OC-0091 An immune gene expression risk score for distant metastases after radiotherapy for cervical cancer

PURPOSE:To develop an immune-based gene expression risk score identifying cervical cancer patients at increased risk of distant metastases (DM). EXPERIMENTAL DESIGN:Tumor biopsies were obtained from 81 patients prior to chemoradiotherapy. Whole transcriptome RNA sequencing was performed (Illumina NextSeq500). Beginning with 4,723 immune-related genes, a 55 gene risk score for DM was derived using Cox modelling and principal component analysis. It was validated in independent cohorts of 274 patients treated at the Norwegian Radium Hospital (NRH) and 206 patients from TCGA. RESULTS:The risk score was predictive of DM (HR 2.7, p<0.0001) and lower cause specific survival (CSS) by univariate analysis (HR 2.0, p=0.0003) and multivariate analysis adjusted for clinical factors (DM HR 3.0, p<0.0001; CSS HR 2.2, p=0.0004). The risk score predicted DM (HR 1.4, p=0.05) and CSS (HR 1.48, p=0.013) in the NRH cohort and CSS (HR 1.4, p=0.03) in the TCGA cohort. Higher risk scores were associated with lower CIBERSORT estimates of tumor infiltrating immune cells, including CD8 T-cells and M1 and M2 macrophages (p<0.001). Higher risk scores were associated with lower expression (all p<0.001) of important chemokines (CXCL12, CXCR4), interferon regulated genes (IRF1, STAT1, IDO1) and immune checkpoint regulators (PD-1, PD-L1, CTLA-4). CONCLUSION:The immune metastatic risk score addresses important challenges in the treatment of cervical cancer - identifying patients at high risk of DM after radiotherapy. The findings of this study indicate that high tumor mutational burden and a 'cold', immune-excluded tumor microenvironment influence distant metastatic recurrence. Further validation of the risk score is needed.