Transcriptional factor regulation network and competitive endogenous RNA (ceRNA) network determining response of esophageal squamous cell carcinomas to neoadjuvant chemoradiotherapy

Background Neoadjuvant chemoradiotherapy (nCRT) followed by surgery benefits survival for patients with esophageal squamous cell carcinomas (ESCC) compared with surgery alone, but the clinical outcomes of nCRT are heterogeneous. This study aimed to elucidate transcriptional factor (TF) regulation network and competitive endogenous RNA (ceRNA) network determining response of ESCC to nCRT. Materials and Methods RNA microarray data of GSE59974 and GSE45670 were analyzed to investigate the significant changes of lincRNAs, miRNAs, mRNAs in responders and non-responders of nCRT in ESCC. Functional and enrichment analyses were conducted by clusterProfiler. The target lincRNAs and mRNAs of miRNAs were predicted by miRWalk. The ceRNA and TF regulatory networks were constructed using Cytoscape. Results Differentially expressed genes between responders and non-responders mainly enriched in biological process including Wnt signaling pathway and regulation of cell development and morphogenesis involved in differentiation. Besides, these genes showed enrichment in molecular function of glycosaminoglycan binding, metalloendopeptidase inhibitor and growth factor activity. KEGG analysis enriched these genes in pathways of neurotrophin signaling pathway, cell adhesion molecules and Wnt signaling pathway. We also constructed ceRNA network and TF network regulating response of ESCC to nCRT. Core regulatory miRNAs were miR-520a, miR-548am, miR-3184, miR-548d, miR-4725, miR-148a, miR-4659a and key regulatory TFs included MBNL1, SLC26A3, BMP4, ZIC1 and ANKRD7. Conclusion We identified significantly altered lincRNAs, miRNAs and mRNAs involved in the nCRT response of ESCC. In addition, the ceRNA regulatory network of lincRNA-miRNA-mRNA and TF regulatory network were constructed, which would elucidate novel molecular mechanisms determining nCRT response of ESCC, thus providing promising clues for clinical therapy.


48
As a common malignant tumour of upper digestive tract, esophageal squamous cell 49 carcinoma (ESCC) represents a significant health burden worldwide due to its aggressive 50 growth (Kang et al. 2015). There is accumulating evidence that surgical resection of locally 51 advanced ESCC is an effective method to control the progression of this disease (Sjoquist 52 et al. 2011). In some cases, however, the recurrence after curative resection and 53 unsatisfactory survival status still pose significant obstacle for surgeons (Rohatgi et al. 54 2006). In recent years, it has been proved that neoadjuvant chemoradiotherapy (nCRT) 55 followed by surgery benefits survival for patients with ESCC compared with surgery alone,

61
Noncoding RNAs (ncRNAs) are transcripts which possess no protein coding 131 Transcription factor regulation network 132 Search Tool for the Retrieval of Interacting Genes (STRING) database was used to 133 find interacting proteins between different genes (von Mering et al. 2005). Interactions with 134 a combined score > 0.4 were defined as significant. Next, we use CentiScape, an 135 application in Cytoscape, to screen for the hub protein. TFCheckpoint database was used 136 to identify the transcription factor among differentially expressed genes. The transcription 137 factor with connection numbers over 5 were considered as the hub transcription factors. 138 Using Cytoscape, we finally built transcription factor regulation network.
139 MiRNA regulatory network 140 The target genes of differentially expressed miRNAs were predicted by miRWalk 141 (http://zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk/index.html), which is a 142 comprehensive online algorithm that provides information on miRNA from Human, Mouse 143 and Rat on their predicted as well as validated binding sites on their target genes (Dweep 144 et al. 2011). Accurate classification of 3p/5p for miRNAs was adopted in target gene 145 prediction. The predicted target genes were matched to the genes whose mRNA 146 expressions were opposite to the miRNA profile because miRNAs are negatively 147 regulated genes. In addition, the miRNAs with connection numbers more than 5 were 148 selected as the hub miRNAs.
149 CeRNA regulation network 150 Using miRWalk, we predicted the interacrtion of lincRNA with miRNA. Then, we 152 expression array. In this study, the top 10 miRNA was used the built the ceRNA regulation 153 network. According to the ceRNA theory that lincRNAs act as natural miRNA sponges to 154 inhibit miRNA functions, the expressions of lincRNA-miRNA and miRNA-mRNA were all 155 negatively correlated. Finally, the ceRNA regulatory network of lincRNA-miRNA-mRNA 156 in nCRT response of ESCC was constructed by Cytoscape software. The significantly 157 altered miRNAs/mRNAs were analyzed for predictive power in separating responders 158 from non-responders.   Figure 3).
183 Transcription factor regulation network 184 According to STRING dataset, totally 49 proteins interacted with each other. Using 185 CentiScape software, we selected BMP4 as the hub transcription factors (Table 3). We 186 then built the transcription factor regulation network. As was shown in Figure 4, the red 187 node represents the up-regulated gene and the blue node represents the down-regulated 188 gene. The darker the color, the higher the expression level was. Purple border 189 represented transcription factors. The node size increased with degree.  By means of miRWalk database, we constructed miRNA regulation network involved 249 in the nCRT response of ESCC including 13 genes regulated by up-regulated hub 250 miRNAs and 31 genes regulated by the hub miRNAs with decreased expression. The 251 miRNAs with connection numbers over 5 were defined as hub miRNAs, which were hsa-252 miR-520a-3p, hsa-miR-548am-3p, hsa-miR-3184-5p, hsa-miR-548d-5p, hsa-miR-4725-253 3p, hsa-miR-148a-5p, and hsa-miR-4659a-3p. Previously, miR-520a has been proved to