A Tumor-Infiltration CD8+ T Cell-Based Gene Signature for Facilitating the Prognosis and Estimation of Immunization Responses in HPV+ Head and Neck Squamous Cell Cancer

Abstract

Background: CD8+ T cells, which play a vital role in response to adaptive immunity, are closely related to the immunization responses to kill tumor cells. Understanding the effects exerted by tumor-infiltrated CD8+ T cells in HPV+ and HPV- head and neck squamous cell carcinoma (HNSCC) patients is critical for predicting their prognosis as well as their responses towards immunization-related therapy.

Materials and methods: HNSCC single cell transcriptome was used to screen for differentially expressed genes (DEGs) based on CD8+ T cells. A gene signature associated with CD8+ T cells was built and verified with the cancer genome atlas dataset with a view to predicting the prognosis of HNSCC patients. Risk scores were calculated for HNSCC cases and categorized into either high- or low-risk cohorts. The prognosis-correlated data of the risk scores were analyzed by using Kaplan-Meier survival curves and multi-variate Cox regression plots. In addition, the possibility of using the genetic profiles to predict responses toward immunization-related therapy was explored.

Results: From the DEGs screened from the sequencing of single-cell RNA, a gene signature of 4 genes (ACAP1, ANKRD28, C12orf75, and M6PR) were identified. It was seen that these genes could predict overall survival in HPV+ HNSCC patients. In addition, high- and low-risk HPV+ HNSCC patients showed marked differences in their CD8+ T-cell infiltration due to immunization when clinical characteristics were taken into consideration. This correlated with their immunization therapy responses.

Conclusions: Our work provides insights into explaining the restricted responses of current immunization checkpoint inhibiting substances in HPV+ HNSCC patients. A novel genetic signature to predict the prognosis and immunization-correlated therapeutic responses is presented. This will provide potential new therapeutic opportunities for HPV+ HNSCC patients.

Keywords: CD8+ T cells; HPV; differentially expressed genes; head and neck squamous cell carcinoma; immunization-correlated genes; immunization-correlated therapy; predicted prognosis.

Figure 1

Prognosis-correlated data of CD8+ T cells from patients with HPV− and HPV+ HNSCC. (A) A heat map of data from the multiple-variate Cox proportional risk model in terms of CD8+ T cells from patients with HPV− versus HPV+ HNSCC. The z-scores represent the risk scores. (B) Kaplan-Meier survival analysis showing the levels of CD8+ T cells from patients with HPV− (right column) and HPV+ (left column) HNSCC by using MCPcounter, CIBERSORT, CIBERSORT-ABS, XCELL and QUANTISEQ approaches.

Figure 2

Identification of HPV− and HPV+ HNSCC tumor-infiltration of CD8+ T cell-correlated genes. (A) View of single cell samples from HPV− and HPV+ TIL. The annotated UMAP plot identifying 11 distinctive cell types. (B) The annotated UMAP plot of HPV− and HPV+ HNSCC TIL. (C) A pie chart of the seven cell types that make up the TIL of HNSCC. (D) Bar graphs of the cell proportions in the TIL of HPV− and HPV+ HNSCC. (E) Violin plots illustrating the expression of CD8A and exhausting T cell marker LAG3 in different TIL cell types.

Figure 3

A gene signature using infiltrated CD8+ T cells. (A) A volcano plot of the differentially expressed genes (DEGs) between HPV+ and HPV− HNSCC tumor-infiltrated CD8+ T cells. (B) A volcano plot showing the DEGs obtained from Cox regression analysis of survival-related HPV+ HNSCC-infiltrated CD8+ T cells. (C) Forest plot lines of the top 4 genes screened by using random survival forest analysis of HNSCC patients.

Figure 4

Expression of the signature genes. (A) Violin plots illustrating the expression of the survival related signature genes in different TIL cell types of scRNA data. (B) Bulk seq data were deconvoluted using CIBERSORT method. Bar graph illustrating the proportion of infiltrated immune cells of HPV+ and HPV− HNSCC. (C) Expression of signature genes were analyzed using TIMER2 website. Box charts showed the expression of the signature genes of the TCGA data. “**” refers p < 0.01, “***” refers p < 0.001 and “****” refers p < 0.0001.

Figure 5

Validation of prognosis gene labels for HNSCC cases. (A) Kaplan-Meier (KM) analysis of the risk group that were defined with CD8+ T cell-correlated gene tags in the TCGA dataset for HPV+ HNSCC. (B) KM analysis of the risk group that were defined with CD8+ T cell-correlated gene tags in the TCGA dataset for HPV- HNSCC. (C) KM analysis of the risk group that were defined with CD8+ T cell-correlated gene tags in the TCGA provisional dataset for HPV+ HNSCC. (D) Three- and five-year ROC survival curves from the TCGA dataset for HPV+ HNSCC. (E) Three- and five-year ROC survival curves from the HPV- HNSCC TCGA dataset. (F) Three- and five-year ROC survival curves from the TCGA provisional dataset for HPV+ HNSCC.

Figure 6

Correlation between risk scores and clinically related characteristics of HNSCC patients. (A) Distribution of risk scores obtained when patients’ data obtained from the HPV+ HNSCC TCGA dataset were separated by age and sex. (B) Risk score distributions for alcohol and smoking in the HPV+ HNSCC TCGA dataset. (C) Distribution of risk scores obtained when patients’ data obtained from the HPV- HNSCC TCGA dataset were separated by age and sex. (D) Risk score distributions for alcohol and smoking in the HPV- HNSCC TCGA dataset. (E) Multi-variate Cox regression forest plots of the risk scores and clinically related characteristics in the HPV+ HNSCC TCGA dataset. (F) Multi-variate Cox regression forest plots of risk scores and clinically related characteristics in the HPV- HNSCC TCGA dataset. Alco and Smok refer to alcohol consumer and smoker.

Figure 7

Profiling the gene expression of the HPV+ HNSCC risk groups. (A) A volcano plot of the DEGs between high and low risk groups of HPV+ HNSCC TCGA samples. (B) Bar graphs showing the enriched gene ontology (GO) Biological Process of the risk DEGs and (C) the enriched KEGG pathways.

Figure 8

Correlation of the signature genes with the checkpoint related genes. Correlation of the signature genes with (A) CTLA4 (B)LAG3 and (C) PDCD1. (D) S100A8 was used as control. (E) A corr plot showing the signature genes with the checkpoint related genes.