Environ. Sci. Technol. 2005, 39, 8014-8022
Ozonation of Carbamazepine in Drinking Water: Identification and Kinetic Study of Major Oxidation Products DEREK C. MCDOWELL,† MARC M. HUBER,‡ MANFRED WAGNER,§ URS VON GUNTEN,‡ AND T H O M A S A . T E R N E S * ,† Federal Institute of Hydrology (BFG), Am Mainzer Tor 1 D-56068 Koblenz, Germany, Swiss Federal Institute for Aquatic Science and Technology (EAWAG), U ¨ berlandstrasse 133, CH-8600 Duebendorf, Switzerland, and Max Planck Institute for Polymer Science, Ackermannweg 10, D-55128 Mainz, Germany
Kinetics and product formation of the anti-epileptic drug carbamazepine (CBZ) were investigated in lab-scale experiments during reactions with ozone and OH radicals. Ozone reacts rapidly with the double bond in CBZ, yielding several ozonation products containing quinazolinebased functional groups. The structures for three new oxidation products were elucidated using a combination of mass spectrometric and NMR techniques. The three products were determined to be 1-(2-benzaldehyde)-4-hydro(1H,3H)-quinazoline-2-one (BQM), 1-(2-benzaldehyde)(1H,3H)-quinazoline-2,4-dione (BQD), and 1-(2-benzoic acid)(1H,3H)-quinazoline-2,4-dione (BaQD). Additional kinetic studies of the ozonation products showed very slow subsequent oxidation kinetics with ozone (second-order rate constants, kO3 ) ∼7 M-1 s-1 and ∼1 M-1 s-1 at pH ) 6 for BQM and BQD, respectively). Rate constants for reactions with OH radicals, kOH, were determined as ∼7 × 109 M-1 s-1 for BQM and ∼5 × 109 M-1 s-1 for BQD. Thus, mainly reactions with OH radicals lead to their further oxidation. A kinetic model including ozone and OH radical reactions allows a prediction of the time-dependent product distribution during ozonation of natural waters. In Rhine River water, CBZ spiked at 500 ng/L was completely oxidized by ozone with applied doses g0.3 mg/L. To confirm that the two major ozonation products BQM and BQD are produced as a result of the ozonation of a CBZ-containing natural water, Lake Zurich water samples were spiked with CBZ (1 µM, 236 µg/L). The oxidation products were identified via LC-UV. Concentrations of 0.48 and 0.15 µM for BQM and BQD, respectively, were measured for an ozone dose of 1.9 mg/L. BQM and BQD were also identified in ozonated water from a German waterworks containing CBZ in its raw water with 0.07-0.20 µg/L. Currently, there are no data available on the biological effects of the formed oxidation products.
* Corresponding author phone: +49 261-1306 5560; fax: +49 2611306 5363; e-mail:
[email protected]. † Federal Institute of Hydrology. ‡ Swiss Federal Institute for Aquatic Science and Technology. § Max Planck Institute for Polymer Science. 8014
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005
Introduction A great deal of interest has been generated in recent years with regard to the environmental fate and behavior of pharmaceutical drugs, because many industrialized countries have discovered drug products in water resources often used for drinking water purposes (1-3). Detection of pharmaceutical residues in drinking water has to date been quite rare (4-6). A few studies have been published recently that concentrated on the removal of pharmaceutical compounds during drinking and wastewater treatment (7-9). Even though the toxicological effects from low concentrations of pharmaceuticals in drinking water are largely unknown, due to the precautionary principle, the number and concentrations of anthropogenic compounds should be as low as possible considering the commonly accepted technological rules as described in the German drinking water directive (10). Ozone is a chemical oxidant that is used extensively in drinking water and to a lesser extent in wastewater treatment. It is applied for disinfection, taste and odor control, color removal, and oxidation of micropollutants (11). Recently, it has been shown that the oxidation of selected pharmaceuticals by ozone and OH radicals can lead to an effective transformation of many drugs during bench-, pilot-, and fullscale drinking and wastewater treatment (7, 8, 12, 13). Similar to many other organic compounds, the ozone-derived oxidation products formed from parent pharmaceutical compounds may be more susceptible to biological degradation (14). Ozone is a very selective oxidant, whereas the OH radical, a secondary oxidant produced during decomposition of aqueous ozone, is very unspecific. The ultimate effectiveness of ozone and OH radicals in oxidizing a micropollutant largely depends on the oxidant concentrations (ozone and OH radicals) and the second-order rate constants for its reaction with the oxidants (14). In the case of ozone, these rate constants are heavily influenced by the electronic properties of the molecule. High rate constants with ozone are typically encountered for chemicals containing olefinic groups, deprotonated amines, and activated aromatic systems such as phenols (15). These functional groups are particularly susceptible to ozone attack and are common constituents in many pharmaceuticals. Therefore, very fast kinetics can be expected for many drugs undergoing ozonation. Findings from a recent study supported this hypothesis by measuring second-order rate constants for the reactions of ozone and OH radicals (kO3 and kOH, respectively) with many pertinent pharmaceuticals (13). Half-lives calculated for numerous pharmaceuticals in this laboratory scale study were shown to be less than 60 s for direct ozone reactions using ozone concentrations of 1 mg/L. This was shown for 17R-ethinylestradiol (EE2) in combination with a product study (16). It was revealed that aqueous solutions containing EE2 exhibit a significant loss (>99%) in estrogenic activity after ozone treatment. Carbamazepine (CBZ), a double bond-containing molecule, was one of the pharmaceuticals identified as having a high rate constant with ozone. It passes sewage treatment plants (STPs) relatively unchanged and hence can be found in many rivers and streams at concentrations averaging 250 ng/L in Germany (17), at 185 ng/L in the Detroit River, and in the Hamilton Harbor of Lake Ontario, Canada, at a median concentration of 120 ng/L (1). To test possibilities of CBZ removal from water, the kinetics and mechanisms of product formation during oxidative treatment were investigated in the present study. 10.1021/es050043l CCC: $30.25
2005 American Chemical Society Published on Web 09/08/2005
Materials and Methods Chemicals and Instrumentation. CBZ, tert-butyl alcohol (tBuOH), and 1-butene-3-ol were all purchased from Sigma Aldrich (Deisenhofen, Germany) at >99% purity (1-butene3-ol 97%). Indigo blue (Riedel-de Hae¨n), 4-chlorobenzoic acid (Merck, Darmstadt, Germany), and all other chemicals (solvents, buffers, etc.) were of the highest grade commercially available. Water used for ozonation experiments and in making stock solutions was Milli-Q water (Millipore). Mass spectra and analytical measurements were generated using a Saturn ion trap GC-MS system (Varian GC 3400 coupled to a Varian Saturn 4D mass spectrometer) and on the PE Sciex API 365 and API 4000 LC-MS/MS instruments (PE Applied Biosystems, Langen, Germany). Samples of 1-(2benzaldehyde)-(1H,3H)-quinazoline-2,4-dione were measured by NMR on both 500 (1H NMR) and 700 MHz (1H and 13 C) Bruker Avance NMR instruments (Rheinstetten, Germany). Isolation of 1-(2-Benzaldehyde)-(1H,3H)-quinazoline2,4-dione (BQD). Isolation of BQD for NMR and IR studies and its use as a purified standard was accomplished by ozonating 1 L of Milli-Q water spiked with 15 mg of CBZ. The stock solution used for spiking was mixed in a 10 mL volumetric flask by dissolving 75 mg of CBZ with tert-butyl alcohol and ∼5% acetone. The t-BuOH concentration was made up to 30 mM. An approximate ozone dose of 400 µM was added and allowed to react for 20 min. Excess ozone was then purged with helium, and the sample was freeze-dried to yield the ozonation products. The freeze-dried product was then dissolved in approximately 2 mL of acetonitrile: water (50:50, v/v). To isolate BQD from the other oxidation products, fractions were collected from 300 µL injections of the dissolved material after separation on a LC-UV (λ ) 278) Discovery RP-Amidec 16 column (12.5 cm × 2.1 mm ID, 5 µm - Supelco, Germany). The LC mobile phase was isocratic using acetonitrile:water (50:50, v/v) and a flow rate of 1 mL/ min. Fractions were combined, and the solvent was removed by rotary evaporation followed by a slow nitrogen stream (T ) 50 °C). The white crystalline residue takes approximately 8 h to dry due to the aqueous content of the eluent. Ozonation of 15 mg of CBZ produced up to 5 mg of BQD with this method. Analytical Methods for Kinetics Experiments. The ozoneresistant probe compound para-chlorobenzoic acid (pCBA), used for the assessment of OH radical exposure, CBZ, and its ozonation products were separated on a 125/4 Nucleosil 100-5 C18 column (4 × 125 mm, 5 µm, Macherey-Nagel AG, Oensingen, Switzerland). Measurements were performed on an Agilent 1100 LC instrument equipped with a UV diode array detector (DAD). Sample injection volumes were 100 µL, and peak areas were integrated from the λ ) 240 nm signals. The LC-eluent was composed of acetonitrile and a 10 mM solution of phosphoric acid in Milli-Q water. The mobile phase gradient started at acetonitrile:water (40:60, v/v) and had the organic phase increased to 70% after 1 min. For the kinetics experiments, ozone stock solutions were produced by sparging ozone containing oxygen into an ice bath cooled Millipore water (∼0 °C) (18). The concentration of the resulting ozone stock solution (∼1.2 mM ozone) was measured directly by UV spectrophotometry at 258 nm using an (O3) ) 3000 M-1 cm-1. Appropriate volumes of the ozone stock solutions, as well as the analyte and pCBA, were added to the reaction vessels to achieve the desired concentrations. The decreases of ozone, the analyte, and pCBA were measured in samples, delivered from a dispenser system (18, 19). All experiments were performed at 20 °C. Determining Rate Constants, kO3, for BQM and BQD Reactions with Ozone. Pure solutions of the oxidation product BQM were produced by a stoichiometric addition
of ozone to CBZ (1 mol:1 mol). A stoichiometric ratio gave high yields of BQM, because its formation is much faster than its degradation. To measure direct ozone reactions, CBZ was dissolved in Milli-Q water (c ) 4.2 µM), which contained t-BuOH (30 mM), and transferred into 500 mL glass bottles to which a dispenser could be attached. A phosphate solution (10 mM, pH ) 6) served as the buffering agent. One minute after addition of ozone (4.2 µM), ozone residual was purged by helium. The absence of CBZ and the product BQD was verified by LC-UV. The BQM solution produced with this procedure was used to determine the second-order rate constant kO3,BQM. A large excess of ozone was then added (50 mM) to work under pseudo first-order conditions. At chosen time intervals, two aliquots of 5 mL were dispensed into 10 mL vials containing two different quenching reagents. The first vial contained an indigo solution to determine the aqueous ozone concentrations (20). The disappearance of BQM was measured after quenching the second aliquot with 1-buten-3-ol (18). The presence of this quenching reagent does not interfere with the LC-UV analysis of BQM. Previous attempts at quenching with sodium thiosulfate altered the decline in BQM concentrations, likely through reduction of the intermediate hydroperoxides. Data were evaluated by plotting the natural logarithm of the relative concentration of BQM (c/c0) versus ozone exposure (ozone concentration integrated over reaction time) (15). The rate constant is extracted from the slope of the straight line according to eq 1:
ln
∫
c ) -k [O3] dt c0
(1)
The same approach was used for the determination of the rate constant for BQD with ozone with the exception that a standard solution was made from the isolated product. Determination of Rate Constants, kOH, for the Reaction of OH Radicals with BQM and BQD. Competition kinetics were used for the determination of rate constants for BQM and BQD using the reference compound pCBA (kOH,pCBA ) 5 × 109 M-1 s-1) (21). To induce reactions with OH radicals, the pH was raised to pH ) 8 using a phosphate buffer (5 mM), and H2O2 was added at a molar ratio of 1:2 for H2O2: ozone. Variable concentrations of ozone were added to the reaction vessels; pCBA (cinitial ) 156 µg/L) and either BQD or BQM (cinit ) 266 and ∼250 µg/L, respectively) were measured by LC-UV. The reactions were halted before LC-UV analysis by purging the solutions for 30 s with helium. A double logarithmic plot of the pCBA concentration versus the BQD or BQM concentration gives a straight line with a slope equivalent to the ratio of kOH,BQM/kOH,pCBA or kOH,BQD/kOH,pCBA. Reaction of Neutral Pharmaceuticals with Ozone in Rhine River Water. Water samples were collected from a drinking water facility close to Frankfurt, Germany, which receives its raw water from the River Rhine. Samples were taken after partial treatment, which consisted of an aeration step (cascade), a sedimentation period in settling ponds, a flocculation step with FeCl3, followed by a rapid sand filtration. Samples of 1 L were spiked with 500 ng of a mixture of neutral pharmaceuticals dissolved in Milli-Q water and 10% (v/v) acetone. Use of a small volume of acetone provided sufficient dissolution of analytes with as little influence on the ozonation process as possible. Samples were ozonated (n ) 3 for each dose) with a lab-scale ozone generator using a method reported elsewhere (7). Ozone doses were in the range of 4.2-43.7 µM, and reactions were stopped after 20 min using a solution of sodium thiosulfate (0.014 M). Unreacted neutral drugs could then be measured by electrospray ionization ESI (positive ion mode) LC-MS/MS after solid-phase extraction at pH ) 7 (22). Quantification of BQM and BQD in Lake Zurich Water. To confirm the BQM and BQD formation as a consequence VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8015
TABLE 1. Structures of Carbamazepine and Its Three Ozonation Products Identified and Their Respective UV Absorbance Data (λmax and Local Maxima)
FIGURE 1. Schematic representation of three common fragmentation processes observed in the GC-MS spectrum of BQD. of the ozonation of CBZ in a natural water system, 1 µM CBZ was spiked into filtered Lake Zurich water (pH ) 7.3, alkalinity ) 2.61 mM, DOC ) 1.6 mg/L), and the products were quantified. Calibration curves were created for the two oxidation products using spiked Lake Zurich water and ozonated samples measured by LC-UV. Concentrations were quantified using the wavelength λ ) 240 nm, and the DAD spectra and retention times confirmed analyte identity. Identification of BQM and BQD in Water of a German Waterworks. A German waterworks that treats surface water using pre-ozonation (ozone dose, 0.7-1.0 mg/L; contact time, ca. 3 min; pH 8.0; TOC, 2.2 mg/L; alkalinity, 2.0 mM), flocculation with Fe(III)Cl3 (5 mg/L), and main ozonation (ozone dose, 1.0-1.5 mg/L; contact time, ca. 10 min; pH 7.2; TOC, 1 mg/L) was sampled after each of the two ozonation steps. Na2EDTA (1 g) was added to a 1 L water sample to facilitate the dissolution of the freeze-dried residues in the phosphate buffer (see below). The water was frozen in several portions in a freezing bath and was then freeze-dried within 3 days. Eventually the solid residues were combined and dissolved in 1 mL phosphate buffer (pH 7, 20 mM KH2PO4/ Na2HPO4) and spiked with the internal standard 10,11dihydrocarbamazepine (DHC) (Alltech, USA). The sample extract (50 µL) was injected onto a 100 × 4.6 mm Chromolith C18ec-column (Merck, Germany) and was analyzed using a binary gradient of 5 mM aqueous ammonium acetate and acetonitrile with a flow rate of 1 mL/min. The detection was conducted using electrospray ionization mass spectrometry in the positive mode at 750 °C with an API 4000 (Applied Biosystems, Germany). LC tandem MS conditions: total run time 17 min, ion spray 5500 V, declustering potentials 66-81 V, collision energy 27-61 eV, collision cell exit potential 128016
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005
16 V. A standard was prepared by adding ozone in a stoichiometric quantity to a 0.10 µmol solution of CBZ in Milli-Q water containing 30 mM t-BuOH. MRM transitions were BQM: m/z ) 250.9 into m/z ) 180.0, 207.9; BQD: m/z ) 267.0 into m/z ) 196.0, 167.0; DHC: m/z ) 239.0 into m/z ) 194.1, 181.1.
Results and Discussion Molecular Weight Determination and Mass Fragmentation of Ozonation Products. The reaction of ozone with CBZ in clean water (Milli-Q) in the presence of the OH radical scavenger t-BuOH produced three primary products visible with LC-UV (λ ) 278 nm, 240 nm), BQM, BQD, and BaQD (Table 1). All three oxidation products had their molecular weights determined using LC-MS/MS in both positive and negative ion mode. Measurement of the residues by LC-MS/MS revealed three major oxidation products showing molecular weights in excess of CBZ’s (m/z increases of 14, 30, and 46). This is indicative of mass to charge increases, equivalent to two amu short of the addition of one, two, and three oxygen atoms. The difference of two, short of consecutive oxygen increases, indicates removal of two protons in an elimination step such as ring closure or a double bond formation. BQD Mass Fragmentation. One oxidation product was sufficiently nonpolar and volatile to obtain a GC-MS spectrum. It was produced at 70 eV by electron impact ionization (EI) and is displayed in Figure 1. The prominent mass fragment peaks are 139, 167, 195 (base peak), 238, and 266 (M+‚). Predominant mass fragment losses are also shown. The fragmentation pattern of the product shows similarities
to fragmentations from three comparable quinazoline compounds. For example, GC-MS spectra of 3-methyl-(1H,3H)quinazoline-2,4-dione, a compound identified as a sex pheromone in the pale-brown chafer, a native Japanese scarab beetle (23), and caffeine (24), have major fragmentations via a two-bond cleavage on their heterocyclic rings. With both examples, the lost fragment is a methyldiformamide moiety (loss of 85 amu instead of the products’ loss of 71 amu, the difference being a methyl group on C-3) and yields important and relatively stable ions. The compound 1H,3H-quinazoline-2,4-dione from Sigma (Deisenhofen, Germany) was also tested for its fragmentation pattern by LC-MS/MS. Its primary losses were 17 (‚OH) and 73 amu, with minor losses of water, carbon monoxide, as well as 71 amu. The major fragmentations of these three comparative compounds suggest nitrogen and carbonyl-oxygen driven R-cleavages at the hetero-ring. Due to their fractionation patterns similar to those of the CBZ ozonation product, a quinazoline type structure was considered a possibility. Low mass ions in Figure 1 (i.e., m/z ) 105, 77, and 51) are part of the low mass, aromatic ion series that can stem from fragmenting benzoyl type ions (24). Additional GC-MS/MS experiments were performed on the ions m/z ) 238, 195, and 167 produced from the EI spectrum of a GC ion trap MS. These ions were trapped and then further fragmented to give their product ions (MS2). It was observed that further fragmentation of the trapped m/z 238 ion produced only the m/z 195 ion, fragmentation of m/z 195 produced only the m/z 167 ion, and fragmentation of m/z 167 produced the m/z 139 ion (100%) and m/z 140 ion (40%). These findings suggest that the mechanism for producing the major fragments of BQD follows a linear pathway and is not the result of many competing pathways. Collision induced dissociation (CID) of the M + H ion of the product produced by atmospheric pressure chemical ionization (APCI) LC-MS/MS (Applied Biosystems, API 365) exhibited all of the major fragments seen in its GC-MS spectrum. In addition, the (M + H)BQD CID promoted loss of water, m/z ) 249, a loss of 43 (HNdCdO) producing m/z ) 224, and loss of 61 for an ion at m/z ) 206. NMR Identification of 1-(2-Benzaldehyde)-(1H,3H)quinazoline-2,4-dione, BQD. The structure of the oxidation product with m/z ) 266, the second ozonation product of CBZ, could not be unambiguously elucidated from the MS data. Positive identification of BQD was assisted through 1H NMR, COSY, 13C NMR, and spin-echo experiments. A significant change to the aromatic hydrogens was immediately observed in the 1H NMR between CBZ and the product (Figure 1S, Supporting Information). The connectivity of the aromatic shifts is observed through vicinal couplings in the two-dimensional COSY experiment (Figure 2S, Supporting Information). The integrity of the two original aromatic rings of CBZ is obviously maintained after ozonation (2-ABCD systems). However, the presence of new functionality distributes the deshielded protons across a wider chemical shift range. The proton NMR spectra did not provide unequivocal evidence of the BQD structure. Further proof was required from a 13C NMR spectrum. Chemical shift assignments, produced from 13C signals of the product dissolved in methylene chloride-D2 and measured at 175 MHz, are labeled in Figure 2. An additional 13C spin-echo experiment was performed to identify the quaternary carbons from which five were observed and fit the proposed structure. The 13C chemical shifts were compared to molecules containing quinazoline and benzaldehyde moieties to provide further support for the proposed structure (26-28). The similarities in chemical shifts coincided extremely well and confirmed the hypothesized structure to be 1-(2-benzaldehyde)-(1H,3H)quinazoline-2,4-dione (BQD).
FIGURE 2. Suggested 13C NMR shifts of BQD (*assignments may be reversed). A pure sample of BQM could unfortunately not be isolated for 1H NMR measurement. Poor separation during the semiprep workup and perhaps the tendency of being photolabile could be the reason for impurities contaminating the BQM sample, one of which was BaQD. It was seen as a later eluting shoulder to the BQM peak under certain eluent concentrations on the LC column. Only on a smaller column using a phosphate buffer and gradient elution was BQM adequately separated. If traces of water are present in the sample, then BQM will also be in equilibrium with structure G in Scheme 1. Despite the contamination, the same pattern of welldistributed chemical shifts in the aromatic region of the proton NMR was still observed. As well, two far downfield hydrogen shifts were apparent, presumable the aldehyde hydrogen and the carboxylic hydrogen from BQM and BaQD, respectively. A singlet was also observed at δ ) 9.27 ppm. A hydrogen bonded to an imine carbon, such as those in unsaturated quinazoline rings, would produce a shift in this range. Therefore, it is likely that this peak is due to the hydrogen on the C-4 of BQM. BQM Mass Fragmentation. The fragmentation pattern of the first ozonation product was studied with LC-MS/MS. Collision induced dissociation of the (M + H) molecular ion of BQM showed strong ions at m/z ) 224 (11%), 208 (67%), 195 (11%), and 180 (base peak). Mass losses from the (M + H) ion were therefore observed to be 28, 43, 56, and 71 amu, respectively. Contrary to the 2,4-dione molecules mentioned above, the primary fragmentation of BQM is not a two-bond cleavage loss of 55 amu from the hetero-ring. Although this fragmentation was observed, it is weak and only approximately 10% of the m/z 180 ion. A combination of lost R cleavage potential resulting from the lack of a carbonyl oxygen at C-4, as well as what would be a more unstable leaving group (1-oxomethanimine), probably discounts this reaction pathway. The base peak is formed rather by a loss of 71 amu. The mechanism for the formation of this product ion is currently unknown. It does, however, include the loss of C-2 and N-3 as confirmed from an isotopically labeled BQM experiment (see below). As with BQD, the isotopically labeled fragmentation of (M + H)BQM by LC-MS/MS showed labeled losses with every major fragment ion except the first loss of 28 amu. Because BQM has no carbonyl group at C-4, the loss of 28 amu has to be either through the benzaldehyde or more likely a consecutive loss of hydrogen cyanide and a hydrogen radical. BaQD Mass Fragmentation. The fragmentation of BaQD was very similar to that of BQD, as detected by positive mode LC-ESI-MS/MS. The exception was a more intensive molecular ion coupled with the classical losses typical of benzoic acid moieties (loss of OH and CO). IR Spectrum of BQD. An FT-IR spectrum was recorded for BQD (Figure 3S, Supporting Information). Noteworthy are the three carbonyl vibrations at 1668, 1627, and 1617 cm-1 denoting the 3 CdO groups in BQD. An N-H stretch is seen at 3199 cm-1 with its bend and the C-N vibrations in the 1530-1100 cm-1 range. An FT-IR spectrum was also recorded for BQM, whereby a wide and intense signal at 1647 cm-1 was observed that would represent the CdN imine group and the carbonyl stretches. VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8017
SCHEME 1. Proposed Mechanism for the Formation of BQM (H) via Direct Ozone Attack of Carbamazepine
The proposed structures of all three oxidation products of CBZ, using the above structural information, are shown in Table 1. Isotopically Labeled CBZ. Fragment ion losses of oxidation products were further confirmed by the ozonation of an isotopic standard of CBZ (Campro Scientific GmbH, Berlin) containing a 13C- and 15N-labeled carboxamide group (MW ) 238.3 g/mol). Ozonation of this compound translates to having the labeled isotopes at the 13C-2 and 15N-3 positions in the quinazoline rings of the oxidation products (see Scheme 1, structure (H) for quinazoline ring numbering). In fact, when the isotopically labeled CBZ was ozonated, the expected M + H signals were observed. That is, oxidation products having ions m/z 252, 268, and 284 were present for isotopically labeled BQM, BQD, and BaQD, respectively. Product ion fragmentation patterns of the isotopically labeled BQD correlated well with the unlabeled BQD and gave some additional insight into the mechanism of fragmentation. In particular, the first loss of 28 amu seen in both the GC and the LC-MS/MS of BQD was also observed as a loss of 28 amu in the labeled oxidation product. In terms of BQD ion behavior, this indicates that the carbonyl group formed in the ozonation (becoming C-4 of the quinazoline ring) is lost first and not the C-1 carbonyl moiety. The labeled (M + H)BQD ion showed the same loss of water that BQD did. However, loss of 73 amu in the isotopically labeled ozonation product instead of 71 amu confirmed the two-bond cleavage of the heterocyclic ring to form the base peak. As expected, all other major fragments produced resulted in 13C- and 15N-labeled losses ((M + H) - 45, 63, and 102 amu). Mechanism for Production of Quinazoline Moiety from the Ozonation of Carbamazepine. The mechanism of ring closure to a quinazoline moiety can be explained via ozone attack at the double bond of CBZ. The widely accepted Criegee mechanism (reaction steps 1-5) of Scheme 1, if applied to CBZ, would result in the production of two new aldehyde groups after the final loss of hydrogen peroxide (structure E). A more detailed description of the Criegee mechanism is given in the literature (29, 30). Steps 6-8 of Scheme 1 rationalize the ring closure by a suspected intramolecular 8018
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005
attack by nitrogen on the aldehyde. It is well documented in organic chemistry that ammonia or primary amines can attack aldehydes or ketones to form imines (31). In this case, the hydrogen-bearing nitrogen from the urea group (E) attacks the aldehyde to eventually form a hemi-aminal moiety in G. Support for this structure was observed during electrospray LC-MS/MS measurement when an ion at m/z ) 268 was produced from a 50:50 water:methanol solution, which contained the first oxidation product of CBZ. The mechanism for the loss of water from the hemi-aminal moiety is initiated by protonation of the hydroxyl group. Dehydration can then occur followed by deprotonation of the intermediate iminium ion to form H (BQM). The BQM accumulates first until a stoichiometric amount (1:1) of ozone is added to CBZ. Determination of Rate Constants for the Reaction of BQM and BQD with Ozone and OH Radicals. The secondorder rate constants for the reactions of ozone and OH radicals with CBZ and several other pharmaceuticals have been determined previously (13). Carbamazepine’s rate constants are shown in Table 2, together with the rate constants determined for its ozonation products BQM and BQD. The proposed product formation is shown in Scheme 2. The oxidation product BQM reacts with ozone with an overall second-order rate constant of kO3 ) ∼7 M-1 s-1 (pH ) 6.0, T ) 20 °C) (reactions 2 and 3 in Table 2). Accumulation of BQD occurs when CBZ is completely consumed, and its oxidation kinetics are governed by a measured second-order rate constant of kO3 ) ∼1 M-1 s-1 (pH ) 6.0, T ) 20 °C) (reaction 4 in Table 2). A slow reaction of BQD with ozone is in agreement with the rate constant for the reaction of ozone with benzaldehyde (kO3 ) 2.5 ( 0.5; pH ) 1.7) (19). The reaction of CBZ (3.6 µM) in Milli-Q water (10 mM phosphate buffered, pH ) 6.0) in the presence of tert-butyl alcohol (30 mM) with an excess initial ozone dose of 50 µM shows that BQD increases with the ozone exposure in parallel to a decrease of BQM. However, it is only formed as a fraction of 40% of the oxidized BQM (Figure 3). Therefore, the corresponding rate constant for the reaction with ozone (reaction 2, Table 2) accounts for 40% of the overall secondorder rate constant for the reaction of BQM with ozone.
SCHEME 2. Proposed Reaction Pathways for the Oxidation of Carbamazepine with Ozone and OH Radicalsa
a
The numbers of the reactions refer to the numbers in Table 2.
TABLE 2. Rate Constants for Carbamazepine and Two Oxidation Products BQM and BQD for the Reactions with Ozone and OH Radicals number
reaction
kOH (109 M-1 s-1) pH 8 kO3 (M-1 s-1)
1 2
CBZ + O3 f BQM BQM + O3 f BQD
3× ∼3
3
BQM + O3 f Prod1
∼4
4 5 6
BQD + O3 f BaQD CBZ + OH f Prod2 BQM + OH f BQD
1
7
BQM + OH f Prod3
8
BQD + OH f Prod4
105
source
ref 12 this study Figure 6 this study Figure 6 this study 8.8 ( 1.2 ref 12 ∼2 this study Figure 7 ∼5 this study Figure 7 5 this study
Furthermore, a third oxidation product, BaQD, is formed. The rate constants kOH for the reaction of OH radicals with BQM and BQD were determined to be ∼6.8 × 109 and
FIGURE 3. Product evolution of three oxidation products after ozonation of 3.6 µM carbamazepine in Milli-Q water containing [tert-butyl alcohol] ) 30 mM. The initial ozone dose was 50 µM. Symbols represent concentrations, and the lines represent the peak areas with UV detection at λ ) 240 nm. For BaQD, no concentrations could be determined, because an appropriate reference compound could not be isolated.
∼5.1 × 109 M-1 s-1, respectively (pH ) 8). Therefore, despite relatively slow reactions with ozone, BQM and BQD can be quickly oxidized during ozonation by OH radicals or even more effectively in advanced oxidation processes. The rate constants for BaQD were not determined as it was not possible to isolate it in appreciable quantities and no commercial standard was available. Based on the kO3 value determined for BQD, the reaction of ozone with BaQD is expected to be quite low (
