Selecting and engineering monoclonal antibodies with drug-like specificity

Abstract

Despite the recent explosion in the use of monoclonal antibodies (mAbs) as drugs, it remains a significant challenge to generate antibodies with a combination of physicochemical properties that are optimal for therapeutic applications. We argue that one of the most important and underappreciated drug-like antibody properties is high specificity - defined here as low levels of antibody non-specific and self-interactions - which is linked to low off-target binding and slow antibody clearance in vivo and high solubility and low viscosity in vitro. Here, we review the latest advances in characterizing antibody specificity and elucidating its molecular determinants as well as using these findings to improve the selection and engineering of antibodies with extremely high, drug-like specificity.

Figure 1.. Drug-like antibodies have low levels of non-specific interactions.

Poor antibody specificity manifests itself in a number of observable phenomena both in vitro and in vivo. The impacts of high levels of non-specific (heterotypic) interactions are observed in vivo via fast antibody clearance, which has been linked to off-target binding and aberrant interactions with the neonatal Fc receptor (FcRn). Problems with high levels of antibody self- (homotypic) interactions arise during antibody formulation where it is often necessary to achieve exceedingly high protein concentrations. High levels of antibody self-interactions can lead to high solution viscosity, high aggregation, and/or low solubility. While heterotypic and homotypic interactions are often discussed independently, it is becoming increasingly clear that these undesirable behaviors often share similar molecular determinants, including excessively hydrophobic and/or charged interfaces.

Figure 2.. Approved mAb drugs are more specific on average than mAbs in clinical trials.

Jain and colleagues performed a broad study of the biophysical properties of clinical-stage mAbs using twelve assays to identify the most important biophysical determinants of drug-like antibodies [7^••]. Notably, measurements of antibody specificity (non-specific binding and self-association) were the only ones to statistically differentiate between approved antibody drugs and antibodies in clinical trials. This finding suggests that approved antibody drugs are, on average, more specific than those in clinical trials and highlights the importance of using specificity assays to guide the selection of antibodies with drug-like properties. The reported assays are: (A) baculovirus particle (BVP), (B) ELISA (panel of six biomolecules), (C) polyspecificity reagent (PSR), (D) clone self-interaction by biolayer interferometry (CSI-BLI), (E) affinity-capture self-interaction nanoparticle spectroscopy (AC-SINS), (F) salt gradient affinity-capture self-interaction nanoparticle spectroscopy (SGAC-SINS), (G) expression titer in HEK cells, (H) aggregation (accelerated stability), (I) folding stability (melting temperature of Fab), (J) cross-interaction chromatography (CIC), (K) hydrophobic interaction chromatography (HIC), and (L) standup monolayer adsorption chromatography (SMAC). It is important to note that poor performance in a single biophysical assay does not preclude the successful development of a therapeutic antibody.

Figure 3.. Preclinical assays for evaluating antibody specificity are important tools in drug development.

The rise in awareness of the importance of antibody specificity has prompted the development of several assays that report on different types of non-specific interactions. Antibody self-interactions can be probed using methods such as nanoparticle-based (e.g., AC-SINS) or biolayer interferometry (e.g., CSI-BLI) methods [41,53]. Non-specific interactions are often assessed using ELISA methods but now are also being evaluated using flow cytometry-based (PSR) methods [32,33^•]. FcRn binding assays are also useful for identifying antibodies with increased risk of displaying poor pharmacokinetic properties due to abnormal interactions mediated via the antibody variable regions with FcRn [37,38]. Pharmacokinetic studies in model organisms (e.g., mouse) are a mainstay in drug development, but the creation of transgenic mice with humanized FcRn has been a major advance for the assessment of therapeutic antibodies [54]. Finally, cell binding assays provide a simple yet valuable assessment of potential antibody behavior in vivo [10]. These assays are most powerful when used in combination for preclinical antibody characterization and lead selection [8^••].