Identification of Key Genes Related to CD8+ T-Cell Infiltration as Prognostic Biomarkers for Lung Adenocarcinoma

Abstract

Background: CD8+ T cells are one of the central effector cells in the immune microenvironment. CD8+ T cells play a vital role in the development and progression of lung adenocarcinoma (LUAD). This study aimed to explore the key genes related to CD8+ T-cell infiltration in LUAD and to develop a novel prognosis model based on these genes.

Methods: With the use of the LUAD dataset from The Cancer Genome Atlas (TCGA), the differentially expressed genes (DEGs) were analyzed, and a co-expression network was constructed by weighted gene co-expression network analysis (WGCNA). Combined with the CIBERSORT algorithm, the gene module in WGCNA, which was the most significantly correlated with CD8+ T cells, was selected for the subsequent analyses. Key genes were then identified by co-expression network analysis, protein-protein interactions network analysis, and least absolute shrinkage and selection operator (Lasso)-penalized Cox regression analysis. A risk assessment model was built based on these key genes and then validated by the dataset from the Gene Expression Omnibus (GEO) database and multiple fluorescence in situ hybridization experiments of a tissue microarray.

Results: Five key genes (MZT2A, ALG3, ATIC, GPI, and GAPDH) related to prognosis and CD8+ T-cell infiltration were identified, and a risk assessment model was established based on them. We found that the risk score could well predict the prognosis of LUAD, and the risk score was negatively related to CD8+ T-cell infiltration and correlated with the advanced tumor stage. The results of the GEO database and tissue microarray were consistent with those of TCGA. Furthermore, the risk score was higher significantly in tumor tissues than in adjacent lung tissues and was correlated with the advanced tumor stage.

Conclusions: This study may provide a novel risk assessment model for prognosis prediction and a new perspective to explore the mechanism of tumor immune microenvironment related to CD8+ T-cell infiltration in LUAD.

Keywords: CD8+ T cell; bioinformatics analysis; immune microenvironment; lung adenocarcinoma; multiplex immunohistochemistry.

Figure 1

Flowchart of the study.

Figure 2

Analysis of DRGs and WGCNA. (A) Volcano plot of differential genes. (B) Heatmap of DEGs. (C) Sample clustering of WGCNA. (D) Screening with soft threshold. (E) Clustering of DEGs. DRG, differentially regulated gene; WGCNA, weighted gene co-expression network analysis; DEG, differentially expressed gene.

Figure 3

Correlation between infiltration of subtypes of T cells and different gene modules.

Figure 4

Function enrichment analysis of genes in pink module, including three types of GO enrichment analysis (A–C) and KEGG enrichment analysis (D). GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Figure 5

Acquisition of hub genes related to immune cell infiltration. (A) Correlation scatter diagram between pink genes and phenotypes of immune cell infiltration. (B) Pink gene co-expression network (red: the screened hub genes). (C) PPI network (red: the screened hub proteins). (D) Venn map from the two types of screening. PPI, protein–protein interaction.

Figure 6

Lasso regression with the Cox single-factor regression results. (A) Top 20 genes of HR obtained from batched Cox single-factor regression. (B) Results of lambda screening. (C) Statistics of regression coefficients with the significantly related genes obtained from Lasso regression. Lasso, least absolute shrinkage and selection operator.

Figure 7

Validation of model by TCGA cohort. (A) Kaplan–Meier curve of overall cohort. (B) ROC curve of overall cohort. (C) Calibration curve of overall cohort. (D) Kaplan–Meier curve of mutation-type cohort. (E) ROC curve of mutation-type cohort. (F) Calibration curve of mutation-type cohort. (G) Kaplan–Meier curve of wild-type cohort. (H) ROC curve of wild-type cohort. (I) Calibration curve of wild-type cohort. TCGA, The Cancer Genome Atlas; ROC, receiver operating characteristic.

Figure 8

Validation of model by GEO cohort. (A) Kaplan–Meier curve of overall cohort. (B) ROC curve of overall cohort. (C) Calibration curve of overall cohort. (D) Kaplan–Meier curve of mutation-type cohort. (E) ROC curve of mutation-type cohort. (F) Calibration curve of mutation-type cohort. (G) Kaplan–Meier curve of wild-type cohort. (H) ROC curve of wild-type cohort. (I) Calibration curve of wild-type cohort. GEO, Gene Expression Omnibus; ROC, receiver operating characteristic.

Figure 9

The expression of ATIC, GPI, GAPDH, ALG3, and MZT2A. (A) The expression of five proteins in tumor tissues. (B) The expression of five proteins in adjacent tissues.

Figure 10

Validation of model by mIHC cohort. (A) Kaplan–Meier curve of overall cohort. (B) ROC curve of overall cohort. (C) Calibration curve of overall cohort. (D) Kaplan–Meier curve of mutation-type cohort. (E) ROC curve of mutation-type cohort. (F) Calibration curve of mutation-type cohort. (G) Kaplan–Meier curve of wild-type cohort. (H) ROC curve of wild-type cohort. (I) Calibration curve of wild-type cohort. ROC, receiver operating characteristic.

Figure 11

(A) The correlation between risk score and CD8+ T cell in mIHC cohort. (B–E) The distribution of CD8+ T cell in tumor tissues and adjacent tissues. High positive rate in tumor tissues (A) and adjacent tissues (B). Low positive rate in tumor tissues (C) and adjacent tissues (D).

Figure 12

Correlation analysis between genes in the TIMER database and subtypes of immune cells.

Figure 13

Subgroup survival analysis of TCGA cohort. (A) Age ≥ 60. (B) Age < 60. (C) Female. (D) Male. (E) Stages T1 and T2. (F) Stages T3 and T4. (G) Stage N0. (H) Stages N1, N2, and N3. (I) Stage M0. (J) Stage M1. (K) TNM stages I and II. (L) TNM stages III and IV. TCGA, The Cancer Genome Atlas.

Figure 14

Subgroup survival analysis of GEO cohort. (A) Age ≥ 60. (B) Age < 60. (C) Female. (D) Male. (E) TNM stages I and II. (F) TNM stages III and IV. GEO, Gene Expression Omnibus.

Figure 15

Subgroup survival analysis of mIHC cohort. (A) Age ≥ 60. (B) Age < 60. (C) Female. (D) Male. (E) TNM stages I and II. (F) TNM stages III and IV. (G) Stage N0. (H) Stages N1, N2, and N3. (I) Stages T1 and T2. (J) Stages T3 and T4.