Identification of differentially expressed SAPs in LGG Patients
The study design is illustrated in Figure 2. The R software was adopted to identify the differentially expressed SAPs from a sum of 201 SAPs. A total of 112 SAPs was screened out (P<0.05, | log2FC) | >0.5), which carried 44 downregulated and 68 upregulated SAPs (Fig.3, Table 2).
Functional enrichment analysis of the differently expressed SAPs
To scrutinize the bio-function of identified SAPs, the SAP genes were grouped in accordance with their expression level. Subsequently, the online tool DAVID and WebGestalt was used to conduct a functional enrichment analysis of these groups. The upregulated differentially expressed SAPs were enriched in biological processes (BP) related to the regulation of trans-synaptic signal, and synapse organization and post synapse organization (Fig.4A and 4E). While the downregulated differentially expressed SAPs were significantly enriched in modulation of chemical synaptic transmission, synapse organization, and dendritic spine organization (Fig.4B and 4F). The molecular function (MF) analysis revealed that the upregulated SAPs were markedly enriched in ionotropic glutamate receptor activity, neurexin family protein binding, and glutamate receptor activity (Fig.4A and 4E), whereas the downregulated SAPs were significantly enriched in ion gated channel activity, gated channel activity, and substrate-specific channel activity (Fig.4B and 4F). In terms of the cellular component (CC), the downregulated SAPs were primarily enriched in the presynapse, cell junction, and ion channel complex, and upregulated SAPs were significantly enriched in the postsynaptic membrane and glutamate synapse (Fig.4B and 4F). Besides, we found that upregulated SAPs were enriched in cell adhesion molecules pathway, glutamatergic synapse, and neuroactive ligand-receptor interaction pathway (Fig.4C and 4G). Furthermore, downregulated differentially expressed SAPs were mainly enriched in the calcium signaling and Nicotine addiction (Fig.4D and 4H).
Construction of PPI network and key modules
Based on the STRING database and Cytoscape software, the PPI network was constructed, which included 94 nodes and 563 edges. Subsequently, the PPI network was analyzed to screen for potential key modules using MODE in Cystoscope. Module 1 contained 23 nodes and 102 edges, module 2 included 12 nodes and 48 edges, module 3 consisted of 5 nodes and 10 edges. Then, the KEGG pathway and GO analyses demonstrated that the SAPs in module 1 were enriched in the glutamatergic synapse, glutamate receptor signaling pathway, synaptic membrane, neurexin family protein binding. The SAPs genes in module 2 were enriched in neuroactive ligand-receptor interaction, regulation of ion transmembrane transporter activity, synaptic membrane, and glutamate receptor activity. The SAPs in module 3 were enriched in GABAergic synapse, chloride transmembrane transport, chloride channel complex, and chloride channel activity (Fig. 5).
Selection of a prognostic-related SAPs
To analyze the prognostic significance of the key SAPs in key modules, univariate Cox regression analysis was used and obtained 20 prognostic-associated SAPs (Fig.6A). These candidate SAPs were analyzed by lasso regression analysis and multiple stepwise Cox regression analysis and four hub SAPs were identified as the independent predictors in LGG patients (Fig.6B-6D).
Copy-number alteration and mutation analysis of SAPs in LGG patients
Copy-number alteration (CNA) analyses and mutation of SAPs were conducted using the GISTIC algorithm and segmentation analysis in cBioPortal6. The results showed that hub SAP genes were altered in 82 samples out of 2200 LGG patients (4%) (Fig. 7E).
Validation of Expression and Prognostic Value of hub SAPs in LGGs
Compared with that in normal brain tissue, the expression of GRIK2, GRID2 and ARC were higher and GABRD was lower in LGGs (Fig. 7A and 7B). The result of quantitative real-time PCR confirmed the outcome mentioned above (Fig. 7C). From the Human Protein Atlas database, immunohistochemistry results showed that GIRD2 and GRIK2 were upregulated in gliomas than normal brain tissue. On the contrary, GABRD were downregulated in glioma tissue (Fig. 7D). Furthermore, we used the Kaplan Meier-plotter method to probe the relationship between OS and hub SAPs. Results indicated that the expression of hub SAPs was associated with the prognosis of LGG patients (Fig. 8A-8D). The CGGA cohort was used to validate the prognostic value of hub SAPs (Supplementary Fig. 1). The expression of hub SAPs seems to correlate with LGG grades and patients’ age (Fig.8E-8L). Besides, GABRD was upregulated in seizure LGG patients (Fig.8N).
Construction and analysis of a prognostic score model
The prognostic score model was constructed following the hub SAPs. The risk score of each patient was calculated according to the following formula:
Risk score= (-0.19806*ExpGRID2) + (0.248830*ExpARC) + (-0.267494*ExpGRIK2) + ( -0.503747*ExpGABRD)
Based on the median risk score, 524 LGG patients in the TCGA cohort were divided into high-risk and low-risk subgroups. Compared with patients in the low-risk subgroup, those in the high-risk subgroup exhibited a poorer OS (Fig. 9A). Then, a time-dependent ROC analysis was performed, which suggested that it has a good diagnostic performance (Fig. 9B). Besides, scatter plots were created to display the survival status of patients and the risk score of the signature in the high- and low-risk subgroups (Fig. 9D). Additionally, the prognostic value of the four-SAPs signature predictive model was assessed, and a similar formula was used to the CGGA cohort. Results revealed that LGG patients with high-risk scores have a significantly lower OS than those with low-risk scores in the CGGA cohort (Fig. 10A-10D).
Construction of nomogram based on the hub SAPs
Multivariate Cox regression analyses were used to estimate the prognostic significance of different clinical characteristics of glioma patients. Results indicated that the risk score was an independent prognostic factor associated with OS in glioma patients (P<0.05, Fig. 11A). Due to the clinical relevance and prognostic value of other clinicopathologic factors, we constructed a nomogram and applied it in evaluating survival rates for glioma patients at 1, 3, and 5 years, which could aid clinicians in setting clinical plans for glioma patients (Fig. 11B). The calibration plots presented good conformity between the predicted and observed outcomes in both the TCGA and CGGA cohorts (Fig. 11C and 11D).