Construction of co-expression modules related to survival by WGCNA and identification of potential prognostic biomarkers in glioblastoma

Abstract

Glioblastoma (GBM) is a malignant brain tumour with poor prognosis. The potential pathogenesis and therapeutic target are still need to be explored. Herein, TCGA expression profile data and clinical information were downloaded, and the WGCNA was conducted. Hub genes which closely related to poor prognosis of GBM were obtained. Further, the relationship between the genes of interest and prognosis of GBM, and immune microenvironment were analysed. Patients from TCGA were divided into high- and low-risk group. WGCNA was applied to the high- and low-risk group and the black module with the lowest preservation was identified which could distinguish the prognosis level of these two groups. The top 10 hub genes which were closely related to poor prognosis of patients were obtained. GO analysis showed the biological process of these genes mainly enriched in: Cell cycle, Progesterone-mediated oocyte maturation and Oocyte meiosis. CDCA5 and CDCA8 were screened out as the genes of interest. We found that their expression levels were closely related to overall survival. The difference analysis resulted from the TCGA database proved both CDCA5 and CDCA8 were highly expressed in GBM. After transfection of U87-MG cells with small interfering RNA, it revealed that knockdown of the CDCA5 and CDCA8 could influence the biological behaviours of proliferation, clonogenicity and apoptosis of GBM cells. Then, single-gene analysis was performed. CDCA5 and CDCA8 both had good correlations with genes that regulate cell cycle in the p53 signalling pathway. Moreover, it revealed that high amplification of CDCA5 was correlated with CD8⁺ T cells while CDCA8 with CD4⁺ T cells in GBM. These results might provide new molecular targets and intervention strategy for GBM.

Keywords: Weighted Gene Co-expression Network Analysis (WGCNA); cell cycle; glioblastoma; hub gene; the Cancer Genome Atlas (TCGA).

FIGURE 1

The detection of outlier samples, the selection and validation of optimal soft threshold power to construct gene co‐expression networks. A, The clustering dendrogram of high‐risk samples to detect outliers. B, The clustering dendrogram of low‐risk samples to detect outliers. C, The scale independence and the mean connectivity of the WGCNA analysis of the high‐risk samples. D, The scale independence and the mean connectivity of the WGCNA analysis of the low‐risk samples. E, The histogram of k and the correlation coefficient between k and p (k) of the high‐risk samples. F, The histogram of k and the correlation coefficient between k and p (k) of the low‐risk samples

FIGURE 2

Co‐expression modules identified and characterized by WGCNA. Clustering dendrograms of (A) High‐risk samples and (B) Low‐risk samples. (C) The preservation median rank of ten co‐expression modules. (D) The preservation Z‐summary score of ten co‐expression modules

FIGURE 3

Drawing of the hub genes heat map and performing GO enrichment analysis in the black module. A, The heat map of 50 hub genes in the black module. B, The bubble plot showed the top enriched biological process of these genes. C, The scatter plot showed the distribution of hub genes involved in specific links in the cell cycle. D, The Up‐Set plot showed the interactions among the ten links of cell cycle

FIGURE 4

Validation with external databases and functional analysis of CDCA5 and CDCA8 in GBM cells. A, The overall survival rate of high‐ and low‐risk patients based on the expression levels of CDCA5 and CDCA8 in the GSE4412 dataset. B, The difference in expression of CDCA5 and CDCA8 between normal and tumour group based on the TCGA database. C, qRT‐PCR experiments were performed to detect the expression of CDCA5 and CDCA8 after siRNA transfection. D, Cell proliferation ability (E) clonogenicity and (F) apoptosis of U87‐MG cells after transfection by si‐CDCA5 and si‐CDCA8 was determined. *P < .05, **P < .01 vs si‐NC

FIGURE 5

Single‐gene analysis to study the relationship between CDCA5 and CDCA8 and cell cycle related genes in the p53 pathway. A and B, The heatmaps of CDCA5 and CDCA8 and cell cycle related genes expression profiles in the p53 pathway. C, The scatter plot showed the correlation between CDCA5 and cell cycle related genes in the p53 pathway. D, The scatter plot showed the correlation between CDCA8 and cell cycle related genes in the p53 pathway

FIGURE 6

Single‐gene analysis to study the relationship between CDCA5 and CDCA8 and immune microenvironment. A, The differential expression of CDCA5 and CDCA8 in tumours and normal tissues of multiple cancer species. B, The relationship between the expression levels of CDCA5 and CDCA8 and the six types of immune cells. The relationship between expression levels and copy number variations of (C) CDCA5 and (D) CDCA8 and the levels of infiltration of six immune cells