ABSTRACT
Emerging immunotherapies such as immune checkpoint blockade (ICB) and chimeric antigen receptor T-cell (CAR-T) therapy have revolutionized cancer treatment and have improved the survival of patients with multiple cancer types. Despite this success many patients are unresponsive to these treatments or relapse following treatment. CRISPR activation and knockout (KO) screens have been used to identify novel single gene targets that can enhance effector T cell function and promote immune cell targeting and eradication of tumors. However, cancer cells often employ multiple genes to promote an immunosuppressive pathway and thus modulating individual genes often has a limited effect. Paralogs are genes that originate from common ancestors and retain similar functions. They often have complex effects on a particular phenotype depending on factors like gene family similarity, each individual gene's expression and the physiological or pathological context. Some paralogs exhibit synthetic lethal interactions in cancer cell survival; however, a thorough investigation of paralog pairs that could enhance the efficacy of cancer immunotherapy is lacking. Here we introduce a sensitive computational approach that uses sgRNA sets enrichment analysis to identify cancer-intrinsic paralog pairs which have the potential to synergistically enhance T cell-mediated tumor destruction. We have further developed an ensemble learning model that uses an XGBoost classifier and incorporates features such as gene characteristics, sequence and structural similarities, protein-protein interaction (PPI) networks, and gene coevolution data to predict paralog pairs that are likely to enhance immunotherapy efficacy. We experimentally validated the functional significance of these predicted paralog pairs using double knockout (DKO) of identified paralog gene pairs as compared to single gene knockouts (SKOs). These data and analyses collectively provide a sensitive approach to identify previously undetected paralog pairs that can enhance cancer immunotherapy even when individual genes within the pair has a limited effect.
ABSTRACT
Radiotherapy (RT), is a fundamental treatment for malignant tumors and is used in over half of cancer patients. As radiation can promote anti-tumor immune effects, a promising therapeutic strategy is to combine radiation with immune checkpoint inhibitors (ICIs). However, the genetic determinants that impact therapeutic response in the context of combination therapy with radiation and ICI have not been systematically investigated. To unbiasedly identify the tumor intrinsic genetic factors governing such responses, we perform a set of genome-scale CRISPR screens in melanoma cells for cancer survival in response to low-dose genotoxic radiation treatment, in the context of CD8 T cell co-culture and with anti-PD1 checkpoint blockade antibody. Two actin capping proteins, Capza3 and Capg, emerge as top hits that upon inactivation promote the survival of melanoma cells in such settings. Capza3 and Capg knockouts (KOs) in mouse and human cancer cells display persistent DNA damage due to impaired homology directed repair (HDR); along with increased radiation, chemotherapy, and DNA repair inhibitor sensitivity. However, when cancer cells with these genes inactivated were exposed to sublethal radiation, inactivation of such actin capping protein promotes activation of the STING pathway, induction of inhibitory CEACAM1 ligand expression and resistance to CD8 T cell killing. Patient cancer genomics analysis reveals an increased mutational burden in patients with inactivating mutations in CAPG and/or CAPZA3, at levels comparable to other HDR associated genes. There is also a positive correlation between CAPG expression and activation of immune related pathways and CD8 T cell tumor infiltration. Our results unveil the critical roles of actin binding proteins for efficient HDR within cancer cells and demonstrate a previously unrecognized regulatory mechanism of therapeutic response to radiation and immunotherapy.