Predicting immunotherapy toxicity through single-cell sequencing of peripheral immune cells: A pilot study.

Abstract

e14538 Background: Immune checkpoint inhibitors are widely used in oncology. However, immune-related adverse events (iRAEs) are idiosyncratic and can be severe. The overall incidence is 27%, with severe iRAEs occurring in 6% of patients. We hypothesised that distinct immune cell phenotypes (gene expression) correlate with immune-related toxicity from checkpoint inhibitor immunotherapy. Single-cell RNA sequencing (scRNA-seq) is a powerful tool that can assay the gene expression of thousands of individual cells. RPS26 expression mediates CD4/CD8 T cell activation through a known trans-expressive quantitative trait locus at rs1131017 and is linked with autoimmunity. FOS and JUN encode for c-Fos and c-Jun, which dimerise to create the activator protein-1 complex, a well-established transcriptional factor of the immune system. c-Fos is an immunoregulator known to suppress pro-inflammatory cytokines, suggesting that c-FOS deficiency plays a crucial role in developing autoimmunity. We aimed to analyse the transcriptomic differences of peripheral immune cells between patients who experienced severe checkpoint inhibitor therapy iRAE and those who did not. Methods: We identified a pilot cohort of eight patients who developed grade three or above toxicity due to immune checkpoint inhibitors and eight controls who were matched for age, cancer and treatment type. Peripheral blood mononuclear cells were collected prospectively at baseline and two subsequent treatment cycles. Cells were captured on the 10x Chromium platform, and libraries were sequenced to 30,000 reads per cell. Samples were demultiplexed, and the data was analysed with the Seurat package on R statistical software (version 4.3.0). Results: Five patients had baseline samples only, two patients had two samples from two cycles, and one patient had samples from three cycles. After quality control filtering, we identified 36,288 cells from 16 individuals and 24 samples. Immune cells were annotated with the Azimuth package into 29 subtypes. We identified differences in cell proportions of immune cell subsets in the toxicity vs control group, including B naive cells (5.57% vs 3.71% p=0.01) and reduced cytotoxic CD4 T lymphocyte cells (0.66% vs 2.26%, p=0.005). Differential gene expression analysis identified a significantly high expression of ribosomal genes, particularly RPS26, in multiple cell types at baseline (p<0.01) and reduced expression of FOS and JUN-related genes in CD14 and CD16 monocytes, CD4 and CD8 T cells in the toxicity cohort (p<0.01). Conclusions: Overexpression of ribosomal genes and reduced expression of FOS/ JUN genes amongst PBMCs at baseline are potentially predictive biomarkers of severe checkpoint inhibitor immunotherapy toxicity. Validation of these findings in a larger cohort is in progress.