Epithelial-Mesenchymal Transition Gene Signature Is Associated with Neoadjuvant Chemoradiotherapy Resistance and Prognosis of Esophageal Squamous Cell Carcinoma

Abstract

Background: Neoadjuvant chemoradiotherapy (neo-CRT) in combination with surgery increases survival compared to surgery alone, as indicated by the esophageal squamous cell carcinoma (ESCC) treatment recommendations. However, the benefits of neo-CRT are diverse among patients. Consequently, the development of new biomarkers that correlate with neo-CRT might be important for the treatment of ESCC.

Methods: The differentially expressed genes (DEG) between responsive and resistant samples from the GSE45670 dataset were obtained. On the TCGA dataset, survival analysis was performed to identify prognosis-related-EMT-genes. For EMT score model construction, lasso regression analysis in the TCGA cohort was used to identify the genes. In the TCGA-ESCC cohort, age, stage, and EMT score were used to construct a nomogram.

Results: In total, 10 prognosis-related-EMT-genes were obtained. These 10 genes consisted of 6 risky genes and 4 protective genes. Based on the lasso analysis and univariate Cox regression, an EMT score model consisting of 7 genes (CLEC18A, PIR, KCNN4, MST1R, CAPG, ALDH5A1, and COX7B) was identified. ESCC patients with a high EMT score have a worse prognosis. These genes were differentially expressed between responsive and resistant patients and had a high accuracy for distinguishing resistant and responsive patients.

Conclusions: The identified genes have the potential to function as molecular biomarkers for predicting ESCC patients' resistance to neo-CRT. This research may aid in the elucidation of the molecular processes driving resistance and the identification of targets for improving the prognosis for ESCC.

Figure 1

Principal component analysis (PCA) in resistant versus responsive samples. (a) The flowchart of this study. (b) Before removing the outliers, the PCA was performed on the gene expression data. (c) After removing the outliers, the PCA was performed on the gene expression. (d) Volcano plot of DEG by log2 foldchange (FC) > 0.5 and p value < 0.05. (e) Clustering heat map of the DEG. The expression data for DEG was normalized.

Figure 2

Identification of PREMTs in ESCC. (a, b) Venn diagrams for identifying PREMTs. (c) Lasso coefficient profiles of the 10 PREMTs. (d) Selection of the number of genes for EMT score by lasso analysis.

Figure 3

EMT score based on 7 EMT genes. EMT score distribution, survival overview (a), and heat map (b) for patients in the different groups. (c) The survival curves differentiate between groups. (d) The predictive accuracy of the EMT signature for TCGA patients was shown using ROC curves.

Figure 4

The nomogram constructed in the TCGA-ESCC. (a) The nomogram for predicting OS. The calibration plots for predicting 1-year (b), 3-year (c), and 5-year (d) OS.

Figure 5

(a) The expression pattern of genes between responsive and resistant patients. ROC curves of genes and EMT score. (b) ROC curve of CLEC18A. (c) ROC curve of ALDH5A1. (d) ROC curve of PIR. (e) ROC curve of COX7B. (f) ROC curve of CAPG. (g) ROC curve of KCNN4. (h) ROC curve of MST1R. (i) ROC curve of EMT score. AUC > 0.7 indicates good effect.

Figure 6

(a) The expression pattern of genes between normal and tumor samples. ROC curves of genes and EMT score. (b) ROC curves of CLEC18A. (c) ROC curves of ALDH5A1. (d) ROC curves of PIR. (e) ROC curves of COX7B. (f) ROC curves of CAPG. (g) ROC curves of KCNN4. (h) ROC curves of MST1R. (i) ROC curves of EMT score. AUC > 0.7 indicates good effect.