Glioblastoma (GBM) is the most common and aggressive primary brain malignancy in adults [18]. Due to the high infiltration and heterogeneity of GBM, it is a great challenge to improve its poor prognosis, low survival and high recurrence rates. With the advancement of microarray and sequencing technologies, increasing lncRNAs were confirmed to have potential to be novel and effective diagnostic, therapeutic and prognostic targets for GBM patients. For example, Tian et al have reported lncRNA AGAP2-AS1 is significantly upregulated in GBM and high expression levels of AGAP2-AS1 predicts is a marker of poor prognosis for GBM patients [19]. Generally, the initiation and progression of GBM is orchestrated by multiple factors [20]. Therefore, the capacity of a single gene to predict the prognosis is extremely limited. Currently, increasing studies dedicated to develop prognostic signatures composed of various tumor hallmarks-related lncRNAs, providing promising prognostic indicators and personalized treatments for tumor malignancies. Wang et al established a valuable glycolysis-related lncRNA signature to predict the prognosis of diffuse glioma patients through univariate Cox regression analysis, the Least Absolute Shrinkage and Selection Operator (LASSO) analysis and multivariate Cox regression analysis [21]. Luan et al constructed a prognostic signature composed of 10 autophagy-related lncRNAs for glioma patients [22].
In our present study, we firstly explored the key mechanisms responsible for the development of GBM and found out 29 hallmarks exhibited significant different enrichment between non-tumor and GBM tissues. Among these 29 hallmarks, EMT, angiogenesis, glycolysis, hypoxia have been confirmed to be critical processes of cancer progression and metastasis. More importantly, 78 of 195 genes enriched in EMT were upregulated both in GBM patients from TCGA and GSE4290, which further proving the close link between EMT and GBM development. A significant number of lncRNAs have proved to be essential regulators of EMT [23]. In our present study, we analyzed the correlation of lncRNAs and EMT-related mRNAs in GBM and identified 301 EMT-related lncRNAs. Among these 301 EMT-related lncRNAs, LINC00511 and NEAT1 have been reported to promote malignancy of GBM through EMT; while lncRNA DGCR5 was demonstrated to suppress the migration, invasiveness of GBM cells through reversing EMT process [24–26]. In our study, LINC00511 and NEAT1 were positively correlated with upregulated EMT-related mRNAs, while DGCR5 was negatively correlated with upregulated EMT-related mRNAs, suggesting the potential of LINC00511 and NEAT to promote EMT, DGCR5 to reverse EMT process. The consistency of our results and existing studies proved the reliability of our study. However, the regulatory roles and corresponding molecular mechanisms of the majority of these 301 EMT-related lncRNA in EMT of GBM have not been explored, which need substantial work to be carried out in future.
Several studies have considered the prognostic value of EMT-related lncRNAs for tumors. Xiao et al constructed an eight EMT-related lncRNA signature for melanoma to predict individualized prognosis and therapeutic effects [27]. Du et al identified a five stromal EMT-related lncRNA for bladder cancer, which could be used to predict the prognosis and responsiveness to immune checkpoint blockade therapy [28]. Zhang et al established a survival signature based on EMT-related differentially expressed lncRNAs for patients with kidney renal clear cell carcinoma [29]. It is worth mentioning that the all of these EMT-related lncRNA signature can provide accurate and effective prediction for patients with their corresponding cancer types, indicating considerable clinical value of EMT-related lncRNAs for cancers. However, few studies have focused on the relationship between EMT-related lncRNA and the prognosis of GBM patients. Here, we analyzed the prognostic value of EMT-related lncRNAs in GBM through univariate Cox regression analysis, and constructed an EMT-related lncRNAs prognostic signature for GBM through multivariate Cox regression analysis. Then the prognostic value and robustness of the EMT-related lncRNA signature were validated through Kaplan-Meier survival analysis and ROC curve analysis. Recently, multiple prognostic models have been proposed for patients with GBM. The AUC values of a five-lncRNAs signature established by Niu et al reached 0.690, 0.704, and 0.709 for 1-, 2-, and 3-year survival, respectively. Besides, the AUC values of a 5 immune-related lncRNA risk signature constructed by Li et al were 0.671 to predict 1-year survival and 0.809 to predict 2-year survival[30]. Compared with above published signatures, the AUC values of our EMT-related lncRNAs signature are relatively satisfactory both in training set and validating set.
Among the seven EMT-related lncRNAs in the prognostic model, H19, LINC00609, POM121L9P, and SNHG11 were risk factors of GBM, while AC012615.1, LINC00634, and USP32P3 were protective factors for GBM. To date, the oncogenic roles and molecular mechanisms of H19, a maternally imprinted gene, have been extensively studies in multiple cancers, including glioma, breast cancer, and lung cancer, etc [31–33]. H19 participates in regulation of angiogenesis, autophagy, as well as cell proliferation, migration and invasion [34]. POM121L9P, located on chromosome 22 (22q11.23), has been reported to be associated with shorter OS and poor clinicopathological features of epithelial ovarian cancer [35]. SNHG11, a small nucleolar RNA (snoRNA) host gene, is a well-recognized cancer-promoting lncRNA and promotes autophagy, proliferation, migration, and invasion of multiple tumor cells. Of note, the promoting effect of H19 and SNHG11 on EMT have been validated in gastric cancer, hepatocellular cancer, and glioma [36–38]. Consistently, our study showed H19 and SNHG11 were oncogenes and positively correlated with upregulated EMT-related genes in GBM, indicating their potential facilitative roles in regulating the EMT of GBM. Zhang et al have reported LINC00634 is upregulated in esophageal cancer and functions as an oncogene through miR-342-3p/Bcl2L1 axis to promote cell viability and inhibit cell apoptosis [39]. Paradoxically, our study suggested LINC00634 was a tumor suppressor and negatively associated with upregulated EMT-related genes, suggesting the inhibitory roles of LINC00634 in EMT of GBM. It is possible that a gene plays opposite roles in different cancers [40]. Admittedly, regardless of whether the results are consistent or contrary to previous studies about other cancer types, further experimental verification is required. More importantly, no study has reported the roles of LINC00609, AC012615.1, and USP32P3 in cancers. Therefore, even the prognostic performance of the seven EMT-related lncRNAs signature in this study is excellent, the specific regulatory roles and underlying mechanisms of all these seven lncRNAs in EMT of GBM need to be further investigated in-depth.
GO functional annotation and KEGG pathway enrichment analyses showed high-risk scores based on our EMT-related lncRNA signature heralded the EMT process and metastatic phenotype of GBM, which explained why the GBM patients in high-risk subgroup had worse prognosis. Indeed, some studies have elaborated that EMT phenotype confers cancers to be more sensitive to immune targeting strategies whereas others have linked EMT phenotype with immunotherapy resistance [41, 42]. Therefore, it is possible that distinct immune infiltration pattern among GBM patients with different risk scores based the EMT-related lncRNA signature. Currently, immune checkpoint blockade is expected to be one of the next frontiers in cancer immunotherapy [43]. We compared the infiltration of immune cells and immune responses between low- and high-risk subgroups. As a result, degrees of immune cell infiltration and immune responses were largely increased in GBM patients of high-risk subgroup, indicating that even though GBM patients of high-risk subgroup with worse prognosis, their response to immune checkpoint blockade therapies may be batter.
There were several limitations in the present study: Firstly, our EMT-related lncRNA prognostic signature was constructed and validated based on the public dataset, which requires more prospective clinical data for clinical application in future. Secondly, the EMT-related lncRNAs in GBM were identified based on their expression correlation with EMT-related mRNA. Even though several lncRNAs have been validated to participate in regulation of EMT, their roles and molecular functions need to be further explored through in vivo and in vitro experiments. Thirdly, we identified the correlation of the EMT-related lncRNA signature and immune activities in GBM, but the underlying mechanisms remains to be investigated in-depth in future studies.