Identification and validation of EMT-immune-related prognostic biomarkers CDKN2A, CMTM8 and ILK in colon cancer

Abstract

Colon cancer (CC) is a disease with high incidence and mortality rate. The interaction between epithelial-mesenchymal transition (EMT) and immune status has important clinical significance. We aim to identify EMT-immune-related prognostic biomarkers in colon cancer. The GEO2R and GEPIA 2.0 were utilized to calculate the differential expression genes between CC and normal mucosa. Immport, InnateDB and EMTome databases were used to define EMT-immune-related genes. We conducted batch prognostic analysis by TCGA data. The expression patterns were verified by multiple datasets and lab experiments. GEPIA 2.0 and TIMER 2.0 were utilized to analyze the correlation of the hub genes with EMT markers and immune infiltration. GeneMANIA, STRING, and Metascape were used for co-expression and pathway enrichment analysis. Finally, we established a signature by the method of multivariate Cox regression analysis. CDKN2A, CMTM8 and ILK were filtered out as prognostic genes. CDKN2A and CMTM8 were up-regulated, while ILK was down-regulated in CC. CDKN2A was positively correlated with infiltration of macrophages, Th2 cells, Treg cells, and negatively correlated with NK cells. CMTM8 was negatively correlated with CD8+ T cells, dendritic cells, and NK cells. ILK was positively correlated with CD8+ T cells and dendritic cells. Moreover, CDKN2A, CMTM8 and ILK were significantly correlated with EMT markers. The three genes could participate in the TGF-β pathway. The prognosis model established by the three hub genes was an independent prognosis factor, which can better predict the prognosis. CDKN2A, CMTM8 and ILK are promising prognostic biomarkers and may be potential therapeutic targets in colon cancer.

Keywords: Bioinformatics analysis; Biomarker; Colon cancer; Epithelial-mesenchymal transition; Immune; Prognosis.

Fig. 1

Flowchart for screening biomarkers in colon cancer

Fig. 2

Identifying the EMT-immune-related prognostic DEGs. A Venn plot displaying the identification of EMT-immune-related DEGs. B Kaplan–Meier survival analysis of CDKN2A, CMTM8 and ILK in TCGA-COAD data. C Kaplan–Meier survival analysis of CDKN2A, CMTM8 and ILK in GSE39582, GSE24551, GSE29621 respectively. D ROC analysis of CDKN2A, CMTM8 and ILK in CC

Fig. 3

Expression analysis of CDKN2A, CMTM8 and ILK by different database and scRNA-seq data. A GSE10950 dataset. B GEPIA 2.0 database. C Oncomine database (P value < 0.05, Fold Change > 2, Gene rank = All, Data Type = mRNA). D Cluster diagram by cell types in the scRNA-seq data. E Cluster diagram by the cell origin in the scRNA-seq data. F The feature plots of CDKN2A, CMTM8 and ILK by scRNA-seq data. G The violin plots of CDKN2A, CMTM8 and ILK by scRNA-seq data

Fig. 4

The immunohistochemistry images from wet lab. A The representative images of CDKN2A in the tumor and adjacent tissues. B The representative images of CMTM8 in the tumor and adjacent tissues. C The representative images of ILK in the tumor and adjacent tissues. D The statistical results of CDKN2A expression in the tumor and adjacent tissues. E The statistical results of CMTM8 expression in the tumor and adjacent tissues. F The statistical results of ILK expression in the tumor and adjacent tissues

Fig. 5

The immune related analysis in colon cancer. A The situation of immune cell infiltration in each TCGA-COAD sample. B The correlation among different immune cell types. C The heat plot of immune infiltration between colon and normal samples. D The violin plot of immune infiltration between colon and normal samples. E Correlation of CDKN2A expression levels with immune infiltration. F Correlation of CMTM8 expression levels with immune infiltration. G Correlation of ILK expression levels with immune infiltration

Fig. 6

The co-expression and pathway enrichment analysis of hub genes. A The display of co-expressional network by GeneMANIA. B The pathway enrichment analysis of hub genes by Metascape. C The interactions among the enriched pathways

Fig. 7

The competing endogenous RNA (ceRNA) network of lncRNA-miRNA-mRNA in colon cancer

Fig. 8

Verification the accuracy of the riskscore and predicted the survival probability of colon cancer patients combined with other clinical characters. A Kaplan–Meier curves between high-risk group and low-risk group by TCGA-COAD data. B The AUC score of the riskscore and other clinical indicators at 1 year in the TCGA-COAD data. C The patients from TCGA are divided into high-risk group and low-risk group based on the riskscore median values. D The distribution of survival time in the high-risk group and low-risk group in the TCGA data. E The expression heatmap of CDKN2A, CMTM8, ILK in the high-risk group and low-risk group in the TCGA data. F Univariate cox regression analysis of age, gender, stage, and riskscore in the TCGA data. Riskscore is significantly associated with the survival of colon cancer. G Multivariate cox regression analysis of age, gender, stage, and riskscore in the TCGA data. Riskscore is an independent prognostic factor for the survival of colon cancer. H Nomogram for the prediction of 1-, 3-, 5-year survival probability in patients with colon cancer