We studied the nature and antimicrobial activity of ozonolysis transformation products (OTPs) of levofloxacin (LEV), a frequently detected fluoroquinolone antimicrobial in environmental waters. Two bioassays, the Kirby–Bauer test and the broth microdilution assay, were used to measure changes in the antimicrobial activity of solutions at low LEV to O3 molar ratios (2:1, 2:3 and 1:3) compared to solutions without added O3 (LEV:O3 1:0). The Kirby–Bauer test was not sensitive enough to detect significant differences in the growth inhibition zones in samples LEV:O3 2:1 and LEV:O3 1:0; however, the broth microdilution assay showed that bacterial growth inhibition was significantly lower (P < 0.001) in the solutions exposed to O3. Loss of antimicrobial activity in LEV:O3 2:1 solutions of (48 ± 16)% was in agreement with the concentration decrease of LEV of (36 ± 3)% in those same samples. A method of identification of OTPs using XCMS Online was applied to LEV:O3 2:1 and 1:0 samples and indicated the presence of an OTP of LEV of formula C18H20O5N3F, which was identified as LEV-N-oxide. The molecular structure of this compound was partially confirmed by tandem mass spectrometry experiments. This study showed that even at sub-optimal ozone doses, OTPs of higher antimicrobial activity than LEV were not formed.

INTRODUCTION

After almost two decades of research on the occurrence of pharmaceutical active compounds (PhACs) in the environment, it has become clear that these substances originate mainly from excreta of individuals taking medication. Consequently, they are almost ubiquitous at nanogram-per-litre concentrations in urban wastewaters (Monteiro & Boxall 2010). Detrimental effects on aquatic biota have been observed at concentrations between 5 and 100 ng L−1 for a few compounds, e.g. ibuprofen and fluoxetine (De Lange et al. 2006) and 17α-ethinylestradiol (Kidd et al. 2007).

Elimination of PhACs using more effective treatment processes in wastewater treatment plants is one of the most investigated approaches to mitigate this problem. The removal of PhACs by ozone (O3) has been proposed as a viable wastewater treatment since elimination rates are >90% for many PhACs using O3 doses around 2–10 mg L−1 (Joss et al. 2006). However, oxidation of PhACs using those O3 doses not result in the complete mineralization of these compounds (Huber et al. 2003). To date, ozonolysis transformation products (OTPs) of many PhACs have been identified (Ikehata et al. 2006); however, there are far fewer studies on their residual toxicity (Yargeau et al. 2008; Radjenović et al. 2009; Larcher et al. 2012).

Among PhACs, anti-infectives (more commonly known as antibiotics), have received a lot of attention from the scientific community since their occurrence in surface waters and sewage is of concern because of their potential role in the dissemination of antibiotic resistance in bacteria and other harmful effects on non-target species such as algae and shellfish (Binelli et al. 2009; Segura et al. 2009). OTPs of antibiotics could conserve their antimicrobial activity if their pharmacophore, i.e. key functional groups for the biological activity of the substance, are not affected by ozonolysis (Boxall et al. 2004). A study showed that for several types of antibiotics (except some β-lactams) residual antimicrobial activity decreases with increasing doses of O3(aq), which suggested that the formation of toxic OTPs of antibiotics are rarely formed (Dodd et al. 2009). However, a more recent paper showed that OTPs of levofloxacin (LEV) were toxic to the bioluminescent bacterium Vibrio fischeri (El Najjar et al. 2013).

The objective of this study was to identify OTPs formed at low LEV to O3 molar ratios (2:1, 2:3 and 1:3) equivalent respectively to 0.4, 1.4 and 2.8 mg L−1 of ozone, and measure their antimicrobial activity. These conditions of ozonolysis represent situations that could be observed in treatment of water containing high amounts of dissolved organic matter, which would decrease the effective concentration of O3 available to react with antibiotics or other PhACs. Ozonolysis experiments with illicit drugs performed by Rodayan et al. (2013) have shown that the rate of removal of organic micropollutants is affected by the composition of the matrix. Therefore, under conditions of low dose of O3, the formation of biologically active OTPs is more likely since the pharmacophore of LEV could be preserved due to the limited availability of ozone to further react with the transformation products formed.

METHODS

Ozonolysis of levofloxacin-containing water

An Ozomax OZO4VTT O3 generator (Granby, QC, Canada) was used to produce gaseous O3in situ from pure (99.5%) compressed O2(g). An in-house 2 L acrylic reactor (height: 60 cm; diameter: 7 cm) was used to prepare an aqueous O3 stock solution (7.1 mg L−1). Levofloxacin with a purity of 99.4% was purchased from Sigma-Aldrich Canada (Oakville, ON, Canada). Deionized water (di-H2O) was produced by a Milli-Q system from Millipore (Billerica, MA, USA). A stock solution of 29.7 mg L−1 of LEV was prepared in di-H2O (initial pH value of 7.91 ± 0.05). Complete dissolution of the antibiotic was achieved by placing the solution in an Branson Ultrasonics model 2510 ultrasonic bath (frequency: 40 kHz, power: 80 W) for 30 min. In these conditions, sonochemical composition of LEV was not observed (Figure S1, supplementary material, available online at http://www.iwaponline.com/wst/072/369.pdf). Working solutions of 15.4 mg L−1 of LEV in di-H2O were prepared before each experiment from the LEV stock solution. While the concentration used for the ozonolysis experiments is several orders of magnitude higher than environmental concentrations, it was necessary to use such concentration in order to facilitate the identification of OTPs without a preconcentration step prior to analysis. Volumes of 15 mL of the working solution were placed in 40 mL amber vials with a silicone septum to allow for addition of O3(aq). A fresh O3(aq) stock solution was prepared before each experiment by introducing O3(g) from the O3 generator into the reactor at a flow rate of 15 standard cubic feet per minute. The concentration of O3(aq) was determined by measuring the absorbance at 260 nm and a molar absorptivity value of 3,300 M−1 cm−1. Test solutions of LEV were prepared by adding different volumes of the O3(aq) stock solution (2, 7 and 14 mL) using gas-tight syringes to obtain approximately molar ratios of LEV to O3 of 2:1, 2:3 and 1:3, respectively. Therefore, applied O3 doses were in the range 0.4 to 2.8 mg L−1. O3 blanks (molar ratio of LEV to O3 of 1:0) were also prepared during each experiment. After addition of O3(aq), samples were agitated with a vortex mixer for a few seconds and their volume was brought to 35 mL by addition of di-H2O. This last step was necessary to obtain the same volume for all samples. Samples were then uncapped and stirred for 5 min to remove all residual O3. All ozonolysis experiments were done in triplicate and were carried out at 22 °C.

Quantification of levofloxacin

The residual concentration of LEV in the test solutions was determined on the same day that the ozonolysis experiments were done, by liquid chromatography–high resolution mass spectrometry. This system was composed of an Accela ultra-high-performance liquid chromatograph (UHPLC) coupled with a linear ion trap–orbital trap mass spectrometer (LIT-OrbitrapMS) model LTQ Orbitrap XL, both manufactured by Thermo Fisher Scientific (Waltham, MA, USA). The protocol of analysis is described in detail elsewhere (Rodayan et al. 2013). A five-point calibration curve injected in triplicate was used for quantification of levofloxacin.

Kirby–Bauer disk diffusion susceptibility test

The Kirby–Bauer assay was performed to monitor antimicrobial activity of LEV test solutions. The protocol used was based on Hudzicki (2012) with two minor modifications. Escherichia coli strain ATCC 11303 was employed as a replacement for E. coli strain ATCC 25922 and a hole-puncher was used to make paper disks (diameter: 6.0 ± 0.5 mm) using Express Plus filter paper of a pore size of 0.22 μm (Millipore). Aseptic techniques (sterile work area, sterile handling, etc.) were used throughout all bioassay experiments to ensure that bacterial cultures were free of contamination. The disks were soaked in the LEV test solutions described previously using tweezers. Three disks were placed per Mueller-Hinton (MH) agar plate previously inoculated with E. coli. The plates were then incubated at 37 °C and the growth inhibition zone diameters were measured to the nearest millimetre after 24 h of incubation. The magnitude of the diameter of the growth inhibition zone provides a semi-quantitative method of analyzing the residual antibacterial activity after exposure of antibiotics to ozone (Nasuhoglu et al. 2012). In order to determine which diameters of the growth inhibition zones were significantly different, an analysis of variance test followed by Tamhane's T2 post hoc test using SPSS statistics 22 (IBM, Armonk, NY, USA) was performed.

Broth microdilution assay

The protocol described by Dodd et al. (2009) was used to measure the antimicrobial activity of LEV test solutions. Minor modifications to the original protocol were applied to adapt it to the material and instruments available in the laboratory: (i) Escherichia coli strain ATCC 11303 was employed as a replacement for E. coli strain ATCC 25922; (ii) the incubation time of the plates was extended to 20 h instead of 8 h, to obtain sufficient growth in the agar; and (iii) antimicrobial activity was evaluated measuring the absorbance of the microplate wells at 625 nm, and the data obtained was normalized to the maximum absorbance value and plotted as a function of the LEV concentration. The obtained dose–response curve was fitted using the logistic non-linear fit of Origin 7.0 (OriginLab Corp., Northampton, MA, USA). Optimization of the growth medium is described in the supplementary material (online at http://www.iwaponline.com/wst/072/369.pdf).

This assay was used to obtain a more quantitative measure of the subtle differences in the antimicrobial activity of LEV test solutions. For this reason, data from two different ozonolysis experiments performed on two different days (total of six samples and 120 dilution points per test solution) were pooled for analysis. To facilitate the comparison of the broth microdilution assay experiments, normalized growth inhibition was plotted as a function of the dilution factor, instead of LEV concentration, of each solution. Therefore the median effective dilution factor (EDF50) was used as a parameter for comparison. The EDF50 represents the dilution factor necessary to inhibit bacterial growth by 50%. Therefore, substances with low antimicrobial activity have higher EDF50 values than substances with high antimicrobial activity, since the former have to be less diluted than the latter to produce the same inhibitory effect. Student's t-test and an F-test were used to evaluate differences between samples at the 95% confidence level.

Method validation: Identification of transformation products

In order to identify the main OTPs of LEV generated after addition low doses of O3, a method using differential analysis of chromatograms obtained by high resolution mass spectrometry was developed and validated using carbamazepine (CBZ), an anticonvulsant and mood regulator frequently detected in the environment. CBZ was used to validate the identification method since its OTPs are known and have been studied extensively (Hübner et al. 2014). Triplicate samples with a molar ratio CBZ: O3 1:0 (2.3 mg L−1 of CBZ and no ozone added) and 1:4 (same CBZ concentration and 1.8 mg L−1 of O3) were analysed in a UHPLC-QqTOFMS system composed of a Nexera UHPLC manufactured by Shimadzu (Kyoto, Japan) coupled to a Maxis quadrupole-time-of-flight mass spectrometer made by Bruker (Billerica, MA, USA). Ionization was performed by ESI+ and the mass analyzer was used in the full scan mode (m/z 50–800) with a resolution power, measured at full width half-maximum, of 19,725 at m/z 237.1025. Chromatographic separation was done with an Acquity UPLC HSS-T3 column (2.1 × 50 mm, 1.8 μm) at a flow rate of 0.5 mL min−1 and a column temperature of 30 °C. Solvent A (0.1% formic acid in H2O) and solvent B (0.1% formic acid in acetonitrile) were used with the following mobile phase gradient: 0 min (10% B), 1 min (10% B), 7 min (35% B), 9 min (100% B), 12 min (100% B), 13 min (10% B), 16 min (10% B). The ionization conditions were: capillary voltage of 3,000 V, dry gas at 10 L min−1, dry temperature of 200 °C. Nitrogen (99%) was used in the source to assist nebulization and in the collision cell as collision gas.

UHPLC-QqTOFMS data files were uploaded to the web-based metabolomics data processing platform XCMS Online, freely accessible at https://xcmsonline.scripps.edu. XCMS Online compares by differential analysis two sets of samples and identifies peaks that increase or decrease significantly by calculating P-values and fold changes (Tautenhahn et al. 2012). Dataset#1 (control) contained triplicate samples of CBZ:O3 1:0 and dataset#2 contained triplicate samples of CBZ:O3 1:4. The parameters used to perform the differential analysis in XCMS Online are indicated in the supplementary material, Table S1 (available online at http://www.iwaponline.com/wst/072/369.pdf). In order to classify as a potential OTP, a peak must: (i) be present in all ozonated samples and absent in non-ozonated samples; (ii) have P-values <0.05; ii) have fold changes >10; (iii) have a mean intensity >1% of the mean intensity of the parent compound signal in the non-ozonated samples; and (iv) elute at a percentage of organic solvent in the mobile phase <85%. This set of rules ensured that only peaks that appeared in ozonated samples with sufficient intensity and not present in the non-ozonated (control) samples were discriminated from the rest of the peaks detected in the samples. Also, our previous study (Segura et al. 2013) showed that peaks eluted at high percentages of organic solvent in the mobile phase are identified by differential analysis as potential OTPs but in fact they are not related to the parent compound.

To confirm the identification of OTPs of CBZ, the most intense peaks were further studied by tandem mass spectrometry. In order to find the best energy able to completely fragment the precursor ion, collision-induced dissociation (CID) experiments were performed in specific retention time ranges with increasing collision energies in the cell, from 10 to 30 eV.

Identification of ozonolysis transformation products of levofloxacin

The validated method was applied to the identification of OTPs of LEV. Samples were grouped in two different datasets: dataset#1 (control) contained triplicate samples of LEV:O3 1:0 while dataset#2 contained triplicate samples of LEV:O3 2:1. Data files generated after UHPLC-LIT-OrbitrapMS analysis in the raw format of Thermo Scientific were converted to the mzXML format using MSConvert application of the ProteoWizard toolkit (Chambers et al. 2012) and then uploaded to XCMS Online. In order to identify potential OTPs of LEV from the list of results, the set of rules described in the previous section was used to filter the data and reduce the number of false positives. Elemental formulas were assigned to the retained peaks using a mass tolerance ≤10 mmu, ring and double bond equivalents (RDBE) within ±4 units of the value for LEV and the nitrogen rule for even-electron ions. For element limits, only H and O atoms were allowed to be higher than the parent compound. For N and C, the maximum number of atoms was the same as LEV. In order to assist the structural elucidation of the potential OTPs, the most intense precursor ions were studied by tandem mass spectrometry experiments in the UHPLC-QqTOFMS system or by multi-stage tandem mass spectrometry (MSn) experiments performed by flow-injection analysis in a linear ion trap mass spectrometer; model LTQ-XL from Thermo Scientific.

RESULTS AND DISCUSSION

Ozonolysis of levofloxacin and antimicrobial activity measured with the Kirby–Bauer disk diffusion susceptibility test

Analysis of the test solutions by UHPLC-LIT-OrbitrapMS showed that the residual concentration of LEV rapidly decreased as a function of O3 dose (Figure 1). This was expected as LEV has a reaction rate constant of 6.0 ± 0.2 × 104 M−1 s−1 with O3 at pH 7.2 and 20 °C (El Najjar et al. 2013). Using a concentration of 0.4 mg L−1 of O3 (LEV:O3 molar ratio of 2:1), LEV initial concentration decreased by (36 ± 3)% while in the samples with the highest applied ozone dose (2.8 mg L−1, equivalent to a LEV:O3 molar ratio of 1:3), about (96 ± 2)% of the initial LEV concentration was transformed. Next, the antimicrobial activity of the test solutions was measured with the Kirby–Bauer test. In this bioassay, the ability of a compound to inhibit microbial growth is determined by the presence or absence of bacteria around the paper disks impregnated with the test substance, identified by a circular zone around the disk without bacterial growth. The magnitude of the activity is correlated to the diameter of the zone of growth inhibition.

Figure 1

Average inhibition zone diameters (n = 9) and residual LEV concentration of the ozonated solutions. Letters that are different indicate a significant difference (P < 0.05) between the inhibition zones of the respective applied ozone doses according to Tamhane's T2 post hoc test. Length of error bars indicates two standard deviations.

Figure 1

Average inhibition zone diameters (n = 9) and residual LEV concentration of the ozonated solutions. Letters that are different indicate a significant difference (P < 0.05) between the inhibition zones of the respective applied ozone doses according to Tamhane's T2 post hoc test. Length of error bars indicates two standard deviations.

Results of the Kirby–Bauer test (Figure 1) showed that ozonolysis of LEV was also accompanied by a diminution of its antimicrobial activity, as demonstrated by the lower average inhibition zone diameter, and reached a point where it was equal to a blank (about 7–8 mm). However, the trend did not correspond completely to the one observed for the residual LEV concentration. An analysis of variance followed by Tamhane's T2 post hoc test indicated that the difference between the average inhibition zone diameter for the test solutions LEV:O3 1:0 (no O3) and 2:1 (0.4 mg L−1 of O3) was not significantly different (P = 0.997). This result might be explained by either: i) OTPs created in the LEV:O3 2:1 sample had an antimicrobial activity similar to that of LEV or ii) the sensitivity of the Kirby–Bauer is insufficient to detect differences in the antimicrobial activity of those two test solutions. We decided to test the validity of those two hypotheses by using a more sensitive antimicrobial activity assay.

Determination of antimicrobial activity with the broth microdilution assay

Since the Kirby–Bauer test was not able to detect differences between the LEV test solutions with no O3 added and the solution in which 0.4 mg L−1 of O3 was added (LEV:O3 1:0 and 2:1, respectively), experiments were conducted with a more sensitive and quantitative bioassay, the broth microdilution assay. In this bioassay, antimicrobial activity is measured as a result of the diminution of the absorbance at 625 nm of a series of growth media solutions containing a microorganism and several dilutions of the substance to be tested for antimicrobial activity. If the test substance has an antimicrobial activity, bacterial growth will be increasingly inhibited in the wells and absorbance will decrease as the concentration of the test substance increases. Therefore, a dose–response (sigmoidal) curve will be observed when plotting normalized growth inhibition as a function of the test substance concentration or dilution factor. Results of the optimization of the growth medium for the broth microdilution assay are shown in the supplementary material (Figures S2, S3 and S4, online at http://www.iwaponline.com/wst/072/369.pdf). Based on the results obtained, the MH was identified as the growth medium providing the best testing conditions for this work. Logistic fit to the data obtained with the broth microdilution assay experiments with the LEV:O3 1:0 and LEV 2:1 solutions (Figure 2) showed that their EDF50 values were 0.0058 ± 0.0006 and 0.0086 ± 0.0006, respectively. The EDF50 values of both regressions indicated that ozonolysis reduced the antimicrobial activity of LEV:O3 2:1 solution since it had to be less diluted than the LEV:O3 1:0 solution to have the same effect (50% growth inhibition). Comparison of those two EDF50 values with unpaired Student's t-test (degrees of freedom = 232) indicated that they were significantly different with a P-value < 0.001 (Motulsky & Christopoulos 2004). Also we performed an f-test to determine if the LEV:O3 1:0 and LEV 2:1 datasets are significantly different. The calculated test value (f = 4.013) with degrees of freedom 4 and 232 for numerator and denominator, respectively, had a P-value of 0.0035. These results confirmed that the second hypothesis proposed earlier, i.e. that the Kirby–Bauer test is not sufficiently sensitive to detect differences in the antimicrobial activity of the two test solutions, was correct. Additionally, the percent loss of antimicrobial activity of the LEV:O3 2:1 solution is (48 ± 16)%, which is in agreement with the diminution of the concentration of LEV after addition of 0.4 mg L−1 of O3 ((36 ± 3)%). A disagreement between loss of antimicrobial activity and LEV concentration after ozonolysis would indicate the presence of OTPs active against microorganisms (higher activity than accounted for by the concentration of LEV) or capable of inhibiting antimicrobial activity (lower activity). The broth microdilution assay results demonstrated that ozonolysis of LEV, even at very low molar ratios or doses, forms OTPs without either property. However, the emergence of a different type of toxicity after ozonolysis of LEV is not ruled out and these results do not contradict the increased toxicity using bioluminescent bacteria previously reported by El Najjar et al. (2013).

Figure 2

Comparison of dose–response curves for LEV before ozonation (LEV:O3 1:0) and after ozonation (LEV:O3 2:1). Median effective dilution factors (EDF50) were 0.0058 ± 0.0006 and 0.0086 ± 0.0006, respectively.

Figure 2

Comparison of dose–response curves for LEV before ozonation (LEV:O3 1:0) and after ozonation (LEV:O3 2:1). Median effective dilution factors (EDF50) were 0.0058 ± 0.0006 and 0.0086 ± 0.0006, respectively.

Results of the validation of the identification method using carbamazepine

XCMS Online was originally developed to process untargeted metabolomic data, but its unique features such as free access, ease of use and wide range of statistical analyses make it an interesting software for the identification of transformation products of emerging contaminants. Results of the differential analysis performed with CBZ to validate our XCMS Online differential analysis method are shown in Table 1. The table shows that, out of 620 unregulated features that the software originally found, only 24 followed the rules specified in section ‘Method validation: identification of transformation products’.

Table 1

Results of the XCMS Online differential analysis

m/z valuea tRb (min) Fold change P-value Identity Normalized intensityc (%) 
273.0641 4.78 597 0.001 [BQM + Na]+ 48.1 
251.0821 4.78 592 0.002 [BQM + H]+ 84.0 
165.5318 4.78 273 0.001 N.I. 15.4 
274.0673 4.78 163 0.002 (M + 1) peak of [BQM + Na]+ 7.5 
291.0748 4.78 87 0.002 [2OH-CBZ + Na]+ 13.0 
292.0780 4.78 62 0.002 (M + 1) peak of [2OH-CBZ + Na]+ 2.1 
174.5370 4.78 48 0.002 N.I. 2.6 
269.0928 4.78 33 0.0008 [2OH-CBZ + H]+ 3.2 
289.0380 4.78 31 0.001 N.I. 1.5 
252.0852 4.78 29 0.002 (M + 1) peak of [BQM + H]+ 13.1 
335.0352 4.78 27 0.001 N.I. 1.5 
145.5198 4.78 23 0.002 N.I. 1.4 
166.0333 4.78 20 0.002 N.I. 2.8 
154.0234 4.78 19 0.002 N.I. 1.8 
173.5182 4.78 17 0.003 N.I. 0.8 
193.0125 4.78 15 0.002 N.I. 1.0 
157.5429 4.78 13 0.003 N.I. 0.9 
289.0591 5.38 17 0.04 [BQD + Na]+ 1.2 
173.5292 5.40 41 0.03 N.I. 2.4 
267.0771 5.40 40 0.0 [BQD + H]+ 3.9 
237.1025 6.61 −264 0.006 [CBZ + H]+ 0.4 
236.1544 8.03 20 0.00001 N.I. 4.4 
236.6560 8.03 21 0.001 N.I. 1.2 
248.0686 8.42 700 0.006 N.I. 29.7 
m/z valuea tRb (min) Fold change P-value Identity Normalized intensityc (%) 
273.0641 4.78 597 0.001 [BQM + Na]+ 48.1 
251.0821 4.78 592 0.002 [BQM + H]+ 84.0 
165.5318 4.78 273 0.001 N.I. 15.4 
274.0673 4.78 163 0.002 (M + 1) peak of [BQM + Na]+ 7.5 
291.0748 4.78 87 0.002 [2OH-CBZ + Na]+ 13.0 
292.0780 4.78 62 0.002 (M + 1) peak of [2OH-CBZ + Na]+ 2.1 
174.5370 4.78 48 0.002 N.I. 2.6 
269.0928 4.78 33 0.0008 [2OH-CBZ + H]+ 3.2 
289.0380 4.78 31 0.001 N.I. 1.5 
252.0852 4.78 29 0.002 (M + 1) peak of [BQM + H]+ 13.1 
335.0352 4.78 27 0.001 N.I. 1.5 
145.5198 4.78 23 0.002 N.I. 1.4 
166.0333 4.78 20 0.002 N.I. 2.8 
154.0234 4.78 19 0.002 N.I. 1.8 
173.5182 4.78 17 0.003 N.I. 0.8 
193.0125 4.78 15 0.002 N.I. 1.0 
157.5429 4.78 13 0.003 N.I. 0.9 
289.0591 5.38 17 0.04 [BQD + Na]+ 1.2 
173.5292 5.40 41 0.03 N.I. 2.4 
267.0771 5.40 40 0.0 [BQD + H]+ 3.9 
237.1025 6.61 −264 0.006 [CBZ + H]+ 0.4 
236.1544 8.03 20 0.00001 N.I. 4.4 
236.6560 8.03 21 0.001 N.I. 1.2 
248.0686 8.42 700 0.006 N.I. 29.7 

aMedian value; bMedian retention time; cPercentage of CBZ median peak height in CBZ:O3 1:0 samples. N.I.: ion not identified. Values for CBZ are only included to indicate differences in retention time, fold change and P-value.

In XCMS Online an ‘upregulated’ feature is a peak that is significantly more intense in the test than in the control samples (Tautenhahn et al. 2012). Similarly a ‘downregulated’ feature is a peak that has higher intensity in the control samples than in the test samples. The ion with the highest normalized intensity of the list was m/z 251.0821, which corresponds to 1-(2-benzaldehyde)-4-hydro-(1H,3H)-quinazoline-2-one also known as BQM (C15H10N2O2, Δm/z = 0.596 mmu). The presence of BQM was confirmed by the presence in Table 1 of several related ions: m/z 252.0852 (M + 1 isotopic peak, theoretical abundance = 17%, experimental abundance = 16%), m/z 273.0641 (sodium adduct, Δm/z = 0.651 mmu) and m/z 274.0673 (M + 1 isotopic peak of ion BQM + Na+, theoretical abundance = 0.17, experimental abundance = 0.16). Two other OTPs of CBZ identified by XCMS Online were m/z 269.0928 that corresponds to dihydroxy-carbamazepine (C15H12N2O3, Δm/z = 0.731 mmu) and m/z 267.0771 that could correspond to either 1-(2-benzaldehyde)-(1H,3H)-quinazoline-2,4-one (BQD) or to 1-(2-benzoic acid)-4-hydro-(1H,3H)-quinazoline-2-one also known as BaQM. These two OTPs are isomers (C15H10N2O3, Δm/z = 0.681 mmu) but BQD is the result of the attack of one O3 molecule on CBZ while BaQM results from the attack of two O3 molecules (Hübner et al. 2014).

A major OTP of CBZ that was not present in the samples was 1-(2-benzoic acid)-(1H,3H)-quinazoline-2,4-one (BaQD), which could be explained by the low amount of O3 used in the experiments compared to previous studies (Hübner et al. 2014). The absence of BaQD suggests that m/z 267.0771 is BQD rather than BaQM since both BaQD and BaQM are the result of two O3 attacks on the CBZ molecule. This was proven by high resolution–tandem mass spectrometry experiments (Table 2) which showed that a product ion of m/z 196.0755, generated after CID of m/z 267.0771, was observed by Hübner et al. (2014) only in the tandem mass spectrum of BQD. Additionally, neutral losses of H2O or CO2, typically observed when carboxylic acid groups are present as is the case of BaQM, were not observed in the mass spectrum of m/z 267.0771.

Table 2

High resolution–tandem mass spectrometry experiments performed on the precursor ions of CBZ and its OTPs

Precursor ion (m/zNeutral formula (Name) Product ionsa (m/z
237.1027 C15H12N2O (CBZ) 194.0970 (100%) 
192.0814 (19%) 
179.0735 (1%) 
251.0810 C15H10N2O2 (BQM) 223.0863 (4%) 
208.0752 (53%) 
195.0674 (12%) 
180.0805 (100%) 
267.1208 C15H10N203 (BQD) 239.0823 (5%) 
224.0709 (6%) 
206.0589 (3%) 
196.0755 (100%) 
Precursor ion (m/zNeutral formula (Name) Product ionsa (m/z
237.1027 C15H12N2O (CBZ) 194.0970 (100%) 
192.0814 (19%) 
179.0735 (1%) 
251.0810 C15H10N2O2 (BQM) 223.0863 (4%) 
208.0752 (53%) 
195.0674 (12%) 
180.0805 (100%) 
267.1208 C15H10N203 (BQD) 239.0823 (5%) 
224.0709 (6%) 
206.0589 (3%) 
196.0755 (100%) 

aValues in parenthesis represent the relative abundance. Collision energy in the collision cell was 30 eV.

While the developed identification method of OTPs still suffers from the presence of false positives, i.e. only 41.7% of the filtered ions were related to known OTPs of CBZ, the results represent an improvement from our previous method (Segura et al. 2013) in which only 13% of the potential OTPs were good candidates. In summary, the method of identification of OTPs by high resolution mass spectrometry and differential analysis with XCMS Online presented here is a simple and efficient way to identify potential transformation products of organic micropollutants. The method identified three of the major OTPs of CBZ: dihydroxy-carbamazepine, BQM and BQD. Other OTPs, reported elsewhere (Hübner et al. 2014), such as BaQD and BaQM were not identified and could be the result of different ozonolysis setups as well as low intensity.

Identification of ozonolysis transformation products of levofloxacin

The validated method of identification of OTPs was applied to the analysis of triplicate LEV:O3 1:0 (control) and LEV:O3 2:1 test solutions. Results obtained with XCMS Online showed the presence of 277 upregulated features. As expected, LEV was the main downregulated feature, with a fold change of −1.6 and P-value of 0.002. Application of the selection rules specified in section ‘Method validation: identification of transformation products’ reduced the number of upregulated features to only three (Table 3). Among those potential OTPs of LEV, one appears to be an isotopic peak: m/z 379.1481, M + 1 isotopic peak of m/z 378.1461 (theoretical abundance = 21%, experimental abundance = 20%). The most probable elemental formulas presented in Table 3 were then assigned to the two potential OTPs according to mass accuracy and RDBE and the nitrogen rule for even-electron ions.

Table 3

List of potential OTPs of LEV found using XCMS Online

Peaksa (m/ztRb (min) Fold change P-value Normalized intensityc (%) Possible neutral formulas Formula difference with LEV RDBE Δm/z (mmu) 
360.1366 6.71 20 0.001 1.5 C18H18O4N3−2 H 10.5 1.189 
362.1509 6.71 − 1.6 0.002 61.1 C18H20O4N3F (LEV) N.A. 9.5 − 0.211 
378.1461 6.79 198 0.013 23.7 C18H20O5N3+1 O 9.5 0.125 
379.1481 6.79 237 0.014 4.8 M + 1 peak of m/z 378.14606 N.A. N.A. 0.955 
Peaksa (m/ztRb (min) Fold change P-value Normalized intensityc (%) Possible neutral formulas Formula difference with LEV RDBE Δm/z (mmu) 
360.1366 6.71 20 0.001 1.5 C18H18O4N3−2 H 10.5 1.189 
362.1509 6.71 − 1.6 0.002 61.1 C18H20O4N3F (LEV) N.A. 9.5 − 0.211 
378.1461 6.79 198 0.013 23.7 C18H20O5N3+1 O 9.5 0.125 
379.1481 6.79 237 0.014 4.8 M + 1 peak of m/z 378.14606 N.A. N.A. 0.955 

aMedian value; bMedian retention time; cPercentage of LEV median peak height in LEV:O3 1:0 samples. Values for LEV are shown to indicate differences in mass and composition. N.A.: not applicable.

Out of these two potential OTPs, C18H20O5N3F has been previously reported and was identified as LEV-N-oxide (Figure 3) and is one of the major OTPs of LEV identified so far (De Witte et al. 2009). The other potential OTP of LEV, C18H18O4N3F, is puzzling since the observed formula difference with LEV, 2 H atoms, is difficult to interpret according to known O3 chemistry, and further analysis was performed and is discussed in the supplementary material (online at http://www.iwaponline.com/wst/072/369.pdf).

Figure 3

Molecular structure of LEV (left) and LEV-N-oxide (right), the major OTP identified in the LEV:O3 2:1 samples.

Figure 3

Molecular structure of LEV (left) and LEV-N-oxide (right), the major OTP identified in the LEV:O3 2:1 samples.

In order to confirm the structure of LEV-N-oxide (neutral formula C18H20O5N3F), tandem mass spectrometry experiments were performed in the UHPLC-QqTOFMS system. The product ion scan of m/z 378 showed three major ions (Table 4). For product ion m/z 361.1445, only one formula within 10 mmu of mass tolerance atom number equal or inferior to the number elements in LEV-N-oxide is possible: C18H20O4N3F•+m/z = 1.264 mmu). This indicates a loss of OH. This type of loss has been previously observed with N-oxides in ESI+ (Ma et al. 2005) and hints the presence of an O atom bonded to a N atom in the structure of this OTP. Molecular orbital calculations performed by De Witte et al. (2009) showed that the N′4-atom in piperazinyl functional group of LEV (Figure 3) is the most likely site of reaction towards an electrophilic attack by O3; therefore the O atom added after ozonolysis must be bonded to this nitrogen atom. As for the second product ion, m/z 317.1544, two possible formulas within the limits specified previously are possible: C17H20O2N3F•+ or C17H21O4N2+. They indicate losses of or CONF, respectively. Since the F atom in fluoroquinolones is usually lost after CID as HF (Volmer et al. 1997), the first loss is more likely and is probably produced by the combined loss of OH followed by CO2. A loss of CO2 indicates also the presence of a carboxylic acid functional group (Volmer et al. 1997). Finally, among the four formulas possible for the product ion m/z 254.1058, none is characteristic of a functional group of LEV-N-oxide. This product ion is probably formed after gas phase bond cleavages and rearrangement reactions that we did not try elucidate, since it was not in the scope of the present work.

Table 4

Analysis of the tandem mass spectrum of the precursor ion of LEV-N-oxide, m/z 378

Product ion (m/zPossible ion formulas and mass accuracy in mmu Neutral loss Functional groups characteristic of the loss 
361.1445 C18H20O4N3F•+m/z = 1.264 mmu) OH N-oxide 
317.1544 C17H20O2N3F•+m/z = 1.264 mmu)  (CO2 + OH) Carboxylic acid, N-oxide 
C17H21O4N2+m/z = 1.264 mmu) CONF 
254.1058 C15H14O2N2•+m/z = 1.264 mmu) C3H7FNO3 
C15H13N3F+m/z = 1.264 mmu) C3H8O5 
C17H15OF•+m/z = 1.264 mmu) CH6O4N3 
C16H13ONF+m/z = 1.264 mmu) C2H8O4N2 
Product ion (m/zPossible ion formulas and mass accuracy in mmu Neutral loss Functional groups characteristic of the loss 
361.1445 C18H20O4N3F•+m/z = 1.264 mmu) OH N-oxide 
317.1544 C17H20O2N3F•+m/z = 1.264 mmu)  (CO2 + OH) Carboxylic acid, N-oxide 
C17H21O4N2+m/z = 1.264 mmu) CONF 
254.1058 C15H14O2N2•+m/z = 1.264 mmu) C3H7FNO3 
C15H13N3F+m/z = 1.264 mmu) C3H8O5 
C17H15OF•+m/z = 1.264 mmu) CH6O4N3 
C16H13ONF+m/z = 1.264 mmu) C2H8O4N2 

The structure of LEV-N-oxide can be used to predict its antimicrobial activity based on the known relationship between the structure and the antimicrobial activity of quinolones. The mode of action of quinolones is based on the selective inhibition of bacterial DNA synthesis (Mitscher 2005). Inside the bacterial cell, quinolones induce the formation of a molecular ternary complex, composed of DNA, a type II topoisomerase (an enzyme essential for DNA functioning) and two quinolone molecules. This ternary complex, which disturbs basic bacterial DNA functions such as transcription, replication and reparation, is held together by multiple interaction sites between the quinolones, the enzyme and DNA (Mitscher 2005; Drlica et al. 2009). Therefore, slight modifications of the structure of LEV such as the formation of a new bond between oxygen and the N′4-atom, as in LEV-N-oxide, can disrupt the formation of the ternary complex and thus decrease antimicrobial activity, as observed with the bioassays.

CONCLUSION

This study demonstrated that LEV-N-oxide, the major OTP generated after addition of a low dose of O3 (molar ratio LEV:O3 2:1, equivalent to 6.7 mg L−1 of LEV and 0.4 mg L−1 of O3) does not possess the same antimicrobial activity as LEV. After optimization of the growth medium, the broth microdilution bioassay showed that ozonolysis of LEV with that amount of O3 resulted in a significantly lower (P < 0.001) antimicrobial activity compared to LEV samples not exposed to ozone (LEV:O3 1:0) as measured by the EDF50. According to the broth microdilution assay, percent loss of antimicrobial activity of the LEV:O3 2:1 solution was (48 ± 16)%, which is in agreement, considering experimental error, with the diminution of the concentration of LEV after ozonolysis of (36 ± 3)%, measured by UHPLC-LIT-OrbitrapMS.

The presence of LEV-N-oxide in LEV:O3 2:1 samples was detected using a validated and improved method of identification of transformation products based on the differential analysis of liquid chromatography–high resolution mass spectrometry data with XCMS Online. Tandem mass spectrometry experiments with the precursor ion of LEV-N-oxide revealed the presence of two functional groups, N-oxide and carboxylic acid, in its molecular structure that partially confirms the structure proposed by De Witte et al. (2009).

Ozonolysis of other quinolones even at low doses of O3 may give similar results in terms of reduction of antimicrobial activity given the effect that small changes in their molecular structure have on their antimicrobial activity. More studies are necessary to confirm the absence of biological activity of LEV-N-oxide and other OTPs, since a new mode of toxic action could arise as a consequence of the changes caused in the molecular structure of the parent compounds after reaction with O3.

ACKNOWLEDGEMENTS

This study was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) through a Discovery Grant awarded to Prof. Viviane Yargeau and a start-up fund attributed to Prof. Pedro A. Segura by the Faculty of Sciences of Université de Sherbrooke.

REFERENCES

REFERENCES
Binelli
A.
Parolini
M.
Cogni
D.
Pedriali
A.
Provini
A.
2009
A multi-biomarker assessment of the impact of the antibacterial trimethoprim on the non-target organism Zebra mussel (Dreissena polymorpha)
.
Comparative Biochemistry and Physiology Part C: Toxicology & Pharmacology
150
(
3
),
329
336
.
Boxall
A. B.
Sinclair
C. J.
Kolpin
D.
Maund
S. J.
2004
When synthetic chemicals degrade in the environment
.
Environmental Science & Technology
38
(
19
),
368A
375A
.
Chambers
M. C.
Maclean
B.
Burke
R.
Amodei
D.
Ruderman
D. L.
Neumann
S.
Gatto
L.
Fischer
B.
Pratt
B.
Egertson
J.
2012
A cross-platform toolkit for mass spectrometry and proteomics
.
Nature Biotechnology
30
(
10
),
918
920
.
De Lange
H.
Noordoven
W.
Murk
A.
Lürling
M.
Peeters
E.
2006
Behavioural responses of Gammarus pulex (Crustacea, Amphipoda) to low concentrations of pharmaceuticals
.
Aquatic Toxicology
78
(
3
),
209
216
.
De Witte
B.
Langenhove
H. V.
Hemelsoet
K.
Demeestere
K.
Wispelaere
P. D.
Van Speybroeck
V.
Dewulf
J.
2009
Levofloxacin ozonation in water: rate determining process parameters and reaction pathway elucidation
.
Chemosphere
76
(
5
),
683
689
.
Drlica
K.
Hiasa
H.
Kerns
R.
Malik
M.
Mustaev
A.
Zhao
X.
2009
Quinolones: action and resistance updated
.
Current Topics in Medicinal Chemistry
9
(
11
),
981
998
.
El Najjar
N. H.
Touffet
A.
Deborde
M.
Journel
R.
Leitner
N. K. V.
2013
Levofloxacin oxidation by ozone and hydroxyl radicals: Kinetic study, transformation products and toxicity
.
Chemosphere
93
(
4
),
604
611
.
Huber
M. M.
Canonica
S.
Park
G. Y.
Von Gunten
U.
2003
Oxidation of pharmaceuticals during ozonation and advanced oxidation processes
.
Environmental Science & Technology
37
(
5
),
1016
1024
.
Hudzicki
J.
2012
Kirby–Bauer Disk Diffusion Susceptibility Test Protocol. http://www.microbelibrary.org/component/resource/laboratory-test/3189-kirby-bauer-disk-diffusion-susceptibility-test-protocol (accessed 27 March 2013)
.
Ikehata
K.
Naghashkar
N. J.
El-Din
M. G.
2006
Degradation of aqueous pharmaceuticals by ozonation and advanced oxidation processes: A review
.
Ozone: Science & Engineering
28
(
6
),
353
414
.
Joss
A.
Carballa
M.
Kreuzinger
N.
Siegrist
H.
Zabczynski
S.
2006
Wastewater treatment
. In:
Human Pharmaceuticals, Hormones and Fragrances: The Challenge of Micropollutants in Urban Water Management
(
Ternes
T. A.
Joss
A.
, eds).
IWA Publishing
,
London
, pp.
243
292
.
Kidd
K. A.
Blanchfield
P. J.
Mills
K. H.
Palace
V. P.
Evans
R. E.
Lazorchak
J. M.
Flick
R. W.
2007
Collapse of a fish population after exposure to a synthetic estrogen
.
Proceedings of the National Academy of Sciences of the United States of America
104
(
21
),
8897
8901
.
Mitscher
L. A.
2005
Bacterial topoisomerase inhibitors: quinolone and pyridone antibacterial agents
.
Chemical Reviews
105
,
559
592
.
Monteiro
S.
Boxall
A.
2010
Occurrence and fate of human pharmaceuticals in the environment
.
Reviews of Environmental Contamination and Toxicology
202
,
53
154
.
Motulsky
H.
Christopoulos
A.
2004
Fitting Models to Biological Data Using Linear and Nonlinear Regression: A Practical Guide to Curve Fitting
.
Oxford University Press
,
New York, NY
.
Nasuhoglu
D.
Rodayan
A.
Berk
D.
Yargeau
V.
2012
Removal of the antibiotic levofloxacin (LEVO) in water by ozonation and TiO2 photocatalysis
.
Chemical Engineering Journal
189–190
,
41
48
.
Radjenović
J.
Godehardt
M.
Hein
A.
Farré
M.
Jekel
M.
Barceló
D.
2009
Evidencing generation of persistent ozonation products of antibiotics roxithromycin and trimethoprim
.
Environmental Science & Technology
43
(
17
),
6808
6815
.
Rodayan
A.
Segura
P. A.
Yargeau
V.
2013
Ozonation of wastewater: Removal and transformation products of drugs of abuse
.
Science of the Total Environment
487
,
763
770
.
Segura
P. A.
François
M.
Gagnon
C.
Sauvé
S.
2009
Review of the occurrence of anti-infectives in contaminated wastewaters and natural and drinking waters
.
Environmental Health Perspectives
117
(
5
),
675
684
.
Segura
P. A.
Kaplan
P.
Yargeau
V.
2013
Identification and structural elucidation of ozonation transformation products of estrone
.
Chemistry Central Journal
7
(
74
),
1
11
.
Tautenhahn
R.
Patti
G. J.
Rinehart
D.
Siuzdak
G.
2012
XCMS Online: a web-based platform to process untargeted metabolomic data
.
Analytical Chemistry
84
(
11
),
5035
5039
.
Yargeau
V.
Huot
J.
Rodayan
A.
Rouleau
L.
Roy
R.
Leask
R.
2008
Impact of degradation products of sulfamethoxazole on mammalian cultured cells
.
Environmental Toxicology
23
(
4
),
492
498
.

Supplementary data