The low response rate to immunosuppressant is mainly due to the lack of adequate knowledge about the tumor microenvironment (TME) and screening biomarkers for gliomas. We aimed to identify the promising immune biomarkers and new immune classification of glioma. In this study, multiple‐immune algorithms were used to calculate immune‐infiltration scores. Unsupervised and supervised machine learning methods were used to perform the classification. We observed that OLFML3 overexpression was indicated in gliomas and linked to poor prognosis. OLFML3 knockdown inhibited proliferation, invasion and increased the sensitivity of glioma cells to temozolomide. OLFML3 expression could also reflect the aberrant immune status. Based on the immune‐related signature, patients were divided into three immune subtypes via consensus clustering. Patients with C2 subtype presented poorer prognosis and shorter progression free survival than patients with other two subtypes. The TME patterns among subtypes were different. C2 and C3 subtypes are the immune‐inflamed and immune‐desert tumors, respectively. Additionally, compared with C3 subtype, patients with C1/C2 subtypes were more likely to respond to immunotherapy. The pRRophetic algorithm indicated patients with C1/C2 subtypes were more resistant to temozolomide, but sensitive to paclitaxel and cisplatin. To conclude, OLFML3 overexpression affects glioma cell proliferation, invasion, and TMZ sensitivity and has been proved to be an independent prognostic‐ and immune‐related biomarker. Additionally, the novel immune subtype's classification could provide the prognostic and predictive predictors for glioma patients and may guide physicians in selecting potential responders.

BackgroundIntraventricular penetration is rare in glioblastoma (GBM). Whether the ependymal region including the ependyma and subventricular zone (SVZ) can prevent GBM invasion remains unclear.MethodsMagnetic resonance imaging (MRI) and haematoxylin–eosin (HE) staining were performed to evaluate the size and anatomical locations of GBM. Binary logistic regression analysis was used to assess the correlation between tumor-ependyma contact, ventricle penetration and clinical characteristics. Cell migration and invasion were assessed via Transwell assays and an orthotopic transplantation model.ResultsAmong 357 patients with GBM, the majority (66%) showed ependymal region contact, and 34 patients (10%) showed ventricle penetration of GBM. GBM cells were spread along the ependyma in the orthotopic transplantation model. The longest tumor diameter was an independent risk factor for GBM-ependymal region contact, as demonstrated by univariate (OR = 1.706, p < 0.0001) and multivariate logistic regression analyses (OR = 1.767, p < 0.0001), but was not associated with ventricle penetration. Cerebrospinal fluid (CSF) could significantly induce tumor cell migration (p < 0.0001), and GBM could grow in CSF. Compared with those from the cortex, cells from the ependymal region attenuated the invasion of C6 whether cocultured with C6 or mixed with Matrigel (p = 0.0054 and p = 0.0488). Immunofluorescence analysis shows a thin gap with GFAP expression delimiting the tumor and ependymal region.ConclusionThe ependymal region might restrict GBM cells from entering the ventricle via a non-mechanical force. Further studies in this area may reveal mechanisms that occur in GBM patients and may enable the design of new therapeutic strategies.