Identification of exosome protein biomarkers in patients with advanced hormone receptor-positive breast cancer treated with palbociclib and tamoxifen.

Abstract

e13014 Background: The combination of a CDK4/6 inhibitor (CDK4/6i) plus endocrine therapy (ET) doubles progression free survival compared with ET alone in hormone receptor (HR)-positive, advanced breast cancer (BC), although not all patients respond and responders eventually develop resistance and disease progression. Exosomes are membrane-bound extracellular vesicles that are released from tumors for cell-to-cell transfer of lipids, proteins, and nucleic acids. Analysis of exosome cargo provides a dynamic and functional read-out of biological pathways that are activated in cancer cells. We performed deep proteomic analysis of plasma exosomes from patients receiving palbociclib/tamoxifen to identify protein networks that predict response to CDK4/6i and ET and that may contribute to drug resistance. Methods: The Big Ten Cancer Research Consortium conducted a phase II trial of palbociclib plus tamoxifen as first line therapy for patients with advanced, HR+/HER2- BC (NCT02668666). Whole blood was collected in Streck tubes from all participants at baseline and at time points during study treatment. Plasma was separated and stored at -80C within 48 hours of collection. Exosome extraction and purification was optimized for maximum proteomic coverage. Proteins were labeled with tandem mass tag 10plex and quantified with ultrasensitive mass spectrometry. Detected proteins were mapped to pathways with the Reactome Pathway Database. An unsupervised machine learning approach with modified graphic neural networks was used to determine whether differential expression of protein networks in plasma exosomes predicts treatment response. Results: We detected more than 700 exosome proteins from100 μl plasma in 16 study participants (responders, n = 11; non-responders, n = 5). Significant enrichment of exosome-specific markers was observed when comparing patient samples with healthy donor samples. Exosomal protein networks in pretreatment samples predicted treatment response with 95% sensitivity and 85% specificity in unsupervised clustering. The top weighted protein networks in the treatment response model are enriched for membrane attack complex, complement activation and lipoprotein receptor binding pathways. Conclusions: Ultrasensitive proteomic analysis combined with deep learning methods provides a detailed picture of the proteome landscape of plasma exosomes in advanced breast cancer patients and is ideally suited for serial analyses to study emergence of resistance mechanisms. This approach also demonstrated unparalleled accuracy as a predictive biomarker to identify patients unlikely to respond to CDK4/6i and ET. If results are confirmed, this novel approach could hold great promise for identifying protein biomarkers and mechanisms of resistance that emerge during anticancer therapy. Clinical trial information: NCT02668666.