Eukaryotic cells rely on mainly two metabolic processes to maintain cellular hemostasis by eliminating damaged proteins and organelles: the ubiquitin-proteasome system and autophagy [19]. Numerous studies have pointed out autophagic dysfunctions were related with various diseases such as infections, neurodegeneration and tumors [20–22]. Autophagy is a double-edged sword in tumorigenesis, which can suppress or promote tumor development in a context-dependent manner, including the tumor microenvironments and patients’ clinical features. Our study showed 30 SD-ARGs between EAC and normal groups. Among these genes, 6 ARGs were down-regulated and 24 ARGs were up-regulated in EAC. By GO and KEGG enrichment analyses, we discovered these SD-ARGs were mainly enriched in the cellular apoptotic signaling pathway. We also found 8 ARGs had prognostic values, in which the BECN1, DAPK1 and CAPN1 played a protective role in survival. In addition, combination of ARGs and clinical data provided a new method to explore independent survival risk factors, which may be crucial to evaluate prognosis and improve the survival chances under appropriate interventions.
Several studies have strongly supported the associations and their distinct expressions levels between autophagic gene and EAC [23–25]. CDKN2A (also known as p16), an autophagic gene, mainly regulates the G1/S cellular cycle process, which is known as a tumor suppressor gene [26–27]. Hardie L J et al [23] found CDKN2A hypermethylation was ubiquitous in EAC compared with normal tissue and had protective roles in the molecular progression of Barrett’s epithelium to EAC, which implicated that CDKN2A was in active state when tumorigenesis. Gockel I et al [24] provided evidence that higher CXCR4 expression was associated with malignant transformation in EAC. This was consistent with our result that CXCR4 expression was higher in EAC than normal tissue. The interaction of CXCR4 and its ligand SDF-1α, which mediates the activation of phosphatidylinositol 3-kinase (PI3K) and Akt pathways, resulting in cell proliferation [28–29]. These demonstrated the trend to a less favorable outcome associated with an increased expression of CXCR4 in EAC, although the significant survival correlation was not observed in our study. Our results demonstrated HDAC1 expression was higher in EAC compared with normal control. This is in line with previous work by Langer R et al [25], revealing HDAC1 had higher expression, especially HDAC2, and was associated with aggressive tumor behavior in EAC. In addition, in vitro studies have shown that high HDAC activity leads to tumor dedifferentiation and enhanced tumor cell proliferation [30]. Hence fore, high HDAC expression may represent a surrogate marker for aggressive tumor behavior in EAC. A promising aspect from HDAC was that HDAC inhibitors have been shown to act as radiosensitizers in a variety of cancer cell lines [31], so HDAC inhibitors might be extremely useful for chemotherapeutic or radio chemotherapeutic combination therapies for EAC.
Another crucial mechanism of cell death is apoptosis that involves the activation of catabolic enzymes. The intricate details between autophagy and apoptosis trigger a pivotal crosstalk in the tumor suppression [32]. In our study, GO and KEGG analyses demonstrated SD-ARGs were mainly enriched in the cellular apoptotic pathways, and these results can be confirmed in other studies [33–34]. Shimizu S et al [33] reported the BCL2 autophagic family members (BAX, BAK1) not only acted as messages of apoptotic signal converge, but also regulated apoptosis, which showed a good agreement with our results. The interaction between autophagy and apoptosis is mediated by different molecular in EAC microenvironment. In an experiment by Ma Z et al [35], they revealed that BNIP3, an autophagic gene, could induce esophageal cancer cell apoptosis in hypoxia and autophagic inhibitor 3-methyladenine (3-MA) could augment BNIP3-induced cell apoptosis and death. His study suggested autophagy and apoptosis played the opposing roles in esophageal cancer. In addition, cleavage of Bcelin-1 could induce proapoptotic factors release from mitochondria and enhance apoptosis, as well as inhibit autophagic function [36]. These studies added strength to the close correlation between autophagy and apoptosis. However, a further mechanistic understanding of the relation for validation is necessary in EAC.
We conducted cox regressions to identify prognostic ARGs and the results showed ATG5, SIRT1, DAPK1 and other five genes were associated with survival. The ATG5 gene encodes autophagy protein 5 (Atg5), which could combine with Atg12/ Atg16 to form complex involving the process of autophagy [37]. The utility of genes as biomarker has been demonstrated and a number of genes have been reported. In a work by Yang PW et al [38], increased ATG5 expression was observed in esophageal cancer compared with normal tissue. His group also found higher ATG5 expression tumor group had poorer prognosis including the overall survival (OS) and progression-free survival (PFS). This result was consistent with our study, implying the ATG5 was a high-risk gene. BECN1 gene (coding for Beclin 1), exerted a crucial role in autophagosome formation though interacting with Vps34 [39–40]. Weh KM et al [41] found Beclin-1 expression loss occurred more frequently in EAC patients compared with controls (49.0% vs 4.8%). Additionally, Beclin 1 expression level was negatively correlated with EAC histologic grade and stage (P < 0.005). Consistently with our study, they also demonstrated increase in Beclin 1 could cause long-term survival, which implied BECN1 was a tumor suppressor gene and may act as a prognostic biomarker. SIRT1 has function of activating stress response, maintaining genomic integrity and involving the apoptosis and tumorigenesis [42]. Ma M C et al [43] reported SIRT1 was an independently survival risk factor in esophageal cancer and its overexpression was associated with worse OS (HR = 1.776, P = 0.009) and disease-free survival (DFS) (HR = 1.642, P = 0.017). These prognostic ARGs may be useful for early detection and might be a valid strategy to increase the survival chances.
Further analysis of our study showed that risk score was an independent risk factor of prognosis, which suggested autophagy genes could serve as an accurate survival indicator. Patients with high risk score exhibited obvious worse prognosis. ROC curve showed risk score associated with ARGs was the most important variable in predicting the EAC survival, implying its significant potential to be used as a prognostic biomarker. Therefore, more precise individualized treatment strategies for EAC patients with high risk scores should be established. At last, we evaluated the relations between ARGs and patients’ clinical features. The results showed BECN1, DAPK1, VAMP7 and risk score were significantly associated with survival status (all P values ༜0.05). The significant downregulations of BECN1 and DAPK1 in EAC patients compared with normal control implied that they were protective effect. Whereas the VAMP7 was conferred an increased risk. Moreover, the level of SIRT1 increased with the increasing tumor stage and primary tumor, which indicated ARGs were involved in the progression of EAC and it is essential to increase our understanding of the pathways between autophagy and clinical features.
The strength of our study is that we performed a systematic analysis of autophagic genes from national database, which provided a robust statistical support. Meanwhile, there are also some limitations. Firstly, the clinical information downloaded from the TCGA was incomplete. Detailed information such as age, tumor size was available. Secondly, the mechanisms how ARGs modulate the process of EAC were unclear. Lastly, the prognostic model needs to be verified in a large-scale and multicenter clinical cohort. Notwithstanding its limitations, this study does provide a comprehensive overview of ARGs profile in EAC and these limitations can be solved if there are enough data in the future.