Construction of Novel Methylation-Driven Gene Model and Investigation of PARVB Function in Glioblastoma

Glioblastoma multiforme (GBM) is characterized by widespread genetic and transcriptional heterogeneity. Aberrant DNA methylation plays a vital role in GBM progression by regulating gene expression. However, little is known about the role of methylation and its association with prognosis in GBM. Our aim was to explore DNA methylation-driven genes (DMDGs) and provide evidence for survival prediction and individualized treatment of GBM patients.

Use of the MethylMix R package identified DMDGs in GBM. The prognostic signature of DMDGs based on the risk score was constructed by multivariate Cox regression analysis. Receiver operating characteristics (ROC) curve and C-index were applied to assess the predictive performance of the DMDG prognostic signature. The predictive ability of the multigene signature model was validated in TCGA and CGGA cohorts. Finally, the role of DMDG β-Parvin (PARVB) was explored in vitro.

The prognostic signature of DMDGs was constructed based on six genes (MDK, NMNAT3, PDPN, PARVB, SERPINB1, and UPP1). The low-risk cohort had significantly better survival than the high-risk cohort (p < 0.001). The area under the curve of the ROC of the six-gene signature was 0.832, 0.927, and 0.980 within 1, 2, and 3 years, respectively. The C-index of 0.704 indicated superior specificity and sensitivity. The six-gene model has been demonstrated to be an independent prognostic factor for GBM. In addition, joint survival analysis indicated that the MDK, NMNAT3, PARVB, SERPINB1, and UPP1 genes were significantly associated with prognosis and therapeutic targets for GBM. Importantly, our DMDG prognostic model was more suitable and accurate for low-grade gliomas. Finally, we verified that PARVB induced epithelial-mesenchymal transition partially through the JAK2/STAT3 pathway, which in turn promoted GBM cell proliferation, migration, and invasion.

This study demonstrated the potential value of the prognostic signature of DMDGs and provided important bioinformatic and potential therapeutic target data to facilitate individualized treatment for GBM, and to elucidate the specific mechanism by which PARVB promotes GBM progression.