Deriving Organism-Relevant Nanoparticle Safety Exposure Limits from In Vitro Toxicity Data

Abstract

Establishing human safety limits of engineered nanoparticles remains a challenge. In vitro assays have become a major toxicity data resource, driven by the demand for new approach methodologies (NAMs) that reduce animal testing. However, robust tools that bridge in vitro data with in vivo outcomes are still lacking. Here, we present an integrated computational framework using physiologically based pharmacokinetic (PBPK) modeling to derive organism-level BMD10-based dose estimates directly from in vitro liver toxicity data. For 4 nondissolvable nanoparticles (TiO2, SiO2, Au, and graphene oxide), we curated in vitro toxicity data sets spanning 7 liver-relevant cell systems (5 human and 2 mouse) across 24 assays and estimated in vitro effect thresholds using the benchmark dose for a 10% response (BMD10). Mouse in vitro BMD10 values were then reverse-translated into predicted BMD10IVIVE values using a validated mouse PBPK model and benchmarked against available mouse-derived in vivo effect threshold (BMD10eq) data to ensure consistency. The resulting nanoparticle-specific BMD10IVIVE ranges showed qualitative concordance with animal-derived reference doses, supporting the biological plausibility of these screening-level dose estimates. We further applied the same workflow to human liver cell systems to derive human-relevant BMD10-based points of departure. This approach establishes a scalable and prospective animal-free workflow for human nanoparticle risk assessments, overcoming a critical translational gap.