Rate constants of hydroxyl radicals reaction with different dissociation species of fluoroquinolones and sulfonamides: Combined experimental and QSAR studies

Water Research 166 (2019) 115083

Contents lists available at ScienceDirect

Water Research journal homepage: www.elsevier.com/locate/watres

Rate constants of hydroxyl radicals reaction with different dissociation species of fluoroquinolones and sulfonamides: Combined experimental and QSAR studies Xiang Luo a, Xiaoxuan Wei a, b, Jingwen Chen a, *, Qing Xie a, Xianhai Yang a, c, Willie J.G.M. Peijnenburg d, e a

Key Laboratory of Industrial Ecology and Environmental Engineering (MOE), School of Environmental Science and Technology, Dalian University of Technology, Dalian, 116024, China College of Geography and Environmental Sciences, Zhejiang Normal University, Jinhua, 321004, China c Jiangsu Key Laboratory of Chemical Pollution Control and Resources Reuse, School of Environmental and Biological Engineering, Nanjing University of Science and Technology, Nanjing, 210094, China d Institute of Environmental Sciences (CML), Leiden University, Leiden, 2300, RA, the Netherlands e National Institute of Public Health and the Environment (RIVM), Center for Safety of Substances and Products, Bilthoven, 3720, BA, the Netherlands b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 21 June 2019 Received in revised form 9 September 2019 Accepted 11 September 2019 Available online 14 September 2019

Hydroxyl radicals ($OH) initiated degradation is an important process governing fate of aquatic organic micropollutants (OMPs). However, rate constants for aqueous reaction of OMPs with $OH (kOH) are available only for a limited number of OMPs, which complicates fate assessment of OMPs. Furthermore, molecular structures of many OMPs contain ionizable groups, and the OMPs may dissociate into different anionic/cationic species with different reactivity towards $OH. Therefore, it is of importance to determine kOH of ionizable OMPs, and to develop quantitative structure-activity relationship (QSAR) models for predicting kOH of OMPs at different ionization forms. Herein kOH values of 9 fluoroquinolones (FQs) and 11 sulfonamides (SAs) at 3 dissociation forms (FQ±/FQþ/FQ, SA0/SAþ/SA) were determined by competition kinetics experiments. A QSAR model using theoretical molecular structural descriptors was subsequently developed. The QSAR model successfully corroborated previous experimental results, exhibited good statistical performance, and is capable to predict kOH for FQs and SAs with different dissociation forms at environmentally relevant pH conditions. As organic ions have rarely been included in previous QSAR studies, the newly developed model that covers both neutral molecules and ions is of significance for future QSAR development as well as fate assessment of ionizable OMPs. © 2019 Elsevier Ltd. All rights reserved.

Keywords: Antibiotics Dissociation forms Hydroxyl radicals Reaction rate constants Quantitative structure-activity relationship (QSAR)

1. Introduction Chemicals in pharmaceuticals and personal care products (PPCPs) have gained increasing concern as aquatic organic micropollutants (OMPs) in recent years (Oberoi et al., 2019; Schwarzenbach et al., 2006). Among PPCPs, antibiotics that are widely used in medicine, aquaculture and livestock husbandry, have been detected in various aqueous systems [e.g., surface water (Liu et al., 2018), groundwater (Boy-Roura et al., 2018), wastewater treatment plants (Ruan et al., 2019)]. Since antibiotics can induce bacterial resistance at environmental concentrations (Gullberg

* Corresponding author. E-mail address: [email protected] (J. Chen). https://doi.org/10.1016/j.watres.2019.115083 0043-1354/© 2019 Elsevier Ltd. All rights reserved.

et al., 2011) and even result in the emergence of “superbugs” (Woodward, 2010), thus posing a threat to ecosystems as well as humans, environmental fate of antibiotics is becoming an issue of increasing concern. Previous studies indicated that reaction with reactive species, such as hydroxyl radical ($OH), singlet oxygen and triplet excited states of chromophoric dissolved organic matter in aqueous systems, is an important pathway governing fate of OMPs (Vione et al., 2014). In aquatic environments, reactive species are commonly generated as a consequence of photochemical reactions, electrontransfer or energy-transfer reactions (Burns et al., 2012). Among various reactive species, $OH is the most powerful reactant (Rosario-Ortiz et al., 2010). In natural waters, $OH is mainly produced by photochemical reactions of dissolved organic matter,

2

X. Luo et al. / Water Research 166 (2019) 115083

nitrate, nitrite and transition metal complexes (Burns et al., 2012), or by processes occuring in dark [e.g., oxidation by oxygen of reduced dissolved organic matter or oxidative attack of H2O2 on dissolved reduced metals such as Fe(II) in hypolimnion waters (Gligorovski et al., 2015), soils (Page et al., 2013) and subsurface sediments containing oxic-anoxic interfaces (Tong et al., 2016)], with its concentrations ranging from 1017 to 1015 mol/L (Luo et al., 2017). In aritificial treatment (e.g., advanced oxidation processes) facilities, $OH is usually generated at sufficient concentrations by coupled chemical and/or physical techniques (Gligorovski et al., 2015), such as H2O2/Fe(II) or H2O2/Fe(III) (Fenton), H2O2/ catalyst or peroxide/catalyst (Fenton-like), UV/H2O2, ozonation, pulse radiolysis and ultrasound. With relatively low selectivity, $OH is capable to oxidize a wide variety of OMPs (Gligorovski et al., 2015). As to antibiotics, $OH attack (e.g., addition to double bonds or aromatic rings, Habstraction, electron-transfer/H-transfer) can lead to a proportional loss of the primary antibacterial activity for parent compounds (Ge et al., 2015; Keen and Linden, 2013), via interrupting specific hydrogen-bonding/charge interactions or oxidizing functional moieties responsible for the antibacterial potency. For the reaction with $OH, the rate constant (kOH) is a significant parameter in determining environmental persistence and fate of OMPs. Therefore, it is of importance to acquire kOH values for OMPs such as antibiotics. Nevertheless, kOH values for most antibiotics are not available. Moreover, molecular structures of many antibiotics contain ionizable functional groups such as eCOOH, eOH and eNH2. In aquatic environments, these antibiotics may exist at different dissociation forms. It is known that different ionized species of OMPs have distinct reactivity with reactive species (Wei et al., 2013; Xie et al., 2013). Therefore, it is of great importance to determine kOH values of ionizable antibiotics at different ionization forms. Generally, kOH values of OMPs can be determined by different experimental methods (Buxton et al., 1988; Wols and HofmanCaris, 2012), including direct (e.g., pulse radiolysis, flash photolysis) and indirect methods (competition kinetics). As the direct determination is limited by availability of sophisticated equipment, the simpler and cheaper competition kinetics method has received a widespread application (Shemer et al., 2006). There are two commonly used systems to generate $OH in competition kinetics determinations: Fenton (Boreen et al., 2004, 2005) and UV/H2O2 systems (Ge et al., 2019; Xie et al., 2013). The Fenton system is only suitable for acidic conditions, whereas the UV/H2O2 system is applicable to a wide range of pH conditions. To date, kOH values have only been determined for dozens of antibiotics at fixed pH, and are available only for no more than hundreds of OMPs (Boreen et al., 2004; 2005; Buxton et al., 1988; Guo et al., 2018; Li et al., 2018; Santoke et al., 2009; Wols and Hofman-Caris, 2012; Wols et al., 2015; Zhang et al., 2016). It seems impossible to experimentally determine kOH for all OMPs, since the experimental determination is costly and laborious. Instead, quantitative structure-activity relationship (QSAR) models on which the OECD has issued guidelines (OECD, 2007), should be developed for kOH prediction. Currently, several QSAR models for aqueous kOH prediction (Borhani et al., 2016; Huang et al., 2012; Jin et al., 2015; Kusi c et al., 2009; Li et al., 2018; Luo et al., 2017; Monod and Doussin, 2008; Sudhakaran and Amy, 2013; Wang et al., 2009) have been developed. However, most of the existing models cover a limited number of OMPs (especially antibiotics) in the applicability domains and do not consider the dissociation forms, thus cannot be employed to predict the aqueous kOH of dissociated antibiotics. In this study, two groups of commonly used and frequently detected antibiotics, fluoroquinolones (FQs) and sulfonamides (SAs) (Oberoi et al., 2019), were selected as model compounds. kOH

values for 9 FQs and 11 SAs (each antibiotic has two available pKa values) at zwitterionic/neutral, cationic and anionic forms (FQ±/ FQþ/FQ, SA0/SAþ/SA) were determined by the competition kinetics experiments. Based on the experimental kOH data, a QSAR model was constructed using theoretical molecular structural descriptors, and was verified with experimental data. 2. Materials and methods 2.1. Reagents and materials Sources of the chemicals used in the experiment are described in the supplementary material (SM). Molecular structures for the 9 FQs and 11 SAs are shown in Fig. S1. All the chemicals, as received, were at least >98% purity, and their solutions were prepared using ultrapure water filtered with an OKP ultrapure water system (Shanghai Lakecore Instrument Co., China). 2.2. Competition kinetics experiments For the competition kinetics method, $OH initiated degradation is assumed to be the only significant loss pathway of target reactant. The reactant competes for $OH with a reference compound whose kOH value (kOH,P) under identical conditions is known (Einschlag et al., 2003; Xie et al., 2013). The bimolecular rate constant of the reactant (kOH,R) can be calculated by:

kOH;R ¼

 lnð½Rt ½R0 Þ kR   kOH;P ¼  kOH;P kP lnð½Pt ½P0 Þ

(1)

where kR and kP are apparent degradation rate constants for the reactant and reference compound, respectively; [R] and [P] are concentrations of the reactant and reference compound, respectively. A plot of ln([R]t/[R]0) versus ln([P]t/[P]0) gives a straight line passing through the origin, whose slope gives the ratio kR/kP. In this study, acetophenone (kOH,P ¼ 5.9  109 M1s1, Buxton et al., 1988) was adopted as the reference compound, and initial concentrations of the antibiotics and acetophenone were 20 mM. Kinetics experiments were performed with an XPA-1 merry-goround photochemical reactor (Nanjing Xujiang Technology Co., China) using an UV/H2O2 system to generate $OH. A 500 W Hg lamp surrounded by 340 nm cut-off filters was used as the light source (Fig. S2), under which $OH was sufficiently generated and direct photolysis of acetophenone was negligible (Xie et al., 2013). Inside the reactor, quartz tubes rotate around the light source, ensuring homogeneous irradiation of the solution (50 mL) in each tube (Fig. S3). To determine kOH for different dissociation species, three pH conditions (Table S1) were set, under which the antibiotics mainly present at a certain dissociation form. The pH adjustment methods and H2O2 concentrations are detained in the SM. Direct photolysis rate constants (kD) for the antibiotics were determined to correct for direct photolysis. For each antibiotic under a certain pH condition, a control experiment with the same irradiation condition but without H2O2 was performed (Fig. S3) to determine kD values. Moreover, light screening effects of H2O2 and the antibiotics were corrected using a factor f (detailed in the SM), since the light screening can decrease irradiance absorbance of the antibiotics and reduce their direct photolysis rates (Shemer et al., 2006). Finally, the apparent kOH,R for a target antibiotic at a specific pH condition (kOH,A) was obtained by:

kOH;A ¼

ðkR  fkD Þ  kOH;P kP

(2)

Dark controls with the presence of H2O2 but without the

X. Luo et al. / Water Research 166 (2019) 115083

irradiation were also performed to evaluate oxidation of the antibiotics by H2O2. All the experiments were repeated six times at room temperature (22 ± 1  C), and kOH,A values were reported from average of the six replicates, with their experimental errors evaluated by the t-test (detailed in the SM). 2.3. Calculation of kOH for different dissociation forms An SA molecule contains one amine group and one amide group, thus exhibiting two pKa values and three dissociation forms (cationic form SAþ, neutral form SA0 and anionic form SA). An FQ molecule contains one carboxylic group and three nitrogen sites, possibly exhibiting four pKa values at maximum. Many previous studies provided only two pKa values and three dissociation forms (cationic form FQþ, zwitterionic form FQ± and anionic form FQ) due to difficulty in determination of the pKa values (Babic et al., 2007). Therefore, in this study, each antibiotic has two pKa values (Table S1) and three dissociation forms (Fig. S4). Distributions of the different dissociation species (detailed in the SM) as a function of pH are shown in Fig. S5. Three pH values (Table S1) for each antibiotic were selected based on the distribution. For the antibiotics at the lower, medium and higher experimental pH conditions (Table S1), the species that present the highest proportions are their cationic, zwitterionic/neutral and anionic forms, respectively. At a given pH, the kOH,A value can be assumed to be sum of the kOH for the different co-existing species (kOH,S, Wei et al., 2013; Xie et al., 2013), i.e., kOH,A ¼ d1kOH,S(1) þ d2kOH,S(2) þ d3kOH,S(3)

(3)

where d1, d2, d3 are proportions of the three dissociation species (Fig. S5); and kOH,S(1), kOH,S(2), kOH,S(3) are their corresponding kOH,S values. By solving three combined equations of eq. (3) at different pH values, the kOH,S values were obtained, which were also reported from average of the six replicates, with their experimental errors evaluated by the t-test. 2.4. Analytical methods At appropriate time intervals, a sample (1.0 mL) was withdrawn, and totally 8 samples were withdrawn from each tube during the irradiation. According to previous studies (Shemer et al., 2006; Xie et al., 2013), the small volume decrease caused by the sampling has little impacts on degradation kinetics of the reactants. After addition of 0.5 mL methanol as a $OH quencher, the sample was analyzed using an Agilent 1260 HPLC. The analysis was performed on a SB C18 column (2.1 mm  150 mm, 3.5 mm particle size) with a fluorescence detector (FLD) for the FQs, and an XDB C18 column (3.0 mm  150 mm, 3.5 mm particle size) with a diode array detector (DAD) for the SAs. Acetophenone was analyzed on the same column with the corresponding antibiotics and detected by the DAD. The HPLC analytical conditions are detailed in Table S2. 2.5. QSAR modeling of kOH,S A QSAR model was developed on the basis of the 60 experimental kOH,S values for the 20 antibiotics at cationic, zwitterionic (FQ)/neutral (SA) and anionic forms. The data points were randomly split into a training set and a validation set with a ratio of 3:1. The molecular structures of the cationic, zwitterionic/neutral and anionic forms of the antibiotics were optimized at the M06-2X (Zhao and Truhlar, 2008) level in conjunction with the 6e31 þ g(d,p) basis set using the Gaussian 09 program suite (Frisch et al., 2009), and 14 quantum chemical descriptors (Table S3) were

3

calculated accordingly. Based on the optimized structures, DRAGON (DRAGON ver. 6.0, 2014) molecular structural descriptors were calculated and employed after excluding the descriptors with missing values or with high pair-wise correlations (i.e. one of any two descriptors having a correlation coefficient greater than 0.99). Following the OECD guidelines (OECD, 2007), the QSAR model was developed by stepwise multiple linear regression using SPSS (Version 16.0), and the applicability domain was assessed by a Williams plot (Netzeva et al., 2005) as detailed in the SM. 3. Results and discussion 3.1. Degradation kinetics at different pH conditions Fig. S6 shows apparent degradation kinetics of the 20 antibiotics. It can be observed that loss of the antibiotics is not significant in the dark control (<3%), implying that direct oxidation of the antibiotics by H2O2 is negligible. However, the controls with irradiation and absence of H2O2 show that direct photolysis of some antibiotics is significant (Fig. S6), indicating necessity to correct for direct photolysis. The determined G values (Table S4) are 0.84, indicating necessity for correcting for the light screening effects of H2O2 and the antibiotics. The determined kOH,A values under the different pH conditions are listed in Table S5. All the kOH,A values for the FQs except for lomefloxacin and difloxacin (kOH,A >1010 M1s1) are higher than those for the SAs (1010 M1s1 > kOH,A >109 M1s1). As can be seen from Fig. 1, for the FQs, kOH,A is the highest at neutral pH (pH ¼ 7 or 7.5) and the lowest at acidic pH (pH ¼ 3). For the SAs, kOH,A at pH ¼ 3.5e6 (pH2 in Table S1) is the lowest. The kOH,A values of the SAs except for sulfadimethoxine at pH ¼ 9e11 (pH3 in Table S1) are higher than those at the other two pH conditions, while kOH,A of sulfadimethoxine at pH ¼ 2 is the highest. It deserves mentioning that most previous studies determined kOH,A values of FQs and SAs at a single pH (Boreen et al., 2004; 2005; Santoke et al., 2009; Zhang et al., 2016). For most of the antibiotics under study, this work is the first to determine their kOH,A at different pH values. 3.2. kOH,S values for the different dissociation forms of the antibiotics The determined kOH,S values (Fig. 2, Table S6) for most of the SAs are similar to those of some pollutants [e.g., kOH,A ¼ 4.2  109 M1s1 for nitrobenzene (Einschlag et al., 2003) and kOH,A ¼ 2.6  109 M1s1 for atrazine (Buxton et al., 1988)], while the kOH,S values for the cationic form of sulfadimethoxine, the anionic form of sulfaguanidine, and the FQs except for lomefloxacin and difloxacin are 1 order of magnitude higher (kOH,S > 1010 M1s1) than the values mentioned above. For the FQs and most of the SAs, it can be seen that variation of the kOH,S values for the three dissociation forms (Fig. 2) corresponds with variation of the kOH,A values at the different pH conditions (Fig. 1). As shown in Fig. 2, the zwitterionic forms of all the FQs degrade the fastest, while their cationic forms degrade the slowest. For the SAs, the kOH,S values for the neutral species are lower than those for the cationic and anionic forms. Additionally, sulfamerazine, sulfamethoxazole, sulfameter, sulfadiazine, sulfamethizole, sulfamethazine, sulfapyridine and sulfaguanidine show a similar tendency of their anionic forms presenting the highest kOH,S values among the three forms; while the kOH,S values for SAþ forms of sulfachloropyridazine, sulfathiazole and sulfadimethoxine are higher than those for the other two forms. For most of the antibiotics under study, this study is the first to report the kOH,S values for the different protonated states, which enriches the kOH database of

4

X. Luo et al. / Water Research 166 (2019) 115083

Fig. 1. kOH,A values for the fluoroquinolones and sulfonamides under study at different pH conditions (the corresponding kOH,A values are listed in Table S5).

Fig. 2. kOH,S values for different dissociation species of the fluoroquinolones and sulfonamides under study (the corresponding kOH,S values are listed in Table S6).

OMPs and is of importance for the fate assessment of the antibiotics. In previous studies, kOH,A values of several SAs were determined at pH ¼ 3 using the Fenton system (Boreen et al., 2004, 2005) or at pH ¼ 7 using the UV/H2O2 system (Zhang et al., 2016). In this study, the kOH,A values for the SAs were determined neither at pH ¼ 3 or pH ¼ 7. To compare the different sets of kOH,A values, the kOH,A values of the SAs at pH ¼ 3 and pH ¼ 7 were calculated by eq. (3) with the kOH,S values (Table S6). It is known that measured kOH,A values vary in consequence of different experimental conditions (Shemer et al., 2006). As can be seen from Fig. 3 and Table S7, there is a slight difference between the kOH,A values of the SAs at pH ¼ 3 in the current study and the studies of Boreen et al. (2004, 2005): the kOH,A values of sulfamethazine and sulfamerazine in the current study are approximately equal to or a bit higher than those determined by Boreen et al. (2004, 2005), respectively; while the kOH,A values of the other SAs are a bit lower than the values determined previously. It can also be found from previous studies (Huber et al., 2003; Packer et al., 2003; Pereira et al., 2007) that for a same compound (e.g., naproxen, clofibric acid, ibuprofen), kOH,A determined by the UV/H2O2 system differs (either higher or lower) from that determined by the Fenton system. Thus, the difference between the kOH,A values determined in the current and the previous studies may mainly attribute to the

different experimental systems [the Fenton system in the studies of Boreen et al. (2004; 2005) versus the UV/H2O2 system in the current study]. To evaluate potential errors of the determined kOH,A, ratio factors (RF) between the kOH,A values in the current study and in the previous studies were calculated. As can be seen from Table S7, all the RF values at pH ¼ 3 are <2, which is within the most acceptable range of RF values (<3) for measured kOH,A values by competition kinetics from different laboratories (von Sonntag and von Gunten, 2012). Compared with the kOH,A values at pH ¼ 7 determined by Zhang et al. (2016) using the UV/H2O2 system, the kOH,A values in the current study are generally lower (Fig. 3 and Table S7). According to Shemer et al. (2006) and Huber et al. (2003), the selection of reference compounds is also an important factor that induces errors of the competition kinetics results. In the studies of Zhang et al. (2016), nitrobenzene was employed as a reference compound. It can be found from previous studies (Armbrust, 2000; Baeza and Knappe, 2011; Shemer et al., 2006; Zhang et al., 2016) that kOH,A values of many OMPs (e.g., sulfadiazine, molinate, diuron) determined using nitrobenzene as the reference compound with the UV/ H2O2 system, are generally higher than those employing other reference compounds (e.g., acetophenone, molinate, atrazine, parachlorobenzoic acid), as can be concluded from Table S8. Thus, the use of acetophenone as the reference compound in the current

X. Luo et al. / Water Research 166 (2019) 115083

5

Fig. 3. Comparison of kOH,A values in this study and in previous studies (Boreen et al., 2004; 2005; Zhang et al., 2016) for selected sulfonamides at pH ¼ 3 and pH ¼ 7 (the corresponding kOH,A values are listed in Table S7).

study may lead to comparatively lower determined kOH,A values than the corresponding values at pH ¼ 7 determined by Zhang et al. (2016). Nevertheless, all the RF values at pH ¼ 7 are <2 (Table S7) and within the range of RF values for measured kOH,A (von Sonntag and von Gunten, 2012). Moreover, the logkOH,A values at pH ¼ 3 and pH ¼ 7 in the current study (Table S7) fall within the range of the previously determined logkOH,A values ± 0.3 log units (Fig. S7). Therefore, it can be concluded that the kOH,A values of the SAs at pH ¼ 3 and pH ¼ 7 agree with those reported previously (Boreen et al., 2004; 2005; Zhang et al., 2016). 3.3. QSAR model for kOH,S Based on the determined kOH,S values (Table S6), a QSAR model was developed for logkOH,S of the antibiotic species: logkOH,S ¼ 9.724 þ 0.2180  C-006 e 0.6650  qþ C þ 0.01300  m(4) ntr ¼ 45, R2adj ¼ 0.805, RMSEtr ¼ 0.228, Q2LOO ¼ 0.778, Q2BOOT ¼ 0.770 next ¼ 15, R2ext ¼ 0.820, Q2ext ¼ 0.816, RMSEext ¼ 0.197 where ntr and next are the number of the data points in the training set and validation set, respectively; R2adj and R2ext are adjusted and external determination coefficients, respectively; Q2ext is external validation coefficient; RMSEtr and RMSEext are root mean square errors for the training set and validation set, respectively; Q2LOO and Q2BOOT are the leave-one-out cross-validated and bootstrap method (repeated 5000 times, at each step 20% of objects are left out from the training set) coefficients, respectively. Some of the parameters are detailed in the SM. The model has three molecular structural descriptors, the Dragon descriptor C-006 (the number of CH2RX fragments), the most positive net atomic charge on C atoms (qþ C ) and the molecular dipole moment (m). Values of the descriptors are listed in Table S9. Variance inflation factors (Norusis, 1997) for these descriptors are all <10 (Table S10), indicating that these descriptors are free of multicollinearities. This model possesses high values of R2adj, Q2 2 2 2 LOO, Q BOOT, R ext and Q ext, and low values of RMSEtr as well as RMSEext, indicating the QSAR model has high goodness-of-fit, robustness, and predictive performance. Plot of predicted versus experimental logkOH,S values (Fig. S8) shows that the absolute

values of the residuals were <0.6 log units, indicating that the predicted logkOH,S values agree well with the experimental values. The applicability domain of the model is characterized in Fig. S9. For all the species in the training and validation sets, the absolute values of standardized residual (defined in the SM) are smaller than 3, suggesting there are no outliers. In addition, the leverage values of all the species in the training and validation sets are lower than the warning leverage value (h* ¼ 0.27), indicating none of the compounds is particularly influential in the model space and thus the training set has good representativeness. Therefore, the developed QSAR model meets all the criteria set by the OECD for QSAR models to be applied for regulatory purposes (OECD, 2007), and it can be employed to predict aqueous kOH,S for other FQ and SA species within the applicability domain. The t values of the t-test for the three descriptors are listed in Table S10. Among the descriptors, C-006 is the most statistically significant (t ¼ 6.526), with a positive coefficient. Defined as the number of specific atom types in a molecule and calculated on the basis of molecular composition as well as atom connectivities (Ghose et al., 1998; Viswanadhan et al., 1989), C-006 encodes the number of CH2RX (R ¼ group linked through C atom; X ¼ electronegative atoms such as O, N, S, P, Se, halogens) fragments. Since $OH reactions can occur on functional groups with reactive H atoms via H-abstraction (Minakata et al., 2009), chemicals with CH2RX groups may be more easily attacked by $OH, leading to their higher $OH reactivity. The quantum chemical descriptor qþ C is the second most statistically significant (t ¼ 4.153), correlating negatively with logkOH,S. With a more positive formal charge, the C atom tends to possess poorer ability to donate electrons. Since $OH is an electrophile, chemicals with these C atoms can be hardly attacked by $OH, thus leading to low $OH reactivity. m encodes displacement with respect to centers of positive and negative charges and carries information about charge distribution in a molecule (Luo et al., 2017). The positive coefficient of m indicates that a chemical with a higher m value is more inclined to react with $OH. The developed QSAR model is capable to predict the aqueous kOH,S of FQs and SAs within the applicability domain and thus kOH,A under realistic environmental pH conditions. To further verify predictability of the model, sulfisoxazole was selected as a case since the kOH,A values of this compound have been measured at pH ¼ 3 (Boreen et al., 2004) and pH ¼ 7 (Zhang et al., 2016). The kOH,A values that were calculated based on eq. (3) from kOH,S for the 3 dissociation forms (Table S11) predicted by the QSAR model, are

6

X. Luo et al. / Water Research 166 (2019) 115083

1.44  109 M1s1 at pH ¼ 3 and 2.68  109 M1s1 at pH ¼ 7 (Table S12). Prediction accuracy of QSARs relies on data used for training the models. Since most of the kOH,A values determined in the current study are generally lower than the corresponding values determined in the previous studies (Table S7), it is not surprising that the kOH,A values for sulfisoxazole predicted by the QSAR model are lower than those reported previously (Table S12). Nevertheless, it can be seen that the predicted kOH,A values for sulfisoxazole and those reported previously are within a same order of magnitude (109 M1s1), and the predicted kOH,A values show a trend of increase from pH ¼ 3 to pH ¼ 7 which is similar to that for the experimental values. Lee and von Gunten (2012) reported that rate constants predicted by QSARs for advanced oxidation processes have a potential error of a factor of up to 9. Ratio factors between the kOH,A in previous studies and the kOH,A predicted by the current QSAR at pH ¼ 3 and pH ¼ 7 are 4.57 and 2.74, respectively (Table S12), and the differences between the previously reported and the QSAR predicted logkOH,A values are <0.7 log units. Therefore, the current QSAR model has a satisfactory performance for kOH prediction. It deserves mentioning that to date most QSAR models on environmental chemicals have been constructed for non-ionized organic molecules. The QSAR model in this study encompasses both organic non-ionized molecules and anions/cations, which indicates that it is possible to construct QSARs for organic anions/ cations. The development of QSARs that cover organic anions/cations in the training/validation sets is of great importance, since the molecular structures of many emerging OMPs carry various ionizable functional groups and these pollutants can ionize in the aquatic environment or in engineered facilities like waste water treatment plants. From this point of view, this study exemplifies how to develop QSARs that can predict environmental behavior of organic ions. Nevertheless, it needs to be pointed out that the current study investigated only two categories of antibiotics with three protonated states. More efforts are needed to develop similar QSARs that cover more organic ionizable molecules and protonated states.

4. Conclusions The 20 antibiotics investigated can react rapidly with $OH (kOH,A or kOH,S > 109 M1s1), and their reactivity towards $OH varies among the different dissociation forms. For the fluoroquinolones, the zwitterionic forms and the cationic forms possess the highest and lowest kOH,S values, respectively; while for the sulfonamides, the neutral forms possess the lowest kOH,S values. The established QSAR model exhibits satisfactory statistical performance and may pave a new way for predicting the aqueous kOH values of antibiotics with different protonated states. According to the model, the number of CH2RX fragments, the most positive net atomic charge on C atoms and the molecular dipole moment are the main factors governing $OH reactivity of the investigated antibiotics. Increases in the number of CH2RX fragments and the values of molecular dipole moment lead to an increase of the logkOH,S values, while increases in the values of the most positive net atomic charge on C atoms lead to a decrease of the logkOH,S values.

Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments This study was supported by the National Natural Science Foundation of China (21661142001) of China and the National Key R&D Program of China (2018YFC1801604). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.watres.2019.115083. References Armbrust, K.L., 2000. Pesticide hydroxyl radical rate constants: measurements and estimates of their importance in aquatic environments. Environ. Toxicol. Chem. 19, 2175e2180. Babic, S., Horvat, A.J.M., Pavlovic, D.M., Kastelan-Macan, M., 2007. Determination of pKa values of active pharmaceutical ingredients. Trac. Trends Anal. Chem. 26, 1043e1061. Baeza, C., Knappe, D.R.U., 2011. Transformation kinetics of biochemically active compounds in low-pressure UV Photolysis and UV/H2O2 advanced oxidation processes. Water Res. 45, 4531e4543. Boreen, A.L., Arnold, W.A., McNeill, K., 2004. Photochemical fate of sulfa drugs in the aquatic environment: sulfa drugs containing fivemembered heterocyclic groups. Environ. Sci. Technol. 38, 3933e3940. Boreen, A.L., Arnold, W.A., McNeill, K., 2005. Triplet-sensitized photodegradation of sulfa drugs containing six-membered heterocyclic groups: identification of an SO2 extrusion photoproduct. Environ. Sci. Technol. 39, 3630e3638. Borhani, T.N.G., Saniedanesh, M., Bagheri, M., Lim, J.S., 2016. QSPR prediction of the hydroxyl radical rate constant of water contaminants. Water Res. 98, 344e353. Boy-Roura, M., Mas-Pla, J., Petrovic, M., Gros, M., Soler, D., Brusi, D., Mencio, A., 2018. Towards the understanding of antibiotic occurrence and transport in groundwater: findings from the Baix Fluvia alluvial aquifer (NE Catalonia, Spain). Sci. Total Environ. 612, 1387e1406. Burns, J.M., Cooper, W.J., Ferry, J.L., King, D.W., Dimento, B.P., McNeill, K., Miller, C.J., Miller, W.L., Peake, B.M., Rusak, S.A., Rose, A.L., Waite, T.D., 2012. Methods for reactive oxygen species (ROS) detection in aqueous environments. Aquat. Sci. 74, 683e734. Buxton, G.V., Greenstock, C.L., Helman, W.P., Ross, A.B., 1988. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals ($OH/$O-) in aqueous solution. J. Phys. Chem. Ref. Data 17, 513e886. Dragon ver. 6 is software of TALETE srl, Italy. http://talete.mi.it/products/dragon_ description.htm (accessed on October 31, 2014). Einschlag, F.S.G., Carlos, L., Capparelli, A.L., 2003. Competition kinetics using the UV/ H2O2 process: a structure reactivity correlation for the rate constants of hydroxyl radicals toward nitroaromatic compounds. Chemosphere 53, 1e7. Frisch, M.J., Trucks, G.W., Schlegel, H.B., Scuseria, G.E., Robb, M.A., Cheeseman, J.R., Scalmani, G., Barone, V., Mennucci, B., Petersson, G.A., Nakatsuji, H., Caricato, M., Li, X., Hratchian, H.P., Izmaylov, A.F., Bloino, J., Zheng, G., Sonnenberg, J.L., Hada, M., Ehara, M., Toyota, K., Fukuda, R., Hasegawa, J., Ishida, M., Nakajima, T., Honda, Y., Kitao, O., Nakai, H., Vreven, T., Montgomery, J.A.J., Peralta, J.E., Ogliaro, F., Bearpark, M., Heyd, J.J., Brothers, E., Kudin, K.N., Staroverov, V.N., Kobayashi, R., Normand, J., Raghavachari, K., Rendell, A., Burant, J.C., Iyengar, S.S., Tomasi, J., Cossi, M., Rega, N., Millam, J.M., Klene, M., Knox, J.E., Cross, J.B., Bakken, V., Adamo, C., Jaramillo, J., Gomperts, R., Stratmann, R.E., Yazyev, O., Austin, A.J., Cammi, R., Pomelli, C., Ochterski, J.W., Martin, R.L., Morokuma, K., Zakrzewski, V.G., Voth, G.A., Salvador, P., Dannenberg, J.J., Dapprich, S., Daniels, A.D., Farkas, O., Foresman, J.B., Ortiz, J.V., Cioslowski, J., ̀ A.01. Gaussian, Inc., Wallingford CT. Fox, D.J., 2009. Gaussian 09, Revision Ge, L.K., Na, G.S., Zhang, S.Y., Li, K., Zhang, P., Ren, H.L., Yao, Z.W., 2015. New insights into the aquatic photochemistry of fluoroquinolone antibiotics: direct photodegradation, hydroxyl-radical oxidation, and antibacterial activity changes. Sci. Total Environ. 527e528C, 12e17. Ge, L.K., Zhang, P., Halsall, C., Li, Y.Y., Chen, C.E., Li, J., Sun, H.L., Yao, Z.W., 2019. The importance of reactive oxygen species on the aqueous phototransformation of sulfonamide antibiotics: kinetics, pathways, and comparisons with direct photolysis. Water Res. 149, 243e250. Ghose, A.K., Viswanadhan, V.N., Wendoloski, J.J., 1998. Prediction of hydrophobic (lipophilic) properties of small organic molecules using fragmental methods: an analysis of ALOGP and CLOGP methods. J. Phys. Chem. A 102, 3762e3772. Gligorovski, S., Strekowski, R., Barbati, S., Vione, D., 2015. Environmental implications of hydroxyl radicals ($OH). Chem. Rev. 115, 13051e13092. Gullberg, E., Cao, S., Berg, O.G., Ilback, C., Sandegren, L., Hughes, D., Andersson, D.I., 2011. Selection of resistant bacteria at very low antibiotic concentrations. PLoS Pathog. 7, e1002158. Guo, K.H., Wu, Z.H., Yan, S.W., Yao, B., Song, W.H., Hua, Z.C., Zhang, X.W., Kong, X.J., Li, X.C., Fang, J.Y., 2018. Comparison of the UV/chlorine and UV/H2O2 processes in the degradation of PPCPs in simulated drinking water and wastewater: kinetics, radical mechanism and energy requirements. Water Res. 147, 184e194. Huang, X.W., Yu, X.L., Yi, B., Zhang, S.H., 2012. Prediction of rate constants for the reactions of alkanes with the hydroxyl radicals. J. Atmos. Chem. 69, 201e213.

X. Luo et al. / Water Research 166 (2019) 115083 Huber, M.M., Canonica, S., Park, G.Y., von Gunten, U., 2003. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environ. Sci. Technol. 37, 1016e1024. Jin, X.H., Peldszus, S., Huck, P., 2015. Predicting the reaction rate constants of micropollutants with hydroxyl radicals in water using QSPR modeling. Chemosphere 138, 1e9. Keen, O.S., Linden, K.G., 2013. Degradation of antibiotic activity during UV/H2O2 advanced oxidation and photolysis in wastewater effluent. Environ. Sci. Technol. 47, 13020e13030. Kusi c, H., Rasulev, B., Leszczynska, D., Leszczynski, J., Koprivanac, N., 2009. Prediction of rate constants for radical degradation of aromatic pollutants in water matrix: a QSAR study. Chemosphere 75, 1128e1134. Lee, Y., von Gunten, U., 2012. Quantitative structure-activity relationships (QSARs) for the transformation of organic micropollutants during oxidative water treatment. Water Res. 46, 6177e6195. Li, C., Wei, G.L., Chen, J.W., Zhao, Y.H., Zhang, Y.N., Su, L.M., Qin, W.C., 2018. Aqueous OH radical reaction rate constants for organophosphorus flame retardants and plasticizers: experimental and modeling studies. Environ. Sci. Technol. 52, 2790e2799. Liu, X., Lu, S., Guo, W., Xi, B., Wang, W., 2018. Antibiotics in the aquatic environments: a review of lakes, China. Sci. Total Environ. 627, 1195e1208. Luo, X., Yang, X.H., Qiao, X.L., Wang, Y., Chen, J.W., Wei, X.X., Peijnenburg, W.J.G.M., 2017. Development of a QSAR model for predicting aqueous reaction rate constants of organic chemicals with hydroxyl radicals. Environ. Sci.: Processes & Impacts 19, 350e356. Minakata, D., Li, K., Westerhoff, P., Crittenden, J., 2009. Development of a group contribution method to predict aqueous phase hydroxyl radical (HO$) reaction rate constants. Environ. Sci. Technol. 43, 6220e6227. Monod, A., Doussin, J.F., 2008. Structure-activity relationship for the estimation of OH-oxidation rate constants of aliphatic organic compounds in the aqueous phase: alkanes, alcohols, organic acids and bases. Atmos. Environ. 42, 7611e7622. Netzeva, T.I., Worth, A.P., Aldenberg, T., Benigni, R., Cronin, M.T., Gramatica, P., Jaworska, J.S., Kahn, S., Klopman, G., Marchant, C.A., Myatt, G., NikolovaJeliazkova, N., Patlewicz, G.Y., Perkins, R., Roberts, D., Schultz, T., Stanton, D.W., van de Sandt, J.J., Tong, W., Veith, G., Yang, C., 2005. Current status of methods for defining the applicability domain of (quantitative) structure-activity relationships - the report and recommendations of ECVAM Workshop 52. ATLA 33, 155e173. Norusis, M.J., 1997. SPSS 7.5 Guide to Data Analysis. A Simon and Schuster Company, Upper Saddle River, New Jersey. Oberoi, A.S., Jia, Y.Y., Zhang, H.Q., Khanal, S.K., Lu, H., 2019. Insights into the fate and removal of antibiotics in engineered biological treatment systems: a critical review. Environ. Sci. Technol. 53, 7234e7264. OECD, 2007. Guidance Document on the Validation of (Quantitative) StructureeActivity Relationships [(Q)SAR] Models. Organisation for Economic Co-Operation and Development, Paris, France. Packer, J.L., Werner, J.J., Latch, D.E., McNeill, K., Arnold, W.A., 2003. Photochemical fate of pharmaceuticals in the environment: naproxen, diclofenac, clofibric acid, and ibuprofen. Aquat. Sci. 65, 342e351. Page, S.E., Kling, G.W., Sander, M., Harrold, K.H., Logan, J.R., Mcneill, K., Cory, R.M., 2013. Dark formation of hydroxyl radical in arctic soil and surface waters. Environ. Sci. Technol. 47, 12860e12867. Pereira, V.J., Weinberg, H.S., Linden, K.G., Singer, P.C., 2007. UV degradation kinetics and modeling of pharmaceutical compounds in laboratory grade and surface water via direct and indirect photolysis at 254 nm. Environ. Sci. Technol. 41, 1682e1688. Rosario-Ortiz, F.L., Wert, E.C., Snyder, S.A., 2010. Evaluation of UV/H2O2 treatment for the oxidation of pharmaceuticals in wastewater. Water Res. 44, 1440e1448.

7

Ruan, Y.F., Wu, R.B., Lam, J.C.W., Zhang, K., Lam, P.K.S., 2019. Seasonal occurrence and fate of chiral pharmaceuticals in different sewage treatment systems in Hong Kong: mass balance, enantiomeric profiling, and risk assessment. Water Res. 149, 607e616. Santoke, H., Song, W.H., Cooper, W.J., Greaves, J., Miller, G.E., 2009. Free-radicalinduced oxidative and reductive degradation of fluoroquinolones pharmaceuticals: kinetic studies and degradation mechanism. J. Phys. Chem. A 113, 7846e7851. Schwarzenbach, R.P., Escher, B.I., Fenner, K., Hofstetter, T.B., Johnson, C.A., von Gunten, U., Wehrli, B., 2006. The challenge of micropollutants in aquatic systems. Science 313, 1072e1077. Shemer, H., Sharpless, C.M., Elovitz, M.S., Linden, K.G., 2006. Relative rate constants of contaminant candidate list pesticides with hydroxyl radicals. Environ. Sci. Technol. 40, 4460e4466. Sudhakaran, S., Amy, G.L., 2013. QSAR models for oxidation of organic micropollutants in water based on ozone and hydroxyl radical rate constants and their chemical classification. Water Res. 47, 1111e1122. Tong, M., Yuan, S.H., Ma, S.C., Jin, M.G., Liu, D., Cheng, D., Liu, X.X., Gan, Y.Q., Wang, Y.X., 2016. Production of abundant hydroxyl radicals from oxygenation of subsurface sediments. Environ. Sci. Technol. 50, 214e221. Viswanadhan, V.N., Ghose, A.K., Revankar, G.R., Robins, R.K., 1989. Atomic physicochemical parameters for 3 dimensional structure directed quantitative structure-activity relationships. 4. Additional parameters for hydrophobic and dispersive interactions and their application for an automated superposition of certain naturally occurring nucleoside antibiotics. J. Chem. Inf. Comput. Sci. 29, 163e172. Vione, D., Minella, M., Maurino, V., Minero, C., 2014. Indirect photochemistry in sunlit surface waters: photoinduced production of reactive transient species. Chem. Eur J. 20, 10590e10606. von Sonntag, C., von Gunten, U., 2012. Chemistry of Ozone in Water and Wastewater Treatment. From Basic Principles to Applications. IWA, London. Wang, Y.N., Chen, J.W., Li, X.H., Zhang, S.Y., Qiao, X.L., 2009. Estimation of aqueousphase reaction rate constants of hydroxyl radical with phenols, alkanes and alcohols. QSAR Comb. Sci. 28, 1309e1316. Wei, X.X., Chen, J.W., Xie, Q., Zhang, S.Y., Ge, L.K., Qiao, X.L., 2013. Distinct photolytic mechanisms and products for different dissociation species of ciprofloxacin. Environ. Sci. Technol. 47, 4284e4290. Wols, B.A., Hofman-Caris, C.H.M., 2012. Review of photochemical reaction constants of organic micropollutants required for UV advanced oxidation processes in water. Water Res. 46, 2815e2827. Wols, B.A., Harmsen, D.J.H., Wanders-Dijk, J., Beerendonk, E.F., HofmanCaris, C.H.M., 2015. Degradation of pharmaceuticals in UV (LP)/H2O2 reactors simulated by means of kinetic modeling and computational fluid dynamics (CFD). Water Res. 75, 11e24. Woodward, C., 2010. Animal antibiotics under tougher United States scrutiny as consensus grows on “superbug” risk to humans. Can. Med. Assoc. J. 182, E513eE514. Xie, Q., Chen, J.W., Zhao, H.X., Qiao, X.L., Cai, X.Y., Li, X.H., 2013. Different photolysis kinetics and photooxidation reactivities of neutral and anionic hydroxylated polybrominated diphenyl ethers. Chemosphere 90, 188e194. Zhang, R.C., Yang, Y.K., Huang, C.H., Zhao, L., Sun, P.Z., 2016. Kinetics and modeling of sulfonamide antibiotic degradation in wastewater and human urine by UV/ H2O2 and UV/PDS. Water Res. 103, 283e292. Zhao, Y., Truhlar, D.G., 2008. The M06 suite of density functional for main group thermochemistry, thermochemical kinetics, noncovalent interactions, excited states, and transition elements: two new functionals and systematic testing of four M06-class functionals and 12 other functionals. Theor. Chem. Acc. 120, 215e241.