Identification of immune-associated genes for the diagnosis of ulcerative colitis-associated carcinogenesis via integrated bioinformatics analysis

Abstract

Background: UC patients suffer more from colorectal cancer (CRC) than the general population, which increases with disease duration. Early colonoscopy is difficult because ulcerative colitis-associated colorectal cancer (UCAC) lesions are flat and multifocal. Our study aimed to identify promising UCAC biomarkers that are complementary endoscopy strategies in the early stages.

Methods: The datasets may be accessed from the Gene Expression Omnibus and The Cancer Genome Atlas databases. The co-expressed modules of UC and CRC were determined via weighted co-expression network analysis (WGCNA). The biological mechanisms of the shared genes were exported for analysis using the Gene Ontology and Kyoto Encyclopedia of Genes and Genomes. To identify protein interactions and hub genes, a protein-protein interaction network and CytoHubba analysis were conducted. To evaluate gene expression, external datasets and experimental validation of human colon tissues were utilized. The diagnostic value of core genes was examined through receiver operating characteristic (ROC) curves. Immune infiltration analysis was employed to investigate the associations between immune cell populations and hub genes.

Results: Three crucial modules were identified from the WGCNA of UC and CRC tissues, and 33 coexpressed genes that were predominantly enriched in the NF-κB pathway were identified. Two biomarkers (CXCL1 and BCL6) were identified via Cytoscape and validated in external datasets and human colon tissues. CRC patients expressed CXCL1 at the highest level, whereas UC and CRC patients showed higher levels than the controls. The UC cohort expressed BCL6 at the highest level, whereas the UC and CRC cohorts expressed it more highly than the controls. The hub genes exhibited significant diagnostic potential (ROC curve > 0.7). The immune infiltration results revealed a correlation among the hub genes and macrophages, neutrophils and B cells.

Conclusions: The findings of our research suggest that BCL6 and CXCL1 could serve as effective biomarkers for UCAC surveillance. Additionally, they demonstrated a robust correlation with immune cell populations within the CRC tumour microenvironment (TME). Our findings provide a valuable insight about diagnosis and therapy of UCAC.

Keywords: bioinformatics analysis; diagnosis biomarker; immune infiltration; ulcerative colitis; ulcerative colitis-associated carcinogenesis.

Figure 1

Research design flowchart. The RNA sequences of colon tissues from UC, CRC cohorts and the controls were obtained from the GEO and TCGA databases. The datasets were pretreated via R software. WGCNA was used to identify the significant coexpression modules between UC and CRC. Then, the coexpressed genes were visualized with a Venn diagram. GO and KEGG analyses were applied to coexpressed genes. The top 5 hub genes related to UC-CRC were screened via cytoHubba. The expression of the five hub genes was verified via validation cohorts, and the hub genes was estimated via ROC curve and immune infiltration analyses. Finally, experimental verification (HE, WB, IHC and IF analysis) of human colon tissues was applied to affirm the differential expression of hub genes among UC patients, CRC patients and controls.

Figure 2

WGCNA of the UC group (GSE87473) and CRC group (TCGA-COAD). (A) β= 20 is the soft threshold in UC patients according to the combined analysis of scale independence and average connectivity. (B) β = 7 was selected as the soft threshold in CRC. (C, D) Gene dendrograms were obtained by hierarchical clustering. The different colored rows under the dendrogram show the gene coexpression module assignments determined by the dynamic tree cut method. (E, F) Relationships between modules and traits visualized as a heatmap. Correlations and p values are displayed in each cell. Each row corresponds to the gene module, and each column is related to a clinical trait.

Figure 3

PPI network and enrichment analysis results. (A) A total of 33 overlapping genes were identified from the gene intersections in UC and CRC patients via WGCNA. (B) GO circle representing the GO enrichment analysis of the overlapping genes. (C) GO functional enrichment analysis of the overlapping genes, comprising BP, CC, and MF. The different GO terms are displayed on the y-axis. Gene ratios enriched in terms are shown on the x-axis. (D) The 15 most significantly enriched KEGG pathways. (E, F) Sixteen interacting genes and important modules visualized via MCODE. (G) Identification of the top five key genes by multiple MCC, DMNC, MNC, Degree and EPC methods. (H) GeneMANIA was applied to explore internal association of overlapping genes and their coexpressed genes.

Figure 4

Verification and ROC curves of the hub genes. (A–D) The expression verification in the GSE87466 dataset. (E–H) Hub gene expression in the GSE39582 dataset. (I–L) ROC curves of BCL6, CXCL1, LCN2 and CXCL2 in the GSE92415 dataset. (M–P) ROC curves of BCL6, CXCL1, LCN2 and CXCL2 in the GSE20916 dataset. *p<0.05, ***p<0.001.

Figure 5

Immune cell infiltration analysis. (A) According to the Spearman correlation analysis, the hub genes were strongly correlated with immune cells. (B) Correlation results of hub genes and immune cells. (C) Violin plot of the proportion of immune cells infiltrating the UC cohort compared to that in the control cohort. (D) Proportions of 22 immune cells visualized from the bar plot. (E) Heatmap of correlations of different immune cells in UC samples. Red: positive correlation; blue: negative correlation.

Figure 6

Verification of gene expression in colon tissues (A) Images of a normal intestinal tract obtained via white light endoscopy. (B) Intestinal tract images of UC patients via white light endoscopy. (C) Intestinal tract images of CRC patients via white light endoscopy. (D) NBI images of the intestinal tract of CRC patients. (E, H) H&E staining of human normal colon sections. (F, I) H&E staining of intestinal tissues from UC patients. (G, J) H&E staining of intestinal sections from CRC patients. (K–P) Expression of CXCL1 (K–M) and BCL6 (N–P) in the control cohort, UC patient cohort and cancer patient cohort determined by IHC analysis. (S, T) Statistical comparison of the IHC analysis of the mean densities of CXCL1 (S) and BCL6 (T) in the control cohort, UC patient cohort and cancer patient cohort. CXCL1 was expressed at the highest level in the cancer patient cohort. BCL6 was expressed at the highest level in the UC patient cohort. (Q, R) Western blot analysis of CXCL1 and BCL6 expression in human intestinal tissues. In the cancer patients group, the CXCL1 expressed the highest than others. In the UC patient group, the BCL6 expression was greater than that in the other groups. *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001.

Figure 7

Immunofluorescence images of CXCL1 and BCL6. (A–I) Immunofluorescence images of CXCL1 in colon tissues from the control, UC and cancer cohorts. (J) Quantitative analysis of CXCL1 immunofluorescence. (K–S) Immunofluorescence images of BCL6 in colon tissues from the control, UC and cancer groups. (T) Quantitative analysis of BCL6 immunofluorescence. ***p<0.001 and ****p<0.0001.

Figure 8

Effects of hub genes on the transformation from ulcerative colitis to colorectal cancer.