Characterization and Modeling of Reversible Antibody Self-Association Provide Insights into Behavior, Prediction, and Correction

Abstract

Reversible antibody self-association, while having major developability and therapeutic implications, is not fully understood or readily predictable and correctable. For a strongly self-associating humanized mAb variant, resulting in unacceptable viscosity, the monovalent affinity of self-interaction was measured in the low μM range, typical of many specific and biologically relevant protein–protein interactions. A face-to-face interaction model extending across both the heavy-chain (HC) and light-chain (LC) Complementary Determining Regions (CDRs) was apparent from biochemical and mutagenesis approaches as well as computational modeling. Light scattering experiments involving individual mAb, Fc, Fab, and Fab’2 domains revealed that Fabs self-interact to form dimers, while bivalent mAb/Fab’2 forms lead to significant oligomerization. Site-directed mutagenesis of aromatic residues identified by homology model patch analysis and self-docking dramatically affected self-association, demonstrating the utility of these predictive approaches, while revealing a highly specific and tunable nature of self-binding modulated by single point mutations. Mutagenesis at these same key HC/LC CDR positions that affect self-interaction also typically abolished target binding with notable exceptions, clearly demonstrating the difficulties yet possibility of correcting self-association through engineering. Clear correlations were also observed between different methods used to assess self-interaction, such as Dynamic Light Scattering (DLS) and Affinity-Capture Self-Interaction Nanoparticle Spectroscopy (AC-SINS). Our findings advance our understanding of therapeutic protein and antibody self-association and offer insights into its prediction, evaluation and corrective mitigation to aid therapeutic development.

The role of intermolecular interactions on monoclonal antibody filtration through virus removal membranes

“…These values are in good agreement with literature results for other mAbs. [13,16,[19][20][21][22] The slopes of the best fit lines in Figure 3…”

Rapid Estimation of Size-Based Heterogeneity in Monoclonal Antibodies by Machine Learning-Enhanced Dynamic Light Scattering

“…The kD and the sedimentation interacting parameter (kS) are used to determine the virial coefficient (B2). The virial coefficients are indicators of protein–protein interaction (PPI) and/or antibody self-association, , which further has been used to assess the solution viscosities and colloidal stability of highly concentrated mAbs solutions. − Furthermore, kD has been found to be a key biophysical property for the mAbs, exhibiting statistically significant correlations with multiple biophysical attributes of antibody formulations such as apparent solubility, viscosity behavior, and electrostatic properties. , …”

High-throughput profiling of antibody self-association in multiple formulation conditions by PEG stabilized self-interaction nanoparticle spectroscopy

“…Several studies have reported that AC-SINS effectively identifies mAbs with either low solubility or high viscosity. 3 , 5 , 13–15 Considering that viscosity determination typically requires at least 20 mg of protein per condition, whereas AC-SINS consumes only microgram quantities of proteins, the potential of using AC-SINS to screen hundreds of molecules is particularly attractive for early-stage antibody discovery, where large numbers of molecules are available in minute amounts. The previous case studies, such as wild-type (WT) and variants, appear to be supportive of such use.…”