The Relationship between Protein–Protein Interactions and Liquid–Liquid Phase Separation for Monoclonal Antibodies

[Image: see text] Being able to predict and control concentrated solution properties for solutions of monoclonal antibodies (mAbs) is critical for developing therapeutic formulations. At higher protein concentrations, undesirable solution properties include high viscosities, opalescence, particle formation, and precipitation. The overall aim of this work is to understand the relationship between commonly measured dilute solution parameters, the reduced osmotic second virial coefficient b(22) and the diffusion interaction parameter k(D) and liquid–liquid phase separation, which occurs at higher protein concentrations. For globular proteins such as lysozyme or γB-crystallin, the location of the liquid–liquid coexistence curve is controlled by the net protein–protein interaction, which is related to b(22). Because many mAbs undergo reversible self-association due to forming highly directional interactions, it is not known if b(22) can be used as a reliable predictor for LLPS since increasing the anisotropy in the interaction potential causes phase separation to occur at much stonger net protein–protein attractions or lower values of b(22). Here, we map the coexistence curves for three mAbs, referred to as COE-01, COE-07, and COE-19, in terms of b(22) and k(D) values. The measurements are carried out at a low salt condition near the pI, where protein–protein interactions are expected to be anisotropic due to the presence of electrostatic attractions, and under salting-out conditions at high ammonium sulfate concentrations, which is expected to reduce the anisotropy by screening electrostatic interactions. We also show that deviations from a linear correlation between b(22) and k(D) can be used as an indicator of reversible self-association. Each of the mAbs under salting-out conditions follows the correlation supporting the hypothesis that protein–protein interactions are nonspecific, while deviations from the correlation occur for COE-01 and COE-19 under low salt conditions indicating the mAbs undergo reversible self-association. For five out of the six conditions, the onset of phase separation, as reflected by the reduced virial coefficient at the critical point b(22)(c) occurs in a small window −1.6 > b(22)(c) > −2.3, which is similar to what has been observed for lysozyme and for bovine γB-crystallin. Under low salt conditions, b(22)(c) ≈ −5.1 for COE-19, which we previously showed to self-associate into small oligomers. Our findings suggest that under conditions where mAb interactions are weakly anisotropic, such as occur at high salt conditions, phase separation will begin to occur in a small window of b(22). Deviations from the window can occur when mAbs undergo reversible self-association, although this is not always the case and likely depends upon whether or not highly directional interactions are passivated in the oligomer formation. We expect fitting LLPS measurements to simplified interaction models for mAbs will provide additional insight into the nature of the protein–protein interactions and guide their development for calculating concentrated solution properties.