Classification and regression machine learning models for predicting the combined toxicity and interactions of antibiotics and fungicides mixtures.

Antibiotics and triazole fungicides coexist in varying concentrations in natural aquatic environments, resulting in complex mixtures. These mixtures can potentially affect aquatic ecosystems. Accurately distinguishing synergistic and antagonistic mixtures and predicting mixture toxicity are crucial for effective mixture risk assessment. We tested the toxicities of 75 binary mixtures of antibiotics and fungicides against Auxenochlorella pyrenoidosa. Both regression and classification models for these mixtures were developed using machine learning models: random forest (RF), k-nearest neighbors (KNN), and kernel k-nearest neighbors (KKNN). The KKNN model emerged as the best regression model with high values of determination coefficient (R2 = 0.977), explained variance in prediction leave-one-out (Q2LOO = 0.894), and explained variance in external prediction (Q2F1 = 0.929, Q2F2 = 0.929, and Q2F3 = 0.923). The RF model, the leading classifier, exhibited high accuracy (accuracy = 1 for the training set and 0.905 for the test set) in distinguishing the synergistic and antagonistic mixtures. These results provide crucial value for the risk assessment of mixtures.

Mechanism of the Synergistic Toxicity of Ampicillin and Cefazoline on Selenastrum capricornutum.

Ampicillin (AMP) and cefazolin (CZO) are commonly used ß-lactam antibiotics which are extensively globally produced. Additionally, AMP and CZO are known to have relatively high ecotoxicity. Notably, the mix of AMP and CZO creates a synergistic effect that is more harmful to the environment, and how exposure to AMP-CZO can induce synergism in algae remains virtually unknown. To yield comprehensive mechanistic insights into chemical toxicity, including dose-response relationships and variations in species sensitivity, the integration of multiple endpoints with de novo transcriptomics analyses were used in this study. We employed Selenastrum capricornutum to investigate its toxicological responses to AMP and CZO at various biological levels, with the aim of elucidating the underlying mechanisms. Our assessment of multiple endpoints revealed a significant growth inhibition in response to AMP at the relevant concentrations. This inhibition was associated with increased levels of reactive oxygen species (ROS) and perturbations in nitrogen metabolism, carbohydrate metabolism, and energy metabolism. Growth inhibition in the presence of CZO and the AMP-CZO combination was linked to reduced viability levels, elevated ROS production, decreased total soluble protein content, inhibited photosynthesis, and disruptions in the key signaling pathways related to starch and sucrose metabolism, ribosome function, amino acid biosynthesis, and the production of secondary metabolites. It was concluded from the physiological level that the synergistic effect of Chlorophyll a (Chla) and Superoxide dismutase (SOD) activity strengthened the growth inhibition of S. capricornutum in the AMP-CZO synergistic group. According to the results of transcriptomic analysis, the simultaneous down-regulation of LHCA4, LHCA1, LHCA5, and sodA destroyed the functions of the photosynthetic system and the antioxidant system, respectively. Such information is invaluable for environmental risk assessments. The results provided critical knowledge for a better understanding of the potential ecological impacts of these antibiotics on non-target organisms.

Toxic interactions at the physiological and biochemical levels of green algae under stress of mixtures of three azole fungicides.

Assessing the interactions between environmental pollutants and these mixtures is of paramount significance in understanding their negative effects on aquatic ecosystems. However, existing research often lacks comprehensive investigations into the physiological and biochemical mechanisms underlying these interactions. This study aimed to reveal the toxic mechanisms of cyproconazole (CYP), imazalil (IMA), and prochloraz (PRO) and corresponding these mixtures on Auxenochlorella pyrenoidosa by analyzing the interactions at physiological and biochemical levels. Higher concentrations of CYP, IMA, and PRO and these mixtures resulted in a reduction in chlorophyll (Chl) content and increased total protein (TP) suppression, and malondialdehyde (MDA) content exhibited a negative correlation with algal growth. The activity of catalase (CAT) and superoxide dismutase (SOD) decreased with increasing azole fungicides and their mixture concentrations, correlating positively with growth inhibition. Azole fungicides induced dose-dependent apoptosis in A. pyrenoidosa, with higher apoptosis rates indicative of greater pollutant toxicity. The results revealed concentration-dependent toxicity effects, with antagonistic interactions at low concentrations and synergistic effects at high concentrations within the CYP-IMA mixtures. These interactions were closely linked to the interactions observed in Chl-a, carotenoid (Car), CAT, and cellular apoptosis. The antagonistic effects of CYP-PRO mixtures on A. pyrenoidosa growth inhibition can be attributed to the antagonism observed in Chl-a, Chl-b, Car, TP, CAT, SOD, and cellular apoptosis. This study emphasized the importance of gaining a comprehensive understanding of the physiological and biochemical interactions within algal cells, which may help understand the potential mechanism of toxic interaction.

Toxicity interactions of azole fungicide mixtures on Chlorella pyrenoidosa.

It is acknowledged that azole fungicides may release into the environment and pose potential toxic risks. The combined toxicity interactions of azole fungicide mixtures, however, are still not fully understood. The combined toxicities and its toxic interactions of 225 binary mixtures and 126 multi-component mixtures on Chlorella pyrenoidosa were performed in this study. The results demonstrated that the negative logarithm 50% effect concentration (pEC50 ) of 10 azole fungicides to Chlorella pyrenoidosa at 96 h ranged from 4.23 (triadimefon) to 7.22 (ketoconazole), while the pEC50 values of the 351 mixtures ranged from 3.91 to 7.44. The high toxicities were found for the mixtures containing epoxiconazole. According to the results of the model deviation ratio (MDR) calculated from the concentration addition (MDRCA ), 243 out of 351 (69.23%) mixtures presented additive effect at the 10% effect, while the 23.08% and 7.69% of mixtures presented synergistic and antagonistic effects, respectively. At the 30% effect, 47.29%, 29.34%, and 23.36% of mixtures presented additive effects, synergism, and antagonism, respectively. At the 50% effect, 44.16%, 34.76%, and 21.08% of mixtures presented additive effects, synergism, and antagonism, respectively. Thus, the toxicity interactions at low concentration (10% effect) were dominated by additive effect (69.23%), whereas 55.84% of mixtures induced synergism and antagonism at high concentration (50% effect). Climbazole and imazalil were the most frequency of components presented in the additive mixtures. Epoxiconazole was the key component induced the synergistic effects, while clotrimazole was the key component in the antagonistic mixtures.

Integration of Zeolite Membrane Bioreactor With Granular Sludge-Based Anammox in High-Efficiency Nitrogen Removal From Iron Oxide Red Wastewater.

Acquisition of stable nitritation and efficient anammox play a crucial role in partial nitritation (PN) combined with anammox for nitrogen removal from ammonium-rich wastewater. Due to the limitation of ammonia-oxidizing bacteria (AOB) enrichment and nitrite-oxidizing bacteria (NOB) control in traditional membrane biological reactor (MBR), it can result in a lower nitrite production rate (NPR) and unstable PN, eventually reducing the nitrogen removal rate (NRR) via PN-anammox. In this study, we developed a zeolite membrane biological reactor (ZMBR) to enhance the PN of iron oxide red wastewater (IORW), in which the biofilm derived from the zeolite surface can provide free ammonia (FA)-containing microenvironment for AOB enrichment and NOB inhibition. The results showed that ZMBR can tolerate a higher influent nitrogen loading rate (NLR) of 2.78 kg/(m3⋅day) in comparison to the traditional MBR [2.02 kg/(m3⋅day)] and the NPR in ZMBR and traditional MBR were 1.39 and 0.96 kg/(m3⋅day), respectively. The mass concentration ratio of NO 2 - -N/ NH 4 + -N ranged from 1.05 to 1.33 in ZMBR, suggesting a suitable condition for nitrogen removal via anammox. Subsequently, the domesticated granular sludge obtained from a paper-making wastewater treatment was used as the carrier of anammox bacteria to remove nitrogen. After 93 days of operation, the NRR was observed to be 2.33 kg/(m3⋅day) and high-throughput sequencing indicated that the relatively higher abundance (45.0%) of Candidatus Kuenenia stuttgartiensis was detected in the granular sludge of the bottom part of the reactor, which can produce more proteins and lipids, suggesting a good settleability. Overall, this study provides a high-efficient method to control PN and domesticate anammox for nitrogen removal from IORW.

Concentration Addition, Independent Action, and Quantitative Structure-Activity Relationships for Chemical Mixture Toxicities of the Disinfection By products of Haloacetic Acids on the Green Alga Raphidocelis subcapitata.

The potential toxicity of haloacetic acids (HAAs), common disinfection by products (DBPs), has been widely studied; but their combined effects on freshwater green algae remain poorly understood. The present study was conducted to investigate the toxicological interactions of HAA mixtures in the green alga Raphidocelis subcapitata and predict the DBP mixture toxicities based on concentration addition, independent action, and quantitative structure-activity relationship (QSAR) models. The acute toxicities of 6 HAAs (iodoacetic acid [IAA], bromoacetic acid [BAA], chloroacetic acid [CAA], dichloroacetic acid [DCAA], trichloroacetic acid [TCAA], and tribromoacetic acid [TBAA]) and their 68 binary mixtures to the green algae were analyzed in 96-well microplates. Results reveal that the rank order of the toxicity of individual HAAs is CAA > IAA ≈ BAA > TCAA > DCAA > TBAA. With concentration addition as the reference additive model, the mixture effects are synergetic in 47.1% and antagonistic in 25%, whereas the additive effects are only observed in 27.9% of the experiments. The main components that induce synergism are DCAA, IAA, and BAA; and CAA is the main component that causes antagonism. Prediction by concentration addition and independent action indicates that the 2 models fail to accurately predict 72% mixture toxicity at an effective concentration level of 50%. Modeling the mixtures by QSAR was established by statistically analyzing descriptors for the determination of the relationship between their chemical structures and the negative logarithm of the 50% effective concentration. The additive mixture toxicities are accurately predicted by the QSAR model based on 2 parameters, the octanol-water partition coefficient and the acid dissociation constant (pKa ). The toxicities of synergetic mixtures can be interpreted with the total energy (ET ) and pKa of the mixtures. Dipole moment and ET are the quantum descriptors that influence the antagonistic mixture toxicity. Therefore, in silico modeling may be a useful tool in predicting disinfection by-product mixture toxicities. Environ Toxicol Chem 2021;40:1431-1442. © 2021 SETAC.

Toxic mechanism of three azole fungicides and their mixture to green alga Chlorella pyrenoidosa.

Currently, few studies have investigated the joint toxicity mechanism of azole fungicides at different exposure times and mixed at the relevant environmental concentrations. In this study, three common azole fungicides, namely, myclobutanil (MYC), propiconazole (PRO), and tebuconazole (TCZ), were used in studying the toxic mechanisms of a single substance and its ternary mixture exposed to ambient concentrations of Chlorella pyrenoidosa. Superoxide dismutase (SOD), catalase (CAT), chlorophyll a (Chla), and total protein (TP), were used as physiological indexes. Results showed that three azole fungicides and ternary mixture presented obvious time-dependent toxicities at high concentrations. MYC induced a hormetic effect on algal growth, whereas PRO and TCZ inhibit algal growth in the entire range of the tested concentrations. The toxicities of the three azole fungicides at 7 days followed the order PRO > TCZ > MYC. Three azole fungicides and their ternary mixture induced different levels of SOD and CAT activities in algae at high concentrations. The ternary mixture showed additive effects after 4 and 7 days exposure, but no effect was observed at actual environmental concentrations. The toxic mechanisms may be related to the continuous accumulation of reactive oxygen species, which not only affected protein structures and compositions but also damaged thylakoid membranes, hindered the synthesis of proteins and chlorophyll a, and eventually inhibited algal growth. These findings increase the understanding of the ecotoxicity of azole fungicides and use of azole fungicides in agricultural production.

Ecological and human health risk of sulfonamides in surface water and groundwater of Huixian karst wetland in Guilin, China.

Sulfonamide antibiotics are contaminants of emerging concern (CEC). These CECs raise considerable alarm because they are commonly present in water environments. Studies on the environmental existence of CECs in karst areas of Guilin (Southern China) have yet to be reported. Thus, this study aims to investigate the presence, temporal and spatial distributions of sulfonamides in surface water and groundwater of four major aquatic environments (i.e., aquafarm water, ditch water, wetland water, and groundwater) in the Huixian karst wetland system of Guilin. Furthermore, this study aims to determine the ecological and human health risks of individual sulfonamides and their mixtures. Ten sulfonamides (i.e., sulfadiazine, sulfapyridine, sulfamerazine, trimethoprim, sulfamethazine, sulfamethoxypyridazine, sulfachloropyridazine, sulfamethoxazole, sulfadimethoxine, and sulfaquinoxaline) were observed in the study area. The highest average concentrations of aquafarm water, ditch water, wetland water, and groundwater were those of sulfadiazine (48.24 µg/L), sulfamethoxypyridazine (1281.50 µg/L), sulfamethoxazole (51.14 µg/L), and sulfamethazine (20.06 µg/L), respectively. The potential ecological risks of the detected compounds were much higher in ditch water than in aquafarm water, wetland water, and groundwater. The most ecological risks were observed for sulfachloropyridazine with a risk quotient (RQ) reaching 335.5 to green algae and 152 to Daphnia magna in ditch water. Similarly, sulfachloropyridazine posed the highest ecological risks to green algae among the ten sulfonamides in aquafarm water (RQ = 3.39), wetland water (RQ = 2.98), and groundwater (RQ = 3.6). Human health risk for age groups<12 months was observed from sulfonamide in drinking groundwater. Ecological and human health risks caused by sulfonamide mixtures were larger than the individual risks. Overall, ecological and human health risks caused by sulfonamides were observed in the study area.

Synergetic effects of novel aromatic brominated and chlorinated disinfection byproducts on Vibrio qinghaiensis sp.-Q67.

Aromatic halogenated chemicals are an unregulated class of byproducts (DBPs) generated from disinfection processes in the water environment. Information on the toxicological interactions, such as antagonism and synergism, present in DBP mixtures remains limited. This study aimed to determine the toxicological effects of aromatic halogenated DBP mixtures on the freshwater bacterium Vibrio qinghaiensis sp.-Q67. The acute toxicities of seven DBPs and their binary mixtures toward V. qinghaiensis sp.-Q67 were determined through microplate toxicity analysis. The toxicities of single DBPs were ranked as follows: 2,5-dibromohydroquinone > 2,4-dibromophenol > 4-bromo-2-chlorophenol ≈ 2,6-dibromo-4-nitrophenol > 2,6-dichloro-4-nitrophenol > 2-bromo-4-chlorophenol > 4-bromophenol. The percentages of synergism (experimental values higher than the predicted concentration addition) on the levels of 50%, 20%, and 10% effective concentrations reached 61%, 41%, and 31%, respectively. These results indicated that the probability of synergism decreased as concentration levels decreased. The synergetic effects of the compounds were dependent on concentration levels and concentration ratios. The proposed quantitative structure-activity relationship model can be used to predict the interactive toxicities exerted by 105 binary DBP mixture rays of 21 DBP mixture systems.

Predicting the cytotoxicity of disinfection by-products to Chinese hamster ovary by using linear quantitative structure-activity relationship models.

A suitable model to predict the toxicity of current and continuously emerging disinfection by-products (DBPs) is needed. This study aims to establish a reliable model for predicting the cytotoxicity of DBPs to Chinese hamster ovary (CHO) cells. We collected the CHO cytotoxicity data of 74 DBPs as the endpoint to build linear quantitative structure-activity relationship (QSAR) models. The linear models were developed by using multiple linear regression (MLR). The MLR models showed high performance in both internal (leave-one-out cross-validation, leave-many-out cross-validation, and bootstrapping) and external validation, indicating their satisfactory goodness of fit (R2 = 0.763-0.799), robustness (Q2LOO = 0.718-0.745), and predictive ability (CCC = 0.806-0.848). The generated QSAR models showed comparable quality on both the training and validation levels. Williams plot verified that the obtained models had wide application domains and covered the 74 structurally diverse DBPs. The molecular descriptors used in the models provided comparable information that influences the CHO cytotoxicity of DBPs. In conclusion, the linear QSAR models can be used to predict the CHO cytotoxicity of DBPs.

Joint toxicity of six common heavy metals to Chlorella pyrenoidosa.

Six common heavy metals (Ni, Fe, Zn, Pb, Cd, and Cr) in the water environment were selected to present five groups of binary mixture systems (Ni-Fe, Ni-Zn, Ni-Pb, Ni-Cd, and Ni-Cr) through a direct equipartition ray design. Microplate toxicity analysis based on Chlorella pyrenoidosa measured the 96-h joint toxicities of the binary mixtures. Toxicity interaction of the binary mixture was analyzed by comparing the observed toxicity data with the reference model (concentration addition). The results indicated that Ni-Fe, Ni-Pb, and Ni-Cr mixtures showed additive effects at concentration tested. It was indicated that Ni-Zn and Ni-Cd mixtures presented additive effects at low concentrations whereas synergistic effects were seen at high concentrations.

QSAR prediction of additive and non-additive mixture toxicities of antibiotics and pesticide.

Antibiotics and pesticides may exist as a mixture in real environment. The combined effect of mixture can either be additive or non-additive (synergism and antagonism). However, no effective predictive approach exists on predicting the synergistic and antagonistic toxicities of mixtures. In this study, we developed a quantitative structure-activity relationship (QSAR) model for the toxicities (half effect concentration, EC50) of 45 binary and multi-component mixtures composed of two antibiotics and four pesticides. The acute toxicities of single compound and mixtures toward Aliivibrio fischeri were tested. A genetic algorithm was used to obtain the optimized model with three theoretical descriptors. Various internal and external validation techniques indicated that the coefficient of determination of 0.9366 and root mean square error of 0.1345 for the QSAR model predicted that 45 mixture toxicities presented additive, synergistic, and antagonistic effects. Compared with the traditional concentration additive and independent action models, the QSAR model exhibited an advantage in predicting mixture toxicity. Thus, the presented approach may be able to fill the gaps in predicting non-additive toxicities of binary and multi-component mixtures.

Two-Stage Prediction on Effects of Mixtures Containing Phenolic Compounds and Heavy Metals on Vibrio qinghaiensis sp. Q67.

Two-stage prediction (TSP) model had been developed to predict toxicities of mixtures containing complex components, but its prediction power need to be further validated. Six phenolic compounds and six heavy metals were selected as mixture components. One mixture (M1) was built with equivalent-effect concentration ratio and four mixtures (M2-M5) were designed with fixed concentration ratio. In M1-M5, the toxicities were well predicted by TSP model, while CA overestimated and IA underestimated the toxicities. In M1-M5, compared with the actual mixture EC50 value, the prediction errors of TSP model (13.9%, 17.9%, 19.2%, and 17.3% and 15.8%, respectively) were significantly lower than those in the CA (higher than 30%) and IA models (20.9%, 33.0%, 20.6%, 21.8% and 12.5%, respectively). Thus, the TSP model performed better than the CA and IA model.

Linear regression model for predicting interactive mixture toxicity of pesticide and ionic liquid.

The nature of most environmental contaminants comes from chemical mixtures rather than from individual chemicals. Most of the existed mixture models are only valid for non-interactive mixture toxicity. Therefore, we built two simple linear regression-based concentration addition (LCA) and independent action (LIA) models that aim to predict the combined toxicities of the interactive mixture. The LCA model was built between the negative log-transformation of experimental and expected effect concentrations of concentration addition (CA), while the LIA model was developed between the negative log-transformation of experimental and expected effect concentrations of independent action (IA). Twenty-four mixtures of pesticide and ionic liquid were used to evaluate the predictive abilities of LCA and LIA models. The models correlated well with the observed responses of the 24 binary mixtures. The values of the coefficient of determination (R (2)) and leave-one-out (LOO) cross-validated correlation coefficient (Q(2)) for LCA and LIA models are larger than 0.99, which indicates high predictive powers of the models. The results showed that the developed LCA and LIA models allow for accurately predicting the mixture toxicities of synergism, additive effect, and antagonism. The proposed LCA and LIA models may serve as a useful tool in ecotoxicological assessment.