Biomarker identification and trans-regulatory network analyses in esophageal adenocarcinoma and Barrett's esophagus

Abstract

Background: Esophageal adenocarcinoma (EAC) is an aggressive disease with high mortality and an overall 5-year survival rate of less than 20%. Barrett's esophagus (BE) is the only known precursor of EAC, and patients with BE have a persistent and excessive risk of EAC over time. Individuals with BE are up to 30-125 times more likely to develop EAC than the general population. Thus, early detection of EAC and BE could significantly improve the 5-year survival rate of EAC. Due to the limitations of endoscopic surveillance and the lack of clinical risk stratification strategies, molecular biomarkers should be considered and thoroughly investigated.

Aim: To explore the transcriptome changes in the progression from normal esophagus (NE) to BE and EAC.

Methods: Two datasets from the Gene Expression Omnibus (GEO) in NCBI Database (https://www.ncbi.nlm.nih.gov/geo/) were retrieved and used as a training and a test dataset separately, since NE, BE, and EAC samples were included and the sample sizes were adequate. This study identified differentially expressed genes (DEGs) using the R/Bioconductor project and constructed trans-regulatory networks based on the Transcriptional Regulatory Element Database and Cytoscape software. Enrichment of Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) terms was identified using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) Bioinformatics Resources. The diagnostic potential of certain DEGs was assessed in both datasets.

Results: In the GSE1420 dataset, the number of up-regulated DEGs was larger than that of down-regulated DEGs when comparing EAC vs NE and BE vs NE. Among these DEGs, five differentially expressed transcription factors (DETFs) displayed the same trend in expression across all the comparison groups. Of these five DETFs, E2F3, FOXA2, and HOXB7 were up-regulated, while PAX9 and TFAP2C were down-regulated. Additionally, the majority of the DEGs in trans-regulatory networks were up-regulated. The intersection of these potential DEGs displayed the same direction of changes in expression when comparing the DEGs in the GSE26886 dataset to the DEGs in trans-regulatory networks above. The receiver operating characteristic curve analysis was performed for both datasets and found that TIMP1 and COL1A1 could discriminate EAC from NE tissue, while REG1A, MMP1, and CA2 could distinguish BE from NE tissue. DAVID annotation indicated that COL1A1 and MMP1 could be potent biomarkers for EAC and BE, respectively, since they participate in the majority of the enriched KEGG and GO terms that are important for inflammation and cancer.

Conclusion: After the construction and analyses of the trans-regulatory networks in EAC and BE, the results indicate that COL1A1 and MMP1 could be potential biomarkers for EAC and BE, respectively.

Keywords: Barrett’s esophagus; Differentially expressed genes; Esophageal adenocarcinoma; Microarray; Transcription factors.

Figure 1

The numbers of differentially expressed genes in different comparison groups (esophageal adenocarcinoma vs normal esophagus, Barrett's esophagus vs normal esophagus, and esophageal adenocarcinoma vs Barrett's esophagus) in the GSE1420 dataset. The GSE1420 dataset consists of three different groups [normal esophagus (NE), Barrett's esophagus (BE), and esophageal adenocarcinoma (EAC)]. The differentially expressed genes (DEGs) in different comparison groups (EAC vs NE, BE vs NE, and EAC vs BE) were identified using the R/Bioconductor software (|FC| > 1.5, P-values < 0.05), and the numbers of DEGs in different comparison groups were summarized.

Figure 2

The trans-regulatory networks of differentially expressed transcription factors and their regulatory differentially expressed genes. Circles represent the differentially expressed genes (DEGs) (red for up-regulated DEGs and green for down-regulated DEGs) regulated by the predicted transcription factor (TFs) (yellow for TFs). The direction of arrows is from the predicted TFs to their target DEGs. A: The trans-regulatory network of differentially expressed TFs and their regulatory DEGs in the comparison group of esophageal adenocarcinoma (EAC) vs normal esophagus (NE); B: The trans-regulatory network of differentially expressed TFs and their regulatory DEGs in the comparison group of Barrett's esophagus (BE) vs NE; C: The trans-regulatory network of differentially expressed TFs and their regulatory DEGs in the comparison group of EAC vs BE.

Figure 3

Bi-cluster analysis of the differentially expressed genes in the trans-regulatory networks existing in both the GSE1420 and GSE26886 datasets. Each row represents one of the differentially expressed genes (DEGs) in the trans-regulatory networks, and each column represents a tissue sample from the two datasets (GSE1420 and GSE26886). The column “normal esophagus (NE)” represents normal esophagus tissue, the column “Barrett's esophagus (BE)” represents Barrett's esophagus tissue, and the column “esophageal adenocarcinoma (EAC)” represents tissue from esophageal adenocarcinoma. The heat map was constructed using each DEG expression value in every tissue sample. A: The heat map for the comparison group of EAC vs. NE in the GSE1420 dataset; B: The heat map for the comparison group of EAC vs NE in the GSE26886 dataset; C: The heat map for the comparison group of BE vs NE in the GSE1420 dataset; D: The heat map for the comparison group of BE vs NE in the GSE26886 dataset.

Figure 4

Receiver operating characteristic curve analysis for tissue discrimination between esophageal adenocarcinoma vs normal esophagus and Barrett's esophagus vs normal esophagus in the GSE1420 dataset. A: Receiver operating characteristic (ROC) curve of TIMP1; B: ROC curve of COL1A1; C: ROC curve of MMP1; D: ROC curve of REG1A; E: ROC curve of CA2; F: ROC curve of ANPEP.

Figure 5

Receiver operating characteristic analysis for tissue discrimination between esophageal adenocarcinoma vs normal esophagus and Barrett's esophagus vs normal esophagus in the GSE26886 dataset. A: Receiver operating characteristic (ROC) curve of TIMP1; B: ROC curve of COL1A1; C: ROC curve of MMP1; D: ROC curve of REG1A; E: ROC curve of CA2; F: ROC curve of ANPEP.