Deciphering immune-genomic markers of hyperprogression in patients (Pts) with metastatic urothelial cancer (mUC) treated with immune checkpoint inhibitors (ICIs).

Abstract

840 Background: ICIs offer modest response rates (~20%) in mUC and there remains a critical need for validated biomarkers to predict patient benefit. Hyperprogressive disease (HPD), characterized by rapid tumor progression following ICI initiation, is a poorly understood phenomenon in mUC. No robust biomarkers currently exist to identify patients at risk for HPD, a condition that leads to unnecessary treatment and exacerbated toxicity. While patient factors associated with HPD have been studied, little is known about the tumor’s immune-genomic landscape in these cases. We investigate the immune-genomic features of mUC tumors associated with HPD to better understand and predict this response to ICI. Methods: We retrospectively analyzed 132 pts with mUC treated with atezolizumab or pembrolizumab between 2015-2023. All patients received at least 2 ICI cycles, and archival FFPE tumor samples were available for immune-genomic analysis. Pts were classified as responders (N=75, defined as complete/partial response or stable disease) or non-responders (N=57, progressive disease). HPD was defined as progression per RECIST within 14 weeks of ICI initiation. RNA was extracted from macrodissected FFPE samples, and transcriptomic profiling was performed using the Nanostring PanCancer IO 360 panel. Data were processed using nSolver 4.0 and nanostring data analysis service analyzed for differential gene expression. Receiver operator characteristic (ROC) analysis was conducted to assess the predictive power of immune signatures for HPD. Results: Of the 132 pts, 48 (36%) were classified as having HPD. Compared to non-responders without HPD, HPD tumors exhibited significant downregulation of immune signatures, including IFN Gamma, PD-L1, myeloid cells, and tumor inflammation signatures (TIS). Key chemokines and IFN downstream signaling were also downregulated (p-adj = 0.041). Single-gene analysis highlighted downregulation of CCL4 (logFC = -0.947, p = 0.00012), GBP1 (logFC = -0.8105, p = 0.00017), and NECTIN2 (logFC = -0.4792, p = 0.00032) in HPD tumors, though statistical significance was not reached after adjustment. ROC analysis demonstrated modest predictive power for HPD with AUCs of 0.66 for TIS and IFN Gamma, and 0.658 for the immunoproteasome. Conclusions: Our findings reveal a distinct immune-genomic profile in mUC tumors associated with HPD following ICI treatment, including reduced IFN Gamma, TIS, and myeloid cell signaling. These immune signatures may serve as potential biomarkers to identify pts at risk of HPD, warranting further validation in prospective trials. Ongoing analyses will continue to explore the genomic underpinnings of HPD and therapeutic resistance in mUC.