Water treatment processes are essential for drinking water production but may produce unintended transformation products. Identifying these products and their potential toxicity (hazards) is necessary to guarantee safe drinking water. Predictive toxicology can speed up and improve the assessment of these substances.
Drinking water production involves a series of processes to remove pathogens and other contaminants from the water. Chemical contaminants in drinking water sources, especially surface waters, include substances produced by humans such as pharmaceuticals, plant protection products, and biocides. These substances can be transformed (WHO, 2022) during drinking water production. Transformation product (TPs) formation is, to some extent, unavoidable (Anagnostopoulou, 2022), and the physicochemical and toxicological properties of most TPs are unknown (Gassman, 2021). Analytical techniques can identify TPs, but these techniques are expensive and time-consuming. When identified, concentrations of individual TPs and other chemical contaminants are usually below health-related thresholds. Since exposure to TPs generally occurs in the form of low-level mixtures and may last a lifetime, it is vital that we learn more about TPs’ formation and toxicological activities. Predictive toxicology can help in this respect.
New horizons for the assessment of TPs
Predictive toxicology is an innovative approach that uses computerized (in silico) tools, that have shown a promising ability to predict the properties of substances (Raies, 2016). in silico tools are based on algorithms that can predict the formation and toxicity (hazard) of compounds based on their chemical structure. They can contribute to filling data gaps on contaminants, such as TPs, that may be present in drinking water at low concentrations. in silico approaches provide a relevant and cost-effective contribution to the exposure and risk assessment for substances that lack toxicological data, including TPs. In this way, predictive toxicology can direct water quality monitoring, prioritize further experiments and research needs, and reduce the costs and time needed. This article gives an overview of the applicability of in silico tools to the evaluation of TPs in drinking water.
Formation of TPs
The first step in assessing TPs in drinking water using in silico tools is identifying which compounds are expected to be formed. The composition of the mixture of TPs that can occur in drinking water depends on chemical and biological reactions related to the water treatments. Transformations in the environment and organisms, including humans, are not considered here.
In Europe, the most common water treatment processes are chlorination, ozonation, UV treatment, rapid sand filtration, filtration over biologically active carbon, and advanced oxidation. Different combinations of these techniques are used to remove pathogens and other contaminants, depending on the quality of the water sources (WHO, 2022).
Based on knowledge from experimental data (reaction libraries), it is possible to predict TPs potentially formed due to the water treatments. This provides a starting point for further chemical analysis, which can save costs and time when assessing water quality. Simulation by computerized methods can thus be a valuable addition to data on TPs actually found in water. Additional drinking water analyses can validate the predictions and verify the actual presence of TPs.
Toxicity of TPs
The second step is the assessment of the toxicity of TPs (hazard assessment). The toxicity of substances can be assessed not only in vivo (animal experiments) and in vitro(biochemical and cell testing), but also in silico (computerized tools). All these approaches have their fundamental limitations that result in uncertainties in the outcome.
in vivo experiments, by definition, inherently consider the adsorption, distribution, metabolism, and excretion (ADME) of substances in an organism but are generally expensive and time-consuming. Their use should be reduced, replaced, or refined from an ethical point of view.
In vitro experiments examine specific interactions of a substance with biological structures, which may, for example, lead to DNA damage. ADME aspects are only partially addressed, leading to uncertainties in translating the results into effects in humans.
In silico tools work with algorithms that predict toxicity based on chemical structure. They are time and cost-effective, but they are based on and strictly linked to the availability and quality of experimental data. Predictive toxicology only gives reliable predictions of chemicals’ in vitro and in vivo responses if the algorithms are based on high-quality data. A specific in silico model is reliably applicable only for substances that are comparable to experimentally analyzed molecules whose data were used to build the algorithm. In addition, the quality of the prediction depends on the specific toxicological effect (endpoint) evaluated (Figure 1).
The most relevant endpoints for assessing drinking water quality include mutagenicity/genotoxicity (DNA damage), carcinogenicity (tumor formation), reproductive toxicity, developmental toxicity, and endocrine disruption (disturbance of the hormone balance). The better the mechanism of action responsible for the toxicity is understood, and the higher the quality of the available data, the more reliable the in silico models will be. The state of knowledge is now such that in silico tools guarantee good predictions for assessing mutagenicity/genotoxicity. The challenge is significantly greater for endpoints with less clear mechanisms of action, primarily due to the scarcity of experimental data and differences between experimental protocols. This is especially true for complex endpoints such as carcinogenicity, and reproductive and developmental toxicology. In these cases, predictive toxicology is less suitable for assessing drinking water quality. Nevertheless, in silico tools can also contribute to understanding specific effects for these endpoints and can play a role in an integrated approach with other information sources.
Different strategies for the hazard assessment, their relations, and advantages (+) and disadvantages (-).
In silico toxicity assessment of TPs
In silico approaches for toxicity assessment are based on (but not limited at) (Figure 1):
(a) the quantitative structure-activity relationship (QSAR): the recognition of chemical substructures that are predictive of specific toxicological effects (endpoints);
b) the ‘read-across’ approach: the extrapolation of available information on toxicologically known substances to chemicals with a similar structure for which no data are available;
(c) ‘expert judgment’: essential for assessing and adequately interpreting the reliability of the predictions.
Even though most software is user-friendly, the output must be critically evaluated to detect outliers, inconsistencies, or errors in the model. The prediction should be backed up by reasoning and mechanistic interpretations as much as possible to increase the reliability of the results.
Several in silico tools for the prediction and toxicity assessment of chemicals are freely available and there are others that require purchase (Figure 2)
The potential of these techniques is increasingly recognized. Authorities such as the European Food Safety Authority (EFSA), the European Chemicals Agency (ECHA), and the Organization for Economic Cooperation and Development (OECD) suggest the use of these techniques to replace and complement animal testing. Nevertheless, authorities have not yet agreed upon an internationally recognized methodology for applying in silico tools specifically for the evaluation of TPs.
Figure 2. in silico approaches and examples of in silico tools for the prediction of TPs formed during different treatment processes, and specific hazards. 1. Chlorination; 2. Ozonation; 3. UV treatments; 4. Biodegradation; 5. Oxidation processes; 6. Genotoxicity; 7. Carcinogenicity; 8. Reproductive and developmental toxicology; 9. Endocrine disruption.
An integrated approach
Suppose that one scientific source of information does not provide sufficient answers about the toxicity of TPs. In this case, an integrated approach that considers all available evidence may be the solution. On one hand, an (internal) statistical validation of each methodology separately is needed to justify their integration into the approach and, on the other hand, an (external) comparison between different methods to assess the significance and concordance of the results has to be conducted.
In a recent study, Hensen et al. (2020) propose a tiered approach to evaluate TPs’ toxicological effects derived from pesticides. Their first step was a combination of literature review on experimental data and in silico methods, followed by targeted in vitroand in vivo experiments to verify the previous results. This study showed that most TPs (94%) were correctly predicted. Despite the researchers’ warning that this methodology needs further evaluation and development, this is a promising result.
Acknowledgments
This article was produced thanks to research by the Dutch and Flemish drinking water companies (BTO). The authors thank Remi Hoencamp for his contribution.
References
– Anagnostopoulou, K., Nannou, C., Evgenidou, E., & Lambropoulou, D. (2022). Overarching issues on relevant pesticide transformation products in the aquatic environment: A review. Science of the Total Environment. Elsevier B.V. https://doi.org/10.1016/j.scitotenv.2021.152863
– Gassmann, M. (2021). Modelling the Fate of Pesticide Transformation Products From Plot to Catchment Scale—State of Knowledge and Future Challenges. Frontiers in Environmental Science. Frontiers Media S.A. https://doi.org/10.3389/fenvs.2021.717738
– Hensen, B., Olsson, O., & Kümmerer, K. (2020). A strategy for an initial assessment of the ecotoxicological effects of transformation products of pesticides in aquatic systems following a tiered approach. Environment International, 137. https://doi.org/10.1016/j.envint.2020.105533
– Raies, A. B., & Bajic, V. B. (2016). in silico toxicology: computational methods for the prediction of chemical toxicity. Wiley Interdisciplinary Reviews: Computational Molecular Science, 147–172. https://doi.org/10.1002/wcms.1240
– WHO. (2022). Guidelines for drinking-water quality, 4th edition, incorporating the 1st and 2nd addendum. Geneva: World Health Organization. Geneva, Switzerland. World Health Organization
Authors: Ferrario A.S. (KWR Water Research Institute, Utrecht University) M.M.L. Dingemans
(KWR Water Research Institute, Utrecht University) Reus A. (KWR) Hofman-Caris R.
(KWR, Wageningen University)