Abstract
Developability assessment facilitates the selection of antibody drug candidates with desirable pharmaceutical properties. However, it remains uncertain whether agitation-induced aggregation can be predicted from standard developability parameters. Here, we investigated whether key biophysical parameters predict agitation-induced aggregation of monoclonal antibodies (mAbs). To this end, we generated a benchmark data set by characterizing the aggregation upon agitation in the presence of an air–liquid interface of ten approved mAbs reformulated in a common surfactant-free buffer. The extent of aggregation varied substantially among mAbs and was primarily dependent on antibody identity. Flow imaging microscopy combined with machine learning revealed micrometre-sized aggregates with distinct morphologies, consistent with aggregation at air–liquid interfaces. Examination of thin liquid films and foams confirmed the presence of aggregates directly at the air–liquid interface and, therefore, the critical role of this interface for antibody aggregation during agitation. We then applied fluorescence-based, light scattering, and chromatographic techniques to determine standard developability parameters for each mAb, including apparent melting temperature (Tm), nonreversibility onset temperature (Tnr), aggregation onset temperature (Tagg), diffusion self-interaction parameter (kD), hydrophobic interaction chromatography retention time, and relative monomer yield after isothermal refolding from chemical denaturants. Notably, none of these parameters correlated with agitation-induced aggregation. Finally, we assessed the surface properties of the mAbs via drop shape analysis and found that the combination of surface pressure and elastic modulus yields a good correlation with the concentration of micrometre-sized aggregates formed due to agitation. Overall, these findings highlight limitations in predicting mAb interfacial stability using standard developability assays and underscore the importance of studying antibody behavior at interfaces.