Investigating the tumor immune infiltrate for populations that predict immune-related adverse events (irAEs) in patients receiving PD-1 inhibitors.

Abstract

3116 Background: The mechanistic relationship between clinical benefit and immune-related adverse events (irAEs) in response to immune checkpoint inhibitors (ICIs) remains unclear, with several clinical studies reporting that irAEs are biomarkers of responses. Single-cell RNA sequencing (scRNAseq) analysis of tumors from patients with advanced melanoma before and after treatment with ICIs have identified immune cells that correlate with response to ICIs. We sought to evaluate if these populations were also associated with irAEs. Methods: A published scRNAseq data set generated with the Smart-Seq2 protocol (Sade-Feldman M, et al. Cell 2018.) was re-analyzed, stratified by two definitions of irAEs: (1) toxicity requiring systemic immunosuppression (prednisone > 10mg/day) or (2) systemic immunosuppression and/or endocrinopathy. Unbiased single-cell analysis was performed, followed by sub-clustering of T cell populations. The percentage of cells in each cluster was determined on a per sample basis. Results: 13,184 immune cells from 39 samples collected from 25 patients were re-analyzed. 27 samples were from patients who did not respond to ICIs, while 12 samples came from responding patients. 21 samples came from patients who required immunosuppression, 5 samples from patients with isolated thyroiditis, and 13 samples from patients who met neither irAE criteria. Unsupervised scRNAseq analyses focused on ICI efficacy re-capitulated published associations between response and populations that included B-cells (p < 0.01) and TCF7 expressing T-cells (p < 0.01). While these cell populations were not associated with either definition of toxicity, we observed a non-Treg CD4 expressing T cell population (0.8-10.5% cells/sample) that positively correlated with either definition of toxicity (p < 0.05) but not efficacy. Conclusions: In a patient cohort with advanced melanoma, tumor-infiltrating immune cell populations associated with response to ICI therapy were not associated with irAEs. This suggests that biomarkers of ICI response may not function as biomarkers of irAEs, and ongoing analysis will seek to validate this result. Understanding the differences between ICI response and irAEs may identify new therapeutic targets for maximizing efficacy while mitigating toxicity.