A 37‐year‐old man was admitted to our hospital with sudden dizziness and headache for 5 days (Figure 1). The first magnetic resonance imaging (MRI) showed that there was a lump of abnormal signal shadow with a size of about 4.5*3.0*3.0 cm in the left temporal lobe. Severe edema regions around the lesion were present. To enable a histological diagnosis and treatment decision, tumor resection was performed. The operation was performed from the posterior part of the inferior temporal gyrus, and the old hematoma and grayish red tumor tissue were seen. Extended resection of the tumor was performed along the lateral side of the relative boundary, deep to the temporal horn of the lateral ventricle. After the surgery, the pathological diagnosis was E-GBM. The surgical specimen showed that cancer cells were distributed diffusely, and the cells were arranged closely, which were polygonal, round or triangular, with dense nuclei, visible local nucleoli, partial nuclear deviation, and rich red cytoplasm. Besides, there were obvious cell atypia, visible mitotic cells, obvious proliferation of vascular endothelial cells, vasodilatation and congestion, and flaky necrosis and hemorrhage. Notably, tumor tissue was found in the local subarachnoid space. The immunohistochemical results were listed as the following: Ki-67 (label index: 15%), GFAP (positive), MGMT (unmethylated), EMA (negative), EGFR (negative), VEGF (positive), ATRX (wild-type), Olig-2 (positive), IDH1 (wild-type), BRAF V600E (mutant), CIC (positive), FUBP1 (positive), H3K27M (negative), Desmin (negative), and CK (negative). Fluorescence in situ hybridization (FISH) detection showed that chromosome (chr) 1p was deleted and 19q was intact. Eight days after operation, the second MRI showed that the left temporal lobe tumor changed after resection, with effusion and a little hematocele in the operation area, and the ependyma of the right lateral ventricle body was slightly thickened.
On the eleventh day after operation, temozolomide (TMZ, Merck Sharp & Dohme Ltd) 75mg/m2 was used as neoadjuvant chemotherapy for 17 days. Compared with the second MRI, the third MRI (28 days after operation) showed a new slice-like enhancement in the operation area and tumor recurrence was not ruled out. The fourth MRI of cervical, thoracic and lumbar showed GBM cerebrospinal fluid spread and metastasis, and the dura mater in spinal canal and spinal cord surface was widely thickened and strengthened (35 days after operation). Cerebrospinal fluid liquid-based cytology showed that individual malignant tumor cells could be seen (35 days after operation). The third and fourth MRI showed that the patient's condition was still progressing after TMZ treatment. Therefore, the patient received 36Gy total central radiotherapy (37 days after operation) for eighteen days, and then locally pushed for 24Gy for seven days. During concurrent radiotherapy and chemotherapy, the patient developed third-degree bone marrow suppression. Besides, there are persistent low back pain, a sharp deterioration of mental and dietary conditions, and bed rest and inability to go to the fields. Therefore, the patient stopped radiotherapy and chemotherapy. Since all other conventional treatment options had been exhausted, and to find a target for an experimental salvage therapy, we adopt genomic technology and whole exome sequencing (WES) analysis to study molecular characteristics of this case. The WES sequencing analysis results revealed a V600E mutation of the BRAF kinase. Consecutively, vemurafenib (Roche, BRAF inhibitor) therapy was initiated (960 mg twice daily).
Following vemurafenib treatment, the patient was in stable condition and moves freely. However, seven gastrointestinal reactions (malignant, vomiting, etc.) occurred two weeks after the start of dose treatment. Therefore, we changed the dose of vemurafenib to 480mg, twice daily. So far, patients have been well-tolerated and the clinical follow-up was stable. Compared with the fourth MRI, the fifth MRI found that the extensive thickening and enhancement of the endocranium on the spinal canal and spinal cord surface were significantly reduced. WES sequencing analysis and RNA-seq results of formalin-fixed and paraffin-embedded (FFPE) tissue revealed nine mutant genes in both DNA-seq data and RNA-seq data (Figure 2A). The protein-protein interaction analysis (PPI) of these nine mutant genes suggested that NFATC3, mTOR, and BRAF interact with each other (Figure 2B). The GBM cohort was divided into a high expression group and a low expression group based on each mutated gene’s median expression level, and then univariate and multivariate Cox regression analysis was performed to determine the prognostic value of the nine mutated genes. The results showed that the four mutant genes were regarded as independent prognostic factors in CGGA data. Independent prognostic analysis showed that hazard ration (HR) was 1.2 (95% confidence interval, CI: 1.05 ~ 1.3) for CNTNAP3, 1.5 (95% CI: 1.30 ~ 1.546) for mTOR, 1.3 (95% CI: 1.19 ~ 1.5) for NFATC3 and 1.3 (95% CI: 1.19 ~ 1.5) for NOM1 Meanwhile, both age and gender are prognostic risk factors (Figure 3A). Thereafter, the Kaplan–Meier method was used to determine the effect of the CNTNAP3, mTOR, NFATC3, and NFATC3 expression on patient’s survival. The data showed that patients with high expression of the four mutant genes had significantly shorter overall survival (OS) time than others (P < 0.05, log-rank test) (Figure 3B). Previous reviews have addressed the evidence behind IDH1 as a major prognostic feature for gliomas [29-33]. The gene expression of CNTNAP3, mTOR, NFATC3, and NOM1 in IDH1-wild type (WT) was higher than that in the IDH1-mutant (MUT) group (Figure 3C). The result of CNV analysis revealed deletion of copy number on chromosome 1 and 6, and amplification of copy number of chromosome 19q and chromosome 21 in this case of E-GBM, as shown in Supplementary Figure 1.
According to the median of each gene expression level, it was divided into high and low expression groups, then taking |log2 FC| > 0.5 and p < 0.05 as the screening criteria, analyzing 966 samples of mRNACGGA data. CNTNAP3, mTOR, NFATC3 and NOM1 respectively screened out 2489 related differential genes (down 986, up 1503), 4650 related differential genes (down 93, up 4557), 5930 related differential genes (down 3853, up 2077), and 4143 related differential genes (down 478, up 3665) (Figure 4A-D). The results of the heat map showed that the most relevant twenty up-regulated genes and twenty down-regulated genes which might be most relevant to each mutant genes (Figure 4A-D).
The case was associated with phenotypes of cell cycle and focal adhesion by GO analysis and KEGG enrichment pathways analysis. GO analysis showed that the related differential genes of CNTNAP3, mTOR, NFATC3, and NOM1 were enriched in biological processes related to cell cycle. For example, the GO terms of cell cycle include regulation of cell cycle G2/M phase transition (GO:1902749, NEDD1, DYNC1H1, NDE1, TAOK1, ATM, FOXN3, DCTN1, PSMD4, CLSPN, PSMB10, PSMC5, etc.), regulation of cell cycle G1/S phase transition (GO:1902806, ADAM17, CNOT6L, RFWD3, GIGYF2, TAF1, STXBP4, KMT2E, EP300, SENP2, etc.), G2/M transition of mitotic cell cycle (GO:0032496, BACH1, TAF2, NEDD1, DYNC1H1, NDE1, FBXL15, TAOK1, ATM, PPM1D, etc.), and regulation of G1/S transition of mitotic cell cycle (GO:2000045, ADAM17, CNOT6L, RFWD3, GIGYF2, KMT2E, EP300, SENP2, ATM, PKD2). Important genes involved in cell cycle in these pathways include ATM, ADAM17, and PKD2. These genes are involved in the cell cycle of GBM [34-36]. GO analysis confirmed that the four mutant genes affected E-GBM through regulation of cell cycle and cell adhesion (Figure 5A-D). Moreover, related differential genes of CNTNAP3, mTOR, NFATC3, and NOM1KEGG were mainly enriched in KEGG enrichment pathways including Glioma, focal adhesion, MAPK signaling pathway, and P13K-Akt signaling pathway (Figure 5A-D). The common KEGG enrichment pathways of the related differential genes of CNTNAP3, mTOR, NFATC3, and NOM1 were the pathways of focal adhesion (hsa04510, HRAS, BRAF, BAD, SHC2, MYL5, PIP5K1A, VCL, MAPK3, GSK3B, MAPK1, etc.), FoxO signaling pathway (hsa04068, SMAD4, HRAS, PRKAA1, STK4, EP300, BRAF, ATM, MAPK3, MAPK11, MAPK1, MAP2K2, STAT3, etc.), and EGFR tyrosine kinase inhibitor resistance (hsa01521, HRAS, BRAF, BAD, SHC2, RPS6KB1, MAPK3, GSK3B, MAPK1, MAP2K2, STAT3, AKT3, etc.). From GO analysis and KEGG pathway enrichment analysis, it is not difficult to find that the BRAF gene may play an important role in the molecular mechanism of poor prognosis in patients with E-GBM.
Construction of prognostic risk model
In order to verify the cooperative relationship between CNTNAP3, mTOR, NFATC3, and NOM1, we established a multivariate prognostic risk model. The ROC curve was used to evaluate the efficacy to predict CNTNAP3, mTOR, NFATC3, and NOM1, and their interaction in glioma patients. The areas under curve (AUC) for CNTNAP3, NOM1, NFATC3, mTOR and their interaction were 0.572, 0.590, 0.623, 0.659 and 0.687, respectively (Figure 6A).
Constrcution of a nomogram intergating the prognostic risk model and clinicopathological factors
The prognostic risk model (CNTNAP3, mTOR, NFATC3, and NOM1) were integrated to establish a prognostic risk model. The risk scores were calculated using the formula mentioned in the glioma methods, as follows: risk score = (1.2 * expression level of CNTNAP3) + ( 1.5 * expression level of mTOR) + ( 1.3 * expression level of NFATC3) + ( 1.2 * expression level of NOM1), in GBM methods, the risk score = (1.30 * expression level of CDCP1) + ( 1.12 * expression level of CD44). Glioma patients were respectively divided into low-risk (n= 485) and high-risk (n= 485) groups according to the median risk score. The survival curve showed poorer prognosis in the high-risk group than low-risk group (Figure 6B, P < 0.001).
To confirm the prognostic value of risk signature, we constructed a nomogram based on risk signature, and the clinical relevance and prognostic value of age and glioma type (primary glioma, secondary glioma, and recurrent glioma), gender, radiotherapy, TMZ chenmotherapy, and IDH status. The incidence of 1-year, 3-year, and 5-year survival rates can be estimated from the total scores, which are the sum of the scores for each item, as shown in Nomotu (Figure 6C).