FNDC3B and BPGM Are Involved in Human Papillomavirus-Mediated Carcinogenesis of Cervical Cancer

Abstract

Human papillomavirus (HPV)-mediated cervical carcinogenesis is a multistep progressing from persistent infection, precancerous lesion to cervical cancer (CCa). Although molecular alterations driven by viral oncoproteins are necessary in cervical carcinogenesis, the key regulators behind the multistep process remain not well understood. It is pivotal to identify the key genes involved in the process for early diagnosis and treatment of this disease. Here we analyzed the mRNA expression profiles in cervical samples including normal, cervical intraepithelial neoplasia (CIN), and CCa. A co-expression network was constructed using weighted gene co-expression network analysis (WGCNA) to reveal the crucial modules in the dynamic process from HPV infection to CCa development. Furthermore, the differentially expressed genes (DEGs) that could distinguish all stages of progression of CCa were screened. The key genes involved in HPV-CCa were identified. It was found that the genes involved in DNA replication/repair and cell cycle were upregulated in CIN compared with normal control, and sustained in CCa, accompanied by substantial metabolic shifts. We found that upregulated fibronectin type III domain-containing 3B (FNDC3B) and downregulated bisphosphoglycerate mutase (BPGM) could differentiate all stages of CCa progression. In patients with CCa, a higher expression of FNDC3B or lower expression of BPGM was closely correlated with a shorter overall survival (OS) and disease-free survival (DFS). A receiver operating characteristic (ROC) analysis of CIN and CCa showed that FNDC3B had the highest sensitivity and specificity for predicting CCa development. Taken together, the current data showed that FNDC3B and BPGM were key genes involved in HPV-mediated transformation from normal epithelium to precancerous lesions and CCa.

Keywords: BPGM; FNDC3B; WGCNA; cervical cancer; human papillomavirus.

Figure 1

WGCNA on the dynamic progression of HPV-associated CCa. (A) Hierarchical clustering analysis of mRNAs differentially expressed in cervical samples spanning from normal, to CIN and CCa (fold change > 1.5 and p < 0.05). Eight normal, 12 CIN, and 10 CCa tissues in GSE138080 were used to conduct bioinformatics analysis after PCA. (B, C) Network topology analysis at different soft-thresholding powers. R2 > 0.8 and mean connectivity < 100 when β = 12, indicating that the co-expression network was a scale-free topology. (D) Cluster dendrogram based on module eigengenes. (E) The cluster dendrogram of co-expression network modules. The mRNAs with highly similar expression patterns were placed in the same module using the adaptive branch pruning algorithm. (F) An adjacency heatmap showed the topological overlap matrix among all 12,524 genes.

Figure 2

GO functional enrichment analysis. GO terms associated with each of the 16 modules.

Figure 3

The relationship between modules and clinical traits. (A) Module-trait relationship analysis. Each row indicated a module eigengene, and each column indicated a clinical trait. Each cell included the corresponding correlation and p value. (B) A scatterplot of gene significance (GS) vs. module membership (MM) in the purple_M1 module (correlation = 0.78, p < 1e-200). (C) A scatterplot of GS vs. MM in the green_M7 module (correlation = 0.62, p < 3.9e-141).

Figure 4

Functional enrichment analysis. (A) KEGG enrichment analysis of genes in the purple_M1 module. (B) GO enrichment analysis of genes in the purple_M1 module. (C) KEGG enrichment analysis of genes in the green_M7 module. (D) GO enrichment analysis of genes in the green_M7 module.

Figure 5

FNDC3B and BPGM expression analysis. (A, B) The expression of FNDC3B analyzed in the test dataset GSE138080 and in the validation dataset GSE63514. (C) The expression of FNDC3B assessed using qRT-PCR in clinical specimens. (D, E) The expression of BPGM analyzed in the test dataset GSE138080 and in the validation dataset GSE63514. (F) The expression of BPGM assessed using qRT-PCR in clinical specimens. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.001, ns, not significant.

Figure 6

Mutation analysis of FNDC3B and BPGM in CCa. There were 1.9% genomic alterations of FNDC3B in CCa, and there were 0% genomic alterations of BPGM in CCa.

Figure 7

The relationships between FNDC3B (or BPGM) expression and the clinical outcome in CCa patients. (A) A higher expression of FNDC3B exhibited a significantly worse OS. (B) A higher expression of FNDC3B exhibited a worse DFS. (C, D) A lower expression of BPGM exhibited a significantly worse OS and DFS.

Figure 8

Receiver operating characteristic curve analysis in CIN and CCa. (A) ROC of FNDC3B in CIN. (B) ROC of FNDC3B in CCa. (C) ROC of BPGM in CIN. (D) ROC of BPGM in CCa.