The 3’ non-coding sequence negatively regulates PD-L1 expression and its regulators are systematically identified in pan-cancer

Abstract Objective Comparing to the coding sequences (CDS), the 3’-untranslated region (3’-UTR) of PD-L1 is extremely longer but its role and regulators are less explored. Methods The whole 3’-UTR region was deleted by CRISPR-Cas9. Prognostic analysis was performed using online tools. Immune infiltration analysis was performed using Timer and Xcell package. Immunotherapy response prediction and cox regression were performed using R software. MicroRNA network analysis was conducted by Cytoscape software. Results The level of PD-L1 was dramatically and significantly up-regulated in 3’-UTR deficient cells. Furthermore, we found a panel of 43 RNA binding proteins (RBPs) that correlated with PD-L1 in a majority of cancer cell lines and tumor tissues. Among them, PARP14 is widely associated with immune checkpoints, tumor microenvironment and immune infiltrating cells in various cancer types. We also identified 38 MicroRNA that associated with PD-L1 across cancers. The miR-3139, miR-4761 and miR-15a-5p are significantly associated with PD-L1 in most of cancer types. Finally, we revealed 21 m6A regulators that have a strong correlation with PD-L1. More importantly, by combing the identified RBPs and m6A regulators, we established a predictive immune signature including RBMS1, QKI, YTHDC1, ZC3HAV1, RBM38 and PPARGC1B to predict the responsiveness of cancer patients upon receiving immune checkpoint blockade. Conclusions We demonstrated the critical role of 3’-UTR in the regulation of PD-L1 and uncovered a large number of potential PD-L1 regulators in pan-cancer. The generated biomarker signature has power to predict patient’s prognosis, but along with the potential PD-L1 regulators should be further biologically investigated.