Background: Anoikis is a speed-limited procedure to inhibit tumor metastasis during epithelial-mesenchymal transition (EMT). Previous studies have explored anoikis-related genes (ARG) in predicting prognosis and distinguishing tumoral immunity in many types of cancer. However, the role of ARGs in regulating NK cell exhaustion (NKE) and in predicting chemotherapy sensitivity is not clear. Therefore, it is necessary to work on it.
Methods: Gene expression profiles and clinical features are collected from TCGA and GEO, and data analysis is performed in R4.2.0.
Results: The ARGs-based no-supervised learning algorithm identifies three ARG subgroups, amongst which the prognosis is different. WCGNA and Artificial intelligence (AI) are applied to construct an NKE-related drug sensitivity stratification and prognosis identification model in digestive system cancer. Pathways association analysis screens out GLI2 is a key gene in regulating NKE by non-classic Hedgehog signaling (GLI2/TGF-β/IL6).
In vitro
experiments show that down-regulation of GLI2 enhances the CAPE-mediated cell toxicity and accompanies with down-regulation of PD-L1, tumor-derive IL6, and snial1 whereas the expression of cleaved caspas3, cleaved caspase4, cleaved PARP, and E-cadherin are up-regulated in colorectal cancer. Co-culture experiments show that GLI2- decreased colorectal tumor cells lead to down-regulation of TIM-3 and PD1 in NK cells, which are restored by TGF-bate active protein powder. Besides, the Elisa assay shows that GLI2-decreased colorectal tumor cells lead to up-regulation of IFN-gamma in NK cells.