The Chemokines Initiating and Maintaining Immune Hot Phenotype Are Prognostic in ICB of HNSCC

Background: The immune checkpoint blockade (ICB) with anti-programmed cell death protein 1(PD-1) on HNSCC is not as effective as on other tumors. In this study, we try to find out the key factors in the heterogeneous tumor-associated monocyte/macrophage (TAMM) that could regulate immune responses and predict the validity of ICB on HNSCC.Experimental Design: To explore the correlation of the TAMM heterogeneity with the immune properties and prognosis of HNSCC, we established the differentiation trajectory of TAMM by analyzing the single-cell RNA-seq data of HNSCC, by which the HNSCC patients were divided into different sub-populations. Then, we exploited the topology of the network to screen out the genes critical for immune hot phenotype of HNSCC, as well as their roles in TAMM differentiation, tumor immune cycle, and progression. Finally, these key genes were used to construct a neural net model via deep-learning framework to predict the validity of treatment with anti-PD-1/PDL-1Results: According to the differentiation trajectory, the genes involved in TAMM differentiation were categorized into early and later groups. Then, the early group genes divided the HNSCC patients into sub-populations with more detailed immune properties. Through network topology, CXCL9, 10, 11, and CLL5 related to TAMM differentiation in the TME were identified as the key genes initiating and maintaining the immune hot phenotype in HNSCC by remarkably strengthening immune responses and infiltration. Genome wide, CASP8 mutations were found to be key to triggering immune responses in the immune hot phenotype. On the other hand, in the immune cold phenotype, the evident changes in CNV resulted in immune evasion by disrupting immune balance. Finally, based on the framework of CXCL9-11, CLL5, CD8(+), CD4(+) T cells, and Macrophage M1, the neural network model could predict the validity of PD-1/PDL-1 therapy with 75% of AUC in the test cohort.Conclusion: We concluded that the CXCL9, 10,11, and CCL5 mediated TAMM differentiation and constructed immune hot phenotype of HNSCC. Since they positively regulated immune cells and immune cycle in HNSCC, the CXCL9-11 and CCL5 could be used to predict the effects of anti-PD-1/PDL-1 therapy on HNSCC.