Identifying immune cell infiltration and effective diagnostic biomarkers in Crohn's disease by bioinformatics analysis

Abstract

Background: Crohn's disease (CD) has an increasing incidence and prevalence worldwide. It is currently believed that both the onset and progression of the disease are closely related to immune system imbalance and the infiltration of immune cells. The aim of this study was to investigate the molecular immune mechanisms associated with CD and its fibrosis through bioinformatics analysis.

Methods: Three datasets from the Gene Expression Omnibus data base (GEO) were downloaded for data analysis and validation. Single sample gene enrichment analysis (ssGSEA) was used to evaluate the infiltration of immune cells in CD samples. Immune cell types with significant differences were identified by Wilcoxon test and Least Absolute Shrinkage and Selection Operator (LASSO) regression analysis. Differentially expressed genes (DEGs) were screened and then subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) functional correlation analysis, as well as protein-protein interaction (PPI) network analysis. The cytoHubba program and the GSE75214 dataset were used to screen for hub genes and plot Receiver operating characteristic (ROC)curves to screen for possible biomarkers of CD based on diagnostic efficacy. The hub genes of CD were correlated with five significantly different immune cells. In addition, validation was performed by real time quantitative PCR (RT-qPCR) experiments in colonic tissue of CD intestinal fibrosis rats to further identify hub genes that are more related to CD intestinal fibrosis.

Results: The DEGs were analyzed separately by 10 algorithms and narrowed down to 9 DEGs after taking the intersection. 4 hub genes were further screened by the GSE75214 validation set, namely COL1A1, CXCL10, MMP2 and FGF2. COL1A1 has the highest specificity and sensitivity for the diagnosis of CD and is considered to have the potential to diagnose CD. Five immune cells with significant differences were screened between CD and health controls (HC). Through the correlation analysis between five kinds of immune cells and four biomarkers, it was found that CXCL10 was positively correlated with activated dendritic cells, effector memory CD8⁺ T cells. MMP2 was positively correlated with activated dendritic cells, gamma delta T cells (γδ T) and mast cells. MMP2 and COL1A1 were significantly increased in colon tissue of CD fibrosis rats.

Conclusion: MMP2, COL1A1, CXCL10 and FGF2 can be used as hub genes for CD. Among them, COL1A1 can be used as a biomarker for the diagnosis of CD. MMP2 and CXCL10 may be involved in the development and progression of CD by regulating activated dendritic cell, effector memory CD8⁺ T cell, γδ T cell and mast cell. In addition, MMP2 and COL1A1 may be more closely related to CD intestinal fibrosis.

Keywords: Crohn’s disease; biomarker; fibrosis; immune cells; ssGSEA.

Figure 1

Flow chart of the analysis process conducted in this study.

Figure 2

Data preprocessing. Box plot and uniform manifold approximation and projection were performed to remove batch correction of GSE36807 and GSE16879. (A, B) before batch correction and (C, D) after batch correction.

Figure 3

Analyzing immune cell infiltration and identifying the significantly different infiltrates of immune cells in CD and normal tissues. The composition of 28 species of immune cells in each sample was showed in a heatmap (A). (B) Wilcoxon test and (C, D) LASSO regression were used to analyze the different infiltrates of immune cells.

Figure 4

Identification of DEGs and functional correlation analysis by GO and KEGG. DEGs were visualized by volcano map (A) and heat map (B). The results of GO were presented using bar plot (C) and circle graph (D). The results of KEGG were showed by bubble plot (E) and circle graph (F).

Figure 5

The PPI network and cluster modules. PPI network of DEGs was conducted by STRING software (A). Visualization of PPI networks of DEGs using Cytoscape software (B) and screening of the 3 most important gene cluster modules by MCODE program (C–E).

Figure 6

Identification of hub genes and a co-expressed network of mRNAs and target miRNAs. Hub genes were screened by the R package “Upset” (A). Visualization the degree of interaction of 9 hub genes by STRING software (B). The mRNA-miRNA co-expressed network was constructed by Cytoscape including 181 nodes and 192 edges (C).

Figure 7

Validation of hub genes. The expression of 9 hub genes were validated by dataset GSE75214. The expression levels of COL1A1, CXCL10, and MMP2 were significantly higher in CD samples and active CD samples than in HC (p < 0.01) and the FGF2 was elevated in active CD compared to HC (p< 0.1) (A–I). The expression levels of COL1A1, CXCL10, MMP2 and FGF2 were presented as a heat map (J). *p< 0. 1, ***p< 0.01, ns, no significant difference.

Figure 8

ROC curves of the 4 specifically expressed hub genes. Diagnostic value of COL1A1、CXCL10、MMP2 and FGF2 in CD using non-inflammatory tissues as controls by dataset GSE75214 (A). Diagnostic value of COL1A1 and CXCL10 in CD (GSE75214 dataset) (B). Diagnostic value of COL1A1 and MMP2 in CD (GSE75214 dataset) (C).

Figure 9

Correlation between hub genes and differential immune cells in CD. Correlation of 4 hub genes and 5 significantly differential immune cells (A). CXCL10 was positively correlated with both activated dendritic cells and effector memory CD8⁺ T cells (B, C). MMP2 was positively correlated with activated dendritic cells, gamma delta T cells and mast cells (D–F).

Figure 10

Examination of biomarkers in CD fibrosis rats. Representative HE stained and Masson stained images in normal (A, C) and model groups (B, D), scale bar = 200µm. Statistical analysis of the expression levels of COL1A1、MMP2、CXCL10 and FGF2 in the control and model groups based on the RT-PCR assay (E–H). Control: blank control rats (n=6); CD fibrosis, Crohn’s disease fibrosis model group rats (n=6). Data are presented as the mean ± SD. **p< 0. 05, ***p< 0.01, ns, no significant difference.