A total of 286 meningioma samples were included in this retrospective single-center study. All the tissue samples were resected at Sun Yat-sen University Cancer Center (SYSUCC, Guangzhou, China) from June 2008 to June 2018. Follow-up was last done in June 2022. The clinical data sets included gender, age at diagnosis, histopathological diagnosis (Goldbrunner et al., 2016), tumor location, time of death, time to radiographic tumor recurrence, and radiotherapy treatment between surgery and tumor recurrence. Hematoxylin and eosin (H&E) slides were reviewed from each case to confirm the diagnosis of meningioma by two pathologists (an experienced senior neuropathologist and a resident pathologist). The outcomes of each sample depend on tumor recurrence and time to recurrence. Recurrence was defined as tumor growth following the time of operation, and the time to recurrence was determined by calculating the duration from the date of surgery to the first post-operative imaging documenting tumor recurrence. The extent of resection was extracted from the surgeon’s operative report and checked with post-operative magnetic resonance imaging (MRI). All samples were ethically approved for use based on informed consent and the Ethics Committee of SYSUCC.

Archived formalin-fixed and paraffin-embedded tissue samples were used for the extraction of two biopsy punches of 2 mm diameter each to construct tissue microarrays with a conventional tissue microarrayer (Beecher Instruments, Sun Prairie, WI, USA). Tumor cylinder extraction was done after evaluation of H&E slides for suitable areas. The newly formed tissue blocks were cut into 4 μm slices with a microtome. After subsequent drying at 80°C for 15 min, subsequent immunohistochemical staining for H3K27me3 (1:200, rabbit monoclonal antibody C36B11, Cell Signaling, Danvers, MA, USA) and EZH2 (1:100, rabbit monoclonal antibody D2C9, Cell Signaling, Danvers, MA, USA) were done with a Ventana BenchMark immunostainer (Ventana Medical Systems, Tucson, AZ, USA). Analysis of immunohistochemical staining were scored independently by two pathologists.

H3 lysine 27 trimethylation staining was assessed for nuclear expression on tumor cells. At least 1,000 tumor cells were counted under ×400 magnifications starting with the hot spot. The vascular endothelial cells and inflammatory cells were carefully excluded while counting morphologically. We calculate the cutoff value by the receiver operating characteristic (ROC) curve, and the labeling index of EZH2 was further sub-grouped as 0 for no positive cells, 1 for ≤25% of positive cells, and 2 for >25% of positive cells. In the case of heterogeneity of coloring intensity, the intensity of the major proportion of cells was considered. For EZH2, score 1 and 2 were considered low expressions and high expressions, and both score 1/2 were considered positive expressions. The expression of H3K27me3 was scored according to the positive proportion of the tumor cells. The percentage of positively stained tumor cells was scored as follows: 0 (<20% positive tumor cells) (Schaefer et al., 2016), 1 (≥20 and <50% positive tumor cells), and 2 (>50% positive tumor cells). The detailed protocol and assessment standard of other antibody markers including programmed cell death protein 1 (PD-1), programmed cell death 1 ligand 1 (PD-L1), CD4, CD8, CD20, FOXP3, CD68, and CD163 were described in a previous study (Du et al., 2015). We assessed and counted all immunohistochemical profiles of immune cell expression and divided them into two groups of high and low expression using the mean value (Table 1) as the cut-off value. The only difference was PDL1, which we divided it into expression and non-expression (<1%) groups by TPS (tumor cell proportion score), where the expression/positive group was divided into low expression (≥1–49%) and high expression group (≥50%) (Han et al., 2016).

The meningioma RNA-seq data were downloaded from https://www.ncbi.nlm.nih.gov/geo/. We analyzed 181 gene expression omnibus (GEO) RNA-seq cohorts of meningioma, ranging from WHO grade 1–3. The meningioma and meningeal RNA-seq data were from GEO datasets, including GSE43290, GSE74385, and GSE16581. Limma package in R software was conducted to screen out the differentially expressed genes (DEGs) with the cut-off criterion of adjusted P < 0.05 and | log2FC| > 1. The dataset GSE16581 with survival information we used to perform survival scores, compare the survival differences between high and low expression groups, and Kaplan–Meier (KM) plot survival curves by the Survival package and Survminer package in R. We divided all meningioma data into high and low expression groups based on the cutoff value of the median EZH2 mRNA expression, and compared and analyzed the differences in immune cell expression between high and low expression groups and plotted the graphs. To explore the functions and pathways of related genes, we performed gene ontology (GO) and Kyoto encyclopedia of genes and genomes (KEGG) analyses on ClueGo and Metascape websites.

GraphPad Prism 8 and SPSS 25 software were performed for statistical analyses. The measurement data was represented as mean ± SD. The Chi-square test was conducted to explore the correlations between the expression of EZH2, H3K27me3, and clinicopathological or immune features. Categorical factors between EZH2 groups were compared using Fisher’s exact tests. Survival distributions of groups were compared using KM estimates and log-rank tests. The Cox proportional hazards regression model was used for univariate and multivariate analyses to evaluate the independence of EZH2, H3K27me3, and other markers in predicting prognosis. A P-value < 0.05 was considered to be statistically significant.

We collected clinical information from 286 meningioma patients and 10 were lost of follow-up. Finally, 276 meningioma patients were included in our retrospective study. Out of the 276 cases, 129 (47%) cases were WHO grade 1, 108 (39%) cases were WHO grade 2, and the remaining 39 (14%) cases were WHO grade 3. The mean age of the patients was 49.58 ± 13.08, with a median age of 50.50, and the majority was the elderly, and only three patients being underage. A total of 179 cases (65%) of patients were women. We also counted Ki67 expression, and based on previous reports in the literature that high Ki67 expression correlates with poor prognosis, we divided the Ki67 expression into low and high expression groups (<5 vs. ≥5%) (Olar et al., 2015) based on immunohistochemical results. In addition, we divided into two groups (Skull base vs. not skull base) according to the location of tumorigenesis. The details were showed in Table 1. By evaluating the expression of EZH2 (Figures 1A–C) and H3K27me3 (Figures 1D,F) by immunohistochemistry, we found some correlations between expression and clinical information. Most H3K27me3 positive cases showed a low level of EZH2 expression, and strong expression of EZH2 was detected in only 11 cases, 4 of which were WHO grade 2 and 6 of which were WHO grade 3 (Table 1 and Figure 2A). EZH2 expression was correlated with increased Ki67 index (p < 0.001), for immunohistochemical positive expression was more prominent in areas of high proliferative activity. To further explore the mRNA expression of EZH2 in meningiomas, we analyzed three datasets of meningioma cases from the GEO database and found that mRNA expression of EZH2 was higher in WHO grade 2–3 meningiomas than in WHO grade 1. This is consistent with our results of the immunohistochemical exploration of EZH2 expression (Figure 2B).

The samples included in this cohort showed the expected distribution patterns of recurrence-free survival (RFS) when stratified by WHO grade, with WHO grade 2–3 tumors showed significantly shorter times to recurrence when compared to WHO grade 1 tumors (Figure 2A, log-rank test P < 0.001). Mean RFS was 108 months (95% CI 102.5−114.2), 88 months (95% CI 77.6−99.1), and 42 months (95% CI 28.2−56.6) for WHO grade 1–3 tumors (Figure 2F), respectively. We analyzed the survival analysis of two groups of meningioma patients with EZH2 positive and negative, and we found that there was a statistically significant difference in the RFS, demonstrating that expression of EZH2 was associated with poorer RFS in meningioma [Mean RFS 48.9 (95% CI 35.9−62.0) months vs. 102.1 (95% CI 94.1−110.1) months, P < 0.001, Figure 2D]. Similarly, we found that there was a worse prognosis in the high EZH2 expression groups than low EZH2 expression groups which divided into two groups by cutoff value (>25 vs. ≤25%) [Mean RFS 26.6 (95% CI 12.9−40.4) months vs. 61.3 (95% CI 46.1−76.4) months, P = 0.022, Figure 2E]. In addition, although meningioma patients have a lower mortality rate, we still did overall survival (OS) survival analysis and the result also showed EZH2 expression predicted a worse prognosis (Figure 2C). However, we cannot yet consider it as an independent prognostic factor (Table 2). The dataset GSE16581 in the GEO database contains survival information, and the OS prognostic analysis also showed EZH2 high expression group had a shorter survival time (Figure 3A).

We analyzed the survival analysis of two groups of meningioma patients with H3K27me3 loss and retained, we found that there was a statistically significant difference in the RFS, demonstrating that loss of trimethylation was associated with poorer RFS in meningioma (Figure 2G). In addition, we also analyzed the EZH2 expression positive and negative groups and we found a statistically significant difference in RFS, demonstrating that EZH2 positive expression is associated with poorer RFS in meningiomas (Figure 2D). However, there was no connection between EZH2 expression and the deletion of H3K27me3. Only two cases expressed EZH2 with concomitant deletion of H3K27me3, and both were EZH2 low expression.

Infiltrating immune cells are important components of the tumor microenvironment and are frequently associated with tumor behavior and patient outcomes. Since several studies and GO analysis revealed that EZH2 was related to the immune response, we further explored the infiltration of immune cells in meningioma (Figure 3D). In addition, we evaluated several immune cell markers, including CD4, CD8, CD20, CD68, CD163, PD-1, PD-L1, and FOXP3 in meningioma samples (Figures 4A–H), and found that the lymphocyte infiltrates in the meningioma tissue was comprised predominantly of T cells with infrequent B cells (Table 3).

Based on the scores for microarray immunohistochemistry, with a mean cutoff value, we divided the expression of each maker into two groups, high and low expression. We found that the density of immunohistochemical expression of infiltrating lymphocytes was significantly different in WHO grade 2–3 meningiomas compared to WHO grade 1 meningiomas. CD4 and CD8 lymphocytes were all more highly expressed in WHO grade 1 meningiomas, compared to WHO grade 2–3 (Figures 5A–H). This is also consistent with previous studies of reduced T-cell infiltration in WHO grade 2–3 meningiomas. CD20, PD1, and FOXP3 expressions were not found to be different between the different grades of meningioma (Table 3). In addition, we counted the differences in survival analysis between high and low expression of each immune cells (Figures 2J–L) and found that the prognosis of the CD8 low expression group was significantly worse than that of the high expression group (p = 0.003; Figure 2I). We further counted the expression pattern of PD-L1+/CD68− and divided into positive and negative groups in this way. Although the overall positive rate of PD-L1 was not high, the PD-L1 positive group was mostly distributed in WHO grade 2–3 meningiomas, especially grade 3 cases, and was statistically significantly different (p < 0.001). Survival analyses of the positive and negative groups were counted according to group expression and were found to be statistically significantly different (p < 0.001). Positive expression of PD-L1 represented a poorer prognosis (Figure 2H).

To study the correlation between EZH2, H3K27me3, and different markers of immune infiltrates in meningioma, we counted the expression of immune cells in the H3K27me3-loss samples and found that the expression of CD8 was lower in the H3K27me3-loss samples compared to H3K27me3-preserved samples. However, there was no significant difference in the expression of CD4, CD20, CD68, CD163, and PD-L1 (Table 4). Interestingly, in the comparison of EZH2 positive and negative, we were surprised to find a significant correlation between EZH2 positive samples and PD-L1 positive samples (<0.001; Figure 5P). In addition, CD68 expression was higher in EZH2-positive specimens compared to EZH2-negative specimens (p = 0.010; Figure 5L) and no correlation has been found in other immune cell groups (Figures 5I–K, 5M–O and Table 4).

To further verify the results of our study, we collected a total of 181 samples from the GEO dataset. We identified overlapping DEGs between EZH2 high expression group and low expression group. The top hub genes were screened via the plug-in molecular complex detection and cytoHubba of Cytoscape. GO analysis was performed to show the overlapping DEGs were involved in several biological processes, including kinase activator activity, and CXC chemokine receptors (CXCR) chemokine receptor binding, etc. (Figure 3B). The KEGG pathways are enriched in several classic signaling pathways, such as cell cycle pathway (Figure 3C). We analyzed the difference in the proportion of immune cells according to the gene EZH2 into high and low expression groups and determined the correlation between the expression of the target gene and the content of immune cells. We found that T cells CD4 naive and plasma cells had a higher proportion in the EZH2 high expression group, while T cells CD8 and macrophages M2 had a lower proportion (Figure 3E).