification by GC-MS as compared with
t. Calibration curves and confidence intervals were produced an authentic sample of 2-methyl-3-furanthiol (m/z: 114
with JMP statistical software version 3.1.6.2 (SAS Institute [M+], 85, 69, 45, 43). Purge and trap (Tenax as adsor-
5428 J. Agric. Food Chem., Vol. 49, No. 11, 2001 Bezman et al.

Figure 1. NIF and SNIF responses for selected RI values in the headspace of Valencia and Shamuti orange juices. Results are
derived from 10 panelists in the Valencia experiment and 12 panelists in the Shamuti experiment. Numbers at the top of each
peak represent SNIF values. Panels A and D, B and E, and C and F designate fresh, pasteurized, and stored juices, respectively.
Peaks at RI 877, 918, and 1366 are, putatively 2-methyl-3-furanthiol, methional, and â-damascenone, respectively. Different
superscript letters within each odor peak indicate significantly (at least p < 0.03) different SNIF responses between juice treatments.

bant) coupled to GC-MS did not result in any MS were 70, 70, and 50 for 2-methyl-3-furanthiol, me-
signals for the three compounds under the experimental thional, and â-damascenone-like odors, respectively.
conditions. To the best of our knowledge, the potent These results appear to indicate that the detection
odorant 2-methyl-3-furanthiol has not been previously frequency (NIF) due to juice treatment did not change
reported in citrus juice. Together, these results suggest for 2-methyl-3-furanthiol, increased for methional dur-
the identification of 2-methyl-3-furanthiol, methional, ing storage, and increased for â-damascenone during
and â-damascenone in the treated orange juice. pasteurization (remained the same for storage). Statisti-
In addition to the contribution of the NIF/SNIF cal analyses performed on the raw SNIF values (peak
procedure to the identification (as other GC-O proce- areas) indicate a significant increase only for methional
dures), the actual raw NIF/SNIF values may serve as following storage.
an indication of quantitative changes in odorant con- No quantitative determinations were performed with
tents. Furthermore, obtaining a calibration curve and the Valencia orange juice, but the relative SNIF values
achievement of linearization (transformation of SNIF (as indicated by the peak areas) essentially followed the
values into probits against log concentration) can lead NIF responses, supporting the increase in content of the
to quantitative results. In Figure 1 we provide the NIF/ â-damascenone-like odor during pasteurization and of
SNIF values for Valencia and Shamuti orange juices, the methional-like odor during storage.
whereas quantification was performed for var. Shamuti Quantification of 2-Methyl-3-furanthiol, Me-
only (Table 2). thional, and â-Damascenone in Shamuti Orange
Relative Levels of 2-Methyl-3-furanthiol, Me- Juice as Determined by GC-O Analysis. Although
thional, and â-Damascenone in Valencia Orange a different sniffing panel was used for the Shamuti var.,
Juice as Determined by GC-O Analysis. In the the NIF/SNIF results were qualitatively the same as
fresh juice, two out of 10 panelists (NIF ) 20) identified those for Valencia (Figure 1). This may support the
methional- and â-damascenone-like odors at Kovats RI reliability of the obtained results. In Shamuti, SNIF
values of 917 and 1368, respectively (Figure 1A), values for both 2-methyl-3-furanthiol and methional did
whereas eight out of 10 panelists (NIF ) 80) identified not change during pasteurization (compared with fresh
a 2-methyl-3-furanthiol-like odor at an RI value of 875. juice), but increased following storage (Figure 1D, E, and
After pasteurization, NIF values were 90, 10, and 50 F). SNIF values for â-damascenone increased signifi-
for 2-methyl-3-furanthiol, methional, and â-damascenone- cantly during pasteurization; the increase during stor-
like odors, respectively. Following storage, NIF values age was not statistically significant (Figure 1E and F).
Analyses of Off-Flavors in Stored Orange Juice J. Agric. Food Chem., Vol. 49, No. 11, 2001 5429

greater than the corresponding increase in the raw

SNIF values. Using the same computations, no clear
dependence was found between SNIF values and â-dam-
ascenone concentrations (correlation was less than 0.3).
The reasons for this lack of correlation are not known.
It is possible that the very high potency of â-dama-
scenone (a concentration of 10 ppt resulted in a 100%
NIF value using the GC-O) produced difficulties for the
sniffing panel in terms of indicating various signals
which could result in a normal calibration curve (e.g.,
for each increase in concentration, obtaining the ap-
propriate increase in SNIF value). Therefore, although
the increased formation of â-damascenone during pas-
teurization and storage was evident, a quantitative
determination of this odorant in each juice treatment
remains to be measured.
Results obtained from parametric (ANOVA) tests
applied to the raw SNIF data were in agreement with
the Friedman and Wilcoxon analyses performed to
verify the statistical significance between treatments.
These quantitative analyses indicated the content of
2-methyl-3-furanthiol in stored juice to be about 130-
fold higher (resulting in a log OAV of 1.73) than that in
fresh orange juice (Table 2). Pasteurization produced
no change in 2-methyl-3-furanthiol content. OAV values
for 2-methyl-3-furanthiol are based on a 5 ppt threshold
value (19). The use of this threshold value resulted in
Figure 2. Calibration curves for 2-methyl-3-furanthiol (A) and some negative OAVs, but one should note that the
methional (B) as expressed by probit transformations of SNIF
responses. Odorant concentrations are indicated in Materials threshold value for 2-methyl-3-furanthiol ranged from
and Methods section. The same 12 panelists participating in 0.5 to 7 ppt (20, 21). Under the same conditions,
the Shamuti experiment (Figure 1 D, E, and F) were used to pasteurization increased methional content by about
obtain the calibration curves. 50% (resulting in a log OAV of 1.22) compared with that
As discussed above,the SNIF values are derivatives in fresh juice, whereas storage induced a 21-fold in-
of the NIF responses but include the time during which crease (resulting in a log OAV of 2.36) in methional
panelists were able to detect a given odor, and therefore, content relative to that of fresh juice.
may be used for quantification (13). In general, the nose Sensory Similarity Experiments. The OAV values
cannot be considered an objective instrument for quan- indicated in Table 2 suggest that 2-methyl-3-furanthiol
tification. However, the “subjective” problem can be and methional are impact odorants in stored orange
overcome by using more subjects, as has been elegantly juice. However, a strong GC-O peak for a potent
shown previously (12). Unlike other GC-O procedures, odorant does not necessarily mean that this odorant
it is not the subject as an individual which is considered plays a significant role in the overall aroma quality of
the measurement, but the sniffing panel as a whole. The a given product (22). It is the composition of a mixture
use of the SNIF procedure for quantification lacks of odorants in the food system, rather than individual
empirical proof of an odor’s persistence being correlated odorants alone, that leads to specific sensations and
with its intensity. However, the present results (e.g., hedonics. The objective of this experiment was, there-
the increases in the concentrations of 2-methyl-3- fore, to assess the significance of 2-methyl-3-furanthiol
furanthiol and methional, which were highly correlated and methional as storage off-flavors in orange juice.
with SNIF responses, see below) clearly suggest high Pasteurized orange juice samples were fortified with
reliability of the SNIF experiments, at least under the 2-methyl-3-furanthiol and methional at the exact doses
present experimental conditions. With the current dif- found in stored orange juice (Table 2), and their aroma
ficulties in quantifying ppt levels of various odorants qualities were tested for their similarity to the aroma
by physical instruments, use of the SNIF procedure is of actual stored orange juice. In the aroma-similarity
a relevant choice. space diagram (Figure 3), the proportion of variance
The SNIF (peak areas) data were used to quantify explained was 99.2% (r2 ) 0.996), indicating that
the actual content of each compound. The same panel- distances in the clustering structure are highly cor-
ists who participated in obtaining the experimental data related with the original aroma similarity. Sets of
shown in Figure 1 (D, E, and F) were used to determine aromas located at the ends of branches and connected
a calibration curve for each of the three compounds to the same node at relatively short (horizontal) dis-
(Figure 2). A saturation curve was obtained. After the tances from each other are highly similar. Thus, there
probit data transformation vs log concentration, a linear are two criteria to evaluate similarity between treat-
dependence with high correlation (r2 ) 0.99 with ments. The first is whether such treatments are con-
relatively small 95% confidence intervals) between the nected to the same cluster and more specifically to the
SNIF values and the concentrations of either 2-methyl- same node: location on the same node indicates high
3-furanthiol or methional (Figure 2) was obtained. The similarity. The second criterion is the horizontal dis-
logarithmic transformations using probit results in a tance between such treatments. A short distance indi-
SNIF-odor concentration dependence in which the in- cates a high degree of similarity. For example, the
crease in relative content of each odorant is much relative distance value of 15.4 (arbitrary units) between
5430 J. Agric. Food Chem., Vol. 49, No. 11, 2001 Bezman et al.

Figure 3. Cluster analysis by ADDTREE of aroma-similarity values. Branches: 1, stored orange juice (35 °C for 21 days); 2,
pasteurized orange juice; 3, pasteurized juice fortified with 11.5 ppb methional; 4, pasteurized juice fortified with 0.27 ppb 2-methyl-
3-furanthiol; and 5, pasteurized juice fortified with both 11.5 ppb methional and 0.27 ppb 2-methyl-3-furanthiol. Each value in
the original matrix represents the mean of 10 panelists (each tested twice). Bold dots represent the location of each aroma in the
hierarchical tree structure. Numbers in parentheses are horizontal distances from branch 1 (stored orange juice).

branches 1 and 5 (Figure 3) is the sum of the dashed than the similarity rating given to the comparison
segment of branch 1 and the dashed segment of branch between the aromas of unfortified pasteurized juice and
5. The relative distance value of 22.3 between branches stored juice.
1 and 2 is the sum of the dashed segment of branch 1, These results suggest that pasteurized orange juice
the dashed segment of the main cluster (holding branches fortified with both 2-methyl-3-furanthiol and methional
1 and 5), and the dashed segment of branch 2. Three produces an aroma which is much more similar to that
major clusters were found in the space diagram de- of stored juice compared to that of other treatments, and
scribed in Figure 3, two of which were further divided thus, suggest that methional and 2-methyl-3-furanthiol
into two branches. Branch 1 (stored orange juice) and at doses found in stored orange juice are possible storage
branch 2 (pasteurized juice) were located in two differ- off-flavors. Among the previously reported storage off-
ent clusters and with the relatively longest horizontal flavors (4-vinylguaiacol, Furaneol, and R-terpineol) (3),
distance (22.3) from each other compared to other hedonic sensory experiments (1, 15) suggest that 4-vi-
aromas, indicating that these two aromas were the most nylguaiacol contributes significantly to the reduced
disparate. Fortification of pasteurized juice with me- acceptability of orange juice, whereas GC-O analyses
thional (branch 3) and 2-methyl-3-furanthiol (branch 4) suggested that R-terpineol and Furaneol may be less
resulted in a different cluster than either pasteurized significant (7, 23). Nevertheless, the aroma similarity
(branch 2) or stored (branch 1) juice. The distances of experiments suggest that, in addition to 2-methyl-3-
branches 3 and 4 from branch 1 were shorter (19.3 and furanthiol and methional, the previously proposed off-
20.8 for methional and 2-methyl-3-furanthiol, respec- flavors and possibly â-damascenone, as well as addi-
tively) than that between branches 1 and 2. This result tional unidentified compounds, should be added to the
suggests that fortification of pasteurized orange juice pasteurized juice to produce an aroma identical to that
with either 2-methyl-3-furanthiol or methional produces of stored orange juice.
aromas which are closer in similarity to the aroma of In summary, GC-MS analysis provided additional
stored juice than that of the unfortified pasteurized evidence for the occurrence of 2-methyl-3-furanthiol in
juice. Most importantly, fortification of pasteurized orange juice. The presence of methional (which was
orange juice with both 2-methyl-3-furanthiol and me- previously identified in fresh orange juice), and â-dama-
thional at doses found in stored juice produced an aroma scenone is also supported. Although these volatiles occur
(branch 5) which was located on the same cluster as that in fresh juice, their content increases significantly
of stored juice (branch 1) and with the shortest hori- during pasteurization and/or storage. The precursors for
zontal distance between them (15.4), indicating high these ppt and ppb levels of such volatiles in orange juice
similarity. remain to be investigated. The presence of methional
ANOVA of the raw data in the original matrix derived in fresh fruits such as tomatoes, grapefruits and oranges
from the results presented in Figure 3 was further already has been proposed (24, 25) and may involve
conducted to directly compare the similarity ratings unknown biosynthetic pathways. To the best of our
between stored juice (branch 1) and other treatments. knowledge, 2-methyl-3-furanthiol can be produced in
Here, higher rating values indicated higher similarity. fruit after fermentation (e.g., grapes (20)) but has not
These direct comparisons indicated a similarity rating been identified in fresh fruits. 2-Methyl-3-furanthiol can
of 7.8 ( 1.1 for the comparison between branch 1 and be produced by the hydrolysis of thiamine under acidic
branch 5, 5.4 ( 1.0 for the comparison between branch conditions (26). â-Damascenone may be released enzy-
4 and branch 1, 4.95 ( 1.1 for the comparison between matically from putative glycosides, as in apples (27).
branch 3 and branch 1, and 4.35 ( 0.6 for the compari- Evidently, these compounds are further produced during
son between branch 2 and branch 1. These comparisons pasteurization and storage by chemical means. Me-
indicated that the similarity rating given for the com- thional can be produced from methionine by Strecker
parison between the aromas of stored orange juice and degradation (28), whereas â-damascenone can be re-
pasteurized juice fortified with both 2-methyl-3-furanthi- leased by acid hydrolysis of glycosides (which is also
ol and methional was significantly (p < 0.002) higher expected to increase with heat). Thus, the acidic condi-
Analyses of Off-Flavors in Stored Orange Juice J. Agric. Food Chem., Vol. 49, No. 11, 2001 5431

tions inherent to citrus juice may favor both 2-methyl- (9) Acree, T. E.; Barnard, J.; Cunningham, D. G. A proce-
3-furanthiol and â-damascenone formation. On the other dure for the sensory analysis of gas chromatographic
hand, one may assume a lower likelihood of methional effluents. Food Chem. 1984, 14, 273-286.
formation via Strecker degradation of methionine under (10) Miranda-Lopez, M.; Libbey, L. M.; Watson, B. T.;
the relatively mild pasteurization and storage conditions McDaniel, M. R. Odor analysis of Pinot Noir wines from
of citrus products. Nevertheless, the minor (ppb and ppt) grapes of different maturities by a gas chromatography-
levels found in this study can probably be formed, even olfactometry technique (Osme). J. Food Sci. 1992, 57,
under such unfavorable conditions. At the level found, 985-993 and 1019.
only 2-methyl-3-furanthiol could be identified by GC- (11) Ulrich, F.; Grosch, W. Identification of most intense
MS under the experimental conditions. In fact, the NIF/ volatile flavor compounds formed during autoxidation
of linoleic acid. Z. Lebensm.-Unters. Forsch. 1987, 184,
SNIF procedure has been proposed to be more sensitive
277-282.
than MS/MS analysis (13).
(12) Pollien, P.; Ott, A.; Montigton, F.; Baumgartner, M.;
The present results strongly support the suggestion Munoz-Box, R.; Chaintreau, A. Hyphenated headspace-
that the NIF/SNIF procedure is an important tool for gas chromatography-sniffing technique: Screening of
both identification and quantitative determination of impact odorants and quantitative aromagram compari-
selected aroma compounds that are found in minor sons. J. Agric. Food Chem. 1997, 45, 2630-2637.
amounts in a food system, but which nevertheless play (13) Pollien, P.; Fay, L. B.; Baumgartner, M.; Chaintreau,
a role in its acceptance. However, quantification may A. First attempt of odorant quantitation using gas
not always be possible (e.g., â-damascenone in the chromatography-olfactometry. Anal. Chem. 1999, 71,
present study). Most significantly, the aroma-similarity 5391-5397.
experiments clearly indicate that 2-methyl-3-furanthiol (14) Buttery, R. G.; Teranishi, R.; Ling, L. C. Fresh tomato
and methional, at the levels found in stored orange juice, aroma volatiles: A quantitative study. J. Agric. Food
are very likely storage off-flavors. Chem. 1987, 35, 540-544.
(15) Naim, M.; Schutz, O.; Zehavi, U.; Rouseff, R. L.; Haleva-
Toledo, E. Effects of orange juice fortification with thiols
ACKNOWLEDGMENT on p-vinylguaiacol formation, ascorbic-acid degradation,
browning, and acceptance during pasteurization and
We thank Dr. J. Kroeze of Utrecht University for
storage under moderate conditions. J. Agric. Food Chem.
useful comments regarding application of the NIF/SNIF 1997, 45, 1861-1867.
procedure and Dr. I. Bilkis for helpful comments re- (16) Sattath, S.; Tversky, A. Additive similarity trees. Psy-
garding chemical identifications. We also thank Mr. Z. chometrika 1977, 42, 319-345.
Amoyal of the Elite Coffee Co. for conducting the GC- (17) O’Mahony, M. Sensory Evaluation of Food: Statistical
MS analyses, Dr. G. Takeoka of the USDA for providing Methods and Procedures; Marcel Dekker: New York,
selected odorants, and Ms. H. Zamir and Dr. I. Peri for 1986.
excellent technical assistance. The statistical advice by (18) Araki, C.; Ito, O.; Sakakibara, H. Changes of volatile
Dr. H. Voet is also acknowledged. flavor compounds in sweet orange juices by heating. J.
Japan. Soc. Food Sci. Technol. 1992, 39, 477-482.
LITERATURE CITED (19) Mottram, D. S.; Nobrega, I. C. C.; Dodson, A. T.
Extraction of thiol and disulfide aroma compounds from
(1) Naim, M.; Striem, B. J.; Kanner, J.; Peleg, H. Potential food systems. In Flavor Analysis: Developments in
of ferulic acid as a precursor to off-flavors in stored Isolation and Characterization; Mussinan, C. J., Morello,
orange juice. J. Food Sci. 1988, 53, 500-503 and 512. M. J., Eds.; ACS Symposium Series 705; American
(2) Rymal, K. S.; Wolford, R. W.; Ahmed, E. M.; Dennison, Chemical Society: Washington, DC, 1998; pp 78-84.
R. A. Changes in volatile flavor constituents of canned (20) Bouchilloux, P.; Darriet, P.; Dubourdieu, D. Identifica-
single-strength orange juice as influenced by storage tion of a very odoriferous thiol, 2-methyl-3-furanthiol,
temperature. Food Technol. 1968, 22, 1592-1595.
in wines. Vitis 1998, 37, 177-180.
(3) Tatum, J. H.; Nagy, S.; Berry, R. E. Degradation
products formed in canned single-strength orange juice (21) Schieberle, P.; Hofmann, T. Characterization of key
during storage. J. Food Sci. 1975, 40, 707-709. odorants in dry-heated cysteine-carbohydrate mix-
(4) Marin, A. B.; Acree, T. E.; Hotchkiss, J. H.; Nagy, S. tures: Comparison with aqueous reaction systems. In
Gas chromatography-olfactometry of orange juice to Flavor Analysis: Developments in Isolation and Char-
assess the effects of plastic polymers on aroma charac- acterization; Mussinan, C. J., Morello, M. J., Eds.; ACS
ter. J. Agric. Food Chem. 1992, 40, 650-654. Symposium Series 705; American Chemical Society:
(5) Haleva-Toledo, E.; Naim, M.; Zehavi, U.; Rouseff, R. L. Washington, DC, 1998; pp 320-330.
4-Hydroxy-2,5-dimethyl-3(2H)-furanone formation in (22) Mayer, F.; Czerny, M.; Grosch, W. Sensory study of the
buffers and model solutions of citrus juice. J. Agric. Food character impact aroma compounds of a coffee beverage.
Chem. 1997, 45, 1314-1319. Eur. Food Res. Technol. 2000, 211, 272-276.
(6) Moshonas, M. G.; Shaw, P. E. Quantitative analysis of (23) Bazemore, R.; Goodner, K.; Rouseff, R. Volatiles from
orange juice flavor volatiles by direct-injection gas unpasteurized and excessively heated orange juice
chromatography. J. Agric. Food Chem. 1987, 35, 161- analysed with solid-phase microextraction and GC-
165. Olfactometry. J. Food Sci. 1999, 64, 800-803.
(7) Naim, M.; Rouseff, R. L.; Zehavi, U.; Schutz, O.; Haleva-
(24) Buettner, A.; Schieberle, P. Characterization of the most
Toledo, E. Chemical and sensory analysis of off-flavors
odor-active volatiles in fresh, hand-squeezed juice of
in citrus products. In Flavor Analysis: Developments in
Isolation and Charcterization; Mussinan, C. J., Morello, grapefruit (Citrus paradisi Macfayden). J. Agric. Food
M. J., Eds.; ACS Symposium Series 705; American Chem. 1999, 47, 5189-5193.
Chemical Society: Washington, DC, 1998; pp 303-319. (25) Krumbein, A.; Auerswald, H. Characterization of aroma
(8) Hinterholzer, A.; Schieberle, P. Identification of the most volatiles in tomatoes by sensory analyses. Nahrung
odor-active volatiles in fresh, hand-extracted juice of 1998, 42, 395-399.
Valencia late oranges by odor dilution techniques. (26) Belitz, H. D.; Grosch, W. Food Chemistry, 2nd ed.;
Flavour Fragrance J. 1998, 13, 49-55. Springer: Berlin and Heidelberg, 1999.
5432 J. Agric. Food Chem., Vol. 49, No. 11, 2001 Bezman et al.

(27) Roberts, D. D.; Acree, T. E. Developments in the (30) Guth, H.; Grosch, W. Evaluation of important odorants
isolation and characterization of â-damascenone precur- in foods by dilution techniques. In Flavor Chemistry:
sors from apples. In Fruit Flavors: Biogenesis, Char- Thirty Years of Progress; Teranishi, R., Wick, E. L.,
acterization and Authentication; Rouseff, R. L., Leahy, Hornstein, I., Eds.; Kluwer Academic/Plenum Publish-
M. M., Eds.; ACS Symposium Series 596; American ers: New York, 1999; pp 377-386.
Chemical Society: Washington, DC, 1995; pp 190-199.
(28) Chan, F.; Reineccius, G. A. Kinetics of the formation of
methional, dimethyl disulfide, and 2-acetylthiophene via
the Maillard reaction. In Sulfur Compounds in Foods;
Mussinan, C. J., Keelan, M. E., Eds.; ACS Symposium Received for review June 4, 2001. Revised manuscript received
series 564; American Chemical Society: Washington, August 31, 2001. Accepted September 3, 2001. This work was
DC, 1994; pp 127-137. supported by grant US-2914-97 from BARD, the United
(29) Buttery, R. G. Quantitative and sensory aspects of flavor States-Israel Binational Agricultural Research and Develop-
of tomato and other vegetables and fruits. In Flavor ment Fund.
Science: Sensible Principles and Techniques; Acree, T.
E., Teranishi, R., Eds.; American Chemical Society:
Washington, DC, 1993; pp 259-286. JF010724+