Upregulation of Centromere Proteins as Potential Biomarkers for Esophageal Squamous Cell Carcinoma Diagnosis and Prognosis

Abstract

Esophageal squamous cell carcinoma (ESCC) has a high incidence and low survival rate, necessitating the identification of novel specific biomarkers. Centromere-associated proteins (CENPs) have been reported to be biomarkers for many cancers, but their roles in ESCC have seldom been investigated. Here, the potential clinical roles of CENPs in ESCC patients were demonstrated by a systematic bioinformatics analysis. Most CENP-encoding genes were differentially expressed between tumor and normal tissues. CENPA, CENPE, CENPF, CENPI, CENPM, CENPN, CENPQ, and CENPR were upregulated universally in the three datasets. Survival analysis demonstrated that high expression of CENPE and CENPQ was positively correlated with the outcomes of ESCC patients. The CENPE-based forecast model was more accurate than the tumor-node-metastasis (TNM) staging-based model, which was classified as stage I/II vs. III/IV. More importantly, the forecast model based on the commonly upregulated CENPs exhibited a much higher area under the curve (AUC) value (0.855) than the currently known TTL, ZNF750, AC016205.1, and BOLA3 biomarkers. The nomogram model integrating the CENPs, TNM stage, and sex was highly accurate in the prognosis of ESCC patients (AUC = 0.906). Besides, gene set enrichment analysis (GSEA) demonstrated that CENPE expression is significantly correlated with cell cycle, G2/M checkpoint, mitotic spindle, p53, etc. Finally, in validation experiments, we also found that CENPE and CENPQ were significantly overexpressed in esophageal cancer cells. Taken together, these results clearly suggest that CENPs are clinically promising diagnostic and prognostic biomarkers for ESCC patients.

Figure 1

Aberrant expression of CENPs in ESCC. (a–c) Heatmaps showing expression differences of CENP-encoding genes between the tumor and normal samples in order of descending logFC in TCGA, GSE38129, and GSE20347. Blue and red colors represent low and high expression, respectively. ^∗∗∗P < 0.001; ^∗∗P < 0.01; ^∗P < 0.05; and ^NSP > 0.05. Genes with FDR < 0.05 and P < 0.05 were identified as DEGs. (d) Venn diagram displaying the overlapped DEGs in the three datasets, including CENPA, E, F, I, M, N, Q, and R. (e–g) Boxplots representing the different expression levels of the overlapped genes in tumor and normal samples according to TCGA, GSE38129, and GSE20347.

Figure 2

The Kaplan–Meier survival analysis of TCGA data. (a) OS curves of the overlapped DEGs, including CENPE, Q, A, R, F, N, M, and I in order of descending P values. (b) OS curves of three clinicopathological risk factors, N stage, sex, and TNM stage. The number of samples in (a) was 81. The total number of patients in (b) was 95. In N stage, N0 + N1 = 84, N2 + N3 = 9, and missing = 2. In gender, male = 80 and female = 15. In TNM stage, stage I + II = 63, stage III + IV = 31, and missing = 1.

Figure 3

Forecast models to forecast the prognosis of ESCC patients. (a) Forecast model according to the TNM stage (I+II vs. III+IV) in ESCC. (b) CENPE-based forecast model. (c) Multigene forecast model and survival curve based on the expression of CENPs. (d) Nomogram integrating TNM stage, sex, and CENPs-based risk score to forecast individual OS for ESCC patients. (e) ROC curve to evaluate the nomogram's performance in forecasting patients' OS. (f) ROC curves showing the sensitivity and specificity of CENPE, CENPs, and other known biomarkers in forecasting ESCC patients' survival. The number of samples in (a) was 94, in (b, c) and (f) was 81, and in (d, e) was 79.

Figure 4

Significantly enriched gene sets in GSEA. The number of samples in GSEA was 81. NES: normalized enrichment score.

Figure 5

The coexpressed network of CENPs and their coexpressed genes. The blue circles represented the coexpressed genes. The red circles represented CENPs, of which the bigger ones represented the overlapped DEGs and the biggest one represented CENPE. The number of samples in WGCNA was 81.

Figure 6

Validation of the mRNA expression differences of CENPE and CENPQ at the cell line level. (a) Histogram showing CENPE expression difference between ESCC cell lines and normal cells based on the GSE23964 dataset (2 normal esophageal normal epithelium normal cell lines and 14 ESCC ones). Normal cell lines and ESCC ones were filled in black and grey colors, respectively. (b) Boxplot representing CENPE expression difference based on the GSE23964 dataset. logFC = 1.86, P < 0.001. (c and d) RT-qPCR results of CENPE and CENPQ, respectively. Normal cell lines and ESCC cells were filled in black and grey colors, respectively. ^∗P < 0.05.