Engineering unnatural amino acids in peptide linkers enables cathepsin-selective antibody-drug conjugates for HER2-positive breast cancer

Abstract

Antibody-drug conjugates (ADCs) are a rapidly evolving class of targeted cancer therapeutics that combine the specificity of monoclonal antibodies with the potent cytotoxicity of small-molecule payloads. Their clinical success has led to significant advances in oncology, positioning ADCs as a transformative modality in cancer treatment. Most clinically approved ADCs utilize a protease-cleavable valine-citrulline (Val-Cit) dipeptide linker designed to facilitate intracellular payload release upon proteolytic activation by the lysosomal cathepsins. However, the Val-Cit linker is susceptible to off-target cleavage by proteases expressed in non-malignant tissues, resulting in premature payload release and systemic toxicity. To address this limitation, we established a high-throughput peptide linker discovery platform using Hybrid Combinatorial Substrate Library (HyCoSuL) screening to comprehensively profile protease substrate preferences. By incorporating unnatural amino acids, we identified peptide sequences with high selectivity toward cancer-associated proteases, thereby overcoming the constraints of conventional linker design. As proof of concept, we engineered trastuzumab-based ADCs selectively activated by cathepsin B or cathepsin L and evaluated their cytotoxic efficacy in HER2-positive breast cancer models. Our HyCoSuL-guided linkers display higher selectivity toward cathepsins than Val-Cit and enable faster, protease-dependent activation of peptide prodrugs and ADCs in vitro, while also exhibiting enhanced stability in human plasma to minimize premature payload release. Furthermore, recognizing that efficient protease-activated ADC function requires co-expression of both the target antigen and the activating protease, we performed single-cell mass cytometry analysis of patient-derived breast cancer samples to assess the correlation between HER2 and cathepsin expression. This analysis revealed heterogeneous cathepsin expression, underscoring that pairing the antibody target with the appropriate protease-selective linker might critical for robust, tumor-confined payload release. Our findings highlight the importance of patient stratification based on antigen and protease expression profiles, providing a foundation for developing ADCs with greater protease selectivity and minimized off-target activation.

Significance

This work presents a rational chemistry-driven strategy using unnatural amino acids to engineer peptide linkers with high cysteine cathepsins specificity and enhanced stability in human plasma. By combining comprehensive enzymatic profiling with single-cell CyTOF analysis of patient tumors, we outline a framework for precision-guided ADC design and protease-informed patient stratification in HER2-positive breast cancer.