Circulating extracellular vesicles carrying PD-1, PD-L1, and CTLA-4 to inform resistance to immunotherapy in HCC.

2567 Background: First-line immune checkpoint inhibitor (ICI) therapy has contributed to improved outcomes in hepatocellular carcinoma (HCC). However, only around 30% of patients respond, highlighting the need for biomarker driven patient selection. Considering the lack of tissue acquirement in HCC in clinical routine, liquid biopsy holds great promise. In this study, we aimed at profiling immune checkpoints (IC) on circulating extracellular vesicles (EVs) as potential liquid biopsy-based biomarkers for prediction of treatment response. Methods: Three distinct cohorts were analyzed in this study: (1) HCC Explorer Cohort (n=40), testing the presence of membrane-bound ICs on EVs; (2) early-stage HCC Cohort (n=37 with paired blood and tissue), assessing the interplay between EV and tissue ICs alongside clinicopathological parameters and (3) Treatment Cohort (n=161 with 548 sequential blood samples), comprising a training (n=79,n=402 sequential samples) and a validation group (n=82, n=146 sequential samples), to identify predictors of immunotherapy response through IC profiling of circulating EVs via multiplex immunoassay. Results: PD-1, PD-L1 and CTLA-4 were enriched in EVs compared to EV-depleted serum. Baseline and early dynamics in EV-IC levels significantly discriminated between responders and non-responders in the training and validation cohort, while simultaneously predicting PFS and OS. In addition, dynamic changes of EV-IC levels during therapy in patients with initial response predicted acquired therapeutic resistance, approximately 36-42 weeks before progression was traceable on imaging. High serum EV-IC levels correlated with “cold” tumor features on paired tissue samples and single-EV profiling revealed various cells of origin of our EV-IC, e.g., PD-L1+ EV mainly derived from hepatocyte/HCC cells and antigen presenting cells, and PD-1+ EV mainly from T-cells and macrophages, as expected. Conclusions: EVs carrying PD-1, PD-L1 and CTLA-4 represent readily quantifiable, non-invasive biomarkers to predict response and survival in advanced HCC patients undergoing ICI therapy and serve to identify biological ICI resistance much earlier than current standards with imaging. Dataset on diverse clinical endpoints in training and validation cohorts. Biomarker Cohort Objective Response p-value mPFS High vs Low (months) mOS High vs Low (months) EV-PD-L1 Training 0.022 2.8 vs 10.8 (p<0.001) 13.7 vs 25.7 (p=0.021) Validation <0.001 3.1 vs 5.8 (p=0.008) 10.4 vs 12.5 (p=0.97) EV-PD-1 Training <0.001 2.9 vs 15.9 (p<0.001) 6.4 vs 25.7 (p=0.002) Validation <0.001 2.9 vs 8.4 (p=0.001) 7.3 vs 12.7 (p=0.056) EV-CTLA-4 Training <0.001 3.3 vs 16.1 (p<0.001) 20.2 vs 21.9 (p=0.081) Validation <0.001 2.9 vs 9.0 (p<0.001) 8.8 vs 16.3 (p=0.003) High vs low -> based on ideal cut-off. NR = not reached.