Extending a ciprofloxacin bioavailability model to predict fluoroquinolone ecotoxicity to aquatic cyanobacteria: the role of pH and dissolved organic carbon.

Realistic environmental risk assessment of ionizable chemicals requires modelling frameworks that extend beyond total concentrations, because environmental factors may affect their bioavailability. For fluoroquinolones (FQs) such as ciprofloxacin (CIP), pH and dissolved organic carbon (DOC) are known drivers of bioavailability, but whether such effects can be extrapolated to structurally related FQs remains unclear. We investigated the influence of pH and DOC on the bioavailability of three FQs, i.e. ciprofloxacin-hemi (CIP-hemi), moxifloxacin (MOX) and ofloxacin (OFX). Bioavailability was quantified using changes in total 50% effect concentrations (EC50total) derived from cyanobacterial growth inhibition tests. The resulting data of tests conducted under various combinations of pH and commercially available DOC was used to recalibrate for these three FQ's an existing CIP-based mechanistic bioavailability model and two associated empirical regression modelling approaches based on either DOC light absorbance at 350 nm (A350) or DOC concentration alone. Model performance was subsequently evaluated using independent ecotoxicity tests in freshwater containing naturally occurring DOC. Our results confirmed that both pH and DOC regulate the bioavailability of CIP-hemi, MOX and OFX, with patterns in EC50total reflecting pH-driven speciation and DOC-mediated bioavailability reduction. Evaluation of various DOC-binding scenarios (i.e. which FQ species are assumed to bind to DOC) revealed that the optimal binding scenario (i.e., best model performance) was compound-specific: (i) interactions of DOC with the positively charged (FQ+) and the zwitterionic (FQ±) species for CIP-hemi, (ii) with FQ+, FQ±, and the negatively charged species (FQ-) for MOX, and (iii) with only FQ± for OFX. When parameterized using the optimal-binding scenario for each FQ, the mechanistic framework consistently outperformed the empirical approaches. Model selection based on the Akaike Information Criterion indicated that the 50% effect concentration of the zwitterion (EC50(FQ±)) and the strength of interaction between the zwitterion and DOC (Kd,FQ±) are the minimum, critically required parameters allowing the mechanistic model to balance complexity with good predictive performance within the tested pH window (7.5-9.0), while the best parameter set required for accurate prediction under the best-performing scenario is compound-specific. Our findings suggest that two key assumptions - i.e. (i) the dominance of the zwitterionic species (FQ+/-) in driving bioavailability, and (ii) the approximation of FQ interaction strength with DOC using the light absorption coefficient ε350 as a predictor variable - are transferable to the FQs investigated.