Identification and validation of autophagy-related gene expression for predicting prognosis in patients with idiopathic pulmonary fibrosis

Abstract

Background: Idiopathic pulmonary fibrosis (IPF) is a chronic, progressive, and fatal fibrotic pulmonary disease with unknow etiology. Owing to lack of reliable prognostic biomarkers and effective treatment measures, patients with IPF usually exhibit poor prognosis. The aim of this study is to establish a risk score prognostic model for predicting the prognosis of patients with IPF based on autophagy-related genes.

Methods: The GSE70866 dataset was obtained from the gene expression omnibus (GEO) database. The autophagy-related genes were collected from the Molecular Signatures Database (MSigDB). Gene enrichment analysis for differentially expressed genes (DEGs) was performed to explore the function of DEGs. Univariate, least absolute shrinkage and selection operator (LASSO), as well as multivariate Cox regression analyses were conducted to identify a multi-gene prognostic model. Receiver operating characteristic (ROC) curve was applied to assess the prediction accuracy of the model. The expression of genes screened from the prognostic model was validated in clinical samples and human lung fibroblasts by qPCR and western blot assays.

Results: Among the 514 autophagy-related genes, a total of 165 genes were identified as DEGs. These DEGs were enriched in autophagy-related processes and pathways. Based on the univariate, LASSO, and multivariate Cox regression analyses, two genes (MET and SH3BP4) were included for establishing the risk score prognostic model. According to the median value of the risk score, patients with IPF were stratified into high-risk and low-risk groups. Patients in high-risk group had shorter overall survival (OS) than low-risk group in both training and test cohorts. Multivariate regression analysis indicated that prognostic model can act as an independent prognostic indicator for IPF. ROC curve analysis confirmed the reliable predictive value of prognostic model. In the validation experiments, upregulated MET expression and downregulated SH3BP4 expression were observed in IPF lung tissues and TGF-β1-activated human lung fibroblasts, which is consistent with results from microarray data analysis.

Conclusion: These findings indicated that the risk score prognostic model based on two autophagy-related genes can effectively predict the prognosis of patients with IPF.

Keywords: MET; SH3BP4; autophagy; idiopathic pulmonary fibrosis; prognostic model.

Figure 1

Flow chart of the study.

Figure 2

GO and KEGG enrichment analyses of DEGs. (A) GO enrichment analysis of DEGs, including BP, CC, and MF. (B) KEGG enrichment analysis of DEGs.

Figure 3

Establishment of a risk score prognostic model. (A) LASSO coefficients of the 3 autophagy-associated DEGs. (B) Cross-validation for selecting key genes. (C) The forest plot of 2 autophagy-associated DEGs in the risk score prognostic model.

Figure 4

Expression levels of two genes in Ctrl and IPF groups. (A) MET. (B) SH3BP4. *p<0.05; ***p<0.001.

Figure 5

Principal component analysis. (A) Principal component analysis based upon all autophagy-associated genes in the training test. (B) Principal component analysis based upon two autophagy-associated genes from risk score prognostic model.

Figure 6

A two-gene risk score model predicted the overall survival in patients with IPF in the training set. (A) Distribution of risk score per patient. (B) Survival status of each patient. (C) Expression heatmap of the two genes. (D) Kaplan-Meier survival curve analysis of IPF patients divided into high-risk and low-risk groups.

Figure 7

Identification of independent prognostic factors in patients with IPF in the training set. (A) The univariate Cox regression analysis for risk score model and clinical parameters. (B) The multivariate Cox regression analysis for risk score model and clinical parameters.

Figure 8

The prognostic value of risk score model in the training set. (A) ROC curves of risk score model at 1-, 3-, and 5-year overall survival. (B) ROC curves of clinical parameters and risk score model at 1-year overall survival.

Figure 9

The relationship between risk scores and clinical parameters in the training set. (A) age; (B) sex; (C) GAP index.

Figure 10

Immune cells and immune-related functions of the two risk groups in the training set. The proportion of 22 types of immune cells (A) and 13 immune-related functions (B) were analyzed in the high-risk and low-risk group. *p<0.05; **p<0.01; ***p<0.001.

Figure 11

GSVA enrichment analysis between the two risk groups in the training set.

Figure 12

Validation of risk score model in the test set. (A) Distribution of risk score per patient. (B) Survival status of each patient. (C) An expression heatmap of the two genes. (D) Kaplan-Meier survival curve analysis of IPF patients divided into the high-risk and low-risk groups.

Figure 13

The prognostic value of risk score model in the test set. (A) ROC curves of risk score model at 1-, 3-, and 5-year overall survival. (B) ROC curves of clinical parameters and risk score model at 1-year overall survival.

Figure 14

The expression of two model genes in HC and IPF lung tissues. (A–C) The protein expression of MET and SH3BP4 in healthy control and IPF lung tissues was detected by western blot assay. HC, healthy control. *p<0.05. NS, not significant. n=6.

Figure 15

The expression of two model genes in human lung fibroblasts. Human lung fibroblasts were treated with 10ng/ml TGF-β1 for 48h. (A, B) The mRNA expression of MET and SH3BP4 was detected by qPCR. (C–E) The protein expression of MET and SH3BP4 was detected by western blot assay. *p<0.05; **p<0.01. n=5.