Risk assessment and mixture toxicity of organic micropollutants and their transformation products
Organic micropollutants, such as pesticides, biocides and pharmaceuticals that are widely used in agriculture and/or households, are released into the environment where they are subject to various transformation processes. The number of possible transformation products is overwhelming and many are not even identified yet. In addition to the parent compounds, the emerging transformation products may pose risks to human health and the environment due to their bioaccumulation potential and their toxicity. Thus, classical approaches for their risk assessment are not appropriate and new ways are required to evaluate transformation products and to prioritize them for further testing and refined risk assessment. Predictive models can be used for comparative risk assessment, where the risk of transformation products is evaluated in relation to their parent compound.
A prediction model based on the combination of quantitative-structure-activity relationships (QSAR) for baseline toxicity and a subsequent toxic ratio (TR) analysis was used to account for bioaccumulation and specific effects of transformation products relative to their parent compounds. To account for changes in mode of action due to transformation of a parent compound, structural units of the molecules were identified that are responsible for a specific mode of action (toxicophores). Two case studies were performed to validate the predictive model and to refine it.
In an initial step to prepare the case studies, an experimental mode-of-action classification scheme was developed. To this end, two algae tests were used, both of which yield information on the physiological mode of action by combining the time and effect pattern of different structural and functional endpoints, e.g. photosynthesis inhibition and growth rate inhibition, with a subsequent QSAR/TR-analysis. Mixture toxicity was introduced as a diagnostic tool to substantiate the mode of action analysis.
During the first case study with the antidepressant fluoxetine and its human metabolites these combined experimental and computational tools were shown to be useful. The metabolite p-trifluoromethylphenol was clearly less toxic than fluoxetine and was classified as a baseline toxicant. Fluoxetine and its demethylated metabolite norfluoxetine were equally toxic, and exhibited unusually high TR values, which pointed to a specific mode of action. However, their effect pattern with respect to the different endpoints was similar to the effect pattern of baseline toxicants. A subsequent investigation on the pH-dependent toxicity of five basic pharmaceuticals, all containing an aliphatic amine group like fluoxetine and norfluoxetine, and the development of a simple toxicokinetic ion-trapping model showed that the apparent high TR was caused by a toxicokinetic effect due to speciation and not by a toxicodynamic effect due to a specific mode of action. When speciation of fluoxetine and norfluoxetine was taken into account, they were also classified as baseline toxicants.
The second case study focused on the herbicide diuron and its transformation products. Effects on algae and daphnids were investigated to answer the question of whether transformation processes lead to differences in the speciesȀ sensitivities depending on changes in mode of action. Indeed, for algae, a decrease in toxicity from diuron to 3,4-dichloroaniline with concomitant decrease in the specific effect could be observed. In contrast, for daphnia, an increase in toxicity from diuron to 3,4-dichloroaniline with subsequent formation of a specific mode of action was observed. The decrease of the specific effect in algae could be related to the presence of the methylurea group. In daphnia, cleavage of the methylurea group led to the change from baseline toxicity of diuron to the specific effect of 3,4-dichloroaniline. Mixture toxicity experiments confirmed this mode of action analysis.
Both case studies demonstrated the importance of accounting for transformation products in the risk assessment of organic micropollutants. This is especially relevant for the primary transformation products, which often still possess the same mode of action as the parent and will add to the overall risk of the parent. Further, the formation of new toxicophores may lead to higher toxicity and might even change the target species and thus species sensitivity.
With the findings of the case studies the prediction model was improved. First, the case of fluoxetine clearly demonstrated that the prediction of the bioaccumulation potential of ionizable chemicals posed problems. The difference in liposome-water partitioning between the neutral and the ionized species (Āmw) varied strongly. However, the underlying database of measured liposome-water partitioning coefficients was too small to derive more precise rules for the Āmw prediction of ionizable compounds. Consequently, ionization was implemented into the prediction model by using the earlier suggested generic Āmw of 1. Second, the case study of diuron demonstrated how important it is to account for speciesspecific toxicophores. Therefore, screening for species-specific toxicophore structures was implemented into the model.
In conclusion, the improved prediction model is a systematic and integrated approach to identify transformation products of relevance. Further research should focus on applying the developed prediction model to a broader range of chemicals to prove its predictive power and applicability.