So far, the GC is the one of the top-ranking digestive cancer and has become a worldwide public concern. Thus, it is worthwhile to investigate the potential biomarkers and promising tumour promoting or suppressor genes in GC. In our study, we identified lncRNA-associated ceRNA network involving in GC tumorigenesis, which was based on analysis of gene expression data obtained from TCGA databases. Then, we identified 9 hub lncRNA accompanied with the sub ceRNA network related to OS. Among those 9 lncRNA, we partially validated the effect of INHBA-AS1 and CCDC144NL-AS1 in vitro and found they were promising oncogene in GC.
As mentioned above, lncRNA can influence the expression of mRNA via competitively binding to shared miRNAs, which is defined as ceRNA and may play critical role in regulation of the cancer development and progression (8). This lncRNA-miRNA-mRNA regulatory mechanism has been identified in GC. For instance, lncRNA LINC01133 can inhibit GC progression by sponging hsa-mir-106a-3p and then effect on APC expression (10). In addition, lncRNA PTENP1 can regulate PTEN expression via binding to hsa-mir-106b and hsa-mir-93 in GC (11). Except those individual cases, some other studies carried out integrated analysis for ceRNA. Arun et al. employed 13 GCs and paired adjacent normal tissues and found the crucial ceRNA network for GC (12). Similarly, Liu et al. integrative analysis of lncRNA-associated ceRNA network and validated their findings in vitro in GC (24). However, one of the cell lines they used, BGC823, is considered as HeLa derivative contamination (25). Our study first investigated the differential expressed lncRNA, miRNA and mRNA between normal and GC, respectively, and then constructed a GC development related ceRNA network. This network is involving in up-regulation of MET activates PTK2 signalling, MET promotes cell motility and non-integrin membrane-ECM interactions. Based on REACTOME annotation, the MET activates PTK2 signalling is related to MET receptor activates the focal adhesion kinase FAK1, which plays crucial roles in focal adhesions (FAs). Specifically, FAs are large macromolecular complexes of integrins that mediate cell-ECMs interactions and facilitate the metastatic process (26, 27). Metastasis is a well-known aggressive feature for cancer. Previous studies identified that FAs is strongly associated with metastasis and lower survival rates (26, 28–30). Moreover, FAs can impact on various tumour behaviours, such as migration, invasion and proliferation (31). This may partially explain the influence of ceRNA network on GC. In addition, MET promotes cell motility may also contribute GC development (32). Non-integrin membrane-ECM interactions, such as dystroglycan and 37/67 laminin receptor, is found to be related to variety of epithelial cancers (33).
Subsequently, we further filtered 9 hub lncRNA, including LINC02731, MIR99AHG, INHBA-AS1, CCDC144NL-AS1, VLDLR-AS1, LIFR-AS1, A2M-AS1, LINC01537 and LINC00702. The hub lncRNA associated ceRNA subnetwork involved in actin filament binding and MAPK signalling pathway. Filament is a form of dense meshwork generated by lamellipodia, at the leading edge of a migrating or extending cell, which facilitate cellular movement and play essential roles in tumour cell metastasis (34). MAPK signaling pathway is involved various promoting-cancer mechanisms, such as anti-drug, inflammation and immune evasion(35–38). In term of individual lncRNA, most of them have already found to be related to development and progression of various cancers. For instance, MIR100HG has been validated as an oncogene in the development of myeloid leukemia in vitro (39). In addition, it was also positively related to poorer prognosis in GC based on dataset other than TCGA (40). LncRNA INHBA-AS1 can promote multiple invasion features, including cell growth, migration, and invasion in oral squamous cell carcinoma, which targeting on hsa-mir-143-3p (41). The INHBA-AS1 in GC plasma was found higher expressed compared to the one in heavy controls (42). Knockdown of lncRNA CCDC144NL-AS1 attenuates migration and invasion phenotypes in endometrial stromal cells (43). The expression of VLDLR-AS1 was independently related to prognosis in thymoma (44). The LIFR-AS1/hsa-mir-29a/TNFAIP3 axis played effect on resistance of pohotodynamic therapy in colorectal cancer (45). High expression of lncRNA LIFR-AS1 was correlated with poor survival in GC (46). Up-regulated lncRNA, A2M-AS1, was associated with invasion and migration in breast cancer (47). Except that, LINC00702 enhanced the progression of ovarian cancer through increasing EZH2 expression (48). Then, LINC00702/hsa-mir-4652-3p/ZEB1 axis is promising to promote the progression of malignant meningioma through activating Wnt/β-catenin pathway (49). Taken together, most of those 9 hub lncRNAs were promising tumour-promoting genes in diverse cancer, and were worthwhile for further investigation in GC.
Then, given the regulatory direction of those 9 lncRNA, INHBA-AS1 and CCDC144NL-AS1 and related axis were further verified in vitro and showed the potential promoting influence on proliferation, migration and invasion. This indicated those two lncRNA were promising to play oncogenic roles in GC. In term of their targeted mRNA, INHBA-AS1-regulated COL5A2 and CCDC144NL-AS1-regulated MATN3 have been found to play critical role in GC prognosis (50, 51). MATN3 is a member of the matrilin protein family, a noncollagenous extracellular matrix. It has been widely investigated in bone and cartilage related fields and broadened to malignant tumours in recent years (52–54). Specifically, it can induce the expression of MMP1, MMP3, MMP13, pro-inflammatory cytokines, iNOS, and COX2, indicating MATN3 can regulate extracellular matrix degradation (53). The gene COL5A2, collagen type V alpha 2 chain, encodes an alpha chain for one of the low abundance fibrillar collagens. It has been reported to play critical role in the pathological process in multiple cancers including the colorectal cancer, ovarian cancer and bladder cancer (55, 56). Moreover, COL5A2 was highly correlated with cell extracellular matrix organization, vascularization and EMTs process function, and those function were known to be involved in cancer invasion and metastasis (55).
There are several limitations in our study. Firstly, we only employed TCGA dataset that is comprehensive cancer database. Then, although we combined well-designed bioinformatics study and partially in vitro validation, there was no in-depth in vitro evidence, such as dual-luciferase reporter assay and mice model. Thirdly, the current study was of a retrospective nature, as it was based on data from TCGA dataset without validating it in a prospective clinical trial. Therefore, some vital experiments such as luciferase reporter systems and co-immunoprecipitation assays as well as mice model will be helpful to further validate the correlation among lncRNA-associated ceRNA network. Besides, a large clinical sample size needed to be collected in our future study to verify the ceRNA network.