Transcriptomic profiling in advanced hepatocellular carcinoma to identify biomarkers for immunotherapy response and mechanisms of acquired resistance.

Abstract

4146 Background: Immune checkpoint inhibitors (ICI) are the standard of care for advanced HCC, yet validated biomarkers for response and mechanisms of acquired resistance remain undefined. We utilized transcriptomic profiling to identify predictive mRNA expression signatures and characterize evolution of the tumor microenvironment (TME) under therapeutic pressure. Methods: RNA-sequencing was performed on 35 HCC samples, comprising a cohort of 20 patients (16 pre-treatment + 4 treatment-naive tumors) and an independent cohort of 11 post-progression tumors (Post-IO). Patients received first-line atezolizumab/bevacizumab (n = 9), durvalumab/tremelimumab (n = 4), or nivolumab/ipilimumab (n = 3). 29 validated immune and stromal gene signatures were quantified (Bagaev, Cancer Cell 2021). Differential expression was assessed using Welch’s t-test and permutation testing. Survival outcomes were correlated with gene signatures using Cox regression and Kaplan-Meier estimation. Results: Comparison of the pre-IO exposure (n = 20) vs post-IO (n = 11) cohorts revealed significant TME remodeling. Acquired resistance was driven by the evolution of a "hypovascular immune desert," characterized by statistical downregulation of antigen presentation (MHC-II, P = 0.04) and angiogenesis (P = 0.04). Regarding baseline prediction to universal IO regimen (all 3 regimens), high tumor proliferation rate (HR 4.79, P = 0.007; mPFS 2.6 vs 10.0 m), Th2 signature (HR 22.5, P = 0.01), and Treg traffic (HR 15.6, P = 0.01) strongly correlating with inferior outcomes; high macrophage/DC traffic predicted favorable outcomes (HR 0.18, P = 0.04; mPFS 7.4 vs 3.4 m). Distinct from these universal IO predictors, we utilized permutation testing to differentiate regimen-specific signals, identifying higher myeloid cell traffic (P = 0.007) and higher angiogenesis (P = 0.02) at baseline as unique biomarkers of response exclusively in anti-CTLA4 regimens. Transcriptomic profiles of post-transplant recurrences (n = 4) were statistically indistinguishable from the pre-IO cohort. Conclusions: Advanced unresectable HCC shows distinct tumor states linked to immunotherapy benefit versus early resistance, and these states evolve with treatment. Our findings support a biology-driven framework for predicting response, understanding resistance, and prioritizing rational combination strategies to improve durable disease control.