Bladder cancer exhibits molecular heterogeneity that complicates early diagnosis and prognosis, and drives confounding clinical outcomes. Non-muscle invasive and muscle-invasive subtypes, especially for intermediate to high grade, carry a 25 - 50% progression-free survival rate, underscoring the need for high precision prognostic strategy. Urinary extracellular vesicles (uEVs) are promising carriers of tumor-derived RNAs and proteins. However, significant challenges in studying uEVs arise from the diverse cellular origin of uEVs associated with the dynamic composition of urine, which presents roadblocks for developing the clinical utility of uEVs. We introduced an AI-driven EV liquid biopsy pipeline that integrates (1) standardized EV isolation via NanoPom magnetic beads, (2) transcriptomic profiling for molecular subtyping, and (3) prognostic scoring algorithm. In a discovery cohort of 16 bladder cancer patients including both MIBC and NMIBC, we compared NanoPom isolated uEVs with ExoEasy and Fujifilm MagCapture isolated uEVs, for identifying bladder cancer subtype-specific gene signatures, and externally validated them using UCSC Xena. The result outperformed currently reported bladder cancer diagnostic biomarkers from assays including Galeas, CxBladder, and Xpert. In a validation cohort of matched 7 patient plasma samples, we confirmed with plasma derived EVs for correlating with urinary EV biomarkers from NGS sequencing. The prognostic score stratified patients into low-, intermediate-, and high-grade risk groups based on Xena's bladder cancer survival outcomes. Our AI-driven uEV liquid biopsy pipeline proves the concept for high precision bladder cancer subtyping and prognosis, which could potentially facilitate treatment decision and lead to advanced profiling of bladder tumor biology using uEV liquid biops
Greenberg et al. (Fri,) studied this question.