Multi-omics comprehensive analyses of programmed cell death patterns to regulate the immune characteristics of head and neck squamous cell carcinoma

Abstract

Head and neck squamous cell carcinoma (HNSCC) is a heterogeneous cancer with high morbidity and mortality. Triggering the programmed cell death (PCD) to enhance the anti-tumor therapies is being applied in multiple cancers. However, the limited understanding of genetic heterogeneity in HNSCC severely hampers the clinical efficacy. We systematically analyzed 14 types of PCD in HNSCC from The Cancer Genome Atlas (TCGA). We utilized ssGSEA to calculate the PCD scores and classify patients into two clusters. Subsequently, we displayed the genomic alteration landscape to unravel the significant differences in copy number alterations and gene mutations. Furthermore, we calculated the IC50 values of targeted drugs to predict the differences in sensitivity. To identify the immune-related prognostic types, we comprehensively estimated the relationship between immune indicators and all prognostic PCD in three datasets (TCGA, GSE65858, GSE41613). Finally, 7 regulators were filtered. Subsequently, we integrated 10 machine learning algorithms and 101 algorithm combinations to test the clinical predictive efficacy. Using WGCNA as a basis, we built a weighted co-expression network to identify modules involved in the immune landscape with different colors. Meanwhile, our results indicated that blue and red modules containing crucial regulators closely related to the CD4+, CD8+ T cells, TMB or PD-L1. FCGR2A from blue module, CSF2, INHBA, and THBS1 from the red module were determined. After verifying in vivo experiments, FCGR2A was identified as hub gene. In conclusion, our findings suggest a potential role of PCD in HNSCC, offering new insights into effective immunotherapy and anti-tumor therapies in HNSCC.

Keywords: Genetic heterogeneity; Head and neck squamous cell carcinoma; Immune-infiltrating characteristics; Machine learning; Programmed cell death.

Graphical abstract

Fig. 1

Innate correlation of cell death types in HNSCC. (A) The Spearman correlation analysis was used to evaluate the relationship among 14 types of PCD. (B) The PCD network of cell type interaction based on the correlation strength. (C) Consensus clustering matrix for k = 2. (D) Kaplan-Meier curves of OS for two clusters. (E) The heatmap of race, sex, T, N, M stage, gene mutation with high frequency and survival status determined by two clusters.

Fig. 2

Tumor-related microenvironment between programmed cell death subtypes. (A) Comprehensive heatmap showing the genomic alterations of two cell-death subtypes. (B) The difference in amplification of copy number alterations between two clusters (***P<0.001; **P<0.01; *P<0.05). (C) The difference in deletion of copy number alterations btween two clusters (***P<0.001; **P<0.01; *P<0.05). (D) The meaningful difference in sensitivity of targeted drugs and IPS scores between two clusters.

Fig. 3

Machine learning-based integrative procedure for testing clinical efficacy. (A) The PCD network of novel cell clusters. (B) The Venn diagram illustrating the overlap of prognostic PCD genes in GSE41613, GSE65858, and TCGA. (C) A total of 101 kinds of prediction models were calculated using the LOOCV framework to determine the C-index actoss all datasets.

Fig. 4

Construction the weighted co-expression network. (A) Kaplan-Meier curves of OS for two clusters based on the 7-PCD gene model. (B) The volcano plot of DEGs from two clusters. (C) Dendrogram and trait heatmap for HNSCC patients. (D) Visualization of a cluster dendrogram through the dynamic tree cut algorithm. (E) Assessing the module-trait relationship of co-expression modules and clinical features.

Fig. 5

Identification of immune-related PCD modules. (A) The GSEA analysis of DEGs based on the Kyoto Encyclopedia of Genes and Genomes. (B) The GSEA analysis of DEGs based on the MSigDB C5 collection. (C) The GSEA analysis of DEGs based on the MSigDB C7 collection. (D) Visualization of the module-trait relationship using co-expression modules and ESTIMAE, CIBERSORT.

Fig. 6

Pan-cancer analysis of FCGR2A. (A) Prognostic performance of the NETosis-related gene, FCGR2A in pan-cancer analysis. (B) The correlation between FCGR2A and immune cell types. (C) The GSEA analysis of FCGR2A based on the Kyoto Encyclopedia of Genes and Genomes pan-cancer analysis.

Fig. 7

The biological roles of FCGR2A in tongue cancer cells. (A) Verification of FCGR2A knockdown efficiency in Cal27 and HN6 using qRT-PCR. CCK-8 (B), EdU assay (C), colony formation assay (D), and transwell assay (E) were performed to confirm the biological function of FCGR2A in tongue cancer cells. ***P<0.001.