Model-optimized bispecific antibodies improve the selectivity of antibody–drug conjugates for tumors

Abstract

Introduction: Antibody–drug conjugates (ADCs) direct chemotherapeutic payloads to tumors using antibodies targeting tumor-associated proteins. Despite this, ADC efficacy remains constrained by toxicity, some of which is driven by ADC binding to non-tumor tissues expressing the ADC’s target. For example, anti-human epidermal growth factor receptor 2 (HER2) ADCs can induce serious cardiac toxicity related to HER2 expression in the heart, and glembatumumab vedotin causes severe rash plausibly due to expression of its target, glycoprotein non-metastatic melanoma protein B (gpNMB), in the skin. Improving the selectivity of ADCs for tumors relative to non-tumor tissue could widen the therapeutic window of these drugs. Here we show that with careful optimization, bispecific ADCs (bsADCs) can improve tumor selectivity. Materials and methods: To investigate the impact of dual targeting on improving the tumor selectivity of bsADCs, we developed a mechanistic pharmacokinetic/receptor occupancy (PK/RO) model that accounts for antibody affinity and avidity. This model quantitatively predicts on-tumor and off-tumor target binding for bsADCs as a function of bsADC pharmacokinetics, target properties, and antibody binding affinity. Results: Bispecific antibodies can only enhance tumor selectivity of ADCs compared to healthy target-expressing tissues when the affinities for both targets are properly optimized. The extent of bsADC tumor selectivity depends on the on- and off-rates for target binding. This four-dimensional search space is best explored using a computational model, as we have developed here. Improvements in ADC tumor selectivity remain robust to an order of magnitude of variation in target expression levels. Conclusions: Bispecific antibodies may improve targeting specificity of ADCs, but it is critical to optimize affinities for the two tumor-associated antigens with the aid of mathematical modeling. This can improve the ratio of target-mediated on-tumor to off-tumor ADC engagement compared to single-target ADCs. The approach presented here provides a practical method of model-guided optimization of dual targeting, which may have broader applicability across antibody-based therapeutics.