The mode of toxic action of ionic liquids: Narrowing down possibilities using high-throughput, in vitro cell-based bioassays

• Modes of toxic action of ionic liquids were investigated using reporter gene assays and the Tox21 database. • Many ionic liquids exerted cytotoxicity higher than the predicted by baseline toxicity model. • Most ionic liquids were not metabolized by cells, but benzalkonium chlorides underwent side-chain oxidation. • The effects of ionic liquids on aromatase activity and on mitochondrial membrane potential were non-specific. Growing concerns about the environmental impact of ionic liquids (ILs) have spurred research into their (eco)toxic effects, but studies on their mode of toxic action (MOA) still remain limited. However, understanding the MOA and identifying structural features responsible for enhanced toxicity is crucial for characterising the hazard and designing safer alternatives. Therefore, 45 ILs, with systematically varied chemical structures, were tested for cytotoxicity and two specific endpoints in reporter gene assays targeting the Nrf2-ARE mediated oxidative stress response (AREc32) and aryl hydrocarbon receptor activation (AhR-CALUX). While none of the ILs activated the reporter genes, cytotoxicity was high and markedly different between cell lines. Seven and 25 ILs proved more cytotoxic than predicted by baseline toxicity model in the AREc32 and the AhR-CALUX assays, respectively. The length of the side chain and headgroup structures of ILs altered the MOA of ILs. Cellular metabolism of the ILs, investigated by LC-MS/MS, showed side-chain oxidation of the long-chain quaternary ammonium compounds in AhR-CALUX cells and, to a lower extent, in AREc32 cells, however, this transformation could not explain the high cytotoxicity. Effect data for 72 ILs for ten endpoints retrieved from the Tox21 database identified the inhibition of aromatase activity and of mitochondrial membrane potential as potential MOAs. However, in vitro fluorimetric assays for these endpoints demonstrated that effects were activated in a non-specific manner, probably through cytotoxicity. Although many of the ILs tested induced cytotoxicity at concentrations lower than baseline toxicity, the specific MOAs responsible could not be identified. Alternatively, we suggest that the descriptors currently used may fail to define the affinity of ILs for cells. Testing of the affinity of ILs for a diverse range of biomolecules is needed to accurately describe their interactions with cells.