Machine Learning-Designed Cell-Penetrating Peptide Boosts Antibody-Drug Conjugate Cancer Killing by Enhancing Lysosomal Delivery

NCR peptide delivered >4x more EGFP into cells than TAT peptide and accumulated in lysosomes. Domain Z-NCR vector non-covalently bound T-DM1 and significantly enhanced its cytotoxicity against HER2-positive tumor cells (MTT assay) while maintaining drug specificity for HER2-expressing cells.

In vitro study: MLCPP2.0 machine learning algorithm screened nuclear localization/export signal databases for CPP candidates. Cell-penetrating activity verified by fluorescent tracing. NCR fused to domain Z (IgG Fc-binding) to create T-DM1 delivery vector. Cytotoxicity assessed by MTT assay in HER2-positive tumor cells.

Antibody-drug conjugates like T-DM1 (trastuzumab emtansine) are important cancer therapies but their effectiveness is limited by inefficient cellular uptake and lysosomal delivery. This peptide-based enhancement strategy could improve ADC potency without changing the drug itself, potentially allowing lower doses or overcoming resistance.

This work sits at the intersection of AI-driven peptide design and cancer drug delivery. As ADCs become a major class of cancer therapeutics, strategies to enhance their intracellular delivery could significantly improve clinical outcomes. The machine learning approach also demonstrates how computational tools can accelerate peptide discovery.

In vitro only — no animal tumor models or pharmacokinetic data. The domain Z-NCR vector binds non-covalently, which may be unstable in blood. Only tested with T-DM1 against HER2-positive cells; generalizability to other ADCs unknown. No toxicity or immunogenicity assessment.