Degradation of sulfamethoxazole by chlorination in water distribution systems: Kinetics, toxicity, and antibiotic resistance genes

Sulfamethoxazole (SMX) is one of veterinary drugs and food additives, which has been frequently detected in surface waters in recent years and will cause damage to organisms. Therefore, SMX was selected as a target to be investigated, including the degradation kinetics, evolution of toxicity, and antibiotic resistance genes (ARGs) of SMX during chlorination in batch reactors and water distribution systems (WDS), to determine the optimal factors for removing SMX. In the range of investigated pH (6.3–9.0), the SMX degradation had the fastest rate at close to neutral pH. The chlorination of SMX was affected by the initial total free chlorine concentration, and the degradation of SMX was consistent with second‐order kinetics. The rate constants in batch reactors are (2.23 ± 0.07) × 102 M−1 s−1 and (5.04 ± 0.30) × 10 M−1 s−1 for HClO and ClO−1, respectively. Moreover, the rate constants in WDS are (1.76 ± 0.07) × 102 M−1 s−1 and (4.06 ± 0.62) × 10 M−1 s−1, respectively. The degradation rate of SMX was also affected by pipe material, and the rate followed the following order: stainless‐steel pipe (SS) > ductile iron pipe (DI) > polyethylene pipe (PE). The degradation rate of SMX in the DI increased with increasing flow rate, but the increase was limited. In addition, SMX could increase the toxicity of water initially, yet the toxicity reduced to the level of tap water after 2‐h chlorination. And the relative abundance of ARGs (sul1 and sul2) of tap water samples was significantly increased under different chlorination conditions.


INTRODUCTION
Since penicillin by Alexander Fleming was discovered in 1928, antibiotics have been widely used to prevent human and animal infectious diseases (Huang et al., 2020;Wang et al., 2020;Zhang, Ying, et al., 2015). However, 60-80% of antibiotics are not absorbed after entering human and animal bodies but directly discharged from the body through feces or urine , which has become the main source of antibiotic residues in the environment. In addition, aquaculture industry (Tian et al., 2013), medical and pharmaceutical wastewater (Gobel et al., 2005;Lindberg et al., 2004) is also an important source of environmental antibiotic residue. Antibiotics are considered to be "pseudo persistent" pollutants because they continue to enter the ecosystem (Gulkowska et al., 2008), which has aroused great concern because they may impose a potential risk to the nontarget species (Chen et al., 2011).
Sulfonamides (SAs) are one of the most widely used antibiotics in the world (Gaffney et al., 2016), with more than 20,000 tons of SAs (excluding drugs used as herbicides) introduced into the biosphere each year (Qiu et al., 2016). As a kind of fairly water-soluble polar compound, SAs are weakly degraded in the nature, thus they can easily enter surface runoff or be leached into the groundwater. Luo et al. (2011) reported that the concentration of trimethoprim, trimethoprim, sulfamethoxazole, and sulfachloropyridazine in the Haihe River was (1.2 ± 0.8) Â 10 2 ng/L, (1.2 ± 0.8) Â 10 2 ng/L, (1.7 ± 1.0) Â 10 2 ng/L, and (1.8 ± 1.4) Â 10 2 ng/L, respectively. A survey (Jiang et al., 2011) of the Huangpu River found that the concentration of trimethoprim, sulfadiazine, sulfamethoxazole, and sulfachloropyridazine was 62. 39, 40.55, 623.27, and 58.29 ng/L, respectively. Li, Ho, et al. (2017) reported that the concentration of oxytetracycline reached 600 ng/L in tap water in Hong Kong. Therefore, the existence of antibiotics is not only limited to the environmental water of nature but also exists in the drinking water related to human beings. SA concentrations in natural water are quite low, ranging from nanograms to micrograms per liter (ng/L to μg/L) , but it still has ecotoxicity at this level (Fent et al., 2006). For example, SAs can damage the human immune system, pose skin allergic reactions (Choquet-Kastylevsky et al., 2002) and even carcinogen risk (Shao et al., 2005), inhibit the germination of plants (Migliore et al., 2003), and case ecological bioaccumulation in aquatic organisms (Peiris et al., 2017). Furthermore, they can promote the microbial genetic variation (Jeong et al., 2010), which may lead to the emergence of ARGs and antibiotic-resistant bacteria (Zheng et al., 2017). Guo et al. (2014) investigated the ARGs levels in the Yangtze River Delta of China and detected the existence of 10 ARGs in the source waters of the seven water treatment plants, among which the relative abundance of ARGs sul1, sul2, tet (c), and tet (g) is the highest, total concentrations of ARGs were above 10 5 copies/ml. The sul1 and sul2 were also the most frequently detected sulfonamide ARGs in surface waters in Germany and Australia (Stoll et al., 2012). ARGs and antibiotic-resistant bacteria could enhance the potential for both new and old diseases to emerge (Mulvey & Simor, 2009).
Various technologies were used to remove or degrade SAs in water, including biodegradation (Chen & Xie, 2018;Yu et al., 2018), activated carbon and porous resin adsorption (Choi et al., 2008;Yang et al., 2011), membrane filtration (Rosman et al., 2018), free chlorination (Gaffney et al., 2016), and advanced oxidation (Beltran et al., 2008;Cai et al., 2022;Guo, Yin, et al., 2015). Among these methods, free chlorination is regarded as one of the most effective technologies due to its low price, continuous disinfection, wide application scope, and high oxidation efficiency for a variety of organic compounds (Dong et al., 2017(Dong et al., , 2019. Nassar et al. (2018) have shown that free chlorine can react with SAs in the first-order. Dodd and Huang (2004) established a degradation model of SMX by chlorine under different pH based on pseudo second-order kinetics, and detected the chlorination products by using GC-MS and LC-MS derivatization methods and proposed the degradation pathway (Dodd & Huang, 2004;Gaffney et al., 2016). However, most SA chlorination studies were generally conducted under labscale static conditions (e.g., beakers with buffered water or deionized). The SA chlorination and the potential influence of chlorination on ARGs in dynamic water distribution system (WDS) has been scarcely studied.
Sulfamethoxazole (SMX, CAS Number 723-46-6) is a typical SA (Yao et al., 2018), which is mainly used as a veterinary drug and feed additive and its selected chemical properties were showed in Figure S1. This study mainly includes the following aspects: (1) Investigate the relationship between the SMX chlorination and the concentrations of free chlorine, pH, flow rate, and pipe materials in batch reactor and WDS; (2) determine the kinetics of SMX chlorination by free chlorine and analyze the difference between its degradation in batch reactor and WDS; (3) evaluate the evolution of toxicity during the SMX chlorination in WDS; (4) investigate the production of antibiotic resistance genes of pipeline bacteria under the exiting condition of only use SMX, chlorination products of SMX, or pretreatment with SMX followed by chlorination.

Experimental materials
SMX (AR, ≥98%) was purchased from TCI (Shanghai, China), and free chlorine stock solutions were prepared by dilution of commercial 5.5-6.5% sodium hypochlorite solution from Aladdin (Shanghai, China). N,N-diethyl-pphenylenediamine (DPD) was purchased from Hach Company (Loveland, Colorado, USA), and it was used to measure chlorine concentration. HPLC-grade acetonitrile (ACN), methanol (MeOH), and phosphoric acid (H 3 PO 4 ) were purchased from Sigma Aldrich (Shanghai, China). Other reagents, including dibasic sodium phosphate (Na 2 HPO 4 , AR grade), sodium dihydrogen phosphate (NaH 2 SO 4 , AR grade), and sodium thiosulfate (Na 2 S 2 O 3 , CP grade), were purchased from Sinopharm Chemical Reagent (Shanghai, China). The bacteria and reactivation reagents were from Hamamatsu Photonics (Beijing, China) and were used to evaluate the toxicity. The purified water was obtained from a Millipore Milli-Q apparatus and tap water obtained from Hangzhou City.

Batch reactor and WDS
The degradation experiments of SMX were conducted in a batch reactor and WDS with the use of sodium hypochlorite (NaClO) as a disinfectant. Batch experiments were conducted in 500-ml brown glass beakers to exclude the influence of potential light ( Figure S2a). The WDS system had three separate loops (each loop is 150 mm in diameter and 80 m in length), made of polyethylene pipe (PE), ductile iron pipe (DI), and stainless-steel pipe (SS). Each loop was equipped with an online monitoring system, circulation pump (which can adjust the flow rate), chemical injection ports, and sampling tap, and each loop could be operated independently ( Figure S2b,c,d).

Degradation experiments of SMX
Using fresh tap water, each WDS loop was washed at a speed of 1 m/s for 20 min before WDS experiments. In the WDS experiments, experimental water was the tap water and its characteristics are shown in Table S1. The initial concentration of SMX was 200 μg/L, and experiment temperature was stabilized at 25 C by an automatic control system. A 5% sodium hypochlorite solution was added in loops through chemical injection ports until ultimately the concentration of free chlorine reached the required values (0.4, 0.6. 0.8, and 1.0 mg/L). The pH levels (6.6, 7.0, 7.4, 7.8, and 8.2) and flow rates (0.7, 1.0, 1.3, and 1.5 m/s) were adjusted by phosphate buffer and circulation pump, respectively. In addition, the effects of pipe materials were examined by using three kinds of pipes (PE, DI, SS) because they were commonly used in drinking water distribution system in China. In the batch reactors experiments, the pH levels were set as 6.3, 6.6, 6.9, 7.2, 7.5, 7.8, 8.1, 8.4, 8.7, and 9.0, and other conditions were consistent with WDS experiments; 10-ml samples were collected at 0, 5, 10, 15, 20, 25, 30, 35, and 40 min, respectively, and then 1 ml of 100 mg/L Na 2 S 2 O 3 (as quenching agent) was immediately added to stop further reaction.

Determination of antibiotic concentration
The SMX in each sample was quantified by an Agilent 1200 Series HPLC, which equipped with a Zorbax Eclipse XDB-C18 column (4.6 mm Â 150 mm, 5 μm) and a UV detector (G1314B, Agilent Technologies, Santa Clara, CA, USA). The column temperature was set at 30 C and injection sample volume was 20 μl. The detection wavelength was set at 270 nm. Pure acetonitrile and 1‰(v/v) phosphoric acid with the ratio of 80:20 was the mobile phase, and the flow was 0.8 ml/min.

Toxicity assessment of SMX degradation
DI pipe material is the most widely used in China, and about 75% of water distribution systems is made of DI pipe material ; thus, the section experiment was carried out in DI. Microtox method (the international procedure ISO 11348-3) (Buysse et al., 1989) was used to evaluate the toxic evolution of a SMX solution during chlorination in WDS (DI), which was based on the inhibition of the luminescence by Vibrio fischeri (Barhoumi et al., 2016) marine bacteria with 15-min exposure. A multiscan spectrum M200 PRO (Tecan, Austria) was used. In order to prevent the inactivation of V. fischeri by free chlorine, the free chlorine in each water sample was quenched with 0.2 mol/L Na 2 S 2 O 3 , and Na 2 S 2 O 3 itself did not inhibit V. fischeri (Wei et al., 2011). The calculation formula of the inhibition of light emissions was as follows: inhibition = (I 0 À I) /I 0 , where I and I 0 was the luminescent intensity of samples at different times and the luminescent intensity of the blank solution, respectively.

DNA extraction
Four group experiments were set up in this section, which were carried out in WDS (DI). A group represented the tap water samples that were placed for 6 days, as blank control; B group represented the tap water samples that were added SMX for 6 days; C group represented the tap water samples that were added the chlorinated products of SMX (obtained by solid phase extraction) for 6 days; D group represented the tap water samples that were pretreatment with SMX followed by chlorination for 6 days.
Five hundred milliliters of water samples were concentrated by 0.22-μm sterile filter membranes until the membrane was clogged to capture bacteria. Then, the membranes were cut up with clean scissors and stored at À20 C in 50-ml centrifuge tubes. The total DNAs were extracted using an EZNA Water DNA Kit (Omega Bio-tek, Georgia, USA) and the extraction steps followed the manufacturer's instructions. The DNA concentration and purity were determined by 1.5% agarose gel electrophoresis and spectrophotometer analysis (NanoDrop ND-2000c, Thermo, USA), as previous reported by Chen and Zhang (2013).
DNA quantification by real-time quantitative polymerase chain reaction (qPCR) Two sulfonamide resistance genes (sul1 and sul2) and an integron gene (intI1, used to reflect the transfer potential of horizontal gene) were selected for quantitative detection by StepOnePlus real-time PCR system (Applied Biosystems, CA, USA). In addition, 16S rRNA was also monitored to reflect relative abundance, and the calculation formula of RAs was as follows: RAs = ARG copies/16S rRNA copies. The detailed primer information and the procedure of qPCR is shown in Table S2. Following previous research (Chen & Zhang, 2013;Li, Sheng, et al., 2017), the plasmids containing target genes were obtained by molecular cloning, and plasmids carrying target genes (serially 10-fold dilution) were used to generate standard curves (including 5 or 6 points).
The qPCRs were conducted in 96-well plates with a final volume of 20-μl mixture, containing 10-μl SYBR Premix Ex Taq (TaKaRa, Bio, Shiga, Japan), 7.2-μl DNA-free water, 2-μl template DNA, and 0.2-μM forward and reverse primers. The temperature procedure was as follows: Initial denaturation step for 180 s at 95 C and then 35 cycles of amplification containing denaturation for 30 s at 94 C, annealing for 30 s at 55 C, and extension for 60 s at 72 C, with a final extension step for 600 s at 72 C.

Kinetics of SMX degradation
Effect of free chlorine Figure 1a,c showed the SMX degradation at different free chlorine (0.4-1.0 mg/L); it can be seen that when free chlorine ≥0.6 mg/L and reaction time was more than 30 min, the SMX was almost completely degraded in both the batch and WDS tests. However, the degradation of SMX in the batch reactor was faster than that in WDS. At the concentration of 1.0 mg/L free chlorine, it took about 15 min for the removal of SMX to reach 100% in WDS and only 10 min in batch test. In addition, the degradation rate of the two reaction systems increased monotonously with the increase of initial free chlorine concentration (Figure 1b,d). At the concentration of 0.4, 0.6, 0.8, and 1.0 mg/L of free chlorine, the removal ratio of SMX in batch reactor increased from 52.9% to 75.3%, 100%, and 100%, and in WDS increased from 56.5% to 66.5%, 85.7%, and 92.6%, respectively, in 10 min.
In order to better describe the correlation between reaction times and ln(SMX t /SMX 0 ) at different free chlorine concentration, the first-order model (Equation 1, Supporting Information) was used. The k 1 and goodness of fit (R 2 ) of each initial free chlorine concentration in batch reactor was bigger than those in WDS (Table S3). The difference was caused by complex conditions of the WDS including pipe contains scale (Lytle & Liggett, 2016) at pipe wall ( Figure S3), which consumed some free chlorine (Dong et al., 2019). On the contrary, the conditions in the batch reactor were relatively stable, thus the linear correlation in backer tests was better than in WDS. However, the linear correlation in WDS was still satisfactory (all R 2 ≥ 0.966); thus, the SMX degradation by various initial free chlorine concentrations in both the batch reactor and WDS were followed the first-order kinetics.
Previous studies have found that the degradation of sulfadiazine the pipe network follows a second-order F I G U R E 2 Degradation of SMX at different pH in batch reactor (a) and WDS (d), and the first-order kinetics plot of SMX degradation in batch reactor (b) and WDS (e), and the second-order rate constants (k app , M À1 s À1 ) as a function of pH for the SMX chlorination by free chlorine in batch reactor (c) and WDS (f). Experimental conditions: ductile iron pipe, 25 C, flow rate = 1.0 m/s, initial free chlorine = 0.8 mg/L, initial SMX = 200 μg/L kinetics (Dodd & Huang, 2004;Dong et al., 2019), so this possibility was also studied. The second-order rate kinetics constants (k 2 ) could be calculated by Equations (2) and (3) (Supporting Information). Figure S5a,b showed that the relationship between k 2 and free chlorine concentration was satisfactory linear correlation in both the batch reactor (R 2 = 0.995) and WDS (R 2 = 0.960). Therefore, the degradation of SMX in both the batch reactor and WDS also could be described by second-order model.

Effect of pH
In the batch reactor (Figure 2a), after 40 min, the degradation amount of SMX by free chlorine remained at 100% during pH 6.3 to pH 7.2 and decreased from 100% to 73.08% during pH 7.2 to pH 9.0. Figure 2d showed that the SMX degradation in the WDS and its results were similar to the batch tests, indicating that their highest efficiency of SMX degradation occurred at pH close to neutral (pH ≈ 7.0). The first-order rate constants (k) of reactions between SMX and free chlorine were obviously dependent on the pH value, as shown in Figure 2b,e. As shown in Table S4, k increased from 0.117 to 0.139 min À1 in batch experiment and from 0.097 to 0.107 min À1 in WDS when the solution pH value increased from 6.3 to 7.0; yet k decreased from 0.139 to 0.027 min À1 in batch experiment and from 0.107 to 0.045 min À1 in WDS when the solution pH value increased from 7.0 to 9.0. Therefore, the fastest rate of SMX degradation also occurred at pH close to neutral. In addition, all R 2 of batch experiment were bigger than that of WDS when the set pH value was similar, and the reason had been discussed in Section 3.1.1.
In the range of investigated pH, the acid-base equilibrium (Equations 1-3) of HClO and the SMX may be used to explain the pH dependence behavior in the reaction of free chlorine and SMX.
HClO ⇌ H þ þ ClO À pK a ¼ 7:54 ð1Þ The two species of free chlorine (i.e., HClO and ClO À ) and the three species of SMX (i.e., HSMX + , SMX, and SMX À ) could be considered to react with each other; thus, the dependence of k app on solution pH can be modeled by Equations (4) and (5): :8 þ 10 À1:8 Ã 10 À5:6 The k i,j was calculated by L-M (Levenberg-Marquardt) regressions of experimental k app according to Equation (5), using MATLAB. While there were six possible reactions from Equation (5), fitting of the experimental data of Figure 2c,f showed that only the following two reactions (Equations 11 and 12) could reasonably fit the experimental data (a solid line), R 2 ≥ 0.974 and R 2 ≥ 0.972 in batch reactor and water distribution systems, respectively. The reason for this is may be that [HSMX + ] and [SMX] have little effect in the reaction system due to their extremely low mole fraction at pH 6.3-9.0 ( Figure S6). In addition, contribution of the individual reactions to the rate constant (k app ) is also shown in Figure 2c,f (dashed lines). This result indicated that the HClO reacts faster with deprotonated SMX (SMX À ) than ClO À (i.e., in batch reactor: k 13 = (2.23 ± 0.07) Â 10 2 > k 23 = (5.04 ± 0.30) Â 10 1 M À1 s À1 ; in WDS: k 13 = (1.76 ± 0.07) Â 10 2 > k 23 = (4.06 ± 0.62) Â 10 1 M À1 s À1 ). At pH 6.3-8.2, reaction 11 mainly contributed to the k app . In contrast, the k app at pH from 8.2 to 9.0 could be attributed to reaction 12. At nearly neutral pH, the comprehensive effect of reaction 11 and reaction 12 is the best; the subsequent experiments can thus be carried out at a nearly neutral pH.
ClO À ½ þ S À ½ !products ð12Þ Effect of flow rate As shown in Figure 3, the change of flow rate had little effect on the SMX degradation. The degradation rate constants for SMX were 0. 129, 0.190, 0.211, and 0.201 min À1 under the condition of 0.7, 1.0, 1.3, and 1.5 m/s, respectively (Table S5). The WDS used in the experiment is a circulating system, each group of experiments is sampled at a fixed reaction time, so their hydraulic retention time is the same. Therefore, the possible reason for this difference is that flow rate could change the hydraulic conditions in WDS, thereby affecting mass transfer coefficient. The Reynolds number (Re) was used to evaluate this behavior, R e = ρνd=μ, where ρ, ν, d, and μ represent the density, flow rate, pipe diameter, and viscosity coefficient of fluid (0.895 Â 10 À3 PaÁs, 25 C), respectively. All Re that are more than 4300 (Table S5) indicate that turbulent flow conditions exist in WDS. Mass transfer coefficient increased with increasing Re under turbulent flow conditions, which ultimately improves mixing and this may make the rate of SMX degradation enhanced when the flow rate is from 0.7 to 1.0 m/s. However, when the flow rate is from 1.0 to 1.5 m/s, the degradation rate hardly increased because the effect of mass transfer on degradation has reached the limit.

Effect of pipe materials
As shown in Figure 4, after 15 min of reaction, the removal of SMX reached 100%, 87.8%, and 94.4% in PE, DI, and SS, respectively. Table S6 summarized the calculated first-order rate constants; these data showed that the order of degradation rate was SS (0.358 min À1 ) > DI (0.191 min À1 ) > PE (0.139 min À1 ). Metal ions and corroded metal might contribute to this phenomenon; DI were better protected because they have lined with a layer of cement, but SS is more prone to rust due to poor welding (Dong et al., 2017). Consequently, compared with DI and PE, the SS has the most metal ions and corroded metal. As discussed in the Section 3.1.3, when the flow rate was 1.0 m/s, these experiments were conducted with turbulent flows (Re > 4300); consequently, the metal ions and corroded metal in solder joints were easily transferred to the water. The concentrations of total iron in the three pipes were detected, which was 0.8 mg/L in SS and less than 0.1 mg/L in DI and PE. Fe 2+ could be oxidized to Fe 3+ by free chlorine, and the process produced some free radicals that enhanced the SMX  (Dong et al., 2019;Perez-Moya et al., 2010). In addition, Fe 3+ might be oxidized to higher valence iron in hypochlorite solution system, such as a way to prepare ferrate (VI) is to oxidize Fe 3+ with sodium hypochlorite solution under alkaline condition (Li et al., 2005). The higher valence iron has stronger oxidation ability, and it could promote the degradation of SMX. In addition, by analyzing the crystal structure of iron precipitation ( Figures S3 and S4), it is found that the main forms of precipitation are Fe 3 O 4 and α-FeOOH, and they have certain adsorption. Therefore, the high reaction rate of SMX in DI was also related to the adsorption of iron precipitation.

Toxicity assessment of SMX degradation
As shown in Figure 5, compared with the inhibition of bioluminescence rate of tap water close to 0, the presence of SMX can significantly increase the toxicity of water. In the first 10 min of reaction, with the degradation of SMX, the toxicity showed a downward trend. Within 10 to 90 min of the reaction, the toxicity showed an upward trend, indicating that the intermediate products of the reaction were also toxic. However, within 90 to 120 min of the reaction, the toxicity showed a downward trend until it was the same as that of tap water, which indicated that SMX could be completely degraded by free chlorine within 2 h, and the toxicity of chlorinated products was low.

Comparison of ARGs in WDS
The comparison of ARGs contents and relative abundance of water samples under different treatments are shown in Figure 6. In general, compared with group A, the number and relative abundance of sul1, sul2, and intI1 genes in group B, C, and D were significantly increased. However, the number of 16S rRNA did not change significantly, indicating that the total number of bacteria changed little. Previous studies have shown that the variation of ARGs relative abundance is mainly caused by two aspects: One is the change of host bacteria abundance and the other is the horizontal transfer of ARGs in bacteria (Jia et al., 2015;Su et al., 2015), and the increase of ARGs in drinking water was the result of two factors (Jia et al., 2015). Therefore, the following analysis mainly consider these two aspects.
In group B, although the total number of bacteria decreased slightly, the number of sul1 and sul2 increased. SMX inhibited the growth of other bacteria and increased antibiotic-resistant bacterial community. In addition, SMX as an environmental pressure could promote the replication of sul1 and sul2, which was the main reason for the increase of sul1 and sul2. On the other hand, the increase of sul1 and sul2 was also caused by the transfer of ARGs (because the number of intI1 has increased). In group C, the increase of 16S rRNA genes number reflected the increase of total bacterial count, indicating that chlorination products had no inhibitory effect on total bacterial count. These results showed that even if SMX was chlorinated in water, its chlorination products would still lead to the production of resistance genes. The reason may be that chlorination cannot completely open the aromatic ring in SMX molecule, and some products F I G U R E 6 Gene copy numbers and relative abundance of antibiotic resistance genes in different treatment conditions F I G U R E 5 Evolution of toxicity of SMX chlorination. Experimental conditions: ductile iron pipe, 25 C, pH = 7.0 ± 0.2, flow rate = 1.0 m/s, initial free chlorine = 0.8 mg/L, initial SMX = 200 μg/L