In the present study, we identified GBM SR genes and uncovered that NSR genes regulate EMT processes in glioblastoma. Among the NSR genes, collagen genes, including COL1A2, COL6A2, and COL8A2, predict high-risk GBM with shorter survival time and are related to enhanced tumour heterogeneity and DC infiltration.
Tumour prognosis involves multiple mechanisms. EMT is an essential process associated with metastasis and drug resistance in cancer[28, 29]. In this study, we found that genes upregulated in short-term survivors were enriched in EMT, which had the highest normalized enrichment score (NES). EMT involves cell-cell and cell-extracellular matrix interactions. Furthermore, we screened 220 SR genes and found that 11 EMT gene set members were integrated with NSR genes, including three matrix proteins. Among SR genes, the NSR genes were mainly associated with multicellular organism development and ECM organization. The ECM is a crucial structure for tumours that can promote the growth, survival, and invasion of tumours and modify fibroblast and immune cell behaviour[31, 32]. Moreover, the matrix and stromal cells can also modulate the efficacy of therapy, and methods that reshape the tumour matrix could improve the outcomes of patients[33, 34]. Thus, the ECM might be an interesting direction for exploring hub SR genes. In this study, the PPI network was constructed and modules were identified with the MCODE plug-in of Cytoscape. The results implied that the collagen proteins COL1A2, COL6A2, COL8A1, and COL8A2 were hub SR genes, and further investigations were carried out.
The collagen superfamily is the most important group of ECM proteins and is characterized by three signature features and comprises 28 members. Studies have revealed that collagens promote the proliferation, metastasis and invasion of cancers[35, 36]. In addition, overexpression of collagens can enhance the resistance of cancers to chemotherapy drugs, leading to poor prognosis[37, 38]. In this study, we found that COL1A2, COL6A2, COL8A1, and COL8A2 were overexpressed in GBM compared to normal tissue and that they were downregulated in long-term survivors compared to short-term survivors, which confirmed that the four mentioned collagen proteins are oncogenes and SR genes. Collagens interact with cancer cells through receptors and play a crucial role. Discoidin domain receptors (DDRs) are a subfamily of tyrosine kinases that are activated by collagens. Research has shown that the activation of DDR2 by COL1 regulates SNAIL1 stability and promotes breast cancer cell invasion and migration. Others further reported that the binding of COL11A1 to DDR2 activates Src-PI3K/Akt-NF-kB signalling and inhibits cisplatin-induced apoptosis in ovarian cancer cells. In addition to DDRs, integrin is also a receptor of collagens, and the binding of integrin to COL1 might enhance the proliferation and invasion of squamous cell carcinoma cells via the MEK/ERK signalling pathway. Collectively, it seems that collagens promote the progression of tumours via multiple approaches.
Collagens bind to receptors that correlate with cancer behaviours and prognosis. Collagens combine with immune cells and comprise the main component of the TME. However, the correlation of collagen expression with immune cell infiltration in GBM is rarely unknown. Studies have shown that immune cell infiltration is highly relevant to antitumor responses and prognosis[43, 44]. In this study, we estimated the association of six immune infiltrates with patient survival. We found that tumour-infiltrating DCs (TIDCs) are significantly associated with clinical outcome in GBM, and a high level of DC infiltration predicts poor cumulative survival. In general, higher survival rates were found in patients harbouring larger populations of DCs and T lymphocytes[45, 46]. However, due to the presence of suppressive immune cells, especially Tregs, tumour cells are able to escape the immune system[47, 48]. In DC-mediated tumour immunity, the prognostic impact is related to the TIDC phenotype; mature DCs have been considered immune stimulatory, whereas immature DCs have been considered suppressive and tolerogenic. In the current study, we found that the expression of COL6A2 was negatively correlated with the abundance of activated DCs. DC activation is suppressed by tumour-derived molecules, including PD-L1 and Tim3[50, 51]. A recent study shown that VEGF can impair the migration capacity and immune function of mature DCs and contribute to immunosuppression. In our study, we found that the expression of COL1A2, COL6A2, and COL8A2 was positively correlated with the DC infiltration level. A previous study demonstrated that collagens could promote DC survival and promote the maturation of monocyte-derived DCs via osteoclast-associated receptors. In this study, COL6A2 increased the level of DC infiltration but decreased the level of activated DC infiltration. Thus, it might be the main factor that leads immune escape and poor prognosis.
The above conclusions indicate that COL1A2, COL6A2, COL8A1, and COL8A2 are oncogenes and SR genes. Recent studies have shown that depleting collagens improves the therapeutic efficacy of antitumour drugs[54–56]. In this study, we found that cisplatin, clofibrate and cytarabine can decrease the expression of COL6A2 mRNA, while valproic acid can increase the methylation of COL6A2. Cisplatin is a first-line chemotherapy drug and is used in various tumours, and valproic acid is a selective inhibitor of histone deacetylase. Barneh’s research shown that valproic acid inhibits stromal cell function and exerts anticancer effects. In GBM, valproic acid is considered a favourable SR drug that acts through multiple mechanisms[58–61].
In conclusion, our results highlight the key roles of collagens in GBM prognosis and immune cell infiltration. Furthermore, we reveal the potential mechanism through which valproic acid regulates GBM progression via COL6A2 methylation. Future studies will aim to investigate additional mechanisms involved in collagen methylation and tumorigenesis.