Investigation of the Synergistic Toxicity of Binary Mixtures of Pesticides and Pharmaceuticals on Aliivibrio fischeri in Major River Basins in South Korea

This work introduces the potential synergistic toxicity of binary mixtures of pesticides and pharmaceuticals, which have been detected in substantial amounts in major river basins in South Korea. Different dose-response curve functions were employed in each experimental toxicity dataset for Aliivibrio fischeri. We tested the toxicity of 30 binary mixtures at two effect concentrations: high effect concentration [EC50] and low effect concentration (EC10) ranges. Thus, the toxicological interactions were evaluated at 60 effected concentration data points in total and based on model deviation ratios (MDRs) between predicted and observed toxicity values (e.g., three types of combined effects: synergistic (MDR > 2), additive (0.5 ≤ MDR ≤ 2), and antagonistic (MDR < 0.5)). From the 60 data points, MDRs could not be applied to 17 points, since their toxicities could not be measured. The result showed 48%-additive (n = 20), 40%-antagonistic (n = 17), and 12%-synergistic (n = 6) toxicity effects from 43 binaries (excluding the 17 combinations without MDRs). In this study, EC10 ratio mixtures at a low overall effect range showed a general tendency to have more synergistic effects than the EC50 ratio mixtures at a high effect range. We also found an inversion phenomenon, which detected three binaries of the combination of synergism at low concentrations and additive antagonism at high concentrations.


Introduction
Conventional chemical risk assessments frequently focus on individual substances rather than mixtures even though previous studies have shown that different mixture toxicities can results from combined effects among chemicals, even at no observed effect concentrations [1,2]. The types of the mixture toxicity are generally explained as additivity, synergism (greater than additivity), and antagonism (less than additivity) [3,4]. Among these combined effects, additive and synergistic toxicity effects can be regarded as more significant than the antagonistic effect (from the aspect of the regulatory chemical risk assessment, which assumes the worst-case scenario as the default) [5]. In this context, the conventional regulatory risk assessment of chemical mixtures is mostly based on Int. J. Environ. Res. Public Health 2019, 16, 208 2 of 16 the concentration addition (CA) model as a default for estimating the mixture toxicity [6][7][8][9]. Although the worst-case scenario can be substantial synergistic toxicity, the CA model, which ignores the synergistic interaction, has been mainly employed since available predictive models for estimating the synergistic effect have been very limited for the purpose of regulatory risk assessment [5]. In addition, some studies have shown that the synergistic interaction could be a relatively rare occurrence, at least within pesticide mixtures and realistic mixtures having low concentrations in mammals (which are comprised of approximately 5% of the tested mixtures) [10][11][12][13]. In contrast, a recently published review showed that less than 25% of research and experiments on the toxicity of chemical mixtures investigated seven or more compounds [14,15]. Some clinical researchers in human toxicology also showed that the probability of synergistic toxicity could be increased in proportion to the number of components, e.g., an 8% toxicity effect was caused by pharmaceutical products with five to ten mixed components, and a 38% toxicity effect was provoked by pharmaceuticals with fourteen or more mixture components [5,16,17]. This issue related to synergism is still controversial and difficult to conclude since most of the studies concerning mixture toxicity have been conducted with a specific binary mixture or simple mixtures having less than ten components [5]. Many studies have found that pesticides and pharmaceuticals were detected in the aquatic environment and thus they might lead to mixture toxicity for aquatic nontarget organisms [14,[18][19][20][21]. Thus, when considering environmental mixtures with complex matrices and different chemicals, any potential risk posed might have been underestimated by the CA model if the synergistic interaction occurred in such mixtures [14].
The objective of this study was to preliminarily investigate the potential synergistic interactions of pharmaceutical and pesticide residues that can be found in the aquatic environment. For this purpose, we tested and evaluated the toxicity of different binary mixtures of pesticides and pharmaceuticals, which had been substantially identified in major river basins in South Korea. To evaluate the toxicity of single and mixed chemicals, a bioassay with a luminescent bacterium, AliiVibrio fischeri, which is widely used in standard toxicity methods, was conducted [22,23]. Based on the original best-fit approach [24], different dose-response curve (DRC) functions were employed to test data sets, and best-fit functions of single and mixed chemicals were used in the mixture toxicity modeling.

Selection of Target Pesticides and Pharmaceuticals
Through previously published studies [25][26][27][28][29][30][31] that investigated the environmental concentration level of 47 pharmaceuticals in four major river basins (Han River, Geum River, Bukhan River, and Yeongsan River) in South Korea, 29 pharmaceuticals could be identified [32]. In addition, based on a study by Lee et al. [33] monitoring 140 pesticide residues in six Korean river basins (Han River, Geum River, Bukhan River, Yeongsan River, Mankyeong River, and Seomjin River), eight pesticides (isoprothiolane > butachlor > prothiofos > chlorpyrifos > hexaconazole molinate > diazinon and alachlor) could be found. From those 29 pharmaceuticals and eight pesticides, six chemicals, including four pharmaceuticals (tetracycline, trimethoprim, sulfamethoxazole, and chlortetracycline) and two pesticides (hexaconazole, and isoprothionlane), were finally selected as target mixutre components in this study (Table S1). The selction was made by considering their toxicities to A. fischeri, solubilities in water and carrier solvents, and detection frequencies in the aquatic environment. To our knowledge, this is the first study to investigate the toxicity of hexaconazole, isoprothiolane, and chlortetracycline and their binary mixtures to A. fischeri.

Test Reagents, Chemicals and Sample Preparation
Six target compounds were purchased from Sigma-Aldrich (Seelze, Germany). According to the physico-chemical properties of these compounds, stock solutions were prepared in either 99.9% dimethyl sulfoxide (DMSO, Sigma-Aldrich, Seelze, Germany) for trimethoprim and chlortetracycline or 99.8% ethanol (EtOH, Carl Roth GmbH, Karlsruhe, Germany) for tetracycline, hexaconazole, isoprothiolane and sulfamethoxazole. All of the stock solutions were kept at −20 • C under dark conditions until the working solutions were prepared. For quality control and quality assurance, the concentrations of the stock solutions were quantified with an HPLC system (Agilent 1290, Agilent Technologies, Santa Clara, CA, USA) connected to an Agilent triple quadrupole mass spectrometry (MS/MS) model 6460. Before the experiments, working solutions (1:25) were prepared by diluting the stock solutions in 2% sodium chloride for a marine bacterium, A. vibrio, according to ISO 11348-3 [22]. The pH values of the working solutions were checked and adjusted to between 6.0 and 8.0 with 1 N NaOH and 1 N HCl.

Testing Organism and Culture
The bioluminescent bacteria A. fischeri (strain NRRL-B-11177 and formerly called Vibrio fischeri) were purchased from MicroTox ® (Lot number 15C4025A, Modern Water, New Castle, UK). The freeze-dried bacteria were activated with the reconstitution solution provided by the MicroTox ® for 30 mins at 15 • C. The activated bacteria were transferred to a photobacteria medium (Sigma-Aldrich) for preculture at 20 • C. For stock culturing, activated A. fischeri were estimated in a 250 ml main culture medium at an initial turbidity of a 1:10 dilution by UV-vis photometric (Ultraspec 3300, Amersham, Buckinghamshire, UK) at OD 578 was 0.02 (10 formazine turbidity units, FNU). A. fischeri were cultured at 20 • C with shaking at 180 rpm/min until the turbidity of the OD 578 was 1.74 (700 FNU to 1800 FNU). The amplified A. fischeri were purified twice with a 2% sodium chloride solution at 4 • C and 20 min at 7000 × g. The bacteria were slowly suspended in protective medium (66 g D-(+)-glucose monohydrate, 4 g sodium chloride, 2 g L-histidine and 0.5 g BSA in 100 mL) at an ice cooled condition until the turbidity of the OD 578 was 2.58 (2000 FNU to 3000 FNU). The suspended stock bacteria were stored at −80 • C.

Single Chemical Toxicity Test
Determination of an effective concentration of samples was performed [34] using the standardized methods of ISO 11348-3, 1998 by luminescent bacteria (A. fischeri) [22]. To activate, the frozen bacteria were suspended using a reconstituted solution (20 g sodium chloride, 0.3 g potassium chloride, HEPES 50 mM and glucose 50 mM for 1 liter) for 30 minutes at 15 • C. The activated luminescence bacteria were mixed with 2% sodium chloride at a 1:25 dilution. Six single compounds were serially diluted at a ratio of 1:1 (100 µL) on a flat-bottomed black 96-well plate (Greiner Bio-one, Kremsmünster, Germany). The bacterial suspensions were exposed to the serially diluted sample at a ratio of 100 µL by 100 µL. Then, reactive samples were measured after 15 min of exposure at 15 • C by a luminescent reader (Tristar2, Berthold technologies, Bad Wildbad, Germany). To assure the quality of the bacteria, a 100 ppm zinc sulfate solution (Sigma-Aldrich) was measured every time.

Mixture toxicity test
An investigation of the synergistic toxicity of all binary mixtures that could be prepared from the six target compounds in Table S1 was conducted in a fixed ratio design based on the equitoxic mixture and the generation of dose-response curves from the mixtures employed in previous studies [35][36][37][38][39]. The advantages of the fixed ratio design are not only the ability to maximize the distribution of the effective dose range but also to minimize the number of experiments [37,[40][41][42]. With the same compounds in Table S1, two different equitoxic mixtures were prepared at a 50% effective concentration for each compound as a high effective concentration ratio mixture (EC 50 ratio mixture) to A. fischeri and at a 10% effective concentration ratio mixture (EC 10 ratio mixture) as a low effective concentration ratio mixture. As shown in Table 1, a total of 30 equitoxic binary mixtures of each combination were tested at high and low effective concentration levels. However, the total doses of the mixtures were systematically different.

Statistical Analysis of the Mixture Toxicity
A "best-fit" approach [24,39] was used to select the best model with the smallest sum of absolute residuals among the different sigmoidal functions because no single function could statistically describe all of the DRCs. Best-fit models with three-parameter sigmoidal equations were finally determined and applied to describe the experimental data of the mixture components tested in this study. The parameters in the sigmoidal regression equations and 95% confidence intervals were estimated using SigmaPlot ® (Ver. 12.5, Systat Software, Chicago, IL, USA). The ECx (e.g., EC 10 and EC 50 ) of the test chemicals was derived from the regression models shown in Table 2 and Table S2. It was assumed that all of the models were confined to the effects range from 0 to 100%. In order to compare between the highest and lowest toxicity without, we expressed the orders of magnitude which were determined as follows: However, in case a test chemical showed low solubility in water, the model for that chemical was assumed to have a range between >0% and <the maximum effect (%), where the chemicals were present at a maximum solubility in water under the test conditions in this study. The ECx for mixtures was calculated by the CA model according to the Loewe equation [11,43]: where ECx mix is the predicted effective concentration of a mixture; Pi and ECx i are the fraction and the individual effective concentration of the component with in the mixture, respectively.
MDR values (Belden et al., 2007 [11]) were used to quantify the interaction between the mixture components. The MDR values were frequently applied to determine the type of interactions of the mixture toxicity [11,13,44]. MDR is defined as: where the predicted ECx indicates the effective concentration of a mixture based on the predictive model, and the observed ECx is the effective concentration of the mixture obtained from experimental toxicity testing. In this study, the CA model, which is recommended as a default approximation for mixtures, was used to predict mixture toxicity [5,10,11,45,46]. Based on the MDR value, the types of combined effects are divided into three groups: synergistic (MDR > 2), additive (0.5 ≤ MDR ≤ 2) and antagonistic (MDR < 0.5) [11,13].

DRCs for Single Compounds
DRCs of all six compounds in Table S1 were empirically determined for A. fischeri, as illustrated in Figure 1. Table 2 summarizes the parameter values of all the best-fitting regression models for the DRCs of single compounds. Table S2 explains the regression models described in Table 2. The best-fitting curves for all of the compounds showed high regression coefficients (r 2 ) ranging from 0.969 to 0.995 and their ANOVA p-values were less than 0.0001. As presented in Table 2, the deviation between the steeper function of the highest slope values (65.25 for chlortetracycline) and the more gradual slope of the lowest (0.0034 for sulfamethoxazole) slope values was approximately 4.28 orders of magnitude. These considerable differences among the slopes of the DRCs of the compounds suggest that the compounds had highly different curve shapes for estimating the toxicity of each compound. The best-fitted DRCs for each compound had high regression coefficients (r 2 ) ranging from 0.971 to 0.995. The EC50 values of A. fischeri were up to 0.82 orders of magnitude and ranged from 51.65 µM for hexaconazole to 338.81 µM for trimethoprim. However, the EC 10 values of A. fischeri were up to 1.66 orders of magnitude and ranged from 1.05 µM for isoprothiolane to 47.79 µM for sulfamethoxazole. That is, hexaconazole and isoprothiolane presented the lowest effective concentrations (or the highest toxicity) at the respective EC 50 and EC 10 values, whereas isoprothiolane and sulfamethoxazole showed the highest effective concentrations (or the lowest toxicity) at the corresponding EC 50 and EC 10 values. These results also show that the toxicological profiles of compounds can be varied according to a given effective concentration level. In the cases of isoprothione, tetracycline, haxaconazole and trimethoprim, more than 80%-effect concentrations couldn't be obtained under the testing conditions because of their water solubility limits.  1 Regression models (C: Chapman, G: Gompertz, H: Hill, L: Logistic); 2 Height; 3 Slope; and 4 Center point

DRCs for Single Compounds
DRCs of all six compounds in Table S1 were empirically determined for A. fischeri, as illustrated in Figure 1.  Table 1 (the data points are geometric means ± standard deviation [SD] of the experimentally observed data and statistical best-fits (solid lines)).  Table 1 (the data points are geometric means ± standard deviation [SD] of the experimentally observed data and statistical best-fits (solid lines)).
The probable reason for the high deviation of those two mixtures is that they were the EC 10 ratio mixture, i.e., an equitoxic mixture based on ratios at 10% effective concentrations for each component but significantly less toxic than the others so that the ECx values could not be appropriately determined. As tested binary mixtures, the effective concentration data ranged from a low of 2.17 µmol/L (EC 10 + EC 10 Table  3).   Table  3). The dose-response curves of 30 binary mixtures of pesticides and pharmaceuticals in Table 1, respectively.  Table 3). The dose-response curves of 30 binary mixtures of pesticides and pharmaceuticals in Table 1, respectively.  Figures 2 and 3 illustrate DRCs for the observed bioluminescent inhibitions and the predicted inhibition of A. fischeri by the CA model for the binary equitoxic mixtures based on ratios at 50% and 10% effective concentrations and following the combinations in Table 1, respectively.

Statistical Analysis of Mixture Toxicity to Investigate Synergism
To quantify the toxicity interactions between mixture components, we calculated MDR values as shown in Table 4. Based on the MDR value, we strictly divided the three types of combined effects into synergistic (MDR > 2), additive (0.5 ≤ MDR ≤ 2) and antagonistic (MDR < 0.5) [13]. As shown in Figure 2 and Table 4, ten binaries of the EC 50 ratio mixtures (i.e., EC 50mix ) showed the same interactions at two effective concentration values of EC 10 and EC 50 . That is, four binaries of the EC 50mix (Mixture 1, 5, 9, and 25) showed the additive effects at the EC 50 and EC 10 ranges. The antagonistic interaction of six binaries of EC 50mix (Mixture 13, 17, 19, 23, 27, and 29) was then detected at the EC 50 and EC 10 ranges. Interestingly, five binaries of the mixture combination detected different interactions from the EC 50 and EC 10 ranges. Two binaries of the EC 50mix (Mixture 3 and 15) found an additive interaction at the EC 50 ranges and it resulted in a synergistic interaction at the EC 10 ranges. At similar trend in one binary combination (Mixture 11) was observed for the antagonistic interaction at the EC 50 ranges and an additive interaction at the EC 10 ranges. In contrast, the two binaries of the EC 50mix (Mixture 7, and 21) resulted in opposite trends in additive interaction at the EC 50 ranges and antagonistic interaction at the EC 10 ranges. These observed inversion phenomena of interaction between synergism at low concentrations and additive or antagonism at high concentrations are difficult to explain. As shown in Table 4, similar phenomena have been reported in previous studies [47][48][49][50]. Wang et al. [47] tested the spiramycin and ampicillin antibiotics on Microcystis aeruginosa at different equitoxic ratios. The study found the equivalent ratio (1:1) of the binary mixture of spiramycin and ampicillin showed a synergistic interaction at low concentrations and an antagonistic interaction at high concentrations. In a study Nica et al. [48], they tested five veterinary pharmaceuticals for the interaction of synergistic, additive and antagonistic effects on A. fischeri by the combination index of the CA and IA models. They found inversion phenomena of antagonism at high concentrations and synergism at low concentrations from six binary combinations and one pentanary mixture with individual predicted no-effect concentrations. The authors assumed that these phenomena seemed to be independent of the mode of action, which are likely complex and mostly unknown in nature. Rodea-Palomares et al. [49] also found interesting results, i.e., opposite interactions between different aquatic organisms of cyanobacteria (Anabaena CPB 4337) and A. fischeri for three pharmaceuticals. The authors reported a tandemly changing interaction from antagonistic and additive effects at low effective concentration levels to synergistic effects at high effective concentration levels in an A. fischeri test for binary and tertiary mixtures.
As shown in Table 5, Anabaena tests showed a converse pattern against A. fischeri toxicity results. Ismael et al. carefully assumed that pharmaceuticals were shared by a common binding motif such as the same target and receptor sites. Because of the structural similarity, these unexpected interactions were shown between different aquatic organisms. Gonzalez-Pleiter et al. [50] also tested cyanobacteria (Anabaena CPB 4337) for levofloxacin and tetracycline. They found synergism at low effective concentration levels and antagonism at high effective concentration levels. Based on these results, we assume that the unexpected interaction was caused by tetracycline. In some cases, the interactions of experimental toxicity screening were different from predictive toxicity results with inversion phenomena of the interaction [47][48][49][50]. This observation suggests that the binary components could make the synergistic effects at low concentrations. Wang et al. [51] reported that the reaction mechanism differs between short-term (acute) and long-term (chronic) exposures because of quorum sensing molecules which known as ain and lux in A. fischeri [51]. So that better understanding the accurate synergistic interaction, chronic toxicity tests are required. In this study, the exposure levels of the tested chemicals were less than their environmentally relevant concentrations. Thus, further studies are needed to determine how substances as synergists interact with biomolecules at low concentrations from different model organisms including A. fischeri.
As shown in Figure 3 and Table 4, the reason for not available (i.e., n.a.) data indicates that they did not reach the EC 50 and EC 10 ranges. Thus, all binaries of the EC 10 ratio mixtures (i.e., EC 10mix ) were not calculated for interactions from the binary mixture at EC 50 ranges. Two binaries of EC 10mix (Mixture 14 and 18) did not reach the experimental data at the EC 10 ranges. Four binaries of the EC 10mix (Mixture 2, 4, 16, and 30) showed synergistic interactions at the EC 10 ranges, whereas seven additive interactions of EC 10mix (Mixture 6, 8, 10, 12, 20, 22, and 26) and two antagonistic interactions of EC 10mix (Mixture 24, and 28) occurred at EC 10 ranges. Figure 4 illustrates the cumulative distribution of MDRs. In total, the binary combinations of pharmaceuticals and pesticides detected in major river basins in Korea showed 48%-additive (n = 20 from Table 4 excluding combinations without MDR), 40%-antagonistic (n = 17), and 12%-synergistic (n = 6) toxicity effects from 43 binaries on the basis of the MDR values at high (EC 50 ) and low effect (EC 10 ) ranges. In this study, the EC 10 ratio mixtures were at a low overall effect range and showed a general tendency to have more synergistic effects than EC 50 ratio mixtures at the high effect range. Table 5, Anabaena tests showed a converse pattern against A. fischeri toxicity results. Ismael et al. carefully assumed that pharmaceuticals were shared by a common binding motif such as the same target and receptor sites. Because of the structural similarity, these unexpected interactions were shown between different aquatic organisms. Gonzalez-Pleiter et al. [50] also tested cyanobacteria (Anabaena CPB 4337) for levofloxacin and tetracycline. They found synergism at low effective concentration levels and antagonism at high effective concentration levels. Based on these results, we assume that the unexpected interaction was caused by tetracycline. In some cases, the interactions of experimental toxicity screening were different from predictive toxicity results with inversion phenomena of the interaction [47][48][49][50]. This observation suggests that the binary components could make the synergistic effects at low concentrations. Wang et al. [51] reported that the reaction mechanism differs between short-term (acute) and long-term (chronic) exposures because of quorum sensing molecules which known as ain and lux in A. fischeri [51]. So that better understanding the accurate synergistic interaction, chronic toxicity tests are required. In this study, the exposure levels of the tested chemicals were less than their environmentally relevant concentrations. Thus, further studies are needed to determine how substances as synergists interact with biomolecules at low concentrations from different model organisms including A. fischeri.

As shown in
As shown in Figure 3 and Table 4, the reason for not available (i.e., n.a.) data indicates that they did not reach the EC50 and EC10 ranges. Thus, all binaries of the EC10 ratio mixtures (i.e., EC10mix) were not calculated for interactions from the binary mixture at EC50 ranges. Two binaries of EC10mix (Mixture 14 and 18) did not reach the experimental data at the EC10 ranges. Four binaries of the EC10mix (Mixture 2, 4, 16, and 30) showed synergistic interactions at the EC10 ranges, whereas seven additive interactions of EC10mix (Mixture 6, 8, 10, 12, 20, 22, and 26) and two antagonistic interactions of EC10mix (Mixture 24, and 28) occurred at EC10 ranges. Figure 4 illustrates the cumulative distribution of MDRs. In total, the binary combinations of pharmaceuticals and pesticides detected in major river basins in Korea showed 48%-additive (n = 20 from Table 4 excluding combinations without MDR), 40%-antagonistic (n = 17), and 12%-synergistic (n = 6) toxicity effects from 43 binaries on the basis of the MDR values at high (EC50) and low effect (EC10) ranges. In this study, the EC10 ratio mixtures were at a low overall effect range and showed a general tendency to have more synergistic effects than EC50 ratio mixtures at the high effect range.  Table 4, excluding combinations without MDR) for A. fischeri (synergism: MDR > 2; additivity: 0.5 ≤ MDR ≤ 2; and antagonism: MDR < 0.5).  Table 4, excluding combinations without MDR) for A. fischeri (synergism: MDR > 2; additivity: 0.5 ≤ MDR ≤ 2; and antagonism: MDR < 0.5).

Conclusions
In this study, the toxicity of the six target compounds (e.g., four pharmaceuticals and two pesticides) detected in major river basins in South Korea and their binary mixtures (30 samples) were tested at high and low effect concentrations (e.g., EC50 and EC10 ratio mixtures) with luminescent bacterium A. fischeri. Thus, their toxicological interactions were evaluated at 60 effect concentration data points in total and based on model deviation ratios (MDRs) between predicted and observed toxicity values. The mixture toxicities of these mixtures were also predicted by the CA model to evaluate their toxicological interactions (e.g., additive, synergistic, and antagonistic effects) based on the MDR value. From the 60 data points, MDRs were not possible for 17 points since their toxicities could not be measured. The result showed 48%-additive (n = 20), 40%-antagonistic (n = 17), and 12%-synergistic (n = 6) toxicity effects from 43 binaries (excluding 17 combinations without MDRs). That is, from the mixture toxicity evaluation and modeling, we found twenty combinations of additive effects, seventeen combinations of antagonistic effects and six combinations of synergistic effects. In addition, we found inversion phenomena such as synergism at low concentrations and additive antagonism at high concentrations. The exposure levels of the tested chemicals were less than their environmentally relevant concentrations. Since the environmentally relevant concentrations of pesticides and pharmaceuticals detected in the aquatic environment can be present at low concentrations, further studies with different species need to be conducted to clarify the mechanisms, which can address what creates these inversion phenomena.
Supplementary Materials: The following are available online at http://www.mdpi.com/1660-4601/16/2/208/s1, Table S1: Selected pesticides and pharmaceuticals, which were identified in major river basins in South Korea, Table S2: The regression models employed in describing the dose-response curves in this study.