Predicting and evaluating monoclonal antibody protein-protein interactions and aggregation: a combined experimental and computational approach
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Over the past decades, therapeutic proteins have become increasingly important tools for improving human health by treating many serious health conditions and diseases. Monoclonal antibodies (MAbs) are currently the largest subset of therapeutic proteins and are effective medicines in part due to their high selectivity towards drug targets. MAbs, like other proteins, are prone to degradation throughout all stages of production, storage, and administration. Non-native irreversible aggregation is one of the most common degradation mechanisms. It reduces the drug potency and can cause detrimental immunogenic responses. ☐ MAbs are typically formulated at precise and optimized solution conditions to minimize degradation and ensure the medicine reaches the patient in an effective form. Accelerated stability studies and biophysical properties, like conformational stability and protein-protein interactions (PPI), are commonly used to aid the development of stable formulations. The present thesis studies the relationships between various biophysical properties, short-term stability, and longer-term stability for three uniquely behaving therapeutic MAbs using both experimental and computational tools. Empirical modelling was utilized to quantify the relationship between biophysical properties and aggregation as well as to predicting longer-term aggregation. The resulting relationships, models, and techniques can aid the formulation development of future molecules. ☐ The electrostatically mediated PPI of these MAbs influenced key product properties such as aggregation rates, solubility, and phase stability. A predictive model of PPIs would allow for candidate molecules / formulations to be screened with little or no protein material usage. The PPI of these distinctly behaving MAbs was experimentally evaluated at a range of typical formulation conditions. PPI parameters (kD, B22, and G22) were obtained from static and dynamic light scattering measurements and spanned from repulsive to strongly attractive net interactions. Coarse-grained (CG) molecular simulations of PPI (specifically the second osmotic virial coefficient, B22) were compared against experimental PPI measurements across multiple pH and salt conditions, using a model that treats each amino acid explicitly. Predicted B22 values with default model parameters matched experimental B22 values semi-quantitatively for some cases; others required parameter tuning for quantitative accuracy due to effects such as ion binding. Experimental PPI values were also analyzed for each MAb within the context of single-protein properties such as net charge, and domain-based and global dipole moments, however PPI predictions were of greater utility. Overall, the results show that B22 predictions by CG molecular modeling can be an effective computational tool for molecule and/or formulation assessment. ☐ Understanding attractive electrostatic protein-protein interactions necessarily involves identifying oppositely charged regions of the protein surface that interact favorably. This cannot be done reliably if one only considers a single protein in isolation unless there are obvious charge “patches” that result in extreme molecular dipoles. A systematic computational approach was developed to identify the origin of diverse PPI, in terms of which sets of amino acids or individual amino acids are most influential, and determining if there are different patterns of pairwise amino acid interaction “maps” that result in different behaviors. The results highlight interaction “maps” that correspond to cases with qualitatively different net electrostatic PPI for the different MAbs and solution conditions, as well as key sets of residues that contribute to strongly attractive PPI. Residue charges were eliminated computationally, one at a time, for each charged residue in the wild-type sequence. The second osmotic virial coefficient of the resulting mutants was simulated and compared to the wild-type PPI. Mutating certain key residues significantly improved PPI as indicated by interaction “maps”. A more computationally efficient method is also shown that identifies key amino acids based on Mayer-weighted interaction energies. ☐ Overall, this thesis studies and characterizes three monoclonal antibodies over typical formulation conditions using both experimental and computational techniques. Computational tools, experimental relationships, and empirical models developed in this thesis could streamline the formulation development of future pharmaceutical antibodies.