Removal of estrogenic activity and formation of oxidation products during ozonation of 17α-ethinylestradiol

Article abstract of DOI:10.1021/es035205x

This study investigated the oxidation of the oral contraceptive 17α-ethinylestradiol (EE2) during ozonation. First, the effect of ozone (O₃) on the estrogenic activity of aqueous solutions of EE2 was studied using a yeast estrogen screen (YES). It could be shown that O ₃ doses typically applied for the disinfection of drinking waters were sufficient to reduce estrogenicity by a factor of more than 200. However, it proved impossible to completely remove estrogenic activity due to the slow reappearance of 0.1-0.2% of the initial EE2 concentration after ozonation. Second, oxidation products formed during ozonation of EE2 were identified with LC-MS/MS and GC/MS and the help of the model compounds 5,6,7,8-tetrahydro-2- naphthol (THN) and 1-ethinyl-1-cyclohexanol (ECH), which represent the reactive phenolic moiety and the ethinyl group of EE2. Additionally, oxidation products of the natural steroid hormones 17β-estradiol (E2) and estrone (E1) were identified. The chemical structures of the oxidation products were significantly altered as compared to the parent compounds, explaining the diminished estrogenic activity after ozonation. Overall, the results demonstrate that ozonation is a promising tool for the control of EE2, E2, and E1 in drinking water and wastewater.

Full text of DOI:10.1021/es035205x

Environ. Sci. Technol. 2004, 38, 5177-5186
formation of oxidation products. Depending on the functional
Removal of Estrogenic Activity and
Formation of Oxidation Products
during Ozonation of
group that is attacked by O3 and its location in the molecule,
changes in the molecular structure might be significant
enough to destroy the pharmacological effects of the parent
compound. However, it cannot be excluded that in some
cases oxidation products still produce the original pharma-
cological effects. To confirm that ozonation is suitable for
the removal of pharmaceuticals, it should be at least shown
that there is high evidence that pharmacological effects are
significantly reduced. So far, the formation of ozonation
products has been investigated for very few pharmaceutical
compounds (7, 8), and the decrease in pharmacological
effects never has been checked in a comprehensive manner.
17r-Ethinylestradiol
M A R C M . H U B E R , †
T H O M A S A . T E R N E S , ‡ A N D
U R S V O N G U N T E N * , †
Swiss Federal Institute for Environmental Science and
Technology (EAWAG), Ueberlandstrasse 133,
CH-8600 Duebendorf, Switzerland, and Federal Institute of
Hydrology (BfG), Am Mainzer Tor 1,
In the present study, the oral contraceptive 17R-ethin-
ylestradiol (EE2) was selected on the basis of its LOEC (lowest
observed effect concentration) of 0.1 ng/ L for vitellogenin
induction in rainbow trout (9). In effluents of sewage
treatment plants (STPs), its concentration typically ranges
from <0.5 ng/ L to 10 ng/ L (10). In vivo tests showed that EE2
is approximately 11-27 times more potent than the female
sex hormone 17â-estradiol (E2) (11). On the basis of in vivo
estrogenic potency, EE2 might be the most important
endocrine disruptor in STP effluents together with the octyl-
and nonylphenols (12).
D-56068 Koblenz, Germany
This study investigated the oxidation of the oral contraceptive
17R-ethinylestradiol (EE2) during ozonation. First, the
effect of ozone (O ) on the estrogenic activity of aqueous
3
solutions of EE2 was studied using a yeast estrogen
screen (YES). It could be shown that O doses typically
3
The estrogenic activity can be considered the primary
pharmacological effect of EE2. With the yeast estrogen screen
(YES) described by Routledge and Sumpter (13), a robust
and easy to handle test was available to quantify the effect
of ozonation on the estrogenic activity of pure aqueous
solutions of EE2. In detail, the effects of substoichiometric
O3 doses, resulting in a partial transformation of EE2, as well
as the effects of higher O3 doses typical for drinking water
treatment have been investigated in bench-scale experiments.
In the second part of this study, a number of oxidation
products formed during the ozonation of EE2, E2, and E1
were identified.
applied for the disinfection of drinking waters were sufficient
to reduce estrogenicity by a factor of more than 200.
However, it proved impossible to completely remove
estrogenic activity due to the slow reappearance of 0.1-
0.2% of the initial EE2 concentration after ozonation.
Second, oxidation products formed during ozonation of
EE2 were identified with LC-MS/MS and GC/MS and the
help of the model compounds 5,6,7,8-tetrahydro-2-naphthol
(THN) and 1-ethinyl-1-cyclohexanol (ECH), which represent
the reactive phenolic moiety and the ethinyl group of
EE2. Additionally, oxidation products of the natural steroid
hormones 17â-estradiol (E2) and estrone (E1) were
identified. The chemical structures of the oxidation products
were significantly altered as compared to the parent
compounds, explaining the diminished estrogenic activity
after ozonation. Overall, the results demonstrate that ozonation
is a promising tool for the control of EE2, E2, and E1 in
drinking water and wastewater.
During ozonation, micropollutants such as EE2 can be
oxidized either by ozone (O3) directly or by hydroxyl radicals
(•OH), which are formed as a consequence of O3 decay. The
two oxidants vary strongly in their reactivity. O3 attacks
selectively certain functional groups, whereas •OH is a
nonselective oxidant that reacts fast with a large number of
moieties. Consequently, most •OH are scavenged by the water
matrix in natural waters. EE2 reacts very fast with O3 and
•OH (5). However, due to the higher selectivity of O3, oxidation
by O3 is normally the predominant process. Therefore, the
present study focused on direct O3 reactions and reactions
with •OH were suppressed by the use of scavenger compounds
such as tert-butyl alcohol (TBA). The performance of ozo-
nation processes with respect to the oxidation of micropol-
lutants can be assessed with O3 and •OH exposure (i.e.,
concentration of oxidant integrated over the reaction time)
(14). When •OH is scavenged, the oxidation of micropollutants
is only a function of O3 exposure. This parameter is also used
to assess the disinfection efficiency of ozonation processes.
Therefore, O3 exposure allows us to compare the removal of
estrogenic activity to the inactivation of microorganisms.
Introduction
In recent years, several studies have reported the occurrence
of a large number of pharmaceuticals in the aquatic
environment (1-3). The presence of pharmaceutical com-
pounds in water resources makes it necessary to assess water-
treatment processes with respect to their removal efficiency
for such compounds. Several studies have recently shown
that ozonation is a very promising technology for the
oxidation of pharmaceuticals during water treatment (4-6).
The results of these studies are based on the disappearance
of the parent compounds. At O3 doses typically used in water
treatment, full mineralization of pharmaceuticals is not
achievable, and consequently ozonation results in the
Experimental Section
Standards and Reagents. Estrone (E1), 17â-estradiol (E2),
1-ethinyl-1-cyclohexanol (ECH), 5,6,7,8-tetrahydro-2-naph-
thol (THN), adipic acid, cyclohexanone, 2-methyl-3-butyn-
2-ol, R-hydroxyisobutyric acid, and the methylester of
1-hydroxycyclohexane-1-carboxylic acid were purchased
from Sigma-Aldrich (see Figure 1 for chemical structures).
* Corresponding author phone: +41-44-8235270; fax: +41-44-
8235210; e-mail: vongunten@eawag.ch.
† Swiss Federal Institute for Environmental Science and Technol-
ogy.
‡ Federal Institute of Hydrology.
9
10.1021/es035205x CCC: $27.50
Published on Web 08/21/2004
2004 Am erican Chem ical Society
VOL. 38, NO. 19, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 5 1 7 7

proximately 2 mL of a water/ acetonitrile (20/ 80, v/ v) mixture
and filtered with a 0.2 µm filter. The samples were analyzed
in the positive as well as the negative mode. Adipic acid,
cyclohexanone, 1,2-cyclohexanedione, and R-hydroxyisobu-
tyric acid were quantified using the MRM (multiple reaction
monitoring) mode after optimizing the LC-MS/ MS for each
compound. For quantification, aqueous solutions were
analyzed directly without any concentration steps.
GC/MS Analysis. After freeze-drying 100-250 mL aliquots
of ozonated model-compound solutions, the residual com-
ponents were resuspended in approximately 2 mL of acetone
and filtered with a 0.2 µm filter to remove insoluble
components. Aliquots of 0.2 mL were derivatized by adding
0.2 mL of a diazomethane/ diethyl ether solution at -20 °C.
The samples were incubated for 1 h at -20 °C before
diazomethane was quenched with 2-3 drops of acetic acid/
acetone solution (1:10, v/ v). Subsequently, the volume of
the samples was reduced to 200 µL by a gentle nitrogen
stream. Diazomethane is highly toxic and carcinogenic. All
work has to be conducted in a fume hood, and skin contact
has to be strongly avoided.
Separation and detection were accomplished with a Varian
GC 3400 coupled to a Varian Saturn 4D mass spectrometer.
The gas chromatograph was equipped with a PTV injector
and a Restek XTI-5 column (30m × 0.25 mm × 0.25 µm). GC
injection parameters: 5 µL, splitless; 50 °C; 100 °C/ min to
300 °C, 300 °C isothermal 10 min. Oven temperatures: 50 °C
isothermal for 2 min, 2.8 °C/ min to 120 °C, 15.4 °C/ min to
290 °C, and 290 °C isothermal for 10 min.
FIGURE 1. Structures of 17r-ethinylestradiol (EE2), 17â-estradiol
(E2), estrone (E1), and the model compounds 5,6,7,8-tetrahydro-2-
naphthol (THN) and 1-ethinyl-1-cyclochexanol (ECH).
The chemicals were of the highest purity available. 17R-
Ethinylestradiol (EE2) was provided by Schering/ Berlin,
Germany. Depending on the experiments, stock solutions of
EE2 and the model compounds ECH and THN were prepared
either in Milli-Q purified water (Millipore), in tert-butyl
alcohol (TBA), or in acetone. Stock solutions of E1 and E2
were prepared in TBA or in acetone. All chemicals used for
solutions (buffer, eluents, etc.) were reagent grade and were
used without further purification. O3 was produced with a
Fischer 500 and a Fischer 502 ozone generator by using pure
oxygen as feed gas. O3 stock solutions (∼1 mM) were produced
by sparging O3-containing oxygen through Milli-Q water that
was cooled in an ice bath (15).
Recom binant Yeast Estrogen Screen (YES). The YES was
conducted as described by Routledge and Sumpter (13).
Briefly, the recombinant yeast strain used in the screen
expresses the human estrogen receptor (hER), which can
interact with estrogenic compounds. Upon binding an active
compound, the hER triggers the expression of the reporter
gene lac-Z, which promotes the production of the enzyme
â-galactosidase. The enzyme metabolizes the yellow dye
chlorophenol red-â-D-galactopyranoside (CPRG) into a red
product that can be measured spectrophotometrically.
Depending on the expected EE2 concentrations, aqueous
samples were freeze-dried and resuspended in ethanol or
directly diluted in ethanol to obtain appropriate concentra-
tions for the YES. Based on these solutions, series of 6 or 12
dilutions (1:1) were prepared in ethanol. Of each dilution,
20 µL was added to a 96-well microtiter plate. Besides the
samples, each plate contained a standard curve with E2 in
ethanol (final concentration dissolved in assay medium:
2 × 10-8 to 1 × 10-11 M) and a row of blanks. After ethanol
was evaporated to dryness, the yeast cells were added together
with the assay medium. The color development was measured
after incubating the microtiter plate for 72 h at 30 °C.
To evaluate the data, the sigmoidal concentration-
response curve of the E2 standard was fitted to a symmetric
logistic function (eq 1) using the software Prism (GraphPad,
San Diego, CA).
Determ ination of EE2. EE2 was determined with a
Hewlett-Packard 1050 series HPLC system equipped with a
Nucleosil-100 C18 column (4 × 125 mm, 5 µm) and a
fluorescence detector (HP 1064A). The mobile phase con-
sisted of 50% 10 mM phosphoric acid and 50% acetonitrile
at a flow rate of 0.6 mL/ min. An excitation wavelength of 229
nm and emission wavelength of 309 nm were used for
fluorescence detection. Under the experimental conditions,
the quantification limit of EE2 was 5 nM (1.5 µg/ L) for an
injection volume of 100 µL. The error of a single measurement
was approximately (5%. Enrichment by freeze-drying low-
ered the quantification limit to 0.2 nM (60 ng/ L). The EE2
recovery for freeze-dried samples was 70-90%.
Determ ination of O3, Hydroperoxides, and Form ic Acid.
Dissolved O3 was determined with the indigo method (15)
or spectrophotometrically by measuring the absorbance at
258 nm (ꢀ ) 3000 M-1 cm-1). Hydroperoxides and hydrogen
peroxide (H2O2) were measured using Allen’s reagent as
described by Flyunt et al. (16). Catalase was applied to
selectively quench H2O2. Formic acid was determined with
ion chromatography using a method adapted from (17).
LC-MS/MS Analysis. The HPLC system consisted of a
Merck-Hitachi pump with an Inertsil ODS-3 column (4.6 ×
100 mm, 3 µm). Elution was performed with 0.1% acetic acid
(A) and acetonitrile (B) at a flow rate of 0.4 mL/ min. The
gradient was as follows: B started at 20% and was increased
with a linear gradient to 80% after 15 min. After 2 min with
B at 80%, the system was reequilibrated for 15 min. Mass
spectrometry was performed using an API 365 (PE Sciex)
triple quadrupole mass spectrometer with turbo-electrospray
ionization (ESI). Nitrogen was used as curtain gas, and
synthetic air was used as nebulizer gas. Curtain and nebulizer
gas flows were both operated with a flow rate of 1.0 L/ min.
The ESI interface was heated to 400 °C. The HPLC flux was
split 1:10 to diminish the flow rate in the ESI.
b - a
1 + 10(log EC50 - log c)‚m
response ) a +
(1)
where a is the baseline response (bottom), b is the maximum
response (top), c is the concentration, m is the Hill slope,
and EC50 is the concentration provoking the half-maximal
response.
With the values received for b (top) and a (bottom), the
response of the standard and the samples was expressed as
the percent of maximum response evoked by E2. To
determine the EC50 and the Hill slope m of the E2 standard
curve, the concentration-response curve was fitted again to
eq 1 with the response expressed in percent whereby a and
b were held constant at 0% and 100%, respectively. To
Full scans as well as precursor and product ion scans
were conducted to determine the quasi-molecular ions and
the structure of major oxidation products. Before analysis,
aliquots of the ozonated solutions (100-250 mL) were freeze-
dried. The residual components were resuspended in ap-
9
5 1 7 8 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004

calculate the EC50 values for the samples, the response in
percent was fitted to eq 1 whereby a (0%), b (100%), and m
were held constant. The concentration c in eq 1 was replaced
by the concentration factor. The estrogenic activity of a
sample expressed in 17â-estradiol equivalents (EEQs) was
then calculated as the ratio of the EC50 for E2 to the EC50
for the sample.
repeated at pH 7 (5 mM phosphate buffer) with the
corresponding ozonation time. Reaction solutions containing
the estrogens consisted of EE2 (0.02 mM), E2 (0.005 mM), or
E1 (0.005 mM) dissolved in Milli-Q water and either acetone
•
or TBA as OH scavenger. Due to the low solubility of the
estrogens in aqueous solutions, only low starting concentra-
tions could be selected. In a first step, estrogen solutions
were ozonated for 0.5 min (E1, E2) to 2 min (EE2). To increase
the product yield, the ozonated solution was spiked again
with the corresponding estrogen and subjected to ozonation
for an identical time interval. Overall, this procedure was
repeated 10 times. The fact that oxidation products were
more polar than the parent compound prevented the
precipitation of oxidation products.
Ozonation Experim ents for YES. If not stated otherwise,
the reaction solutions consisted of Milli-Q purified water
spiked with 1 or 10 µM EE2 and were buffered to pH 8 with
5 mM phosphate buffer. Furthermore, the solutions con-
•
tained 5 mM TBA as an OH scavenger.
Substoichiometric O3 Doses. Experiments with sub-
stoichiometric O3 doses were carried out as follows: To a
series of identical reaction solutions (20 mL) with 10 µM
EE2, O3 doses ranging from 5 to 24 µM were added under
vigorous stirring. Immediately after the addition of O3, 10
mL of the solution was removed from the reaction vessel
and stored for 2-3 days at room temperature. Of the 10 mL,
0.5 mL was taken for HPLC analysis and 9.5 mL was freeze-
dried. After freeze-drying, the residual components were
redissolved with the help of sonication in 0.95 mL of ethanol.
Insoluble phosphate salts were allowed to settle before the
liquid phase was removed and stored at -20 °C until the YES
was conducted. Samples with EE2 > 0.5 µM were diluted
directly with ethanol without freeze-drying.
Reduction of Estrogenicity as a Function of O3 Exposure.
To quantify the reduction in estrogenicity as a function of
O3 exposure, a reaction solution (500 mL) spiked with 1 µM
EE2 was thermostated at 10 °C. After adding 20 µM (1 mg/ L)
O3, a series of 10 mL samples was taken over the course of
the reaction. Thiosulfate was used to quench O3 in the
samples. The corresponding O3 exposures of the samples
were calculated on the basis of an O3 decay curve, which was
determined in a preliminary experiment under conditions
identical to those described in ref 18. In an experiment with
a slightly different setup, 20 µM (1 mg/ L) O3 was added to
a series of 250 mL solutions. Samples of 100 mL were taken
after the desired time intervals and immediately quenched
with thiosulfate. These experiments allowed for the quan-
tification of EE2 concentration < 1 nM with HPLC due to a
higher enrichment during freeze-drying. After quenching O3,
samples of both experiments were stored for 3 days at room
temperature before they were freeze-dried. With the help of
sonication, the residual components were redissolved in 1
and 2 mL of ethanol, respectively.
Kinetics of Reappearance of EE2 after Ozonation. The
reappearance kinetics of EE2 was investigated at two different
O3 concentrations (50 and 100 µM), which were added to
two reaction solutions spiked with 10 µM EE2. After 5 min,
2 mL of the solutions was withdrawn and transferred into an
amber HPLC vial where O3 was immediately quenched with
thiosulfate. Each of the samples was repeatedly analyzed
over the next 165 h.
To investigate the pH dependence of the EE2 reappear-
ance, reaction solutions were buffered to pH 5, 6, 7, 8, and
9. The experiments were performed as described for the
reappearance kinetics except that samples were only analyzed
after 3 days.
Investigation of Product Form ation. The oxygen stream
containing 2% O3 was continuously bubbled through a 1-L
amber glass bottle, which contained the reaction solution.
The solution was stirred with a magnetic stir bar at 80 rpm.
Reaction solutions of the model compounds THN (0.2 mM)
and ECH (5 mM) as well as of 2-methyl-3-butyn-2-ol (5 mM)
were prepared by dissolving the pure chemicals in acetone
or TBA and spiking them to Milli-Q water. The reaction
solutions containing the model compounds were ozonated
during 5-40 min. The experiments that produced the highest
yield with respect to the quantified oxidation product were
Results and Discussion
The second-order rate constant (kO3) for the reaction of 17R-
ethinylestradiol (EE2) with ozone (O3) is extremely high (at
pH 7, kO3 ) 3 × 106 M-1 s-1 (5)), resulting in a half-life of
approximately 10 ms for an O3 concentration of 1 mg/ L. O3
attack must take place on the phenolic moiety of EE2, because
phenols are well known to react fast with ozone at neutral
or basic pH (19). In this pH range, the reaction of O3 with
the phenolate anion of EE2 is the predominant reaction,
whereas the reaction with neutral EE2 is negligible. Conse-
quently, the reaction of EE2 (pKa ) 10.4) with O3 is strongly
pH-dependent (the apparent kO3 at pH 8 is 10 times higher
than at pH 7). The second reactive group of EE2, the ethinyl
group, has a considerably lower reactivity toward O3. Its kO3
is estimated to be close to the kO3 of ECH, which was
determined in this study to be 200 M-1 s-1 at 20 °C. The
corresponding half-life is approximately 3 min for an O3
concentration of 1 mg/ L. On the basis of these kinetic
considerations, it is clear that EE2 disappears very fast during
ozonation processes. However, if ozonation is applied to treat
EE2-containing water, the removal of estrogenic activity has
to be assessed in addition to the disappearance of EE2. To
test whether ozonation reduces estrogenic activity, aqueous
solutions of EE2 were treated with O3 and subsequently tested
for estrogenic activity using a yeast estrogen screen (YES).
Reduction of Estrogenicity with Substoichiom etric O3
Doses. The high kO3 for the oxidation of EE2 refers to the first
transformation step. On the basis of a study on the ozonation
of phenol (20), it can be assumed that some of the important
subsequent oxidation steps of EE2 proceed as fast as the first
transformation step. This sequence of fast transformation
steps will lead to a modification and partly to a cleavage of
the phenol ring in EE2. To test whether this results in a
substantial removal of estrogenicity, experiments were
performed with substoichiometric O3 doses. Under these
conditions, only the fast initial transformation steps take
place.
Figure 2a illustrates the correlation between decrease in
estrogenicity and the concentration of EE2 as a function of
O3 dosages. The estrogenicity expressed in 17â-estradiol
equivalents (EEQs) and EE2 concentrations were normalized
to the values of the solution without O3 ([EE2]0 ) 10 µM).
Approximately 1.8 mol of O3 was consumed per mole of EE2.
The stoichiometry is higher than 1 because oxidation products
of the first oxidation step obviously underwent further
reactions with O3. Adding 19 µM O3 reduced the estrogenicity
of the solution to 1.5%. The reduction in estrogenicity was
proportional to the decrease in EE2. This indicates that the
estrogenicity of the intermediates of the first transformation
step is much lower than that of EE2. In Figure 2b, the
estrogenicity expressed in EEQ is plotted versus the EE2
concentration. The fact that the data form a straight line
with a slope of approximately 1 demonstrates that the parent
compound EE2 is responsible for the observed estrogenicity.
Even at the two lowest EE2 concentrations, accounting for
9
VOL. 38, NO. 19, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 5 1 7 9

FIGURE 2. The effect of substoichiometric O doses on the estrogenicity of aqueous solutions of EE2. Experimental conditions: [EE2]0 )
3
10 µM, pH ) 8, T ) RT, [TBA] ) 5 mM. (a) Relative decrease of estrogenicity and EE2 concentration as a function of O dose. (b) Logarithmic
3
plot of estrogenicity in 17â-estradiol equivalents (EEQs) versus EE2. Data are from Figure 2a and an additional experiment.
0.12% and 0.28% of the initial EE2 concentration, no
significant deviation from the straight line occurs. This means
that the sum of the estrogenic activity of the intermediates
is at least 200 times lower than the estrogenicity of the original
EE2 solution.
These findings were confirmed with an estrogen receptor
competitive-binding assay adapted from Blair et al. (21). In
this assay, the estrogen receptor can directly interact with
the investigated compounds. Because no living organism is
used, biological processes associated with test organisms do
not influence the response of the assay. After ozonation, EE2-
containing solutions did not displace radiolabeled E2 from
the rat estrogen receptor, whereas untreated solutions
displaced it completely. This means that the oxidation
products of EE2 do not bind efficiently to the estrogen
receptor.
The YES experiments described above were designed to
FIGURE 3. Reduction of estrogenicity and the EE2 concentration as
a function of O exposure. Experimental conditions: [EE2]0 ) 1 µM,
3
yield a reduction of estrogenicity by a factor of 1000-10 000.
However, the observed reduction was lower due to a slow
reappearance of EE2 after oxidation. The reappearance of
EE2 will be discussed later. To achieve concentrations lower
than 0.1 µM with nearly stoichiometric O3 doses, the reaction
solutions had to be spiked with a second O3 dose after storing
the solutions for 3 days at room temperature. The second O3
dose was kept as low as possible to avoid significant changes
in the product distribution of the fast initial transformation
steps.
[O ] ) 1 mg/L, pH ) 8, T ) 10 °C, [TBA] ) 5 mM.
3 0
units. A much larger O3 exposure of approximately 9 mgL-1
min is required for a 2-log inactivation of Cryptosporidium
parvum oocysts. On the basis of this inactivation data, it can
be predicted that EE2-based estrogenicity is reduced at least
by a factor of 200, if an ozonation processes achieves a 2-log
inactivation of Giardia muris cysts or by a factor of 1000 if
a 2-log inactivation of Cryptosporidium parvum oocysts is
attained.
Reduction of Estrogenicity as a Function of O3 Exposure.
In drinking-water treatment, O3 exposures are much higher
than in the experiments with substoichiometric O3 doses, in
which O3 was immediately consumed. To test whether higher
O3 exposures result in further removal of estrogenicity,
aqueous solutions of EE2 at 10 °C and pH 8 were subjected
to O3 exposures ranging from 0.5 to 20 mgL-1 min. Figure 3
shows the decrease of the EE2 concentration together with
the EEQ reduction. As expected, the results showed a drop
of the estrogenicity by a factor of 200-500 for the lowest
measured O3 exposure, confirming the results of the experi-
ments with substoichiometric O3 doses. However, at higher
O3 exposures, the further decrease of estrogenicity was very
slow. To reduce estrogenicity by a factor of 1000, an O3
exposure of approximately 10 mgL-1 min was necessary.
The O3 exposures reported above can be related to O3
exposures applied for disinfection, which is often the primary
objective of ozonation. O3 exposures required to achieve
certain levels of inactivation for specific microorganisms were
calculated on the basis of inactivation rate constants and
activation energies compiled by von Gunten (22). At an O3
exposure of 0.5 mgL-1 min and 10 °C, the inactivation of
Giardia muris cysts is approximately 2 log units. Under the
same conditions, inactivation for E. coli is more than 6 log
To quantify how much of the residual estrogenicity was
caused by EE2 itself, EE2 concentrations were measured with
HPLC and fluorescence detection as well as LC-MS/ MS for
confirmation. The results show that up to O3 exposures of
approximately 10 mgL-1 min, residual EE2 accounted for
most of the estrogenicity. At higher O3 exposures, EE2
concentrations were close to the quantification limit of the
HPLC method (0.2 nM). Therefore, on the basis of the current
data, it cannot be excluded that other compounds are
responsible for a part of the estrogenicity observed at higher
O3 exposures. A possible explanation for the formation of
estrogenic products will be given later.
Reappearance of EE2. The presence of EE2 in solutions
that were treated with high O3 exposures was unexpected.
For an O3 concentration of 1 mg/ L, the half-life of EE2 at pH
8 is approximately 1 ms. Consequently, EE2 should disappear
below the detection limit in less than 1 s. To assess the
reappearance of EE2, the kinetics of this process was
investigated. Figure 4 shows the reappearance of EE2 after
ozonation as a function of time. The EE2 concentration rose
quickly during the first hours after the experiments. After
this initial step, the concentrations increased more slowly
over several days. As compared to an O3 dose of 50 µM, the
reappearance at 100 µM O3 was less pronounced but still
9
5 1 8 0 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004

slow decrease of EE2 in Figure 3 corresponds to a second-
order rate constant of approximately 150 M-1 s-1 at 10 °C.
Therefore, it seems to be plausible that the O3 attack on the
ethinyl group of “protected” EE2 contributes to the disap-
pearance of EE2. This reaction would result in the formation
of products for which the phenolic moiety of EE2 can be
reformed after ozonation. It can be expected that these
products still exhibit estrogenic activity. Therefore, we assume
that they could be responsible for the remaining estrogenicity
that seemed not to be caused by EE2.
Despite the fact that a reappearance of EE2 has been
observed, it has to be emphasized that it accounts only for
0.1-0.5% of EE2 oxidized. We assume that this phenomenon
will have little relevance in practice where a removal of 99.5%
estrogenicity should be sufficient in most cases.
Identification of Oxidation Products. In the second part
of this study, oxidation products formed during the ozonation
of EE2 were identified. As mentioned earlier, EE2 has a highly
ozone reactive phenolic moiety (kO3 ) 3 × 106 M-1 s-1 at pH
7, t1/ 2 ≈ 10 ms for 1 mg/ L O3) and a significantly less reactive
ethinyl group (kO3 ≈ 200 M-1 s-1, t1/ 2 ≈ 3 min for 1 mg/ L O3).
On the basis of these rate constants, it can be assumed that
both reactive moieties will be attacked by O3 during drinking
water treatment. To facilitate the identification of oxidation
products, two model compounds were chosen that represent
the two different reactive moieties. The selected compounds
were 5,6,7,8-tetrahydro-2-naphthol (THN) for the phenolic
moiety and 1-ethinyl-1-cyclohexanol (ECH) for the ethinyl
group (Figure 1).
Ozonation of the Model Com pound THN. Aqueous
solutions containing THN were ozonated and subsequently
analyzed for oxidation products with LC-MS/ MS. Scheme 1
shows the suggested reaction mechanism and the identified
products. The chemical structure of two oxidation products
could be determined with high certainty. The first product
was identified as adipic acid (6) using an authentic standard
for LC-MS/ MS analysis. The second oxidation product was
identified as 1-hydroxy-cyclopentanecarboxylic acid (7). No
standard was available for this compound. After derivatization
with diazomethane, the fragmentation in the GC/ MS spec-
trum (Figure 5a) was very similar to the fragmentation found
in the spectrum of the commercially available methylester
of 1-hydroxy-cyclohexanecarboxylic acid (9) (Figure 5b). The
spectrum of methylated 7 corresponded to the spectrum of
methylated 9 shifted 14 Da (CH2) to lower masses, suggesting
that 7 consisted of a five- instead of a six-membered ring.
The proposed structure of 7 is further supported by the loss
of 46 (CH2O2) from the quasi molecular ion [M - H]-
m/ z ) 129 in LC-MS/ MS experiments performed in the
negative mode (Table 2). Normally, carboxylic acids lose 44
FIGURE 4. Reappearance of EE2 as a function of O dose and time.
3
Experimental conditions: [EE2]0 ) 10 µM, pH ) 8, T ) RT,
[TBA] ) 5 mM. O was quenched with thiosulfate after 5 min. An
3
-1
O dose of 50 µM resulted in an O exposure of 4-5 mgL min,
3
3
whereas 100 µM O yielded an approximately 4 times higher O
3
3
exposure of 15-20 mgL-1 min.
significant. The same pattern of reappearance was also
detected for THN and E2 (data not shown). This observation
was not limited to the selected experimental conditions with
pure water. Experiments with pretreated water from the River
Seine in Paris, France, showed the same behavior of EE2.
Furthermore, the dependence on EE2 concentration seemed
to be small. Starting with 10, 1, and 0.1 µM EE2 resulted in
comparable percentages of reappearance (0.05-0.2%) for
similar O3 doses.
The reappearance of EE2 can only be explained if a small
fraction of EE2 is “protected” from O3 attack. We hypothesize
that such “protected” forms of EE2 consist of hydroperoxides
formed by the fast reaction of phenoxyl radicals with
superoxide anion. This reaction was investigated in detail
for radiolytically generated phenoxyl-type radicals by
d’Alessandro et al. (23). It can be expected that such
hydroperoxides would not be highly reactive toward ozone.
Because they revert to phenols in a slow reaction by
eliminating dioxygen, as shown in the study mentioned
above, the formation of hydroperoxides provides a plausible
explanation for the slow reappearance of EE2 observed in
the present study.
These hydroperoxides would be “protected” from the fast
oxidation of the phenolic moiety, but O3 could still attack on
the ethinyl group. On the basis of the reaction of O3 with
ECH, the second-order rate constant for the O3 attack on the
ethinyl group is estimated to be 160 M-1 s-1 at 10 °C. The
SCHEME 1. Reaction of O with the Model Compound THN
3
9
VOL. 38, NO. 19, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 5 1 8 1

tion of the hydroperoxide 4 and/ or 1,2-cyclohexanedione
(5a). Product 4 can be categorized as R-hydroxy or R-keto-
hydroperoxide. It can react through hydrogen peroxide
elimination (26), leading to the formation of 5a. A second
reaction is a rearrangement by cleavage of the bond between
the hydroperoxy and the keto group (27). The latter reaction
results in the formation of adipic acid (6), which was identified
as a major oxidation product. LC-MS/ MS analysis indicated
that if 5a is formed during the reaction, it accounts for less
than 10% of THN transformed. However, the fact that 5a is
almost entirely hydrated in aqueous solution (5c) and
additionally slowly transformed into the enol form 5b (28)
makes analysis difficult, and it cannot be excluded that some
of the species of 5 were missed in the analysis. The formation
of 7 must be the product of a benzilic acid rearrangement
(29) of 5a via 5c. Ring contractions of six-membered cyclic
diketones to R-hydroxy acids have been reported by several
studies dealing with steroid synthesis (e.g., ref 30). Usually,
these experiments were performed under alkaline conditions
in solvent/ water mixtures. In aqueous solution, a study on
the enolization of 5a (31) explained its slow irreversible
disappearance and the formation of an acidic compound by
the formation of 7 as well. On the basis of these consider-
ations, we assume that 5 was at least partly transformed into
7 in a relatively slow reaction after ozonation.
Scheme 1 refers to the initial O3 attack at position 3 of the
phenol ring. O3 can also attack at position 1. The corre-
sponding first step intermediates will be formed. Further
oxidation by O3 might proceed in a different way and yield
different products due to the different positions of the double
bonds. However, it is possible that products 6 and 7 were
partly formed by this way as well.
Ozonation of the Model Com pound ECH. The ethinyl
group represents the reactive moiety of the second model
compound 1-ethinyl-1-cyclohexanol (ECH). Even if acety-
lenes react considerably more slowly with O3 than olefins,
it can be assumed that O3 attack on the ethinyl group proceeds
analogous to the well-established Crigee mechanism for
double bonds (32). According to this mechanism, O3 attack
on the ethinyl group of ECH results in the formation of the
primary ozonide 8, which quickly decomposes into the
hydroxyhydroperoxides 9a and/ or 9b (Scheme 2). The fact
that less than 2 min after the completion of the reaction no
hydroperoxides except hydrogen peroxide could be deter-
mined demonstrates that 9a and 9b are only short-lived
intermediates. The H2O2 yield was 65% relative to the added
O3 concentration. This indicates that the keto aldehyde 10
is the major product of this reaction (in aqueous solution 10
may be hydrated).
FIGURE 5. GC/MS spectra of derivatized oxidation products formed
during the ozonation of THN (a) and ECH (b,c). Methylester of (a)
1-hydroxycyclopentane-1-carboxylic acid (7), (b) 1-hydroxycyclo-
hexane-1-carboxylic acid (9), (c) 1-hydroxycyclohexane-1-car-
boxylic acid, formate (10).
With LC-MS/ MS analysis, three further oxidation products
were identified. The first product was identified as 1-hydroxy-
cyclohexanecarboxylic acid (11). After derivatization with
diazomethane, its GC/ MS spectrum (Figure 5b) was identical
to the spectrum of the commercially available methylester
of the compound. The identification of 11 is in agreement
with the detection of formic acid immediately after comple-
tion of the reaction of O3 with ECH. For the second product
12, no standard was available. The GC/ MS spectrum of
derivatized 12 (Figure 5c) was virtually identical to the
spectrum of 11 except that the molecular ion was m/ z 186
instead of m/ z 158 and that the loss of 31 (OCH3) from the
molecular ion resulted consequently in a peak at m/ z 155
instead of m/ z 127. The difference of 28 Da between 12 and
11 and the associated derivatives can be explained best by
the presence of an additional CO group in 12. In both spectra,
the base peak is at m/ z 141, demonstrating that methylated
12 lost 45 (OCOH) instead of 17 (OH) as did 11. The further
fragmentation of the base peak (m/ z 141) was identical for
both molecules. Obviously, the CO group has to be attached
to the tertiary alcohol group. Additionally, it could be shown
(CO2) in the negative mode (24), whereas the loss of 46 seems
to be characteristic of R and â-hydroxy acids (25).
A third product with MW ) 176 (not shown in Scheme
1) was only detected with LC-MS/ MS. The MS/ MS spectrum
in the negative mode showed the loss of 18, 44, 46, and 62
from [M - H]-. The compound was tentatively identified as
2-hydroxyheptanedioic acid (C7H12O5) because the loss of 44
and 46 suggests that the compound contains a conventional
and an R or â-hydroxy carboxylic acid. A forth major peak
in the LC-MS chromatogram with MW ) 154 could not be
identified.
Scheme 1 shows the suggested reaction mechanism for
O3 attack at position 3 of the phenol ring. On the basis of a
study on the ozonation of phenol (20), it can be assumed
that the fast first reaction step leads to the muconic acid
derivative 1, 5,6,7,8-tetrahydro-2,3-naphthalendiol (2), and
2,3-nathphalendione (3) as major intermediates. These
intermediates are still reactive to O3 and consequently
undergo further reactions, which might result in the forma-
9
5 1 8 2 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004

SCHEME 2. Oxidation Products Formed by the Reaction of O with the Model Compound ECH
3
TABLE 1. Ozonation Products of EE2 Identified with the Model Compounds THN and ECH
with LC/ MS that 12 hydrolyzes at pH 12 to 11. The third
oxidation product was identified as cyclohexanone (13) using
an authentic standard for LC-MS/ MS analysis. Due to
analytical problems, the formation of glyoxylic acid, which
should be produced in an equivalent amount, could not be
confirmed.
Product 10 was not detected with LC-MS/ MS analysis.
Because H2O2 was not destroyed in the samples and the
formation of 10 from the hydroxyhydroperoxides 9a and/ or
9b is a reversible process, 10 was slowly converted into 11,
12, and 13 during sample storage. The half-life time for 10
analogous to 10 exhibited a half-life > 40 d at pH 8 for base-
catalyzed degradation (34). Therefore, 10 might be a more
important product than 12 under realistic treatment condi-
tions where H2O2 concentrations are significantly lower.
Quantification of Product Form ation. The formation of
adipic acid (6) could be quantified with LC-MS/ MS, because
an authentic standard was available. Product 6 accounted
for up to 24% of THN transformed during ozonation. The
second THN product 7 could not be quantified.
Based on measurements of H2O2 and formic acid, ozo-
nation of ECH yielded 65% 10 and 20% 11. Cyclohexanone
(13) accounted for 15%. This compound was quantified with
LC-MS/ MS, and the result might have been influenced by
the storage time. Ozonation of the similar ethinyl compound
2-methyl-3-butyn-2-ol yielded 25% R-hydroxyisobutyric acid
at pH 7. This reaction is equivalent to the formation of 11
from ECH, and the 25% yield agrees with the yield determined
for 11. The mass balance for ECH seems complete, whereas
for THN the identified products probably do not account for
more than 50% of THN transformed.
is estimated to be <40 h for 1 mM H2O2 at pH 7 (kobserved
)
5-7 M-1 s-1 at pH 10 and 20 °C). At pH 12, the base-catalyzed
reaction of 10 with H2O2 yielded less than 30% formic acid,
indicating that 12 and not 11 or 13 is the major product
(glyoxylic acid quickly decomposes to formic acid under these
conditions). The slow conversion of a very similar keto
aldehyde into the corresponding formate by Baeyer-Villiger-
type oxidation in aqueous solution has already been reported
in (33). In the absence of H2O2, 10 seems to be fairly stable.
A degradation product of hydrocortisone with a moiety
9
VOL. 38, NO. 19, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 5 1 8 3

TABLE 2. LC-MS/MS Spectra for Oxidation Products of Model Compounds and Steroid Hormones in the Negative Modea
[M - H]-
-H O
-CO
∆28
-CO
-CH O
-H O-CO
2 2
2
2
2
2
N°
MW
m/z
∆18
∆44
∆46
∆62
THN
7
6
130
146
176
129
145(32%)
175
111(13%)
127(25%)
157(12%)
83(84%)
101
83(95%)
113(57%)
131(54%)
223(44%)
129(13%)
ECH
143
EE2
11
12
144
172
143
171(0%)
97(57%)
14
15
16
17
19
268
314
342b
252
326b
267(7%)
313
341(64%)
251
249(1%)
295(4%)
205
251(13%)
267(10%)
313
297
233(6%)
205(12%)
279(38%)
325(41%)
a
For each spectrum , the m /z values for the quasi-m olecular ion [M - H]- and the m ajor fragm ent ions are given. Bold values correspond to
b
the base peak. The intensity of the ions relative to the base peak is reported in brackets. Full-scan spectra.
Oxidation Products of EE2. Table 1 lists the oxidation
products of EE2 which were derived from the oxidation
products of the model compounds detected with LC-MS/
MS. Five out of six possible oxidation products of EE2 could
also be detected with LC-MS/ MS. Product 18 (Table 1) could
not be unambiguously assigned to a peak of the LC/ MS
chromatogram due to interferences with fragment ions from
other products. Table 2 reports the major mass fragments of
the MS/ MS spectra recorded in the negative mode. The
suggested structures of the oxidation products are supported
by the loss of characteristic masses depending on the
functional groups present in the molecules. Molecule 14
(Table 1) primarily lost 44 Da (CO2) as did the model-
compound product adipic acid (6). In the negative mode,
the loss of CO2 is typical for carboxylic acids (24). In contrast,
the R-hydroxy acid 17 lost 46 Da (CH2O2) instead of 44 Da.
The same behavior was observed for the model-compound
products 7 and 11. As mentioned earlier, the loss of 46 Da
seems to be characteristic for R or â-hydroxy acids (25).
Product 15, which has two conventional carboxyl groups
and one R-hydroxy acid group, lost 46 Da but not 44 Da.
However, the loss of 62 Da (CO2 +H2O) indicated the presence
of the two conventional carboxyl groups, because the same
fragmentation was also observed for 14 and 6. The molecules
16 and 19 exhibiting a formate group primarily lost 28 Da
(CO) and had the base peak at [M - 1 - 28]. The same
fragmentation was observed for the model-compound
product 12.
FIGURE 6. Major intermediates formed by the reaction of O with
3
EE2. These compounds were not detected with LC-MS/MS. However,
mechanistic considerations provide strong evidence that the
detected products in Table 1 were partly formed via 20 and
21. By superimposing the cyclohexanedione moiety of 20 on
products 14, 15, and 16 (Table 1), three further products are obtained,
which also may be formed. Product 20 can also be transformed into
its enol forms as shown in Scheme 1.
ments for the quantification of hydroperoxides and mecha-
nistic considerations demonstrated that some of these
oxidation products derive from less stable intermediates. Due
to time-consuming sample preparation and the presence of
H2O2 in the samples, these less stable intermediates de-
The oxidation products shown in Table 1 represent the
stable oxidation products detected with LC-MS/ MS. Experi-
SCHEME 3. Oxidation Products Formed by the Reaction of O with E2 and E1
3
9
5 1 8 4 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004

composed and were not detected with LC-MS/ MS. Figure 6
depicts two major transient oxidation products which must
have formed but escaped detection. It is expected that these
intermediates persist for a couple of hours to a few days in
the water before they are transformed into the more stable
products shown in Table 1.
The products listed in Table 1 and Figure 6 are formed
at relatively high O3 exposures as applied in ozonation of
drinking water. At lower O3 exposures, O3 does not react with
the ethinyl group. As a consequence, only the phenolic moiety
will be transformed. Despite the lower O3 exposure, it can
be expected that the phenol moiety is, at least partly,
transformed in the same way as for higher doses because the
first reaction steps are fast.
Acknowledgments
We thank the following persons for their assistance: Werner
Angst, Matthias Bonerz, Nadine Bramaz, Derek McDowell,
Beate Escher, Nadine Hermann, Barbara Rutishauser, Lisa
Sahli, Rene´ Schoenenberger, Marc Suter, Daniel Sutter, and
Mischa Zschokke. We also thank the ESWE Institute, Wies-
baden, Germany, where a significant part of this work was
performed. This study was performed within the framework
of POSEIDON, European Union project EVK1-CT-2000-
00047. Financial support by BBW (Bundesamt fu¨ r Bildung
und Wissenschaft) is gratefully acknowledged.
Literature Cited
Oxidation Products of 17â-Estradiol (E2) and Estrone
(E1). The oxidation products formed during the ozonation
of the natural hormones Ε2 and E1 were investigated as well.
Instead of an ethinyl group, E2 and E1 exhibit an alcohol and
a keto group at the 17 position (Figure 1). As carbonyl and
alcohol groups are much less reactive to O3 than an ethinyl
group, E2 and E1 were expected to react only at the phenolic
moiety. Surprisingly, the experiments yielded the same two
major products for E2 and E1. With LC-MS/ MS, they were
identified as 14 and 17, which were also formed during the
ozonation of EE2 (Scheme 3). This was expected for E1,
because these oxidation products of EE2 have a carbonyl
group at the 17 position identical to that of E1. The corres-
ponding alcohol group of E2 was unexpectedly oxidized to
the carbonyl group under the applied conditions and yielded
the same products as E1. Product 17 must be a decomposition
product of the intermediate 22. As shown in Scheme 1, the
cyclohexanedione moiety can be hydrated and is partly
present in the enol form 23.
(1) Ternes, T. A. Water Res. 1998, 32, 3245-3260.
(2) Kolpin, D. W.; Furlong, E. T.; Meyer, M. T.; Thurman, E. M.;
Zaugg, S. D.; Barber, L. B.; Buxton, H. T. Environ. Sci. Technol.
2002, 36, 1202-1211.
(3) Heberer, T. Toxicol. Lett. 2002, 131, 5-17.
(4) Ternes, T. A.; Meisenheimer, M.; McDowell, D.; Sacher, F.;
Brauch, H.-J.; Haist-Gulde, B.; Preuss, G.; Wilme, U.; Zullei-
Seibert, N. Environ. Sci. Technol. 2002, 36, 3855-3863.
(5) Huber, M. M.; Canonica, S.; Park, G.-Y.; von Gunten, U. Environ.
Sci. Technol. 2003, 37, 1016-1024.
(6) Ternes, T. A.; Stu¨ ber, J.; Herrmann, N.; McDowell, D.; Ried, A.;
Kampmann, M.; Teiser, B. Water Res. 2003, 37, 1976-1982.
(7) Andreozzi, R.; Caprio, V.; Marotta, R.; Vogna, D. Water Res. 2003,
37, 993-1004.
(8) McDowell, D.; Huber, M. M.; Wagner, M.; von Gunten, U.;
Ternes, T. A., in preparation.
(9) Purdom, C. E.; Haridman, P. A.; Bye, V. J.; Eno, N.; Tyler, C. R.;
Sumpter, J. P. Chem. Ecol. 1994, 8, 275.
(10) Ternes, T. A.; Stumpf, M.; Mueller, J.; Haberer, K.; Wilken, R.-D.;
Relationship between the Chem ical Structure and the
Estrogenic Activity of Oxidation Products. The identified
oxidation products of EE2 are expected to form at an O3
exposure of 5-10 mgL-1 min at 10 °C. In the experiment
shown in Figure 3, the corresponding O3 exposure resulted
in a reduction of estrogenicity by a factor of 5000-10 000 if
the estrogenicity caused by reappeared EE2 is subtracted.
Most probably, the observed reduction of estrogenicity can
be attributed to the cleavage of the phenolic moiety of EE2
because the 3-hydroxy group and the aromatic ring of the
phenolic moiety is of particular importance for the binding
of estrogens to the estrogen receptor (35). As demonstrated
by our investigation, the phenol rings of EE2, E2, and E1
were cleaved in all of the identified products. However, the
fact that the fast initial transformation steps, which result
only in a partial cleavage of the phenol ring (Scheme 1),
reduced the estrogenicity by a factor of 200-500 indicates
that only small modifications of the phenol ring may be
required to reduce the estrogenicity of EE2 substantially.
Overall, the present study provides direct evidence that the
selective oxidation of the phenolic moiety efficiently reduces
the estrogenicity of EE2-containing solutions.
Servos, M. Sci. Total Environ. 1999, 225, 81-90.
(11) Thorpe, K. L.; Cummings, R. I.; Huchtinson, T. H.; Scholze, M.;
Brighty, G.; Sumpter, J. P.; Tyler, C. R. Environ. Sci. Technol.
2003, 37, 1142-1149.
(12) Johnson, A. C.; Sumpter, J. P. Environ. Sci. Technol. 2001, 35,
4697-4703.
(13) Routledge, E. J.; Sumpter, J. P. Environ. Toxicol. Chem. 1996, 15,
241-248.
(14) Elovitz, M. S.; von Gunten, U. Ozone Sci. Eng. 1999, 21, 239-
260.
(15) Bader, H.; Hoigne´, J. Water Res. 1981, 15, 449-456.
(16) Flyunt, R.; Leitzke, A.; von Sonntag, C. Radiat. Phys. Chem. 2003,
67, 469-473.
(17) Kuo, C.-Y. J. Chromatogr., A 1998, 804, 265-272.
(18) Hoigne´, J.; Bader, H. Ozone Sci. Eng. 1994, 16, 121-134.
(19) Hoigne´, J.; Bader, H. Water Res. 1983, 17.
(20) Mvula, E.; von Sonntag, C. Org. Biomol. Chem. 2003, 1, 1749-
1756.
(21) Blair, R. M.; Fang, H.; Branham, W. S.; Hass, B. A.; Dial, S. L.;
Moland, C. L.; Tong, W.; Shi, L.; Perkins, R.; Sheehan, D. M.
Toxicol. Sci. 2000, 54, 138-153.
(22) von Gunten, U. Water Res. 2003, 37, 1469-1487.
For the first time, the oxidation of a pharmaceutical
compound during ozonation has been investigated in a
comprehensive manner, including oxidation kinetics, product
formation, and the pharmacological effects of the oxidation
products. The results demonstrate that O3 doses applied for
the disinfection of drinking water can efficiently remove EE2,
E2, and E1 and the estrogenicity associated with their
presence in the water. Due to the selective oxidation of
estrogens by O3, ozonation is also a promising tool for the
control of estrogenicity in effluents of sewage treatment
plants. This has already been confirmed by ozonation
experiments using a pilot plant, which was operated with
real wastewater (36). When the disinfection of wastewater is
a legal requirement, ozonation, therefore, may be a viable
alternative to other disinfection processes.
(23) d’Alessandro, N.; Bianchi, G.; Fang, X.; Jin, F.; Schuchmann,
H.-P.; von Sonntag, C. J. Chem. Soc., Perkin Trans. 2 2000, 1862-
1867.
(24) Frauendorf, H.; Herzschuh, R. Eur. Mass Spectrom. 1998, 4, 269-
278.
(25) Kerwin, J. L.; Torvik, J. J. Anal. Biochem. 1996, 237, 56-64.
(26) Zvilichovsky, G.; Zvilichovsky, B. In The Chemistry of Hydroxyl,
Ether and Peroxide Groups; Patai, S., Ed.; John Wiley & Sons
Ltd.: Chichester, 1993; Vol. 2, pp 687-784.
(27) Plesnicar, B. In The Chemistry of Peroxides; Patai, S., Ed.; John
Wiley and Sons Ltd.: Chichester, 1983; pp 521-584.
(28) Bakule, R.; Long, F. A. 1963, 85, 2309-2312.
(29) Selman, S.; Estham, J. Q. Rev. Chem. Soc. 1960, 14, 221-235.
(30) Wendler, N. L.; Taub, D.; Graber, R. P. Tetrahedron 1959, 7,
173-184.
9
VOL. 38, NO. 19, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 5 1 8 5

(31) Schwarzenbach, G.; Wittwer, C. Helv. Chim. Acta 1947, 30, 663-
669.
(32) Dowideit, P.; Von Sonntag, C. Environ. Sci. Technol. 1998, 32,
1112-1119.
(36) Ternes, T. A.; Bonerz, M.; Herrmann, N.; Ried, A.; Teiser, B.
Envirpharma - European conference on human and veterinary
pharmaceuticals in the environment; Lyon, France, 2003; http:/ /
www.envirpharma.org.
(33) Conrow, R. E.; Dillow, G. W.; Bian, L.; Xue, L.; Papadopoulou,
O.; Baker, J. K.; Scott, B. S. J. Org. Chem. 2002, 67, 6835-6836.
(34) Bundgaard, H.; Hansen, J. Arch. Pharm. Chemi 1980, 87, 995-
1013.
Received for review October 29, 2003. Revised manuscript
received May 24, 2004. Accepted June 15, 2004.
(35) Anstead, G. M.; Carlson, K. E.; Katzenellenbogen, J. A. Steroids
1997, 62, 268-303.
ES035205X
9
5 1 8 6 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004