Identification of Transformation Products for Benzotriazole, Triclosan, and Trimethoprim by Aerobic and Anoxic-Activated Sludge

All standards and reagents used were of the highest purity and were commercially available. Benzotriazole, triclosan, trimethoprim, L-arginine, methanol, and ammonium acetate were purchased from Millipore Sigma (Burlington, Massachusetts). The Q-TOF reference stock solution, hexakis (1H, 1H, 3H-tetrafluoropropoxy)–phosphazine (HP-0921), was purchased from Agilent Technologies (Santa Clara, California). Neat materials for transformation products were purchased from either Millipore Sigma or Toronto Research Chemicals (North York, Ontario, Canada), except for 5-methoxy benzotriazole, which was custom synthesized by Frinton Laboratories. Stock solutions of all neat materials were initially prepared in methanol and diluted accordingly. The 3-desmethyl trimethoprim transformation product stock solution was initially diluted in a 1:1 solution of methanol and reagent grade deionized (DI) water for better solubility. Deionized water was obtained using a Milli-Q Ultrapure Water Purification System (Millipore Sigma, Burlington, Massachusetts).

The biodegradability simulation test followed the OECD 314 B Biodegradation in Activated Sludge test protocol, though anoxic activated sludge was also evaluated in addition to aerobic activated sludge. Briefly, activated sludge was sourced from a 30 MGD BNR treatment plant located in Southern Nevada and shipped overnight on ice to the University of Cincinnati. This BNR treatment plant is designed to remove both nitrogen (N) and phosphorus (P), thus it included anaerobic, anoxic, and aerobic selector basins (anaerobic activated sludge was not collected). The BNR system is operated at a solids residence time (SRT) of eight days, and a hydraulic residence time (HRT) of 5.5 h. Within 2 h of receiving the samples in the laboratory, 8 L of activated sludge were placed in four separate stainless-steel batch reactors. Two of the four reactors were operated as biotic reactors (aerobic and anoxic, respectively) and the other two were corresponding abiotic reactors (biocide solution containing 10 mM each of ${\mathrm{NaN}}_3$, ${\mathrm{BaCl}}_2$, and ${\mathrm{NiCl}}_2$). No transformation intermediates were identified in the abiotic controls. All reactors were buffered with ${\mathrm{NaHCO}}_3$ to a final equivalence of 3 mM. A nutrient solution was spiked in the aerobic reactors (5 mg ${\mathrm{NH}}_3\text{-}\mathrm{N}/\mathrm{L}$ and $10\mathrm{mg}/\mathrm{L}{\mathrm{PO}}_4\text{-}\mathrm{P}/\mathrm{L}$) to assess nitrification and phosphorus uptake, and in the anoxic reactors ($30{\mathrm{NO}}_3\text{-}\mathrm{N}\mathrm{mg}/\mathrm{L}$ every 12 h) to ensure denitrification.

Results for conventional parameters (total suspended solids [TSS], volatile suspended solids [VSS], ammonia [${\mathrm{NH}}_3\text{-}\mathrm{N}$], nitrate [${\mathrm{NO}}_3\text{-}\mathrm{N}$], phosphate [${\mathrm{PO}}_4\text{-}\mathrm{P}$], pH, and soluble chemical oxidation demand [sCOD]) in samples from the biotransformation experiments are provided in Fig. S1(a–d) in the Supplemental Data. The temperature in all four bioreactors were maintained at $20°\mathrm{C}\pm 0.5°\mathrm{C}$. Buffer addition held the pH (Standard Method 4500-H B) close to 8 in both aerobic and anoxic biotic reactors though the pH dropped to 7 by the end of the seven-day experiment in the aerobic biotic reactor. Also, pH in both the abiotic control reactors decreased due to the presence of the biocide. This decrease in pH was due to the removal of alkalinity from solution via precipitation reactions with the biocide (as seen visually). Dissolved oxygen (DO) (Standard Method 4500-O G) in the aerobic reactors was maintained above $4\mathrm{mg}/\mathrm{L}$ by bubbling compressed air into the reactors. Anoxic reactors had DO values below $0.1\mathrm{mg}/\mathrm{L}$ by passing nitrogen gas through the mixed liquor. Total and volatile suspended solids (TSS/VSS) (Standard Methods 2540 D) were similar in both the aerobic and anoxic activated sludge with an average VSS of 1,086 and $\mathrm{1,015}\mathrm{mg}/\mathrm{L}$ in the aerobic and anoxic biotic reactors, respectively. Generally, the TSS and VSS values decreased slightly with time in all experiments, indicating a decay of biomass. There was a decrease in soluble COD (sCOD) concentration (Method 8000 COD, Hach Company) in the aerobic reactor indicating biomass activity. In the aerobic biotic reactor, about half the phosphate spike (10 mg ${\mathrm{PO}}_4\text{-}\mathrm{P}/\mathrm{L}$) (Standard Method 4110B) was taken up during the first 24 h, after which the P uptake levelled off. A significant increase in ${\mathrm{NO}}_3\text{-}\mathrm{N}$ (Standard Method 4110B) and decrease in ${\mathrm{NH}}_3\text{-}\mathrm{N}$ (below detection) (Standard Method 4500-NH3D) was seen in the aerobic biotic reactor indicating nitrification by nitrifying bacteria. In the anoxic biotic reactor, 30 mg ${\mathrm{NO}}_3\text{-}\mathrm{N}/\mathrm{L}$ was spiked at 12 h and 24 h to maintain anoxic conditions. Nitrate consumption was seen during the first 24 h, after which the concentration remained stable [Fig. S1(a–c)]. Overall, the trends in nutrients and other conventional parameters demonstrate that not only the relevant redox conditions were maintained, but also relevant microbial populations were active in the experiments.

The reactors were equilibrated for an hour prior to spiking in the target organic compounds. A stock spiking solution, containing a mixture of $80\mathrm{mg}/\mathrm{L}$ each of benzotriazole, triclosan, and trimethoprim, was prepared in a 10% methanol/ultrapure water solution. A concentration of $1\mathrm{mg}/\mathrm{L}$ for each compound was targeted for the initial reactor concentrations to allow for enough sensitivity to detect multiple transformation products during Q-TOF analysis. The experiment was run for seven days with samples collected from each reactor at specific time points (0, 0.25. 0.5, 1, 3, 5, and 7 day). The mixed liquor samples (125 mL) were filtered using a Whatman GF/C glass microfiber filter (1.5 μm pore size) and the filtrate was transferred to amber bottles containing a final equivalence of $1\mathrm{g}/\mathrm{L}$ ${\mathrm{NaN}}_3$ and $0.5\mathrm{g}/\mathrm{L}$ ascorbic acid as preservatives. A filtered activated sludge sample was also collected before spiking to provide a base line for parent compounds and transformation products. For all samples, a small aliquot was transferred to a 2 mL autosampler vial prior to instrumental analysis by direct injection.

All samples were analyzed for parent and transformation products using an HTC PAL autosampler and Agilent liquid chromatography (LC) binary pump system coupled to an Agilent 6520 quadrupole time-of-flight (Q-TOF) mass spectrometer with electrospray ionization (ESI) in both positive and negative modes. Initial LC methods were developed using the parent compounds (benzotriazole, triclosan, and trimethoprim) to allow for adequate separation and peak shape. An example chromatogram and mass error for the three parent compounds is provided in Fig. S2. Separation was performed using a Luna C18, 5 μm LC column ($150\times 4.6\mathrm{mm}$) and a gradient consisting of 5 mM ammonium acetate in DI water (A) and methanol (B) with a constant flow rate of $800\mu \mathrm{L}/\min$. A 50 μL injection volume was used for all analyses. The dual ESI source was set at 350°C, the drying gas was set at $10\mathrm{L}/\min$, and the nebulizer was set at 55 psig. A stock reference mass solution containing L-arginine and HP-0921 was prepared according to manufacturer specifications. Individual reference mass solutions were then made with final concentrations adjusted for optimal signal in each of the positive and negative ionization methods. Reference masses 121.050873 and $922.009798\mathrm{m}/\mathrm{z}$ were used for positive mode, and reference masses 119.03632 and $980.016375\mathrm{m}/\mathrm{z}$ were used for negative mode. The Q-TOF acquisition was operated in MS only with a mass range of $50–\mathrm{1,100}\mathrm{m}/\mathrm{z}$ for both positive and negative ionization methods. Due to the use of a defined Q-TOF database to rapidly scan and identify samples for unknowns, MS2 fragmentation was not employed.

The analytical method was initially evaluated using a variety of matrix spikes to determine expected sensitivity of parent compounds in DI water, wastewater effluent, and previous BNR sludge samples (liquid portion only), as well as their identification in the presence of these matrices as compared to standards in DI water. A 1-point calibration curve was created for the three parent compounds using a $200\mu \mathrm{g}/\mathrm{L}$ standard in DI water and a regression line forced through the origin. This allowed for a semi-quantitative estimate of the initial spiked concentrations of the parent compounds at time zero and permitted the monitoring of their degradation in the experimental samples. In addition, the calibration standard was analyzed periodically during sample analysis to act as a continuing calibration verification (CCV) standard to monitor the performance of the instrument. When expected concentrations of the CCV were low ($>30\%$ drop in signal), sample analysis was stopped for instrument cleaning and maintenance.

For targeted screening of predicted transformation products, a compound database was first created that included the compound name, molecular formula, and calculated exact mass. The database was populated by including transformation products reported in the literature and by using the EAWAG Pathway Prediction System software to identify potential biotransformation intermediates that were deemed suitable for LC-QTOF analysis. However, this prediction model was developed for aerobic transformations, so its applicability toward anoxic intermediates is potentially limited. The pathway chart for each parent compound is shown in Figs. S3–S5. Additional information, such as Chemical Abstract Service numbers and structures, was entered when available. The final database included 40 possible transformation products for the three parent compounds. A full list of the transformation products included in the compound database can be found in Fig. S6. International Union of Pure and Applied Chemistry (IUPAC) names were included for structures proposed by the predictive software, which does not return a name for each predicted intermediate. The compound database was used to screen experimental BNR samples using the MassHunter Qualitative software. The raw peak area for each identified transformation product was integrated for a qualitative comparison to observe increasing and decreasing trends between samples.

In addition to using the compound database for targeted screening of transformation products, select samples in which the most abundant transformation products were detected were also screened for unknowns using a molecular feature algorithm technique built into the MassHunter Qualitative software. This technique identified possible unknown compounds by identifying ions that were related through a rise and fall in chromatographic pattern, an isotopic distribution, and the presence of adducts as set by the software and within user-defined parameters, such as retention time windows, mass ranges, and possible adducts for both positive and negative ionization modes. This second screening method was performed to ensure that unexpected transformation products were not missed when using the compound database. Potential compounds identified by either qualitative screening method were required to have a mass difference less than 3 ppm. Note that the screening methods applied are limited only to compounds amenable to the described LC-QTOF methods. In addition, no attempts were made to assess or correct for matrix suppression.

Benzotriazole biotransformed more than 99% in aerobic activated sludge and close to 50% in anoxic activated sludge over the seven-day study [Fig. 1(a)]. Four transformation products were identified in both aerobic and anoxic conditions including two hydroxy products of benzotriazole (OH-BTA) and two methoxy products of benzotriazole (MeO-BTA) [Fig. 1(b and c)]. The hydroxylation and methylation pathways of BTA are shown in Fig. 2.

For the OH-BTA products (${\mathrm{C}}_6{\mathrm{H}}_5{\mathrm{N}}_3\mathrm{O}$—mass 135.043 amu), three structural isomers were possible (1-OH BTA, 4-OH BTA, and 5-OH BTA). Huntscha et al. (2014) observed two isomers of OH-BTA and identified them as 4-OH BTA and 5-OH BTA by using reference standards and noted that 1-OH BTA would not be expected. For this study, a 4-hydroxybenzotriazole standard was used and its retention time and mass spectrum matched the second peak observed in the BNR sample chromatograms, indicating the first peak was the 5-OH BTA isomer [Fig. 3(a)]. Peak assignment and retention time difference agreed with isomers identified by Huntscha et al. (2014). In the case of the third possible OH-BTA isomer, 1-OH-BTA, a third peak was not observed and thus this isomer was not expected to be formed (Huntscha et al. 2014).

Under aerobic conditions, both hydroxy benzotriazole biotransformation products increased rapidly during the first day of the experiment. After the first day, they started to decrease, dropping to background levels by the end of the seventh day [Fig. 1(b)]. These trends under aerobic conditions were also observed by Huntscha et al. (2014), who measured higher concentrations of 5-OH BTA when compared to 4-OH BTA, despite the aerobic sludge being from different sources.

Under anoxic conditions, there was a lag phase for the primary biodegradation of benzotriazole [Fig. 1(b)]. Formation of the OH-BTAs started after the first day and continued to increase until approximately the fifth day of the experiment. This lag in degradation of benzotriazole under anoxic conditions suggests that the microbes/enzymes responsible were not available in required amounts for immediate primary degradation, though they were produced during the lag phase. Typical hydraulic retention times in wastewater treatment plants are less than one day and thus these intermediates are likely to pass from the anoxic unit process to the aerobic unit process where they would further degrade. Similar to the aerobic results, 5-OH BTA form at a higher concentration when compared to 4-OH BTA under anoxic conditions.

The MeO-BTA transformation products correspond to the two isomers 4-MeO BTA and 5-MeO BTA (${\mathrm{C}}_7{\mathrm{H}}_7{\mathrm{N}}_3\mathrm{O}$—mass 149.059 amu). A 5-MeO BTA standard was used for peak identification and isomer confirmation. Based on retention time and mass spectrum comparison, the second peak observed in the BNR samples was identified as 5-MeO BTA, indicating the first peak was 4-MeO BTA [Fig. 2(b)]. Both MeO-BTA isomers were observed under both aerobic and anoxic conditions [Fig. 1(c)]. Liu et al. (2011) detected these methoxy intermediates in aerobic conditions. They proposed methylation of the nitrogen atom in the triazole ring to form 1-methyl benzotriazole under aerobic conditions. Huntscha et al. (2014) measured the formation of 1-methyl benzotriazole from benzotriazole in aerobic activated sludge and concluded the methylation pathway was a minor reaction that accounted for only 0.2% of the initial spike. Although 1-methyl benzotriazole was included in the compound database for this study, it was not detected in any of the BNR samples. In both the aerobic and anoxic conditions, 4-MeO BTA was formed at slightly higher concentrations when compared to 5-MeO BTA. Under aerobic conditions, both MeO-BTAs increased until the third day and then leveled off through the seventh day. Under anoxic conditions initial formation of both 4-MeO BTA and 5-MeO BTA lagged until the third day and continued to increase through the end of the experiment.

Trimethoprim biotransformed slowly during the seven-day experiment for both aerobic and anoxic conditions [Fig. 4(a)]. Four transformation products were potentially identified under aerobic conditions [Fig. 4(b)]. Two were products of hydroxylation, namely TMP 306 (${\mathrm{C}}_{14}{\mathrm{H}}_{18}{\mathrm{N}}_4{\mathrm{O}}_4$—mass 306.133 amu) and TMP 324 (${\mathrm{C}}_{14}{\mathrm{H}}_{20}{\mathrm{N}}_4{\mathrm{O}}_5$—mass 324.143 amu). The other two products were isomers 3-desmethyl TMP and 4-desmethyl TMP (${\mathrm{C}}_{13}{\mathrm{H}}_{16}{\mathrm{N}}_4{\mathrm{O}}_3$—mass 276.122 amu), formed by o-demethylation of one of the three methoxy groups on the benzene ring. Trimethoprim pathways are shown in Fig. 5. Interestingly, no transformation products were identified under anoxic conditions even though there was degradation of the parent compound.

To differentiate isomers, a 3-desmethyl trimethoprim standard was compared to the LC-QTOF retention time and mass spectrum. The standard corresponded to the second peak for the pair of desmethyl trimethoprim isomers observed in the BNR samples, suggesting the first peak was the 4-desmethyl trimethoprim isomer (Fig. 6). Standards of the two other observed transformation products, TMP 306 and TMP 324, were not available and could not be confirmed any further, although they have been reported previously in aerobic sludge by Jewell et al. (2016), as well as 4-desmethyl trimethoprim. Both desmethyl-trimethoprim isomers increased until the third day and began to slightly decrease by the seventh day, whereas TMP 306 and TMP 324 both continued to increase over the course of the seven-day experiment [Fig. 4(b)].

A decrease in triclosan concentration was observed under aerobic conditions, but under anoxic conditions the aqueous phase concentration of triclosan increased through the first day before decreasing to a steady concentration at day three [Fig. 7(a)]. This increase under anoxic conditions may be due to poor mixing, slow kinetics of solubility, and desorption from the activated sludge at the high spiking concentrations (spiked for a final concentration of $1\mathrm{mg}/\mathrm{L}$). After a day of mixing, more triclosan may have dissolved and/or desorbed, resulting in an initial increase before degradation is observed. Because degradation occurs faster in the aerobic reactor, triclosan does not exhibit as much of an initial increase.

The transformation product, triclosan-o-sulfate (${\mathrm{C}}_{12}{\mathrm{H}}_7{\mathrm{Cl}}_3{\mathrm{O}}_5\mathrm{S}$—mass 367.908 amu), was observed under aerobic and anoxic conditions using the compound database. A standard for triclosan-o-sulfate confirmed the identification of this transformation product observed in the BNR samples. Chen et al. (2015) performed aerobic degradation experiments and proposed the sulfation pathway through the addition of a ${\mathrm{SO}}_3\mathrm{H}\text{-}\text{group}$ to the triclosan phenolic group to form triclosan-o-sulfate (Fig. 8).

Initially, triclosan-o-sulfate formed rapidly under aerobic conditions, then there was a lag phase between 12 h and five days, after which the primary metabolite started to degrade [Fig. 7(b)]. Under anoxic conditions, triclosan-o-sulfate built up during the first day before being metabolized into secondary products. Under both conditions the concentration of triclosan-o-sulfate decreased over time, contrary to observations by Chen et al. (2015) who found triclosan-o-sulfate to be persistent in aerobic activated sludge even after 10 days.

A possible transformation product, not included in the compound database, was identified for triclosan in the aerobic activated sludge samples using the molecular feature screening method for unknowns in ESI negative mode. It had an observed mass of 241.920 amu, which resulted in one possible molecular formula, ${\mathrm{C}}_6{\mathrm{H}}_4{\mathrm{Cl}}_2{\mathrm{O}}_4\mathrm{S}$. Initial searches revealed that this could potentially be 3,5-dichloro-2-hydroxy-benzene sulfonate (Fig. 9).

A standard for 3,5-dichloro-2-hydroxy-benzene sulfonate had a slightly shifted retention time compared to what was observed for the unknown, where the standard eluted approximately 0.42 min later (Fig. S7). A matrix spike was performed to ensure that the matrix was not the cause of the shifted retention time; however, the matrix spike matched the standard. This suggests that the observed compound in the samples may be a closely related isomer, but not specifically 3,5-dichloro-2-hydroxy-benzene sulfonate. The unknown dichloro-2-hydroxy-benzene sulfonate isomer also appeared to exhibit in-source degradation and underwent a partial loss of ${\mathrm{SO}}_3^{-}$ during ionization to form both the transformation product and the corresponding dichlorophenol degradation product (${\mathrm{C}}_6{\mathrm{H}}_4{\mathrm{Cl}}_2\mathrm{O}$—mass 161.964 amu) at the same retention time. This was detected and identified by the database as 2,4-dichlorophenol, a potential transformation product. However, standards for possible dichlorophenol isomers (Fig. S8) were analyzed and confirmed that the retention times for the standards did not match the dichlorphenol isomer observed in the samples. In addition, this by-product was observed in tests performed using the standard for 3,5-dichloro-2-hydroxy-benzene sulfonate, which further concluded that in-source degradation was occurring.