Development and validation of an immune-related prognostic signature in lung adenocarcinoma

Abstract

Background: Lung adenocarcinomas (LUAD) is the most common histological subtype of lung cancers. Tumor immune microenvironment (TIME) is involved in tumorigeneses, progressions, and metastases. This study is aimed to develop a robust immune-related signature of LUAD.

Methods: A total of 1774 LUAD cases sourced from public databases were included in this study. Immune scores were calculated through ESTIMATE algorithm and weighted gene co-expression network analysis (WGCNA) was applied to identify immune-related genes. Stability selections and Lasso COX regressions were implemented to construct prognostic signatures. Validations and comparisons with other immune-related signatures were conducted in independent Gene Expression Omnibus (GEO) cohorts. Abundant infiltrated immune cells and pathway enrichment analyses were carried out, respectively, through ImmuCellAI and gene set enrichment analysis (GSEA).

Results: In Cancer Genome Atlas (TCGA) LUAD cohorts, immune scores of higher levels were significantly associated with better prognoses (P < .05). Yellow (n = 270) and Blue (n = 764) colored genes were selected as immune-related genes, and after univariate Cox regression analysis (P < .005), a total of 133 genes were screened out for subsequent model constructions. A four-gene signature (ARNTL2, ECT2, PPIA, and TUBA4A) named IPSLUAD was developed through stability selection and Lasso COX regression. It was suggested by multivariate and subgroup analyses that IPSLUAD was an independent prognostic factor. It was suggested by Kaplan-Meier survival analysis that eight out of nine patients in high-risk groups had significantly worse prognoses in validation data sets (P < .05). IPSLUAD outperformed other signatures in two independent cohorts.

Conclusions: A robust immune-related prognostic signature with great performances in multiple LUAD cohorts was developed in this study.

Keywords: biomarker; infiltrated immune cell; lung adenocarcinoma; prognostic signature; tumor immune microenvironment.

Figure 1

Flowchart of developments and validations of IPSLUAD

Figure 2

Associations between immune/stromal/ESTIMATE scores and clinicopathological indicators. A‐C, Differences among stromal/immune/ESTIMATE scores in different pathologic stages. Asterisk (*) indicated a significant difference between two groups (P < .05). D‐F, Survival analyses of stromal/immune/ESTIMATE scores through Kaplan‐Meier curve with log‐rank test

Figure 3

Network constructions and module detections of LUAD. A and B, Analyses of network topologies for various soft‐thresholding powers through scale‐free fit index (A) and mean connectivity (B). C, Clustering dendrogram of genes based on topological overlapping. Different colors were assigned to corresponding modules. A total of 23 modules were identified

Figure 4

Prediction performances of IPSLUAD in TCGA‐LUAD cohort. A, Distributions of risk scores (top), survival statuses of patients in low‐risk and high‐risk groups (middle), and four‐gene expression profiles of each patient (bottom). B‐E, Kaplan‐Meier curves of OS between low‐risk and high‐risk groups based on whole‐TCGA cohort (B), Stage I (C), Stage II (D), and Stage III subgroup (E)

Figure 5

Prediction performances of IPSLUAD in validating data sets. Kaplan‐Meier survival curves of overall survivals in (A) GSE3141, (B) GSE13213, (C) GSE14814, (D) GSE29016, (I) GSE30219, (J) GSE31210, (K) GSE37745, (L) GSE50081, (Q) GSE68465. Receiver operating curve (ROC) analysis of IPSLUAD in (E) GSE3141, (F) GSE13213, (G) GSE14814, (H) GSE29016, (M) GSE30219, (N) GSE31210, (O) GSE37745, (P) GSE50081, (R) GSE68465

Figure 6

Comparisons of IPSLUAD with other published immune‐related models in GSE31210 and GSE68465 data sets. A, Two‐year ROC in GSE31210. B, Five‐year ROC in GSE31210. C, Two‐year ROC in GSE68465. D, Five‐year ROC in GSE68‐465

Figure 7

Compositions of infiltrated immune cells between low‐risk and high‐risk groups in TCGA‐LUAD cohort through ImmuCellAI. A, Abundance of 24 immune cell types in TCGA. B, Comparisons between immune cells in low‐risk and high‐risk groups in TCGA

Figure 8

Different immune statuses between low‐risk and high‐risk groups in TCGA‐LUAD cohort. A and B, Significant enrichments of immune‐related pathways among high‐risk patients were indicated through gene set enrichment analysis (GSEA). C and D, Gene sets between low‐risk and high‐risk groups were analyzed through expression profiles of the two enrichments