Degradation of oligosaccharides in nonenzymatic browning by formation of α-dicarbonyl compounds via a peeling off mechanism

Article abstract of DOI:10.1021/jf9906127

The formation of α-dicarbonyl-containing substances and Amadori rearrangement products was studied in the glycine-catalyzed (Maillard reaction) and uncatalyzed thermal degradation of glucose, maltose, and maltotriose using o-phenylenediamine as trapping agent. Various degradation products, especially α-dicarbonyl compounds, are formed from carbohydrates with differing degrees of polymerization during nonenzymatic browning. The different Amadori rearrangement products, isomerization products, and α-dicarbonyls produced by the used carbohydrates were quantified throughout the observed reaction time, and the relevance of the different degradation pathways is discussed. In the Maillard reaction (MR) the amino-catalyzed rearrangement with subsequent elimination of water predominated, giving rise to hexosuloses with α-dicarbonyl structure, whereas under caramelization conditions more sugar fragments with an α-dicarbonyl moiety were formed. For the MR of oligosaccharides a mechanism is proposed in which 1,4-dideoxyosone is formed as the predominating α-dicarbonyl in the quasi-water-free thermolysis of di- and trisaccharides in the presence of glycine.

Full text of DOI:10.1021/jf9906127

J. Agric. Food Chem. 2000, 48, 6219−6226
6219
Degr a d a tion of Oligosa cch a r id es in Non en zym a tic Br ow n in g by
F or m a tion of r-Dica r bon yl Com p ou n d s via a “P eelin g Off”
Mech a n ism
Anke Hollnagel and Lothar W. Kroh*
Institut fu¨r Lebensmittelchemie, Technische Unversita¨t Berlin, Gustav-Meyer-Allee 25,
D-13355 Berlin, Germany
The formation of R-dicarbonyl-containing substances and Amadori rearrangement products was
studied in the glycine-catalyzed (Maillard reaction) and uncatalyzed thermal degradation of glucose,
maltose, and maltotriose using o-phenylenediamine as trapping agent. Various degradation products,
especially R-dicarbonyl compounds, are formed from carbohydrates with differing degrees of
polymerization during nonenzymatic browning. The different Amadori rearrangement products,
isomerization products, and R-dicarbonyls produced by the used carbohydrates were quantified
throughout the observed reaction time, and the relevance of the different degradation pathways is
discussed. In the Maillard reaction (MR) the amino-catalyzed rearrangement with subsequent
elimination of water predominated, giving rise to hexosuloses with R-dicarbonyl structure, whereas
under caramelization conditions more sugar fragments with an R-dicarbonyl moiety were formed.
For the MR of oligosaccharides a mechanism is proposed in which 1,4-dideoxyosone is formed as
the predominating R-dicarbonyl in the quasi-water-free thermolysis of di- and trisaccharides in the
presence of glycine.
Keyw or d s: Nonenzymatic browning; Maillard reaction; R-dicarbonyl compound; oligosaccharides;
peeling off mechanism
INTRODUCTION
formation of a glycosyl cation and subsequent formation
of either anhydrosugars and differently linked saccha-
rides or, in the presence of water, mainly mono- and
oligosaccharides. Parameters such as water content
obviously play an important role (Ha¨seler and Kroh,
1998) and have not been considered thoroughly in the
investigations of oligosaccharides until now.
Whether degradation of the oligosaccharides also
occurs via interaction of the amino compound with the
free glycosidic OH group is currently under investiga-
tion. Pischetsrieder et al. (1998) published the formation
of amino reductones in aqueous solution from disaccha-
rides with various amino compounds. It is therefore
reasonable to assume that reaction of the free glycosidic
OH group can initialize the breakdown of the sugar
molecule.
In the course of MR, compounds with R-dicarbonyl
structure are reported to play a key role in the initial
and advanced stages (Beck et al., 1988). Those R-dicar-
bonyls are very reactive, which explains their impor-
tance throughout the reaction, but also complicates their
quantification. Although they have been investigated in
mixtures containing monosaccharides (Nedvidek et al.,
1992), it is not known whether the different behavior
of oligosaccharides in MR is caused by differences in the
formation of R-dicarbonyls.
In our research we investigated the formation of
R-dicarbonyls, as nonvolatile key intermediates in cara-
melization and the Maillard reaction, throughout the
thermolysis of saccharides with different degrees of
polymerization (dp) using glycine in the case of the
Maillard reaction, and in the presence of the trapping
reagent o-phenylenediamine (OPD).
The Maillard reaction (MR) has been investigated
thoroughly for monosaccharides and amino acids in
aqueous solution (Ledl and Schleicher, 1990). However,
most real food systems contain oligosaccharides or even
polymeric saccharides; therefore, it should be of great
interest to study how those carbohydrates contribute to
the MR in the formation of flavor and color.
For mono- and disaccharides significant differences
are observed in the spectrum of volatile compounds
formed. Glucose undergoes reaction toward the dihydro-
γ-pyranone or hydroxymaltol, whereas maltose, caused
by the glycosidic substituent, favors pathways toward
maltol or heterocyclic compounds that still possess the
glycosidic substituent, such as 4-(glucopyranosyloxy)-
2-hydroxy-2-methyl-2H-pyran-3(6H)-one or 4-(gluco-
pyranosyloxy)-5-(hydroxymethyl)-2-methyl-3(2)-fura-
none (Pischetsrieder and Severin, 1994). Furthermore,
the position of the glycosidic link is reported to have
consequences for the reactivity of disaccharides in MR
(Kato et al., 1989).
Additionally it was demonstrated by a kinetic ap-
proach that di- and oligosaccharides show principally a
longer induction phase in the formation of brown color
(Wedzicha and Kedward, 1995). They react slower,
which suggests that oligomers first have to be converted
into more reactive, smaller intermediates.
Kroh et al. (1996) reported the breakdown of oligo-
and polysaccharides to nonvolatile reaction products by
* To whom correspondence should be addressed. Phone:
+49-30-31472584.Fax: +49-30-31472585.E-mail: lothar.kroh@
tu-berlin.de.
10.1021/jf9906127 CCC: $19.00 © 2000 American Chemical Society
Published on Web 10/28/2000

6220 J. Agric. Food Chem., Vol. 48, No. 12, 2000
Hollnagel and Kroh
EXPERIMENTAL PROCEDURES
120-5 C18 Macherey-Nagel; column, Nucleosil 5 C18 (250 × 4.6,
5 µm); detector, Kontron 440; flow, 1.0 mL/min; temperature,
30 °C; injection volume, 20 µL; eluent, methanol (solvent A),
water (solvent B) (both HPLC grade); detection wavelength,
320 nm; full scan, 190-440 nm, gradient, 0-5 min 5% A, 5-25
min 5% f 50% A, 25-30 min 50% f 100% A, 30-40 min 100%
A.
GLC/MS a n d GLC/F ID. Extraction and derivatization
were carried out according to Hollnagel and Kroh (1998).
GLC/ MS-Parameters. Analytical GLC was performed on
a Hewlett-Packard 5890 series II gas-liquid chromatograph
equipped with a Hewlett-Packard 5989B mass spectrometer
and a fused-silica capillary column, DB-5HT (J &W; i.d. 0.32
mm × 30 m, 0.1 µm film thickness) (carrier gas, He; detector/
injector temperature, 280 °C; temperature program, initial
temperature 120 °C, initial time 5 min isotherm, 120 f 200
°C, 10 °C/min, 5 min isotherm, 200 f 280 °C, 10 °C/min, 9
min isotherm). Column effluents were analyzed by electron-
ionization mass spectrometry; to verify the identity of the
analytes, they were compared with independently synthesized
standards. For EI spectra, the instrument was scanned from
m/z 40 amu to m/z 800 amu.
F u r osin e Deter m in a tion (Alltech Ned er la n d ). A 2 mL
reaction mixture was mixed with 6 mL of 10.6 M HCl in screw-
capped vessels, the vessel was flushed with nitrogen for 2 min
and then heated for 23 h at 110 °C. After heat treatment an
aliquot was centrifuged, and 1 mL of the supernatant was
purified by being passed through a Sep-Pack cartridge (acti-
vated with 5 mL of methanol and 10 mL of water); additional
elution with 4 mL of 3 M HCl ensured that all furosine was
washed from the cartridge.
HPLC Instrumentation. Pump, Spectraphysics P2000 Bi-
nary Gradient; column, furosine dedicated column (250 × 4.6,
5 µm, Alltech); detector, Spectraphysics UV2000 Dual UV/vis;
flow, 1.2 mL/min; temperature, room temperature; injection
volume, 20 µL; eluent, 0.4% acetic acid (v/v) (a), 0.3% KCl in
A (w/v) (B) (all HPLC grade); detection wavelength, 280 nm,
gradient, 0-11.5 min 100% A; 11.5-18 min 100% f 50% A,
18.5-21 min 50% A, 21-23 min 50% f 100% A, 23-32 min
100% A.
Th er m a l Tr ea tm en t. Caramelization. Mono-, di-, or tri-
saccharide and o-phenylenediamine (OPD), 1 mmol of glucose
(Glc) (Merck, water-free), maltose monohydrate (Mal) (Merck),
or maltotriose (Fluka, >93% HPLC), and 1 mmol of OPD
(recrystallized from water), are mixed thoroughly by use of a
vortex and heated to 480 min at 100 ( 1 °C in sealed tubes by
means of a thermoblock (behrotest ET 1, behr Labor Technik).
Maillard Reaction. Mono-, di-, or trisaccharide, glycine, and
OPD, 1 mmol of glucose, maltose monohydrate, or maltotriose,
1 mmol of glycine (Serva), and 1 mmol of OPD, are mixed
thoroughly and heated for up to 480 min at 100 ( 1 °C in
sealed tubes.
For determination of furosine, 1 mmol of glucose or maltose
was mixed thoroughly with 1 mmol of Boc-lysine and heated
to 90 min at 120 ( 1 °C in sealed tubes. The reaction was
carried out in distilled water; pH was not controlled.
After thermolysis samples were dissolved in 10 mL of a
mixture of methanol (Merck, suprasolv) and distilled water,
for furosine determination only in water, filtered, and diluted
if necessary. Experiments were carried out in duplicate; where
necessary, additional experiments were conducted to obtain
reliable results.
Syn th esis of Qu in oxa lin es. Synthesis of 2-methyl-3-(1,2,3-
trihydroxypropyl)quinoxaline was performed according to Beck
and Ledl (1988). The Amadori rearrangement product 1-deoxy-
1-piperidino-D-fructose was obtained according to Hodge and
Fisher (1963).
The fructosylpiperidine was mixed with twice as much OPD,
solved in phosphate buffer (pH 7, 10 mL), boiled under reflux,
and then concentrated. The resulting fine precipitate was
soluble under heating, and with slow cooling good crystalliza-
tion was achieved. Repeated recrystallization from acetone
gave a yellow-white precipitate. Its identity was proven by
various spectrometric measures. 1H NMR (300 MHz) in CD3OD
(Aldrich): δ 2.85 (s, 3H), 3.85 (dd, 1H), 3.93 (dd, 1H), 5.95 (d,
1H), 7.74 (m, 2H), 7.96 (m, 1H), 8.07 (m, 1H). Mass spectrum
of the acetylated substance: m/z 360 (0.2, M+), 301 (4), 300
(4), 259 (31), 258 (27), 241 (18), 174 (100), 157 (5), 107 (22)
188 (3), 103 (9), 43 (96).
HP TLC (Kr oh et a l., 1996). Monosaccharides. HPTLC
plates (Merck, Kieselgel 60, 20 × 10 cm) were used, the eluent
consisted of chloroform (Merck, HPLC grade), acetic acid
(Merck, 96%), methanol (Merck suprasolv), and water (60:18:
12.5:5, v/v/v/v), and the plates were developed twice.
Di- and Trisaccharides. HPTLC plates were developed twice
in an AMD chamber (CAMAG) using an eluent mixture of
chloroform, methanol, and water (50:40:8, v/v/v) in which 2.5-
2.8 mg of boric acid/98 mL of eluent mixture was dissolved.
Detection was performed in both systems with diphenylamine/
aniline/methanol/phosphoric acid (85%) (1 g/1 mL/100 mL/10
mL).
HP AEC/P AD (P u lsed Am p er om etr ic Detector ). Instru-
mentation. Pump, Dionex GP 40; column, 2 × PA-100 (250 ×
4 mm, Dionex); detector, PAD (Dionex); oven, HPLC column
oven 2155 (Pharmacia); temperature, 25 °C; injection volume,
20 µL; flow, 0.5 mL/min; eluent, 0.15 N NaOH (A), 1 N NaAc
in 0.15 N NaOH (B); gradient, 0-10 min 100% A, 10-60 min
0% f 60% B. Quantification was done by external calibration.
Synthesis of 2-(2,3,4-trihydroxybutyl)quinoxaline was carried
out according to Madson and Feather (1981). Via the forma-
tion of the bishydrazone of 3-deoxyhexosulose free 3-deoxy-
hexosulose was obtained.
To the crude reaction mixture was added an approximately
equimolar amount of OPD; further workup, similar to the one
described for the 2-methyl-3-(1,2,3-trihydroxypropyl)quinoxa-
line, again gave crystals which allowed structure analysis by
means of 1H NMR and GC/MS. 1H NMR (300 MHz) in
CD3OD: δ 3.13 (dd, 1H), 3.42 (dd, 1H), 3.60 (m, 1H), 3.66 (dd,
2H), 3.80 (dd, 1H), 7.77 (m, 2H), 8.03 (m, 2H), 8.83 (M, 1H).
Mass spectrum of the acetylated substance: m/z 360 (0.2, M+),
301 (7), 300 (9), 258 (12), 241 (14), 215 (14), 199 (33), 181 (24),
157 (34), 144 (100), 102 (11), 43 (89).
Synthesis of 2-(arabino-1,2,3,4-tetrahydroxybutyl)quinoxa-
line was performed according to Morita et al. (1981). 1H NMR
(300 MHz) in d6-DMSO (Aldrich): δ 3.34 (s, 4H), 3.66 (dd, 1H),
4.65 (m, 1H), 5.14 (d, 1H), 5.59 (d, 1H), 7.80 (m, 2H), 8.05 (m,
2H), 9.08 (s, 1H);. Mass spectrum of the acetylated sub-
stance: m/z 418 (0.2, M+), 358 (2), 317 (3), 316 (3), 299 (63),
257 (119, 215 (14), 202 (50), 196 (8), 171 (19), 160 (100), 143
(4), 129 (10), 115 (12) 102 (7), 43 (97).
RESULTS AND DISCUSSION
A lot of investigations have been conducted on the
volatile compounds that are formed in nonenzymatic
browning of the various carbohydrates; however, there
is evidence that also the nonvolatile fraction, especially
during the initial phase of MR, contains many reactive
intermediates, influencing the quality and extent of the
reaction, such as color, aroma, and antioxidative activ-
ity.
Synthesis of 2-methyl-3-(2,3-dihydroxypropyl)quinoxaline
was performed according to Morita et al. (1981). 1H NMR (300
MHz) in CD3OD: δ 2.77 (s, 3H), 3.17 (m, 2H), 3.65 (d, 2H),
4.30 (m, 1H), 7.70 (m, 2H), 7.94 (m, 2H). Mass spectrum of
the acetylated substance: m/z 302 (0.2, M+), 243 (16), 201 (42),
183 (100), 171 (40), 143 (7), 117 (9), 102 (9).
Syntheses of the Amadori rearrangement products glucosyl-
glycine and maltosylglycine were carried out according to Kroh
et al. (1992).
Prior to the investigation of the reaction mixtures, it
was confirmed that OPD inhibited the progress (brown-
ing) of the Maillard reaction. Although it is a diamine,
HP LC/DAD (Diod e Ar r a y Detector ). Instrumentation.
Degasser, Degasys DG-13000 (Knauer); pump, Shimadzu LC
10 AT; thermostat, Haake F3 (Fisons); guard column, Nucleosil

Degradation of Oligosaccharides in Browning
J. Agric. Food Chem., Vol. 48, No. 12, 2000 6221
F igu r e 1. Formation of R-dicarbonyl compounds detected as
quinoxalines in thermal treatment of Glc/OPD as a caramel-
ization-resembling system.
F igu r e 2. Formation of R-dicarbonyl compounds detected as
quinoxalines in thermal treatment of Glc/Gly/OPD as a Mail-
lard-resembling system.
it did not react disproportionately in the MR itself
(Hollnagel and Kroh, 1998). It was shown that there
are differences in the formation of sugar fragments with
R-dicarbonyl structure between mono- and disaccharides
as well as between aldoses and ketoses.
Another measure that was taken to reduce the influ-
ence of OPD on the results of the experiments was to
repeat the sets of experiments with and without glycine
so that the effect of the glycine could be measured
independently from that of OPD.
Although questioned in recent papers (Hofmann et
al., 1999), the trapping reagent was added in this
investigation directly to the reaction mixture so that all
formed R-dicarbonyls could be analyzed in their summed
up concentration, giving a clearer insight into their
relevance in the reaction mixture. Taking into account
the varying reactivity of the R-dicarbonyls as reported
by Glomb and Pfahler (1999), this seemed to be even
more important.
With six different reference substances, determination
of a wide range of R-dicarbonyl compounds quantita-
tively was determined. Besides quantification of R-di-
carbonyls by HPLC, the identity of the various quinox-
alines was confirmed by GLC/MS after derivatization.
Additionally the varying reaction mixtures were
characterized with regard to the degradation of the
starting carbohydrate, the formation of the Amadori
rearrangement products (ARPs) and isomerization prod-
ucts, and, if detected, subsequent formation of carbo-
hydrates by degradation of the starting compound.
Rea ction s of Mon osa cch a r id es. To determine the
differences in reaction between mono- and oligosaccha-
rides, the reaction behavior of glucose was investigated
thoroughly.
caramelization R-dicarbonyls are rather formed via
breakdown of the carbon chain (retro-aldolization),
whereas in the MR they are formed by amine-catalyzed
rearrangement and subsequent vinylogous â-elimina-
tion of water. This could also be the result of a lower
pH due to the addition of the amino acid. However,
measurement of the pH revealed that both reaction
mixtures had similar pH values, most probably due to
the OPD.
In MR systems the formation of methylglyoxal was
enhanced; it was formed quickly and after 60 min had
already reached a concentration of 3.6 mol %, and at
180 min was even 5.5 mol %. This phenomenon is
probably based on the additional reaction pathways via
cleavage of the Schiff base and 3-DH (Weenen, 1998;
Hayashi et al., 1986).
Moreover, it is interesting to see that with amino
catalysis besides 1-DH (1.3 mol % after 180 min) and
3-DH (4.1 mol % after 180 min) 1,4-DDH (5.8 mol %
after 180 min) was also formed in comparable amounts.
Nedvidek et al. (1992) attributed the formation of this
substance (1,4-DDH) to Strecker degradation and showed
that its formation was less with non-Strecker-active
amino acids.
Measurement of the pH value after the sample of the
MR mixture was dissolved in water indicated a stabile,
neutral to slightly acidic reaction medium, from 6.8 to
5.5 at the end of the reaction time, due to the pK of OPD.
This for the MR rather high pH suggests that 1-DH
should play an important role (Hodge, 1967). However,
in the investigated mixtures there was 3 times more
3-DH formed than 1-DH, which might be explained by
the high concentration of 1,4-DDH, as the subsequent
reaction product of 1-DH, as described by Nedvidek,
pointing out the essential role of the R-amino acid again.
Degradation of Glucose and Isomerization to Fructose.
Glucose was degraded very fast in the cases of both
caramelization and the MR (Figure 3).
Formation of R-Dicarbonyls. When the spectrum of
R-dicarbonyls formed in thermal treatment from Glc/
OPD and Glc/Gly/OPD was studied, there was clearly
an enhancing effect of glycine (Gly) observed.
In a caramelization-resembling system, glucose alone
produced only minor amounts of R-dicarbonyls. After
180 min methylglyoxal (≈2.5 mol %) was predominating
and less than 1 mol % each 3-deoxyhexosulose (3-DH),
1-deoxyhexosulose (1-DH), and 1,4-dideoxyhexosulose
(1,4-DDH) were formed (Figure 1).
In the MR, glycine induced the formation of R-dicar-
bonyls strongly. It was striking that in the presence of
Gly more deoxyhexosuloses with a complete carbon
chain (Σ 11 mol % at 180 min; see Figure 2) were formed
than sugar fragments with an R-dicarbonyl structure
(Σ 7 mol % at 180 min), which indicates that in
After 30 min more than 60 mol % glucose was
converted in both cases. The fact that, in contrast to MR
mixtures, in caramelization reaction mixtures 1,6-â-
anhydroglucose was detected by HPTLC underlines the
parallelism to caramelization of monosaccharides in the
absence of OPD as investigated earlier (Kroh et al.,
1996). In the MR no glucose could be detected after 120
min, whereas in caramelization there remained about
15 mol % starting material even after 180 min.
Under the chosen reaction conditions degradation of
glucose is fast; glycine even increased the reaction rate
of the degradation.

6222 J. Agric. Food Chem., Vol. 48, No. 12, 2000
Hollnagel and Kroh
Ta ble 1. F or m a tion of r-Dica r bon yls fr om Ma ltose
Detected a s Qu in oxa lin es, m ol %, in Ca r a m eliza tion
(ca r a ; Ma l/OP D) a n d in th e Ma illa r d r ea ction (MR;
Ma l/Gly/OP D)a
methyl-
glyoxal
1-DH
3-DH
1,4-DDH
glyoxal
time,
min cara MR cara MR cara MR cara MR cara MR
30 0.06 0.14 0.00 0.16 0.05 0.96 0.07 0.07 0.03 0.54
60 0.13 0.30 0.02 nr
0.15 4.62 0.14 0.14 0.10 1.29
120 0.08 1.43 0.20 0.96 0.54 12.34 0.31 0.31 0.33 2.06
180 0.15 1.84 0.26 1.44 1.06 17.71 0.42 0.42 0.56 2.70
240 0.17 1.89 0.42 1.35 1.44 18.10 0.56 0.96 0.80 2.94
a
nr ) not resolved.
F igu r e 3. Degradation of glucose in caramelization (Glc/OPD)
and in the MR (Glc/Gly/OPD).
Apparently the formation of 1,4-DDH is an important
pathway in the degradation of maltose in the course of
the Maillard reaction. We propose for its formation a
pathway similar to a “peeling off mechanism” (Figure
4). This reaction principle of degradation of the oligo-
or polysaccharide from the reducing end was published
first by Morita et al. (1983), who investigated the
degradation of differently linked homoglucans in alka-
line solution. In the case of the MR, we think that
glycine acts in the first instance as a catalyst in the
rearrangement reactions at the reducing end of an
R-glucan (1) via the Schiff base 2 toward the 2,3-enediol
structure 3. Either this enediol is now able to eliminate
the glycosidic residue at C4, leading to 1-amino-1,4-
dideoxyhexosulose (4) (Beck et al., 1988), or the amino
residue at C1 is eliminated first. Subsequent recovery
of the enediol structure 5 by Strecker degradation, for
example, or by reaction with oxidized reductones then
allows the elimination of the second residue, leading to
the malto-oligosaccharide 6 and the 1,4-DDH (7), which
was found in high amounts in the reaction mixture.
The likelihood of elimination at C4 or C1 could depend
on steric properties or on electronic conditions in the
molecule. Since 3 most probably does not contain
charged substituents when the pK values (2.34 and 9.6)
of glycine are considered, the steric properties should
influence the further reaction. For instance, on the
orientation of the π-orbitals which will form the double
bond, they should be parallel for smooth elimination.
Therefore, the carbon atoms where the vinylogous
â-elimination will occur should lie in one plane. By
molecular modeling using Alchemy II (Tripos Associates,
Inc.), it was revealed that C1 of the maltose-glycine
adduct lies perfectly in the plane whereas C4 lies 2.4°
out of the plane. However, since the molecule is pro-
posed to exist in the open chain form, only little energy
would be necessary to accomplish this requirement for
C4 as well. Furthermore, the probability of elimination
depends on the size of the substituents; the bigger the
substituent at C4 becomes (i.e., with increasing chain
length), the more elimination at C4 should be favored.
Additional proof for the importance of the elimination
of the glucosidic residue at C4 was the observation of
the formation of furosine. Furosine is formed by the
reaction of the ꢀ-amino group of lysine with saccharides
and possesses an unsubstituted furan ring (Figure 5).
Therefore, furosine should be the reaction product of
an elimination reaction at C4. When the formation of
furosine was compared in the reaction of glucose and
maltose with Boc-lysine at 120 °C in a quasi-water-free
system, there was a clear preference for furosine forma-
tion in the mixtures containing maltose. At a compa-
rable degree of browning (E420 ) 4.12), which was
reached by maltose/Boc-lysine under the chosen condi-
Considerable formation of fructose could only be
observed in reaction mixtures without glycine by means
of HPTLC.
The maximal concentration observed was 6.1 mol %
at 30 min, which decreased to 5 mol % after 1 h and
then further to 4 mol % after 120 min and reached 3.5
mol % after 180 min. It was earlier reported that
isomerization is an important reaction pathway in
caramelization (Kroh et al., 1996). Although it is likely
to occur in MR as well, in our case no fructose could be
detected, because either the concentration of glucose
decreased so fast that there was not enough starting
material available to produce considerable amounts of
fructose or the ketose was immediately converted.
Formation of Fructosylglycine. As fructosylglycine is
reported to be an important intermediate in the MR, it
was monitored as well by HPTLC. There was a maxi-
mum concentration observed; already after 30 min
about 8.9 mol % Amadori rearrangement product was
formed, and its concentration decreased to 7.3 mol %
after 60 min when glucose had been converted virtually
completely. After 120 min the concentration was as low
as 2.6 mol % and decreased more slowly now to 1.9 mol
% after 180 min. This result is consistent with the
observation that the formation of the Amadori re-
arrangement product occurs in the initial phase of the
MR and is readily converted further as described earlier.
Therefore, it can be concluded that under the chosen
conditions the typical reaction pathway for monosac-
charides in the MR leads via the Amadori rearrange-
ment product to the production of the R-dicarbonyls
3-DH, 1,4-DDH, methylglyoxal, and 1-DH.
Rea ction of Disa cch a r id es. Formation of R-Dicar-
bonyls. Under the same reaction conditions as for
glucose, maltose showed a lower reactivity; therefore,
the reaction time was prolonged to 240 min. In cara-
melization only small amounts of 1-DH and 3-DH or
sugar fragments with an R-dicarbonyl moiety were
detectable from the investigated dicarbonyl compounds.
Except 1,4-DDH, none was formed at more than 1 mol
%. 1,4-DDH was also formed (maximum 1.4 mol %, after
240 min of reaction time, Table 1).
As outlined later, the disaccharide was degraded
relatively fast; after 60 min most of the maltose was
degraded, indicating that formation of R-dicarbonyls
plays a minor role in carbohydrate degradation under
caramelization conditions.
When glycine was added, a dramatic increase in the
formation of 1,4-DDH was observed. It was formed at
more than 18.1 mol % at 240 min, whereas the concen-
tration of all other detected R-dicarbonyls, among which
methylglyoxal predominated (2.9 mol %, at 240 min),
remained negligible (Table 1).

Degradation of Oligosaccharides in Browning
J. Agric. Food Chem., Vol. 48, No. 12, 2000 6223
F igu r e 5. Formation of furosine from glycosylated lysine (R
) H for maltose or R ) Glc for maltotriose).
Ta ble 2. Degr a d a tion of Ma ltose a n d F or m a tion of
Ma ltu lose, m ol %, in Ca r a m eliza tion (ca r a ; Ma l/OP D) a n d
th e Ma illa r d Rea ction (MR; Ma l/Gly/OP D) a n d
Ma ltu losylglycin e in Ma l/Gly/OP Da
maltose
cara MR
100 100
maltulose
cara MR
time,
min
maltulosylglycine
MR
0
30
60
120
180
240
0
0
0
16.58
13.85
10.39
13.02
10.47
15.50
15.90
15.17
8.14
0.09
0.13
0.20
0.35
0.44
nd
nd
0.72
0.65
0.41
3.27
6.80
10.63
10.78
8.26
3.41
a
nd ) not detected.
reaction mixture is treated with very strong acid, even
the elimination of water at C3 should be favored,
forming 3-DH. There is no reason elimination at C4 and
subsequent formation of furosine should play a compa-
rable role in glucose degradation as in the case of
maltose.
Degradation of Maltose Formation of Glucose. Maltose
degradation is slower than that of Glc, both in caramel-
ization and in the MR. However, the majority was
readily converted in the first 60 min, but after 240 min
there still remained a small concentration unreacted.
The difference in degradation of the starting compound
between caramelization and the MR was rather small
(Table 2).
The identity of glucose was additionally confirmed by
HPLC/PAD. In caramelization not much more glucose
was detected than was originally present in maltose as
an impurity; apparently breakdown of the disaccharide
into monosaccharides is not important in quasi-water-
free caramelization (Figure 6). The earlier postulated
formation of a glycosyl cation might in this case pref-
erentially lead to the formation of anhydrosugars and
transglycosylation products (Ha¨seler and Kroh, 1998).
In the MR, however, glucose reached with 4.4 mol %
a relatively high concentration at 60 min and was
degraded to 1 mol % at 240 min. It is interesting to see
that the maximum concentration of glucose was ob-
served at a point where the 1,4-DDH concentration
began to increase strongly (Table 1); therefore, the
formation of glucose in the case of the MR can be seen
as additional evidence for the proposed peeling off
mechanism, yielding glucose as the second reaction
product when starting from maltose.
F igu r e 4. Proposed peeling off mechanism for the formation
of 1,4-DDH from oligosaccharide.
tions after 15 min, a furosine concentration of 6 mmol/L
was detected, whereas Glc/Boc-lysine gave an absorb-
ance of E420 ) 3.95 after 30 min, but only 0.2 mmol/L
furosine was produced. This pronounced difference of
more than a degree of magnitude is logical when one
considers that the precursor of furosine, the Amadori
rearrangement product, has, in the case of maltose, only
a limited number of degradation pathways. The forma-
tion of the 1-lysino-1,4-dideoxyosone should be favored
in order to eliminate the rather bulky glucosidic residue.
Cyclization and dehydration of 1-lysino-1,4-dideoxy-
osone easily gives rise to furosine. In the case of glucose
the elimination of water and subsequent formation of
an R-dicarbonyl should, from the steric point of view,
occur more or less equally at C1, C3, or C4. Since the
Isomerization to Maltulose. In all systems containing
maltose/OPD with or without glycine, formation of

6224 J. Agric. Food Chem., Vol. 48, No. 12, 2000
Hollnagel and Kroh
Ta ble 3. F or m a tion of r-Dica r bon yls Detected a s
Qu in oxa lin es, m ol %, in Ca r a m eliza tion (Ma ltotr iose/
OP D) a n d in th e MR (Ma ltotr iose/Gly/OP D)a
methyl-
glyoxal
1-DH
3-DH
1,4-DDH
glyoxal
time,
min cara MR cara MR cara MR cara MR cara MR
30 0.03 0.03 nd 0.00 0.06 0.09 0.07 0.13 0.15 0.32
60 0.07 0.08 nd 0.04 0.19 0.47 0.17 0.49 0.47 1.84
120 0.07 0.27 nd 0.08 0.65 1.57 0.47 1.05 0.96 3.44
180 0.10 0.46 nd 0.12 1.16 3.15 0.62 1.12 1.69 3.87
240 0.14 0.56 nd 0.21 1.57 4.02 0.91 nr
360 0.23 1.57 nd 0.35 2.80 9.76 1.47 nr
480 0.27 1.58 nd 0.37 3.91 11.72 1.67 nr
2.34 3.26
2.92 3.49
2.89 3.24
a
nr ) not resolved, nd ) not detected.
F igu r e 6. Formation of glucose from maltose in carameliza-
tion (Mal/OPD) and in the Maillard reaction (Mal/Gly/OPD).
maltulose was observed. However, the concentration of
maltulose was rather small. There was no significant
difference in the extent of isomerization observable; only
the course of the reaction appeared to differ between
caramelization and the MR (Table 2).
In the latter system, maltulose was detected only after
120 min and was degraded again readily, whereas in
caramelization a small but steady increase of maltulose
was seen. The observation that the maltulose concen-
tration was rising in caramelization throughout the
observed time range can be explained by the limited
number of other pathways for maltulose degradation
compared to the Maillard reaction. Moreover, more
starting compound, maltose, was available throughout
caramelization than in the MR. However, due to a
maximum concentration of maltulose of less than 1 mol
% in both systems, this reaction product was almost
negligible.
Formation of Maltulosylglycine. Maltulosylglycine
was formed somewhat slower than the Amadori re-
arrangement product from glucose; at 60 min when most
of the maltose had already been converted, the concen-
tration of maltulosylglycine was still rising (Table 2).
The highest concentration was formed at about 150 min
with ≈11 mol %; the maximum was not as sharp as in
the case of Glc/Gly/OPD, indicating that neither produc-
tion nor degradation of the product was as fast as in
the case of the monosaccharide. After 240 min the major
part of the Amadori rearrangement product (8.3 mol %)
had not degraded.
Summarizing the obtained results, it is evident that
disaccharides show a different behavior in the MR than
monosaccharides.
Most distinct is the enhanced formation of 1,4-DDH
from disaccharides, by the proposed peeling off mech-
anism, which is most probably caused by the glucosyl
substituent. Additionally it was shown that maltose is
degraded slower and that also the Amadori rearrange-
ment product is formed slower than in the case of the
monosaccharide.
F igu r e 7. Formation of maltose from maltotriose in caramel-
ization (maltotriose/OPD) and in the MR (maltotriose/Gly/
OPD).
by far the most abundant R-dicarbonyl detected, show-
ing clearly the differences from glucose as a monosac-
charide.
In caramelization the maximum concentration of 1,4-
DDH was nearly 4 mol % after 480 min. Further only
sugar fragments possessing R-dicarbonyl structure could
be quantified at that time in considerable amounts, such
as methylglyoxal with almost 3 mol % and glyoxal with
more than 1.5 mol % (Table 3).
In the presence of glycine a phenomenon similar to
that for maltose was observed. Formation of 1,4-DDH
was increased strongly, and even after 8 h, when nearly
12 mol % 1,4-DDH was formed, the final concentration
had not been reached. This could be due to various
reasons. First, maltotriose was degraded slower than
the aforementioned saccharides, lending itself longer to
the peeling off mechanism. Second, the second reaction
product of a peeling off reaction besides 1,4-DDH would
be maltose, which in turn could react again in terms of
a second peeling off step. Indeed the formation of
maltose was observed in MR mixtures; the strong
increase of 1,4-DDH between 240 and 360 min occurred
parallel to enhanced production of maltose. This finding
can be seen as further evidence for the proposed peeling
off mechanism (Figure 7).
Rea ction s of Tr isa cch a r id es. Formation of R-Di-
carbonyls. To confirm the validity of the hypothesis for
the degradation of oligosaccharides with a higher degree
of polymerization, the same set of experiments was
carried out with maltotriose. Looking at the formation
of R-dicarbonyls from the trisaccharide, the pathway
postulated for maltose seems to be repeated.
Formation of other R-dicarbonyl compounds was not
increased as much. Methylglyoxal was now formed
earlier, but remained at a concentration of slightly more
than 3 mol %, and also 1-DH was detected to some
extent but at a maximum of 1.6 mol % only played a
minor role compared to 1,4-DDH.
Production of R-dicarbonyls occurred even slower than
in reaction mixtures with maltose, and therefore samples
were taken over a time range of 8 h. As for maltose in
caramelization as well as in the MR, 1,4-DDH was again
Degradation of Maltotriose. J ust as for glucose and
for maltose, it was found that the major part of the
starting compound was degraded in the first hour.
Nevertheless, maltotriose remained longer in caramel-

Degradation of Oligosaccharides in Browning
J. Agric. Food Chem., Vol. 48, No. 12, 2000 6225
Ta ble 4. Degr a d a tion of Ma ltotr iose, m ol %, in Ca r a m eliza tion (Ma ltotr iose/OP D) a n d th e Ma illa r d Rea ction
(Ma ltotr iose/Gly/OP D)
time, min
maltotriose/OPD
maltotriose/Gly/OPD
0
100
100
30
36.31
25.06
60
30.74
20.77
120
28.28
20.16
180
32.35
18.24
240
31.35
16.44
360
29.65
17.38
480
25.82
9.80
ization and in MR mixtures than the other sugars. As
published elsewhere this is well in line with the finding
that saccharides become less reactive with increasing
degree of polymerization.
In the case of caramelization after 8 h there was still
more than 25 mol % present. In the MR at that time
only 9.8 mol % was present (Table 4). The addition of
equimolar amounts of glycine increased the degradation
markedly.
Degradation of maltotriose in caramelization differs
from that in the Maillard reaction, which is underlined
additionally by the fact that in caramelization carbo-
hydrates larger than maltotriose were detected on the
HPTLC plate in this system. They are probably formed
by transglycosylation.
Formation of Maltose and Maltotriulose. For the
formation of maltose a similar observation was made
as for glucose formation from maltose. In both caramel-
ization and the MR maltose was formed. In caramel-
ization a slow but steady increase of the maltose
concentration was observable. During the long reaction
time, splitting off of a glycosyl cation from the non-
reducing end of the molecule would lead to the formation
of maltose from maltotriose. After 8 h maltose was still
being formed, which is consistent with the finding that
there was still a considerable amount of maltotriose
present (Figure 7).
In the MR the concentration of maltose rose to 5.2
mol % at 360 min and then decreased again to 3.6 mol
% after 8 h. Several pathways could lead to its forma-
tion: besides by formation of a glucosyl cation in MR,
maltose might also be formed by the peeling off reaction
that occurs at a higher rate in the later phase of the
reaction. The decrease of maltose in the advanced stage
of the reaction can be explained by the slower formation
of maltose due to the decreasing concentration of
maltotriose, but also by the fact that maltose could
readily react again with glycine, in terms of a second
peeling off step, competing with the less reactive mal-
totriose.
Maltotriulose could only be assigned tentatively by
color and Rf value on the HPTLC plate, since there was
no reference material available. Comparing the absolute
values on one plate, in the cases of both the MR and
caramelization, the concentration of maltotriulose re-
mained at the initial value.
The Amadori rearrangement product could also only
be assigned by Rf value and color of the spot. Roughly,
one can say that the highest concentration of the
Amadori rearrangement product was reached after 30
min, and at 480 min maltotriulosylglycine could no
longer be detected. No maltulosylglycine or fructosyl-
glycine could be detected on the HPTLC plate.
F igu r e 8. Schematic reactions of malto-oligosaccharides in
caramelization and the Maillard reaction.
comparable to that of maltose after a shorter reaction
time. Degradation of the starting compound led to
R-dicarbonyls, isomerization products, smaller carbo-
hydrates, and, in the case of the reaction with glycine,
the Amadori rearrangement product as important in-
termediates in the Maillard reaction.
Formation of R-dicarbonyls most probably occurs by
reaction of the free glycosidic OH group, catalyzed by
an amino compound, with subsequent enolization and
elimination of water or by retro-aldolization. In contrast
to monosaccharides oligosaccharides generated 1,4-DDH
as the predominating R-dicarbonyl. Most probably this
pathway is favored due to the glycosidic substituent at
C4, which might block other reaction mechanisms. The
reaction via vinylogous â-elimination of the glycosidic
residue at C4 and subsequent formation of the 1,4-DDH
was strongly preferred. Elimination at C4 appears to
be the only pathway that leads to degradation of the
oligosaccharide from the reducing end.
Under MR conditions, formation of R-dicarbonyls from
malto-oligosaccharides is more or less limited to forma-
tion or 1,4-DDH (Figure 8). Other reactions, such as
dehydration and transglycosylation, lead to rather un-
reactive intermediates (Kroh et al., 1996) and therefore
may not be as relevant for the further reaction of malto-
oligosaccharides.
1,4-DDH is a reactive intermediate and a precursor
of various heterocyclic volatile compounds. Due to its
R-dicarbonyl moiety, it is likely to take part in conden-
sation reactions, and therefore might play a role in
formation of melanoidin.
Finally the R-dicarbonyls, and in the case of oligosac-
charides especially 1,4-DDH, influence the production
of flavor and color a great deal.
CONCLUSIONS
ACKNOWLEDGMENT
The reactivity of the di- and trisaccharides under
quasi-water-free conditions decreased in comparison to
that of glucose due to the increasing degree of poly-
merization. However, between the oligosaccharides no
qualitative difference was observed; when the reaction
time was prolonged, maltotriose had reacted to a degree
We express our sincere gratitude to Mrs. Anette
Bergha¨user for excellent technical assistance. We are
also indebted to Dr. Clemens Mu¨gge, Humboldt Uni-
versity Berlin, for NMR measurements. Further grati-
tude is due to Dr. Martinus I. J . S. van Boekel and

6226 J. Agric. Food Chem., Vol. 48, No. 12, 2000
Hollnagel and Kroh
Kroh, L. W.; Schro¨der, R.; Mu¨gge, C.; Westphal, G.; Baltes,
W. Thermisch induzierter Abbau von Disaccharid-Amadori-
Verbindungen in quasi wasserfreiem Reaktionsmilieu. Z.
Lebensm.-Unter.-Forsch. 1992, 194, 216-221.
Ledl, F.; Schleicher, E. Die Maillard-Reaktion in Lebensmitteln
und im menschlichen Ko¨rper- neue Ergebnisse zu Chemie,
Biochemie und Medizin. Angew. Chem. 1990, 102, 597-734.
Madson, M. A.; Feather, M. S. An improved preparation of
3-deoxy-D-erythro-hexos-2-ulose via Bis-(benzoyl)hydrazone
and some related constitutional studies. Carbohydr. Res.
1981, 94, 183-191.
Morita, N.; Hayashi, K.; Takagi, M.; Miyano, K. Gas-Liquid
Chromatography and Mass Spectrometry of Quinoxalines
Derived from Various Homoglucans by Alkaline o-Phenyl-
enediamine Method. Agric. Biol. Chem. 1983, 47, 757-763.
Morita, N.; Inoue, K.; Takagi M. Quinoxalines Derived from
D-Glucose and o-Phenylenediamine in a Weakly Acidic
Medium. Agric. Biol. Chem. 1981, 45, 2665-2668.
Nedvidek, W.; Ledl, F.; Fischer, P. Detection of 5-hydroxy-
methyl-2-methyl-3(2H)-furanone and of R-dicarbonyl com-
pounds in reaction mixtures of hexoses and pentoses with
different amines. Z. Lebensm.-Unters.-Forsch. 1992, 194,
222-228.
Carline Brands, Wageningen Agricultural University,
for determination of furosine.
LITERATURE CITED
Alltech Nederland, Breda. Determination and quantification
of Furosine in heat treated dairy products by means of
HPLC, extracted from Gazetta Ufficiale della Republica
Italiana (24-3-94).
Beck, J .; Ledl, F.; Severin, T. Formation of 1-deoxy-D-erythro-
2,3-hexodiulose from Amadori compounds. Carbohydr. Res.
1988, 177, 240-243.
Glomb, M. A.; Pfahler; C. Synthese eines bedeutenden Mail-
lard-Intermediats beim Abbau von Hexosen. Lebensmit-
telchemie 1999, 53, 16-17.
Ha¨seler, J .; Kroh, L. W. Nonvolatile reaction products of heat-
induces degradation of R-glucans; Part II: Analysis of short-
chain R-glucans by gas chromatography/mass spectrometry.
Z. Lebensm.-Unters.-Forsch. A 1998, 207, 44-49.
Hayashi, T.; Mase, S.; Namiki, M. Formation of Three-carbon
Sugar Fragment at an Early Stage of the Browning Reaction
of Sugar with Amines or Amino Acids. Agric. Biol. Chem.
1986, 50, 1959-1964.
Pischetsrieder, M.; Severin, T. Maillard Reaction of Maltose-
Isolation of 4-(Glucopyranosyloxy)-5-(hydroxymethyl)-2-
3(2H)-furane. J . Agric. Food Chem. 1994, 42, 890-892.
Pischetsrieder, M.; Schoetter, C.; Severin, T. Formation of an
Aminoreductone during the Maillard Reaction of Lactose
with NR-Acetyllysine or Proteins. J . Agric. Food Chem. 1998,
46, 928-931.
Wedzicha, B. L.; Kedward C. Kinetics of the oligosaccharides-
glycine-sulphite reaction: relationship to the browning of
oligosaccharide mixtures. Food Chem. 1995, 54, 397-402.
Weenen, H. Reactive intermediates and carbohydrate frag-
mentation in Maillard chemistry. Food Chem. 1998, 62,
393-401.
Hodge, J . E.; Fisher, B. E. Amadori Rearrangement Products.
In Methods in Carbohydrate Chemistry II; Whistler, W., Ed.;
Academic Press Inc.: New York, 1963.
Hodge, J . E. Origin of Flavour in Food. Nonenzymatic brown-
ing reactions. Chemistry and Physiology of Flavours; AVI:
Wersport, CT, 1967.
Hofmann, T.; Bors, W.; Stettmaier, K. Studies on Radical
Intermediates in the Early Stage of the Nonenzymatic
Browning of Carbohydrates and Amino Acids. J . Agric. Food
Chem. 1999, 47, 379-390.
Hollnagel, A.; Kroh, L. W. Formation of R-dicarbonyl fragments
from mono- and disaccharides under caramelisation and
Maillard reaction conditions. Z. Lebensm.-Unters.-Forsch.
A 1998, 207, 50-54.
Kato, Y.; Matsuda, T.; Kato, N.; Nakamura. Maillard Reaction
of Disaccharides with Protein: Suppressive Effect of Non-
reducing End Pyranoside Groups on Browning and Protein.
J . Agric. Food Chem. 1989, 37, 1077-1081.
Kroh, L. W.; J alyschko, W.; Ha¨seler, J . Non-volatile Reaction
Products by Heat-induced Degradation of R-Glucans. Part
I: Analysis of oligomeric Maltodextrins and Anhydrosugars.
Starch/ Sta¨rke 1996, 48, 426-433.
Received for review J une 8, 1999. Accepted J uly 28, 2000. This
study was carried out with financial support from the Com-
mission of the European Communities, Agriculture and Fish-
eries (FAIR) specific program, FAIR-CT96-1080, Optimization
of the Maillard reaction. A way to improve quality and safety
of thermally processed foods.
J F9906127