A mode-of-action-based approach for predicting cross-species sensitivity of freshwater invertebrates to chemicals

Chemical pollution is a major and growing pressure on freshwater ecosystems, yet it is unfeasible to test the effects of every chemical on every species. As a result, environmental risk assessment depends heavily on extrapolating toxicity from a few model species and chemicals, often without mechanistic justification. Here, we investigate whether grouping chemicals by toxic mode of action, i.e., chemicals acting through the same biological mechanism, can support prediction of cross-species and cross-chemical variation in sensitivity. Specifically, we address three questions: (1) Do species show more consistent rank-order sensitivity across chemicals that share the same toxic mode of action? (2) Within each mode of action, can variation in potency be explained by differences in chemical hydrophobicity? and (3) Can stable interspecific sensitivity patterns be used to predict toxicity for untested species and chemicals using a sensitivity-ratio read-across approach anchored on Daphnia magna? We generated an acute toxicity dataset for 12 chemicals representing four modes of action (demethylation inhibitors, quinone outside inhibitors, sodium-channel modulators, and aryl hydrocarbon receptor agonists) and nine freshwater invertebrates spanning a diverse range of taxa tested under standardized laboratory conditions. Species sensitivity rankings were strongly conserved within modes of action but not between modes of action. Increasing chemical hydrophobicity was associated with higher toxicity for three modes of action, but this relationship was not consistent across all modes of action. The sensitivity-ratio method, anchored on Daphnia magna, successfully reproduced species sensitivity patterns and predicted mean effect concentration values for untested species with median fold-errors of 1.7-2.3 across modes of action. Overall, the stable trait-driven sensitivity patterns revealed within each mode of action offer a mechanistic foundation for more reliable read-across predictions and a pathway toward more efficient and ecologically meaningful chemical risk assessment.