Thermal decomposition of caffeic acid in model systems: Identification of novel tetraoxygenated phenylindan isomers and their stability in aqueous solution

Article abstract of DOI:10.1021/jf950411c

Caffeic acid subjected to mild pyrolysis (225-226 °C) under vacuum resulted in rapid decarboxylation and the formation of simple catechol monomers as well as more complex cyclocondensed dimers and polymers. This reaction yielded the same spectrum of products as did acid-catalyzed cyclization of caffeic acid. The major pyrolysis products were identified by reversed-phase HPLC and LC-tandem mass spectrometry. Two novel compounds, identified by MS, ¹H NMR, and ¹³C NMR as 1,3-cis- and 1,3-trans-tetraoxygenated phenylindans, were present as major products in both the caffeic acid pyrolysate and the acid-treated sample. The stability and reactivity of the pyrolysis products in weakly buffered aqueous solutions were determined concomitantly by measuring hydrogen peroxide generation and by monitoring the concentration of the individual components by reversed-phase HPLC. Such model studies may provide information pertaining to reaction mechanisms and the nature of the compounds involved in hydrogen peroxide formation in coffee.

Full text of DOI:10.1021/jf950411c

898
J. Agric. Food Chem. 1996, 44, 898−905
Th er m a l Decom p osition of Ca ffeic Acid in Mod el System s:
Id en tifica tion of Novel Tetr a oxygen a ted P h en ylin d a n Isom er s a n d
Th eir Sta bility in Aqu eou s Solu tion
Richard H. Stadler,* Dieter H. Welti, Andreas A. Sta¨mpfli,† and Laurent B. Fay
Nestle´ Research Centre, Nestec Ltd., Vers-chez-les-Blanc, P.O. Box 44, CH-1000 Lausanne 26, Switzerland
Caffeic acid subjected to mild pyrolysis (225-226 °C) under vacuum resulted in rapid decarboxylation
and the formation of simple catechol monomers as well as more complex cyclocondensed dimers
and polymers. This reaction yielded the same spectrum of products as did acid-catalyzed cyclization
of caffeic acid. The major pyrolysis products were identified by reversed-phase HPLC and LC-
1
tandem mass spectrometry. Two novel compounds, identified by MS, H NMR, and 13C NMR as
1,3-cis- and 1,3-trans-tetraoxygenated phenylindans, were present as major products in both the
caffeic acid pyrolysate and the acid-treated sample. The stability and reactivity of the pyrolysis
products in weakly buffered aqueous solutions were determined concomitantly by measuring
hydrogen peroxide generation and by monitoring the concentration of the individual components
by reversed-phase HPLC. Such model studies may provide information pertaining to reaction
mechanisms and the nature of the compounds involved in hydrogen peroxide formation in coffee.
Keyw or d s: Pyrolysis; caffeic acid; hydrogen peroxide; phenylindan isomers; model system studies
INTRODUCTION
gen peroxide, whereas roasted beans have this ability
depending on the degree and duration of roasting (Tsuji
et al., 1991). The same authors also propose the
involvement of thermal decomposition products of caffeic
acid, in particular the substituted benzenediol p-vinyl-
catechol. However, no data were presented that could
confirm an active role of p-vinylpyrocatechol in hydrogen
peroxide production either in model pyrolysis systems
or in coffee itself. This dioxystyrene is apparently
formed in trace amounts by decarboxylation of caffeic
acid during pyrolysis or roasting (Clarke and MacRae,
1983; Heinrich and Baltes, 1987) and, as generally
known for o- and p-dihydroxybenzene moieties, can
produce hydrogen peroxide when exposed to atmo-
spheric conditions (Clapp et al., 1990). Furthermore,
this molecule is extremely susceptible to oxidation,
resulting in rapid polymerization reactions in solution
(Tiedke, 1936) and making isolation and identification
difficult.
Even though caffeic acid has been implicated indi-
rectly in the hydrogen peroxide generation process,
there is only limited information on the ability of caffeic
acid oxidation or pyrolysis products to produce hydrogen
peroxide in model systems. Different pathways and
products of chemically induced oxidation as well as
autoxidation of caffeic acid at ambient temperatures
have been described in the literature. The products of
such reactions are diverse and include cyclolignans
(Nahrstedt et al., 1990; Gumbinger et al., 1993), ben-
zodioxane, and naphthalene-1,2-dihydro- (Cilliers and
Singleton, 1991) and tetrahydrofuran-type compounds
(Fulcrand et al., 1994). On the other hand, subjection
of phenolic acids to thermal treatment results in rapid
decarboxylation to furnish substituted styrenes (Klaren
de Wit et al., 1971; Rizzi and Boekley, 1992) and simple
vinylpyrocatechol monomers (Fiddler et al., 1967; Tsuji
et al., 1991).
The formation of hydrogen peroxide in coffee solution
over time has been unequivocally demonstrated by
numerous investigators (Nagao et al., 1986a,b; Rinkus
and Taylor, 1990; Tsuji et al., 1991; Stadler et al., 1994).
Much emphasis has been placed on the relationship
between hydrogen peroxide generation in coffee and the
weak in vitro genotoxic effects that are observed in
bacterial and mammalian mutagenicity test systems
(Fujita et al., 1985; Itagaki et al., 1992). This contribu-
tion of hydrogen peroxide in coffee-mediated mutage-
nicity is undisputed, because in vitro mutagenicity can
be effectively abolished by addition of antioxidative
enzymes such as catalase or peroxidase (Nagao et al.,
1986a; Friederich et al., 1985; Itagaki et al., 1992).
Reports in this field have shown that dissolved oxygen
and water temperature during coffee preparation are
decisive factors in hydrogen peroxide formation. This
may explain the large fluctuations in the levels of
hydrogen peroxide reported in the literature (Fujita et
al., 1985; Rinkus and Taylor, 1990; Stadler et al., 1994).
The oxidative effects portrayed by coffee seem primarily
attributed to polyphenolics which adventitiously form
hydrogen peroxide when exposed to oxygen and metal
catalysts. However, coffee and certain polyphenolic
constituents can also act as potent antioxidants and
antimutagens as shown in in vitro assays (Stich et al.,
1982; Obana et al., 1986; Stich, 1991; Graf, 1992; Stadler
et al., 1994), and coffee as a whole has also been reported
to protect against various carcinogens in animal studies
(Abraham, 1989; Aeschbacher and J accaud, 1990).
However, no detailed work has been done to elucidate
the mechanisms and chemicals responsible for hydrogen
peroxide formation and the antioxidative effects dis-
played by coffee. A report by Tsuji and co-workers
showed that green coffee beans do not generate hydro-
The positive correlation between the degree of roast-
ing of the coffee beans and hydrogen peroxide formation
suggests that the pertinent reductants are formed
during roasting. To gain a better insight into the nature
* Author to whom correspondence should be ad-
dressed (fax +41/21 785 8553).
† Present address: Ciba-Geigy Ltd., K-127.5.02, 4002
Basel, Switzerland.
0021-8561/96/1444-0898$12.00/0
© 1996 American Chemical Society

Thermal Decomposition of Caffeic Acid
J. Agric. Food Chem., Vol. 44, No. 3, 1996 899
at a flow rate of 2.0 mL/min. The compounds were detected
photometrically eluting the trans isomer after tR ) 36 min and
the cis isomer after tR ) 37.1 min. The peaks were collected
manually and rechromatographed under the same HPLC
conditions to ensure complete separation of the isomers.
Thereafter, the fractions were concentrated in vacuo (35 °C)
to remove methanol. Water and ammonium acetate were
removed by lyophilization, affording 29 mg (0.11 mmol) the
1,3-trans (compound I) and 33 mg (0.12 mmol) of the 1,3-cis
isomer (compound II), with a total yield of both isomers of
3.6%. The light orange powders of both the cis and trans
isomers were then subjected to mass spectroscopic and NMR
analyses as described.
of the reducing components, one of the major catecholic
constituents of green coffee, caffeic acid, was subjected
to mild pyrolysis using model systems at temperatures
employed during the actual roasting of coffee beans.
Thus, the first objective of this work is to identify the
major thermal degradation products of caffeic acid.
Second, this study determines the contribution of the
major individual pyrolysis products to hydrogen perox-
ide formation in a model system and assesses to what
extent such model studies can reflect reaction mecha-
nisms in a complex mixture such as coffee.
Syn th esis of th e P h en ylin d a n Isom er s by Acid -Ca ta -
lyzed Cycliza tion . Caffeic acid (1 g, 5.55 mmol) was sus-
pended in 2 N sulfuric acid and refluxed for 2 h. The clear
yellow solution was then cooled to room temperature and
extracted into ether (3 × 30 mL). The organic layer was dried
(solid Na2SO4), filtered, and concentrated in vacuo (40 °C) to
afford 0.85 g of a gummy residue. Thin layer chromatography
on silica gel plates (Merck, 60254, 0.5 mm) with ether as
developing solvent yielded 93 mg (0.34 mmol) of the phenylin-
dan isomer mix (Rf ) 0.75, visualized by UV at 256 nm or
ammonia fumes). The 1,3-cis/trans isomers were purified by
semipreparative HPLC employing conditions identical to those
described in the previous section. Yields from 93 mg of the
isomeric mixture after HPLC were 22 mg (0.081 mmol) of 1,3-
trans isomer and 23 mg (0.085 mmol) of 1,3-cis isomer, giving
a total yield of 5.9% for both isomers together.
An a lytica l HP LC Con d ition s. Analysis of the thermal
decomposition products was carried out by HPLC employing
a Waters 625 LC chromatograph system equipped with a
Waters 994 photodiode array detector and a Waters 464 pulsed
electrochemical detector, with both the UV and the electro-
chemical detectors interfaced to a Berthold HPLC program
integrator. Samples were injected manually (50 µL). Separa-
tion conditions used a Supelco LC-18 DB column (4.6 × 250
mm, 5 µm) at a flow rate of 0.8 mL/min with a gradient going
from 100% solvent A (5% acetonitrile in water with a total of
0.1% trifluoroacetic acid) after 10 min to 20% solvent B (100%
acetonitrile with 0.1% trifluoroacetic acid) over 5 min, then to
40% B over 10 min, remaining at 40% B for 5 min, then to
50% B over 10 min, and finally to 100% B over 5 min.
Compounds were detected photometrically (photodiode array
detector range 200-400 nm) with wavelengths set at 220 and
280 nm. Electrochemical detection (ECD) was with a glassy
carbon electrode with a working potential of 0.9 eV relative to
the AgCl/KCl reference electrode. Pyrocatechol, 4-ethylpyro-
catechol, and the tetraoxygenated phenylindan isomers in the
pyrolysates were quantified by comparison of the peak inte-
gration area to a standard curve which was prepared by
injecting different known amounts of stock solutions of the
commercial standard compounds and the purified phenylin-
dans (1, 5, 10, 20, and 40 µg/mL) under the same HPLC
conditions as described above and integrating the UV signals
at 220 and 280 nm.
Hyd r ogen P er oxid e Mea su r em en ts. Hydrogen peroxide
was determined by the homovanillic acid/horseradish peroxi-
dase (HVA/HRPase) assay initially reported for the measure-
ment of activity of antioxidative enzymes (Guilbeault et al.,
1967; Snyder and Hendley, 1971). The dissolved pyrolysates
(1 mL aliquots) were acidified with hydrochloric acid (1 M, 50
µL) and subjected to solid phase extraction on C18 Chromabond
cartridges (500 mg, Macherey & Nagel, Du¨ren, Germany),
which were preconditioned with 2 bed-volumes each of metha-
nol and water and finally 10 mM hydrochloric acid. After
penetration of the pyrolysate solution, the columns were
washed with 1 mL of 10 mM hydrochloric acid and the pooled
effluents (2 mL total volume) analyzed for hydrogen peroxide
content by the HVA/HRPase assay as described (Stadler et al.,
1994). Measurement of hydrogen peroxide generation of the
individual peaks/regions isolated by semipreparative HPLC
was done by collecting the fractions manually (2 mL) and
removing excess acetonitrile under a stream of nitrogen at 35
°C for 15 min. Thereafter, a solution of boiling 50 mM
potassium phosphate buffer (pH 6.5) was added to each of the
EXPERIMENTAL PROCEDURES
Ch em ica ls. Chlorogenic acid and caffeic acid were pur-
chased from Fluka (Buchs, Switzerland). Pyrocatechol was
from Aldrich (Buchs, Switzerland). 4-Ethylpyrocatechol was
from Lancaster (Strasbourg, France); homovanillic acid (HVA),
horseradish peroxidase type II (HRPase), and quinic acid were
obtained from Sigma (Buchs, Switzerland). All other reagents
and solvents were of at least analytical grade purity.
P yr olysis P r oced u r es. Typically, 44.4 µmol of phenolic
acid was heated on a heating module (Techne Dri Block DB-
4) in a vacuum hydrolysis tube (5 mL volume) either evacuated
(0.02-0.05 mbar) or under atmospheric conditions. Prior to
pyrolysis under vacuum, water (30 µL) was added to create a
slurry that enabled vacuum application without loss of mate-
rial. Excess water was removed under vacuum at room
temperature and the dry residue placed in the thermostated
heater block preheated to 225-226 °C. After 15 min, the tube
was removed and opened after reaching ambient temperature.
The oily residue was taken up in 0.4 mL of a 50% methanolic
solution and transferred into a beaker to which was added 40
mL of boiling potassium phosphate buffer (10 mM, pH 6.5).
This solution (equivalent to 1.1 mM of caffeic acid as its
pyrolysate) was then divided into two fractions each of 20 mL
in separate beakers, to which was added MnCl2 (1 mg/mL
solution in water) to a give a final Mn2+ concentration of 2.2
µg/mL. The samples were then incubated at 35 °C in a
thermostated water bath. Aliquots were removed for hydrogen
peroxide determination (1 mL), HPLC, LC-MS, and LC-
tandem mass spectrometry analyses as described.
La r ge-Sca le P yr olysis of Ca ffeic Acid a n d Isola tion of
P h en ylin d a n Isom er s. Caffeic acid (6 g, 33.3 mmol) was
pyrolyzed in three consecutive 2 g lots under constant evacu-
ation (0.08-0.3 mbar) for a period of 30 min using a Bu¨chi
GKR vacuum oven set at 230 °C. After pyrolysis was complete,
the oily residue was left under vacuum to reach ambient
temperature and then dissolved in methanol (10-20 mL)
which was acidified with a few drops of dilute hydrochloric
acid. The pooled residues were dried in vacuo (35 °C),
affording 2.97 g of crude pyrolysate. Further purification was
achieved by dissolving the pyrolysate in 10 mL of ether and
loading the clear solution onto a silica gel column (80 g of silica
gel 60, 70-230 mesh, Merck, Darmstadt, Germany) packed
and preconditioned with ether. The column was eluted with
ether, and fractions (20 mL) were collected manually. The
phenylindan isomers were detected colorimetrically by remov-
ing aliquots (0.5 mL) of the ethereal effluent and adding one
drop of 1 N NaOH. A very characteristic deep violet color
developed due to phenylindan oxidation, which allowed simple
and rapid detection of the phenylindan-rich fractions. This
step afforded 0.4 g of the crude phenylindan isomer mixture
with a purity as determined by HPLC analysis of approxi-
mately 65%. Yield based on a stoichiometry of 2 mol of caffeic
acid affording 1 mol of phenylindan and 2 mol of CO2 was in
the range of 3-5.5% for both isomers together. A portion (150
mg, ca. 65% purity as determined by HPLC) of the crude
phenylindan 1,3-cis/trans isomer mix was subjected to semi-
preparative HPLC using a Supelco LC-18 DB semipreparative
column (10 × 250 mm) with solvent A (50 mM ammonium
acetate, pH 4.0 with acetic acid) and solvent B (100% metha-
nol), eluting isocratically with 45% solvent B for 20 min and
then going to 55% solvent B immediately for a further 20 min

900 J. Agric. Food Chem., Vol. 44, No. 3, 1996
Stadler et al.
Ta ble 1. Hyd r ogen P er oxid e F or m a tion a (Micr om ola r )
of P yr olyzed Ch lor ogen ic Acid sb over Tim e a n d Effect of
Ad d ition of Mn 2+
fractions to a final volume of 10 mL, to which was also added
MnCl2 to a final Mn2+ concentration of 2.2 µg/mL. Samples
were incubated at 35 °C for 1 h. Aliquots (1 mL) were then
loaded onto preconditioned C18 columns and the effluents
analyzed for hydrogen peroxide content as described above.
caffeic acid
chlorogenic acid
quinic acid
incubation without
time (min) Mn2+
with
Mn2+
without
Mn2+
with
Mn2+
without
Mn2+
with
Mn2+
Ma ss Sp ectr oscop y. Electron impact (EI) and chemical
ionization (CI) mass spectrometric measurements were carried
out using a Finnigan MAT-8430 double-focusing mass spec-
trometer. EI spectra were recorded at 70 eV from 20 to 800
Da and positive CI spectra at 150 eV from 60 to 800 Da using
ammonia as reagent gas. For both experiments the source
temperature was 180 °C. The samples were directly intro-
duced into the source of the mass spectrometer and heated at
2 °C/s. High-resolution measurements were made at a resolu-
tion power of 5000 using perfluorokerosene as reference
compound. Silylation of the cis and trans isomers of the
phenylindans was performed using a mixture of dimethylfor-
mamide/N,O-bis(trimethylsilyl)trifluoroacetamide (1:1 v/v)
heated at 40 °C for 30 min and subsequent EI mass spectros-
copy.
0
30
60
11 ( 1.3 15 ( 1.0 6 ( 1.2 16 ( 2.2 3.8 ( 0.8 11 ( 0.5
33 ( 2 104 ( 2 10 ( 3 24 ( 1 6 ( 0.5 15 ( 3
64 ( 2 205 ( 5 17 ( 1 41 ( 2 3.4 ( 0.3 18 ( 0.7
a
Hydrogen peroxide determined by the HVA/HRPase method
as described under Experimental Procedures. All incubations with
the addition of a final concentration of 2.2 µg/mL Mn2+. Entries
are means ( SD (n ) 4) of two independent experiments with
b
duplicate determinations. Pyrolysis of the acids under vacuum
at 225-226 °C for 15 min; incubation (35 °C) in potassium
phosphate buffer (10 mM, pH 6.5) at 35 °C. Detailed experimental
conditions as described under Experimental Procedures.
LC-MS a n d LC-MS/MS An a lyses. Experiments were
performed using a Finnigan TSQ-700 mass spectrometer
connected via a thermospray interface to a Waters 600-MS
pump and a Waters 717 autosampler. Separation was ac-
complished with a Macherey & Nagel C18 column (4.8 × 250
mm) operated at room temperature at a flow rate of 1 mL/
min, starting at 100% solvent A (5% acetonitrile with 0.1%
trifluoroacetic acid) for 10 min, then going to 20% solvent B
(90% acetonitrile, 0.1% trifluoroacetic acid) over 5 min, then
to 40% B over 10 min, remaining at 40% B for 5 min, then to
50% B over 5 min, and finally to 100% B over 5 min and resting
at 100% B for 10 min. The thermospray was set at a discharge
voltage of 1200 V, the vaporizer at 95 °C, and the source
temperature at 260 °C, scanning over the mass range 50-600
Da. The collision gas was argon at 1.9 mTorr and a collision
offset of 13.3 eV in the laboratory frame.
NMR An a lyses. All measurements were performed on a
Bruker AM 360 narrow bore spectrometer, using a 5 mm QNP
1
quadrinuclear probehead at a H frequency of 360.13 MHz and
a
13C frequency of 90.56 MHz or selective 5 mm 1H
a
probehead. One-dimensional 1H and proton-decoupled 13C
spectroscopy, 1H homodecoupling experiments, NOE difference
experiments, 13C DEPT, and 13C proton-coupled spectra with
NOE, two-dimensional homonuclear proton correlation (COSY),
1H-13C heteronuclear correlation (HETCOR) optimized for a
coupling of 145 Hz, and long-range heteronuclear correlation
optimized for 8 Hz coupling were done for both phenylindans.
F igu r e 1. HPLC analysis (UV detection at λ ) 280 nm) of
vacuum pyrolyzed (225-226 °C, 15 min) caffeic acid dissolved
in potassium phosphate buffer and incubated at 35 °C.
Aliquots (50 µL) were injected, and hydrogen peroxide forma-
tion (HVA/HRPase assay) was determined concomitantly in
the solution, revealing concentrations of 16, 350, and 480 µM
hydrogen peroxide after t0, t120, and t240 min, respectively.
Incubation and HPLC conditions are as described under
Experimental Procedures.
RESULTS
Hydr ogen P er oxide For m ation an d HP LC An aly-
sis of Ca ffeic Acid P yr olysa tes. The first step in the
analysis of the pyrolysates was to determine the hydro-
gen peroxide formation of a pyrolyzed caffeic acid
solution over time. For comparative purposes, equimo-
lar amounts of chlorogenic and quinic acid pyrolysis
solutions were prepared under identical conditions as
described (see Experimental Procedures) and the influ-
ence of Mn2+ on hydrogen peroxide formation deter-
mined. As depicted in Table 1, pyrolyzed caffeic acid
catalyzed the highest levels of hydrogen peroxide forma-
tion over time. The same concentration of nonpyrolyzed
caffeic acid incubated under the same conditions for 30
min produced 5 ( 0.8 and 9 ( 1.5 µM hydrogen peroxide
(n ) 4 ( SD) for caffeic acid solutions (does not fully
dissolve in hot buffer at pH 6.5) without and with
manganese addition (2.2 µg/mL), respectively. There-
fore, manganese promotes autoxidation of both pyro-
lyzed and nonpyrolyzed caffeic acid but shows a much
greater catalytic effect in the heat-treated phenolic
acids. Furthermore, the chlorogenic acid pyrolysate
fortified with manganese shows only about 25% of the
hydrogen peroxide generating capacity as compared to
equimolar amounts of the equally treated caffeic acid
pyrolysate. Levels of hydrogen peroxide in quinic acid
pyrolysates remained very low and did not increase over
a time period of 1 h, with only a very minor increase in
the same solution with added Mn2+
.
The caffeic acid pyrolysate was subsequently analyzed
by HPLC with dual UV and ECD without prior cleanup
procedures. As portrayed in Figure 1, the thermal
degradation products of caffeic acid which was pyrolyzed
under vacuum at 225-226 °C for 15 min showed four
prominent peaks in the HPLC trace (λ ) 280 nm), the
first two of which were identified as pyrocatechol (tR
)
15.4 min) and 4-ethylpyrocatechol (tR ) 26.1 min) by
co-injection of standard compounds and on-line UV
comparisons. The remaining two major components
eluting after tR ) 29.1 and ) 29.8 min were termed
compounds I and II, respectively. These showed identi-
cal UV spectra, characteristic of substituted catechols
(on-line λmax ) 286 nm) but with lack of the conjugated
side chain (no absorbance at λ ) 320 nm).

Thermal Decomposition of Caffeic Acid
J. Agric. Food Chem., Vol. 44, No. 3, 1996 901
Ta ble 2. Qu a n tita tion of th e Ma jor Ca ffeic Acid
P yr olysis P r od u ctsa a n d Hyd r ogen P er oxid e P r od u ction b
of th e P yr olysa te over Tim e
4-ethylpyro- pyro-
incubation
H2O2
catechol
catechol compounds
time (min) formation (µM)
(µM)
(µM) I and II (µM)
0
60
120
240
16
210
350
480
31
24
30
29
75
73
76
73
79
71
61
39
a
Pyrolysis under vacuum at 225-226 °C for 15 min. Pyrolysate
(obtained from 44.4 µmol of caffeic acid) dissolved in potassium
phosphate buffer and reaction products quantified by HPLC
analysis as described under Experimental Procedures. Hydrogen
b
peroxide determined by the HVA/HRPase method as described
under Experimental Procedures. All entries are averages of
duplicate determinations.
To elucidate the major compounds that could be
involved in oxygen reduction, hydrogen peroxide mea-
surements were performed after incubation times t0, t120
,
and t240 min, and the changes in the chemical composi-
tion of the pyrolysates were monitored simultaneously.
As depicted in Figure 1, the weakly buffered pyrolysate
products undergo prominent change in concentration
over time. However, the amount of residual caffeic acid
(tR ) 20.2 min) as well as pyrocatechol and 4-ethylpy-
rocatechol did not change significantly even after a 4 h
incubation period (Table 2), with the level of hydrogen
peroxide reaching up to 500 µM in the same pyrolysate
solution. On the other hand, the levels of compounds I
and II decreased rapidly over 4 h, with the concomitant
increase in UV absorbing material eluting after 40 min.
This polymeric material could not be identified further
due to instability (polymerization) during purification
attempts. The amount of hydrogen peroxide formed
after the specified incubation times shows that the
thermal reaction products eluting around tR ) 29-35
min contribute significantly to hydrogen peroxide for-
mation. Therefore, the aforementioned monomeric 1,2-
diphenols are not as susceptible as, for example, com-
pounds I and II to oxidation and do not contribute
significantly to hydrogen peroxide production in aqueous
caffeic acid pyrolysates.
Caffeic acid that was thermally treated under atmo-
spheric conditions at the same temperature and for the
same duration of time as the vacuum pyrolysates
showed the presence of the same degradation products
but in different ratios. Analysis by HPLC (Figure 2)
revealed higher levels of pyrocatechol and 4-ethylpyro-
catechol but lower levels of compounds I and II, which
were not easily distinguishable due to overlap of other
reaction products in this region of the HPLC trace. As
already observed (Figure 1), incubation of the atmo-
spheric pyrolysate at 35 °C resulted in the rapid
decomposition/polymerization of compounds I and II as
well as the other electrochemically active products
eluting in this region of the HPLC profile.
F igu r e 2. HPLC analysis (ECD) of caffeic acid pyrolyzed
under atmospheric conditions (225-226 °C, 15 min), dissolved
in potassium phosphate buffer and incubated at 35 °C.
Aliquots (50 µL) were injected, and hydrogen peroxide forma-
tion (HVA/HRPase assay) was determined concomitantly in
the solution, revealing concentrations of 88, 560, 725, and 900
µM hydrogen peroxide after t0, t60, t120, and t240 min, respec-
tively. Incubation and HPLC conditions are as described
under Experimental Procedures.
pyrolysates. Furthermore, both pyrocatechol and 4-eth-
ylpyrocatechol concentrations did not change signifi-
cantly over 4 h, in contrast to the hydrogen peroxide
level which reached up to 900 µM after the same time
period. This again corroborates that hydrogen peroxide
is not formed at the expense of monomeric benzenediols.
To assess the contribution of the individual pyrolysate
reaction products to hydrogen peroxide generation, the
crude pyrolysate obtained from 8.3 µmol of caffeic acid
was injected onto a semipreparative HPLC column and
all of the peaks/regions of the HPLC trace were collected
manually (detection at λ ) 280 nm). After removal of
acetonitrile, hydrogen peroxide generation was mea-
sured after a 1 h incubation period (35 °C) in the
individual fractions as described under Experimental
Procedures. As portrayed in Table 3, compounds I and
II contributed up to 56% of the total hydrogen peroxide
formed in all of the collected peaks/regions together. The
significantly lower contribution of pyrocatechol and
4-ethylpyrocatechol to hydrogen peroxide generation is
therefore also demonstrated in the isolated compounds.
LC-MS a n d LC-MS/MS An a lyses. The crude
pyrolysate of caffeic acid was subjected to LC-MS,
revealing the major pyrolysate products compounds I
and II with identical base peaks of 163 Da. In the upper
mass range there were no significant ions that could be
attributed to the compounds of interest. The LC-MS/
MS parent ion scan of the base peak ion at 163 Da gave
a single ion at 273 Da, tentatively assigned to [M + H]+
for both isomers, which were not well resolved under
the HPLC conditions employed and eluted together as
a relatively broad peak (see Experimental Procedures).
An ion trace scan at m/z 407 Da revealed a smaller peak
The apparent absence of polymeric material (which
elutes as a broad region at tR ) 40 min) in the HPLC-
ECD traces (Figure 2), which was clearly visible in the
HPLC UV profiles after 4 h of incubation (Figure 1), is
attributable to the lack of electrochemical activity of this
material at the working electrode potential (0.9 eV).
This polymeric material was clearly detectable at 280
nm and increased concomitantly with an increase in
hydrogen peroxide levels as already observed for the
vacuum pyrolysate. Comparison of the hydrogen per-
oxide generation revealed a 2-fold increase after the
same incubation period as compared to the vacuum

902 J. Agric. Food Chem., Vol. 44, No. 3, 1996
Stadler et al.
Ta ble 3. Hyd r ogen P er oxid e P r od u ction a of th e Ma jor
P yr olysis P r od u cts of Ca ffeic Acid Isola ted by
Sem ip r ep a r a tive HP LCb
HPLC
H2O2
compound/region
retention time (min)
production (µM)
pyrocatechol
region A
region B
4-ethylpyrocatechol
region C
compound I
compound II
region D
24.3
31.8
32.6
33.7
34-35.5
36
37.1
37.2-41
2.6
2.1
2.5
16
15
29
33
10
a
Hydrogen peroxide determined after an incubation period of
1 h at 35 °C in the presence of Mn2+ by the HVA/HRPase method
as described under Experimental Procedures. All entries are
b
averages of duplicate determinations. Pyrolysate (225-226 °C
for 15 min) derived from 8.32 µmol of caffeic acid. Peaks/regions
were isolated by semipreparative reversed-phase HPLC as de-
scribed under Experimental Procedures.
eluting ca. 1 min after compounds I and II that also
showed an ion fragment at m/z 273 Da. This may be
due to a more apolar higher molecular weight polymer
consisting of three caffeic acid units, i.e. the cyclocon-
densation product of three 4-ethylene-1,2-dihydroxy-
benzyl moieties.
Minor peaks attributable to 4-ethylpyrocatechol and
p-vinylpyrocatechol were also identified by their proto-
nated molecular ions [M + H]+ ) 139 and m/z ) 137,
respectively. Daughter ion scans of the molecular ions
of 4-ethylpyrocatechol gave m/z ) 123 and 113 and for
p-vinylpyrocatechol m/z ) 119 and 100.
F igu r e 3. EI-MS spectra of (A) compound I, purified by
semipreparative HPLC and (B) silylated compound I. Mass
spectral conditions are as described under Experimental
Procedures. Spectral characteristics of compound II were
identical in all respects to those of compound I and are
therefore not depicted.
Ch a r a cter iza tion of Com p ou n d s I a n d II. The
UV spectra of the purified compounds I and II were
identical in all respects, with an absorption maximum
at 285.5 nm in 10 mM HCl (log ꢀ ) 3.74), suggesting
that these compounds are stereoisomers with loss of the
caffeic acid conjugated side chain. Preliminary mass
spectral results of both compounds obtained by LC-MS/
MS of the crude pyrolysate revealed a base peak at m/z
) 163 which, after a parent ion scan, resulted in a single
ion at m/z 273. Identical mass spectral data were also
obtained for both compounds in the EI mode, with
molecular ions at m/z ) 272 (M+) and fragments at m/z
) 257 [M - CH3]+, 255 [M - OH]+, and 162 [M -
C6H5O2]+ (Figure 3A). MS-CI revealed an [M + NH4]+
ion at 290 for compounds I and II. High-resolution
Ta ble 4. 360.13 MHz 1H NMR Sp ectr a a of 1,3-cis-b a n d
1,3-tr a n s-Tetr a oxy-1-m eth yl-3-p h en ylin d a n s in CD3OD
1,3-cis
1,3-trans
proton ppm
multiplicity, J c
ppm
multiplicity, J c
1-H
3.002 m; J not extractable 3.176 6 lines, br; av J 6.6
2-HR
2-Hâ
3-H
1.428 d, t; 12.1, 10.4
2.576 d, t; 12.1, 7.0
3.925 d, d br; 10.5, 7.2
6.148 d; 1.0
2.040 d, d, d; 12.5, 8.1, 5.5
2.146 d, d, d; 12.5, 7.6, 6.0
4.122 d, d sl br; 8.0, 6.1
6.375 d; 0.9
4-H
mass spectrometry of the molecular ion gave M+
)
7-H
6.631 d; 1.1
6.635 d; 0.8
8-CH3 1.276 d; 6.7
1.188 d; 6.9
272.10507 amu and 272.10450 amu for compounds I and
II, respectively. These values are in accordance with
the theoretical value of 272.10486 amu and correspond
to the molecular formula C16H16O4. Silylation of both
2′-H
5′-H
6′-H
6.593 d; 2.1
6.691 d; 8.0
6.480 d, (d); 2.1, (?)
6.643 d; 8.0
6.531 d, d, (d); 8.0, 2.1, 0.4 6.434 d, d, (d); 8.0, 2.1, 0.5
a
Shift values in ppm, with TMS as an internal standard.
compounds I and II and subsequent EI-MS gave M+
)
Abbreviations: s ) singlet; d ) doublet; t ) triplet; m ) multiplet;
560 for both isomers, with major fragmentation ions at
m/z 545, 487, 471 [M - OSi(CH3)]+, 383, 307, 203, 147,
and 73 [Si(CH3)3]+, corroborating the presence of four
silyl groups and suggesting a tetraoxygenated dimeric
lignan derivative (Figure 3B).
b
av ) average; sl ) slightly; br ) broad. The cis compound could
not be obtained pure; it always contained some trans-phenylindan
(ca. 10%) and traces of a structurally related compound, possibly
the ring-opened dimeric precursor of the cis/trans phenylindans.
c Coupling constant(s) in Hz, absolute values, measured with at
least 0.12 Hz/point resolution and moderate resolution enhance-
ment. Coupling constants below 0.4 Hz are not listed.
Compounds I and II were procured by three different
approaches, and each of the isomers was subjected to
NMR investigation. The first approach was the initial
small-scale pyrolysis, and the purified isomers were
measured in DMSO-d6 (proton spectra only). Thorough
NMR analysis was done in the solvent CD3OD (better
solubility) on the products derived by sulfuric acid
catalyzed cyclization. Third, NMR measurements were
performed with the isomers obtained from the large-
scale pyrolysis in both solvents to confirm the identity
of the corresponding phenylindans. The 1H and 13C data
in CD3OD of compounds I and II obtained by acid-
catalyzed cyclization are given in Tables 4, 5, and 6,
respectively, the proton data correlating well with
earlier results published on nonoxygenated phenylin-
dans (Taylor et al., 1977). [NMR data of the cis/trans-
phenylindans in DMSO-d6 are available upon request
(D.H.W.).]
1
Common features of both compounds in the H NMR
spectra were five aromatic proton signals in a pattern

Thermal Decomposition of Caffeic Acid
J. Agric. Food Chem., Vol. 44, No. 3, 1996 903
Ta ble 5. 1H Hom on u clea r Cou p lin g Con sta n ts (of a t
Lea st 0.4 Hz) of 1,3-cis- a n d 1,3-tr a n s-
Tetr a oxy-1-m eth yl-3-p h en ylin d a n s in CD3OD
7-H and between 3-H and 4-H, respectively. The facts
that the proton shifts of 2-HR and 2-Hâ are more
different in compound II than in compound I and that
the coupling constants of both 2-H to either methine
proton are equal in compound II point to a cis config-
uration for this isomer. The most strategic of the NOEs
observed in CD3OD are illustrated by arrows in Figure
4. For compound I, they showed that the 8-CH3, 2-HR,
and 3-H hydrogens are all situated on the same side of
the planes spanned by the five-ring. In compound II,
the NOE from 8-CH3 to 2-HR is not observable because
of the immediate vicinity of the two signals. The NOE
of 3.1% from 8-CH3 to 2-Hâ indicates that in the motion-
averaged structure the 2-C atom must be above the
approximate plane formed by carbons 1, 7a, 3a, and 3,
bringing 2-Hâ upward. This is compatible with a 1-H,1-
C,2-C,2-HR torsion angle tending toward 180° and
should reduce steric interaction between the 8-CH3 and
the 3-phenyl substituents. In the trans isomer (com-
pound I), the proton shifts of 2-HR and 2-Hâ are more
equal and the coupling constants of 2-HR (or 2-Hâ) to
the two methine protons are different. Those coupling
constants seem to indicate a motion-averaged position
of the 2-C atom close to the said plane, as corroborated
by the absence of NOE from 8-CH3 to 2-Hâ.
scalar coupling
1,3-cis (Hz)
1,3-trans (Hz)
2J 2-HR,2-Hâ
3J 1-H,2-HR
3J 1-H,2-Hâ
3J 1-H,8-CH3
3J 2-HR,3-H
3J 2-Hâ,3-H
3J 5′-H,6′-H
4J 1-H,7-H
12.1
10.4
7.0
6.7
10.4
7.0
8.0
1.1
1.1
2.1
12.5
5.5
7.6
6.9
8.1
6.0
8.0
0.8
0.9
2.1
0.5
4J 3-H,4-H
4J 2′-H,6′-H
5J 3-H,5′-H
0.4
Ta ble 6. 90.56 MHz 13C NMR Sp ectr a a of 1,3-cis-b a n d
1,3-tr a n s-Tetr a oxy-1-m eth yl-3-p h en ylin d a n s in CD3OD
carbon
1,3-cis, (ppm, multipl)
1,3-trans (ppm, multipl)
1-CH
2-CH2
3-CH
39.03, d
48.45, t
50.98, d
38.96, d
46.78, t
50.04, d
3a-C
4-CH
5-C-OH
6-C-OH
7-CH
7a-C
8-CH3
1′-C
2′-CH
3′-C-OH
4′-C-OH
5′-CH
6′-CH
139.71, s
112.46, d
144.95, s
145.24, s
110.57, d
141.13, s
20.15, q
138.70, s
116.15, d
146.22, s
144.61, s
116.15, d
120.71, d
138.73, s
112.68, d
145.16, s
145.45, s
110.99, d
141.48, s
21.33, q
139.59, s
115.74, d
146.13, s
144.40, s
116.13, d
120.12, d
The assignment of the protonated carbons was
straightforward, while the quaternary carbons were
identified by their long-range proton couplings (mainly
three-bond) derived from long-range HETCOR experi-
ments. For comparison, the 13C spectrum of the phe-
nylindans I and II was calculated by the STN Specinfo
data base SPECAL prediction program (which does not
take into account the stereochemistry), and reasonable
correspondence with the measured spectra was observed
(mean deviations of 4.0 and 3.8 ppm for cis and trans,
respectively).
a
Spectrum with TMS as an internal shift standard. Abbrevia-
tions: s ) singlet (quaternary carbon); d ) doublet (CH); t ) triplet
b
(CH2); q ) quartet (CH3). As in Table 4.
1
The results of the NMR, 13C NMR, and MS experi-
ments confirm the structures of compounds I and II,
respectively, as 1,3-trans-5,6-dihydroxy-3-(3′,4′-dihy-
droxyphenyl)indan-1-methyl and 1,3-cis-5,6-dihydroxy-
3-(3′,4′-dihydroxyphenyl)indan-1-methyl, depicted in Fig-
ure 4.
Syn t h esis of Com p ou n d s I a n d II b y Acid -
Ca ta lyzed Cycliza tion . Caffeic acid refluxed in 2 N
sulfuric acid resulted in the formation of a very similar
spectrum of compounds as compared to the pyrolytic
procedure, revealing also compounds I and II as the
main products with identical retention times and on-
line UV spectra as their thermally generated counter-
parts. Both compounds I and II were present in ca.
equal ratios and were purified by semipreparative
HPLC to give overall yields in the region of 5-6% for
both isomers. The 1,3-cis and 1,3-trans stereochemistry
of the tricyclic dimers was corroborated by thorough
NMR analyses.
F igu r e 4. Structural formulas of compounds I and II showing
the stereochemically significant proton nuclear Overhauser
effects (in percent of the original signal intensity).
typical for two oxygen-bearing aromatic rings substi-
tuted at the 1-, 3-, and 4-, and the 1-, 2-, 4-, and
5-positions. Shifts and coupling topology of the aliphatic
resonances were compatible with a five-ring fused to a
benzene, with a second phenyl attached to one of the
carbons directly bound to the first aromatic ring. In the
small-scale pyrolytic products measured in DMSO-d6,
four slightly broadened phenolic OH signals could be
observed in the range of 8.4-8.8 ppm but not individu-
ally assigned (e.g., no COSY cross-peaks were seen for
compound I). At higher concentration these four signals
coalesced, and in CD3OD they are not observed, due to
exchange with the solvent. To clarify the NMR results,
the global stereochemistry of compounds I and II is
given in Figure 4. COSY cross-peaks and selective
homodecoupling proved the couplings between 1-H and
DISCUSSION
Catechols (1,2-benzenediols) such as caffeic acid readily
autoxidize in the presence of oxygen and transition
metal catalysts to form reactive oxygen species which
include the superoxide radical and hydrogen peroxide
(Marklund and Marklund, 1974; Hanham et al., 1983;
Inoue et al., 1992; Stadler et al., 1995). The rate of
formation of hydrogen peroxide is dependent on numer-
ous factors, for example, oxygen tension, pH, ambient
temperature, and transition metal availability. In
particular, Mn2+ has been shown to catalyze the au-
toxidation of polyphenols and thus increase hydrogen

904 J. Agric. Food Chem., Vol. 44, No. 3, 1996
Stadler et al.
peroxide formation (Tyson and Martell, 1972; Inoue et
al., 1992; Stadler et al., 1995), demonstrated here by
the significant increase in the level of hydrogen peroxide
in Mn2+-fortified pyrolysis solutions of caffeic and chlo-
rogenic acids.
The results presented here show that simple mono-
meric catechols such as pyrocatechol and 4-ethylpyro-
catechol are not the major contributors to hydrogen
peroxide formation in model pyrolysis systems, impli-
cating dimers and higher molecular weight oligomers
that are formed during cyclocondensation processes.
Pyrolysis of caffeic acid under atmospheric conditions
and under vacuum revealed the formation of novel 1,3-
cis- and 1,3-trans-tetraoxygenated phenylindan isomers,
which were progressively depleted over time with the
concomitant formation of hydrogen peroxide. The
amount of hydrogen peroxide formed in the pyrolysate
solution shows that the phenylindans contribute only
17% of the total amount of hydrogen peroxide formed.
This implies that the redox active oligomers/polymers
of caffeic acidseluting shortly after the phenylindans
in the HPLC tracesalso actively participate in the
reduction of molecular oxygen.
It should be noted that the contribution of the
phenylindans to hydrogen peroxide formation in the
pyrolysate mixture is based upon their rate of decom-
position, measured by analytical HPLC. However, the
relative rate of hydrogen peroxide production of the
isolated phenylindans, determined by measuring hy-
drogen peroxide directly in buffered solution, is much
higher than in the pyrolysate mix (56 vs 17%). Thus, if
redox recycling of the phenylindans occurs in the
pyrolysate mix, then calculation of hydrogen peroxide
evolution based solely on the loss of phenylindans will
not reflect the true contribution of these compounds to
the total amount of hydrogen peroxide formed.
It has been reported that in the presence of oxygen
the thermal decomposition of ferulic acid (245 °C) is
accelerated, leading to the incorporation of oxygen into
thermal decomposition products (Fiddler et al., 1967).
The atmospheric pyrolysates of caffeic acid showed more
products of oligomeric/polymeric nature as compared to
thermal treatment under anoxic conditions. Oxygen can
thus catalyze additional oxidative coupling reactions,
leading to both a broader spectrum of products and the
formation of more polymeric material.
Phenylindan-type compounds have been synthesized
from styrene (Taylor et al., 1977), propenylbenzenes
such as isohomogenol and isosafrol (MacMillan et al.,
1969), and the dioxystyrene derivative cinnamic acid
(Mu¨ller et al., 1951). However, in the case of the latter,
stereochemical characterization of the products was
performed simply on chemical grounds and, to our
knowledge, no such reactions have been reported for
caffeic acid. The cis/trans phenylindan stereoisomers
isolated from caffeic acid pyrolysates are thermal reac-
tion products of caffeic acid, the latter having a procliv-
ity to lose carbon dioxide readily due to tautomerization.
The mechanistic route can be envisaged as a cyclization
reaction of the reactive nucleophilic p-vinylpyrocatechol
with a cation (Figure 5), as described for styrene (Taylor
et al., 1977; Rizzi and Boekley, 1992). The intermediary
dimeric cation then eliminates a proton to cyclize
intramolecularily to form the tricyclic dimer with 1,3-
trans and 1,3-cis configuration in a ca. 1:1 ratio.
F igu r e 5. Proposed mechanism of formation of 1,3-cis- and
1,3-trans-phenylindans.
1965, 1966). These compounds are responsible for the
staining of the redwood and are structurally very similar
to the phenylindan isomers isolated from the caffeic acid
pyrolysates. Furthermore, the sesquirins have been
proposed to function as precursors of more complex
lignan structures (Balough and Anderson, 1966), also
observed here for the phenylindan isomers which un-
dergo rapid polymerization in the presence of atmo-
spheric oxygen and transition metal catalysts.
In food processing, model system reactions can give
concrete ideas of possible reaction mechanisms, inter-
mediates, and pathways. The incorporation of phenolic
moieties that were derived from thermally treated
ferulic acid into food macromolecules has been reported
(Rizzi and Boekley, 1992). Moreover, analytical HPLC
analysis with ECD of roasted and ground and instant
coffees has shown that the cis and trans tetraoxygenated
phenylindans described here are indeed present at
levels ranging from 10 to 15 ppm (unpublished results).
Even though such low levels of the phenylindans will
not contribute significantly to the overall production of
hydrogen peroxide in coffee, it is feasible that the
reactive intermediates of caffeic acid pyrolysis interact
with ambient nucleophilic functions such as sugars and
amino acids. Curie point pyrolysis experiments with
high molecular weight coffee melanoidins have already
revealed the presence of covalently bound phenolics of
phenylpropanoid origin (Heinrich and Baltes, 1987).
Such complex oligomers of phenolic nature could un-
dergo keto-enolization and redox cycling, leading to
typical univalent reduction of molecular oxygen and
formation of the superoxide anion radical. Dismutation
of superoxide would thus generate hydrogen peroxide,
a process typical for many di- and triphenolic com-
pounds (Marklund and Marklund, 1974; Kahl, 1991).
Therefore, in the case of coffee, hydrogen peroxide
production is probably attributable to a plethora of
diverse oligomeric and polymeric compounds, possibly
with integrated catecholic moieties that readily autoxi-
dize in the presence of oxygen and transition metal
catalysts.
Preliminary in vitro data on the biological activity of
the cis/trans phenylindans indicates that these tricyclic
dimers display potent antioxidative and antimutagenic
activity (Stadler, 1994; Guillot and Stadler, unpublished
results). Further work is now underway to elucidate
thermal degradation pathways of chlorogenic acids and
Naturally occurring phenylindan-type compounds
termed sesquirins have been isolated from the redwood
tree Sequoia sempervirens (Balough and Anderson,

Thermal Decomposition of Caffeic Acid
J. Agric. Food Chem., Vol. 44, No. 3, 1996 905
Kahl, R. Protective and adverse biological actions of phenolic
antioxidants. In Oxidative Stress: Oxidants and Antioxi-
dants; Sies, H., Ed.; Academic Press: London, 1991.
Klaren-de-Wit, M.; Frost, D. J .; Ward, J . P. Formation of
p-vinylguaiacol oligomers in the thermal decarboxylation of
ferulic acid. Recueil 1971, 90, 906-911.
MacMillan, J .; Martin, I. L.; Morris, D. J . Tricyclic dimers of
propenylphenyl ethers-I. NMR and stereochemistry. Tet-
rahedron 1969, 25, 905-914.
Marklund, S.; Marklund, G. Involvement of the superoxide
anion radical in the autoxidation of pyrogallol and a
convenient assay for superoxide dismutase. Eur. J . Bio-
chem. 1974, 47, 469-474.
the interaction of pyrolysis products with other con-
stituents in foods, as well as the potentially beneficial
antioxidative effects of reaction intermediates and
products.
ACKNOWLEDGMENT
We thank Ms. J . Richoz for excellent technical as-
sistance and Mrs. F. Arce Vera for assistance with NMR
spectroscopy.
LITERATURE CITED
Mu¨ller, A.; Toldy, L.; Halmi, G.; Meszaros, M. Dimeric prope-
nyl phenol ethers. XII. The synthetic stereoisomer of
diisohomogenol, diisoeugenol diethylether, and metanethole.
J . Org. Chem. 1951, 16, 481-491.
Nagao, M.; Wakabayashi, K.; Fujita, Y.; Tahira, T.; Ochiai,
M.; Sugimura, T. Mutagenic compounds in soy sauce,
chinese cabbage, coffee and herbal teas. Prog. Clin. Biol.
Res. 1986a , 206, 55-62.
Nagao, M.; Fujita, Y.; Wakabayashi, K.; Nukaya, H.; Kosuge,
T.; Sugimura, T. Mutagens in coffee and other beverages.
Environ. Health Perspect. 1986b, 67, 89-91.
Nahrstedt, A.; Albrecht, M.; Wray, V.; Gumbinger, H. G.; J ohn,
M.; Winterhoff, H.; Kemper, F. H. Structures of compounds
with antigonadotropic activity obtained by in vitro oxidation
of caffeic acid. Planta Med. 1990, 56, 395-398.
Obana, H.; Nakamura, S.; Tanaka, R. Suppressive effects of
coffee on the SOS responses induced by UV and chemical
mutagens. Mutat. Res. 1986, 175, 47-50.
Rinkus, S. J .; Taylor, R. T. Analysis of hydrogen peroxide in
freshly prepared coffees. Food Chem. Toxicol. 1990, 28,
323-331.
Rizzi, G. P.; Boekley, L. J . Observation of ether linked phenolic
products during thermal degradation of ferulic acid in the
presence of alcohols. J . Agric. Food Chem. 1992, 40, 1666-
1670.
Snyder, S. H.; Hendley, E. D. Sensitive fluorometric and
radiomatric assays for monoamine oxidase and diamine
oxidase. Methods Enzymol. 1971, 17B, 741-746.
Stadler, R. H. Phenylindans, their preparation and utilization.
Eur. Pat. 94109355.1, 1994.
Stadler, R. H.; Turesky, R. J .; Mu¨ller, O.; Markovic, J .; Leong-
Morgenthaler, P. M. The inhibitory effects of coffee on
radical mediated oxidation and mutagenicity. Mutat. Res.
1994, 308, 177-190.
Abraham, S. K. Inhibition of in vivo genotoxicity by coffee.
Food Chem. Toxicol. 1989, 27, 787-792.
Aeschbacher, H. U.; J accaud, E. Inhibition by coffee of ni-
trosourea-mediated DNA damage in mice. Food Chem.
Toxicol. 1990, 28, 633-637.
Balough, B.; Anderson, A. B. Chemistry of the genus Sequoia-
II. Isolation of sequirins, new phenolic compounds from the
coast redwood (Sequoia sempervirens). Phytochemistry 1965,
4, 569-575.
Balough, B.; Anderson, A. B. Chemistry of the genus Sequoia-
III. Structural studies of isosequirin. Phytochemistry 1966,
5, 325-330.
Cilliers, J . J . L.; Singleton, V. L. Characterization of the
products of nonenzymic autoxidative phenolic reactions in
a caffeic acid model system. J . Agric. Food Chem. 1991,
39, 1298-1303.
Clapp, P. A.; Du, N.; Evans, D. F. Thermal and photochemical
production of hydrogen peroxide from dioxygen and tannic
acid, gallic acid and other related compounds in aqueous
solution. J . Chem. Soc., Faraday Trans. 1990, 86, 2587-
2592.
Clarke, R. J .; MacRae, R. Coffee. Volume 1: Chemistry;
Elsevier Applied Science: London, 1983.
Fiddler, W.; Parker, W. E.; Wasserman, A. E.; Doerr, R. C.
Thermal decomposition of ferulic acid. J . Agric. Food Chem.
1967, 15, 757-761.
Friederich, U.; Hann, D.; Albertini, S.; Schlatter, Ch.; Wu¨rgler,
F. E. Mutagenicity studies on coffee: the influence of
different factors on the mutagenic activity in the Salmonella/
mammalian microsome assay. Mutat. Res. 1985, 156, 39-
52.
Fujita, Y.; Wakabayashi, K.; Nagao, M.; Sugimura, T. Implica-
tion of hydrogen peroxide in the mutagenicity of coffee.
Mutat. Res. 1985, 144, 227-230.
Fulcrand, H.; Cheminat, E.; Brouillard, R.; Cheynier, V.
Characterization of compounds obtained by chemical oxida-
tion of caffeic acid in acid conditions. Phytochemistry 1994,
35, 499-505.
Graf, E. Antioxidant potential of ferulic acid. Free Radical
Biol. Med. 1992, 13, 435-448.
Guilbeault, G. G.; Kramer, D. N.; Hackley, E. A new substrate
for fluorometric determination of oxidative enzymes. Anal.
Chem. 1967, 39, 271.
Gumbinger, H. G.; Vahlensieck, U.; Winterhoff, H. Metabolism
of caffeic acid in the isolated perfused rat liver. Planta Med.
1993, 59, 491-493.
Stadler, R. H.; Markovic, J .; Turesky, R. J . In vitro anti- and
pro-oxidative effects of natural polyphenols. Biol. Trace
Elem. Res. 1995, 47, 299-305.
Stich, H. F. The beneficial and hazardous effects of simple
phenolic compounds. Mutat. Res. 1991, 259, 307-324.
Stich, H. F.; Rosin, M. P.; Bryson, L. Inhibition of mutagenicity
of a model nitrosation reaction by naturally occuring
phenolics, coffee and tea. Mutat. Res. 1982, 95, 119-128.
Taylor, A. R.; Keen, G. W.; Eisenbraun, E. J . Cyclodimerization
of styrene. J . Org. Chem. 1977, 42, 3477-3480.
Tiedke, C. Der Gerbstoffkomplex in Kaffee. Z. Unters. Leb-
ensm. 1936, 71, 393-404.
Tsuji, S.; Shibata, T.; Ohara, K.; Okada, N.; Ito, Y. Studies on
the factors affecting the formation of hydrogen peroxide in
coffee. J . Food Hyg. Soc. J pn. 1991, 32, 504-512.
Tyson, C. A.; Martell, A. E. Kinetics and mechanism of the
metal chelate catalyzed oxidation of pyrocatechols. J . Am.
Chem. Soc. 1971, 94, 939-945.
Hanham, A. F.; Dunn, B. P.; Stich, H. F. Clastogenic activity
of caffeic acid and its relationship to hydrogen peroxide
generated during autooxidation. Mutat. Res. 1983, 116,
333-339.
Heinrich, L.; Baltes, W.; Vorkommen von Phenolen in Kaffee-
Melanoiden (Occurrence of phenols in coffee melanoidins).
Z. Lebensm. Unters. Forsch. 1987, 185, 366-370.
Inoue, S.; Ito, K.; Yamamoto, K.; Kawanishi, S. Caffeic acid
causes metal-dependent damage to cellular and isolated
DNA through hydrogen peroxide formation. Carcinogenesis
1992, 13, 1497-1502.
Received for review J uly 5, 1995. Accepted December 29,
1995.X
J F950411C
Itagaki, S. K.; Kobayashi, T.; Kitagawa, Y.; Iwata, S.; Nukaya,
H.; Tsuji, K. Cytotoxicity of coffee in human intestinal cells
in vitro and its inhibition by peroxidase. Toxicol. in Vitro
1992, 6, 417-421.
X Abstract published in Advance ACS Abstracts, Feb-
ruary 15, 1996.