ABSTRACT
Environmental risk assessments of organic chemicals usually do not consider pH as a key factor. Hence, most substances are tested at a single pH only, which may underestimate the toxicity of ionisable substances with a pKa in the range of 4-10. Thus, the ability to consider the pH-dependent toxicity would be crucial for a more realistic assessment. Moreover, there is a tendency in acute toxicity tests to focus on mortality only, while little attention is paid to sublethal endpoints. We used Danio rerio embryos exposed to ten ionisable substances (the acids diclofenac, ibuprofen, naproxen and triclosan and the bases citalopram, fluoxetine, metoprolol, propranolol, tramadol and tetracaine) at four external pH levels, investigating the endpoints mortality (LC50) and heart rate (EC20). Dose-response curves were fitted with an ensemble-model to determine the true uncertainty and variation around the mean endpoints. The ensemble considers eight (heart rate) or twelve (mortality) individual models for binominal and Poisson distributed data, respectively, selected based on the Akaike Information Criterion (AIC). In case of equally good models, the mean endpoint of all models in the ensemble was calculated, resulting in more robust ECx estimates with lower 'standard errors' as compared to randomly selected individual models. We detected a high correlation between mortality (LC50) at 96 hpf and reduced heart rate (EC20) at 48 hpf for all compounds and all external pH levels (r = 0.98). Moreover, the observed pH-dependent effects were strongly associated with log D and thus, likely driven by differences in uptake (toxicokinetic) rather than internal (toxicodynamic) processes. Prospectively, the a priori consideration of pH-dependent effects of ionisable substances might make testing at different pH levels redundant, while the endpoint of mortality might even be replaced by a reliable sublethal proxy that would reduce the exposure, accelerating the evaluation process.
