Meningiomas arise from arachnoid cap cells attached to the inner layer of the dura, which covers the spinal cord and brain. They represent about 37.6% of primary central nervous system tumors, making them the most common histological types of intracranial tumor, with an incidence of 8.83 per 100,000 (1). They primarily occur in elderly individuals, with increased incidence in individuals older than 65 years (1). The incidence of meningiomas has also increased among adolescents and young adults; these tumors now represent about 16% of all intracranial tumors in people aged 15–39 years (2). Meningiomas preferentially affect women, with a female-to-male ratio between 2:1 and 3.5:1 (3–5). According to the 2021 World Health Organization (WHO) classification of central nervous system tumors, approximately 80% of cases are WHO grade 1 meningiomas with benign histology, whereas about 20% of cases are WHO grade 2 and 3 meningiomas showing signs of increased malignancy at histology (6).

RNA methylation, including 5-methylcytosine (m⁵C), N6-methyladenosine (m⁶A), and N1-methyladenosine (m¹A), has become a common phenomenon and a critical regulating factor for transcript expression in different types of cancer (7, 8). N6-methylation (m⁶A), methylated at the N6 position of adenosine, has been regarded as the most pervasive, abundant, and conserved internal transcriptional modification within eukaryotic messenger RNAs (mRNAs), microRNAs (miRNAs), and long non-coding RNAs (lncRNAs) (9, 10). The deposition of m⁶A is encoded by a methyltransferase complex involving three homologous factors: methyltransferases (termed as “writers”), demethylases (termed as “erasers”), and recognition from m⁶A-binding proteins (termed as “readers”) (11). The m⁶A dysregulation, caused by dysregulated expression and genetic changes in m⁶A regulators, is related to the disorders of multiple critical biological processes in human cancer (12, 13). Qi et al. reported that the self-renewal and tumorigenesis of glioma stem cells (GSCs) were regulated by m⁶A RNA methylation, and an m⁶A mRNA demethylase FTO inhibitor could suppress the progression of GSC-initiated tumor (14). Yang et al. demonstrated that FTO played an important role in promoting melanoma tumorigenesis and anti-PD-1 resistance, and the combination of FTO inhibitors with anti-PD-1 blockers could reduce the resistance to immunotherapy in melanoma (15). Miao et al. revealed that m⁶A methyltransferase METTL3 promoted osteosarcoma cell progression by regulating the m⁶A level of LEF1 and activating the Wnt/β-catenin signaling pathway (16). Besides, Vengoechea J. et al. observed that IGF2BP1, one of the m⁶A regulators, could increase the malignant potential of meningiomas by enhancing cell adhesion (17). Hwang M. et al. identified significantly higher expression of HNRNPA2B1 in benign meningioma compared to normal brain tissue (18).

Numerous studies revealed that the tumor microenvironment (TME) was fundamental for tumor survival, growth, and progression. The immune part of TME contained tumor-associated macrophages, tumor-associated neutrophils, dendritic cells, myeloid-derived suppressor cells (MDSCs), and Tie2-expressing monocytes comprising tumor-associated myeloid cells (19). Patients with meningiomas exhibit signs of peripheral immunosuppression, including increased PD-L1 on myeloid cells and elevated MDSC abundance proportional to tumor grade (20). The accumulation of mast cells in meningiomas could contribute to the aggressiveness of tumors (21). Chen et al. revealed that the proportions of tumor-infiltrating immune cells were associated with the prognosis for patients with meningioma (22). Overall, tumor cells elicited multiple biological behavioral changes through direct and indirect interactions with immune cells, such as inducing proliferation and angiogenesis, inhibiting apoptosis, avoiding hypoxia, and inducing immune tolerance (23–26). Therefore, a deeper understanding of the immune infiltration of meningiomas could help parse the TME landscape and finding promising biomarkers for immunotherapy. However, whether immune infiltration in meningiomas is regulated through the modification of m⁶A patterns is still unknown. To address this question, we established a meningioma classification based on m⁶A regulator gene expression and evaluated the associations of m⁶A-deduced subtypes with immune infiltration in meningioma.

A total of 19 m⁶A regulators were identified in the present study, including 6 writers, 2 erasers, and 11 readers. The genes of these regulators were distributed widely on multiple human chromosomes ( Figure 1A ). Several m⁶A regulators were co-expressed, including (a) HNRNPA2B1 and YTHDC2, (b) LRPPRC, CBLL1, and FMR1, (c) FTO, YTHDC1, RMB15, ELAVL1, HNRNPC, and ALKBH5, and (d) YTHDF2 and WTAP ( Figure 1C ). K-means unsupervised clustering based on the expression of the 19 m⁶A regulators segregated meningioma samples into 2 clusters ( Figure 1B and Supplementary Figures 1A, B ). PCA showed that these two different m⁶A clusters could be well separated ( Supplementary Figure 1C ). Eight m⁶A regulators were differentially expressed between both clusters, including WTAP, ALKBH5, ELAVL1, FTO, YTHDC1, YTHDC2, HNRNPA2B1, and METTL3 (P < 0.05, Figure 1D ). Employing CIBERSORT to estimate immune infiltration in both clusters suggested different numbers of plasma B cells, resting mast cells, and neutrophils in the two m⁶A clusters (P < 0.05, Figure 1E ). As for some major inflammatory reaction−related genes, we found that the expression of IL-15 and IL-18 was also significantly different between the two distinct m⁶A clusters (P < 0.05, Figure 1F ).

Next, we performed weighted gene co-expression network analyses to better characterize the biological traits of both clusters. A total of 1,126 DEGs were used for the construction of weighted gene co-expression networks. A soft threshold power of β = 3 was used to construct co-expression modules ( Supplementary Figures 2A, B ). Three modules were identified and designated as the blue, gray, and turquoise modules, comprising 137, 179, and 810 genes, respectively ( Supplementary Figure 2C ). WGCNA was then applied to explore the module-trait relationships of different modules and m⁶A clusters, and the turquoise module was identified as a hub gene set for explaining the difference between both clusters ( Supplementary Figure 2D ). Gene set enrichment analysis of the 810 genes in the turquoise module identified Notch and Wnt signaling as the key differentially expressed oncogenic pathways in either cluster ( Figures 2A – D ).

We chose a combined approach by (i) employing the STRING tool for functional interaction analyses to identify key network nodes ( Supplementary Figure 3A ) and (ii) calculating the Pearson correlation coefficient of the turquoise module and m⁶A cluster characteristics to define hub genes and further characterize each m⁶A cluster ( Supplementary Figure 3B ). After intersecting these two results, 32 key hub genes were identified ( Supplementary Figure 3C ). These key hub genes comprised ACOT2, ALDH16A1, ALKBH7, BAD, C1orf122, C6orf226, CTSD, DNPH1, EPN1, EXOSC4, FAAP20, FAM207A, FZD2, H2AX, LINC00863, LTBP3, MAP1S, MEMO1, MFSD3, MZT2B, NME3, NT5C, PGLS, RPL13, RPL21P28, RPS15, SCAND1, STUB1, TIGD5, UBE2S, WDR18, and ZNF358. All of these 32 hub genes had lower expression in cluster 1 and higher expression in cluster 2 (P < 0.05, Figure 3A ). Moreover, 8 hub genes were associated with WHO grade ( Figure 3B ) and 20 were associated with age ( Figure 3C ), but no gene was related to the sex of patients with meningioma.

Besides, we also analyzed the association of key hub genes with normal meningeal tissues and meningiomas. Further, 18 out of 32 key hub genes were included in the GSE43290 dataset, and 11 key hub genes (CTSD, DNPH1, EPN1, EXOSC4, FZD2, H2AX, MAP1S, MZT2B, RPS15, STUB1, and WDR18) showed significantly different expression between normal meningeal tissues and meningiomas (P < 0.05, Figure 4A ).

Based on the expression profile of these 32 key genes, we divided the meningioma samples into two clusters by K-means unsupervised clustering ( Supplementary Figures 4A, B ). Based on the heatmap drawn by R language, the combination of these 32 key hub genes could help distinguish the meningioma dataset into two subtypes, indicating that these 32 key hub genes were critical for meningioma molecular subtypes ( Figure 4B ). Furthermore, we analyzed the expression of these 32 key hub genes and discovered that all these 32 key hub genes had significantly different expression levels between two different molecular subtypes. Also, all of them had lower expression in subtype 1 and higher expression in subtype 2, which uncovered that these key hub genes might function as the marker genes of different meningioma molecular subtypes ( Figure 4C ).

For a better understanding of the association of m⁶A clusters and molecular subtypes with the sex, age, and WHO grade of patients with meningioma, we conducted a correlation analysis and found that every WHO grade was composed of two m⁶A clusters and two molecular subtypes, and the age of patients was most likely between 30 and 70 years ( Figure 5 ).

Increasing evidence revealed that m⁶A modification played an important role in regulating tumor immunity and shaping TME through interaction with various m⁶A regulators. However, most research focused on a single m⁶A regulator, and the integrated roles of multiple m⁶A regulators were not comprehensively understood, especially for meningioma.

The present study investigated the association of meningioma with multiple m⁶A regulators, established meningioma classification based on m⁶A regulator gene expression, and evaluated its merit with immune infiltration in meningioma. Based on 19 m⁶A regulators, we constructed and segregated meningiomas into two clusters. We found that the genes of 19 m⁶A regulators distributed widely on multiple human chromosomes and different m⁶A regulators [e.g., (a) HNRNPA2B1 and YTHDC2; (b) LRPPRC, CBLL1, and FMR1; (c) FTO, YTHDC1, RMB15, ELAVL1, HNRNPC, and ALKBH5; and (d) YTHDF2 and WTAP] might have common effects. As for the difference between two distinct m⁶A clusters, we revealed that the gene expression of WTAP, ALKBH5, ELAVL1, FTO, YTHDC1, YTHDC2, HNRNPA2B1, and METTL3; the infiltration of some immune cells (plasma cells, resting mast cells, and neutrophils); and the expression of IL15 and IL18 were significantly different. A total of 1,126 DEGs were included in their different modules between the two different m⁶A clusters, and the turquoise module containing 810 DEGs was the key module. Besides, these 810 DEGs played regulatory roles in organ and tissue generation and development and multiple important signaling pathways involved in tumor development. ACOT2, ALDH16A1, ALKBH7, BAD, C1orf122, C6orf226, CTSD, DNPH1, EPN1, EXOSC4, FAAP20, FAM207A, FZD2, H2AX, LINC00863, LTBP3, MAP1S, MEMO1, MFSD3, MZT2B, NME3, NT5C, PGLS, RPL13, RPL21P28, RPS15, SCAND1, STUB1, TIGD5, UBE2S, WDR18, and ZNF358 were identified as key hub genes, and all these genes could be the marker genes to distinguish different m⁶A clusters for their different expression. Among these DEGs, the expression of FAM207A, EXOSC4, ALDH16A1, MZT2B, UBE2S, ACOT2, EPN1, and H2AX was significantly related to WHO grades, while the expression of UBE2S, FAM207A, NT5C, STUB1, MZT2B, SCAND1, EPN1, CTSD, EXOSC4, H2AX, NME3, ALDH16A1, C1orf122, BAD, TIGD5, MAP1S, FAAP20, FZD2, PGLS, and ZNF358 was significantly related to the age of patients with meningioma. Additionally, CTSD, DNPH1, EPN1, EXOSC4, FZD2, H2AX, MAP1S, MZT2B, RPS15, STUB1, and WDR18 showed significantly different expression levels between normal meningeal tissues and meningiomas. Furthermore, the combination of these 32 DEGs could be marker genes to help in segregating meningiomas into 2 subtypes based on their gene expression. The expression of EPN1, EXOSC4, H2AX, and MZT2B was related to both WHO grades and age of patients and was significantly different between normal meningeal tissues and meningiomas.

Several studies revealed the importance of the m⁶A modification pattern, regulated by m⁶A regulators, for the development and progression of a tumor. Liu et al. revealed reduced m⁶A mRNA methylation as an oncogenic mechanism in endometrial cancer and identified m⁶A methylation as a regulator of Akt signaling (32). However, Li et al. found enhanced m⁶A mRNA methylation as an oncogenic mechanism in hepatoblastoma because METTL3 was significantly upregulated and promoted hepatoblastoma development (33). Du et al. suggested two distinct m⁶A modification patterns (an immune-activated differentiation pattern and an immune-desert dedifferentiation pattern) in lower-grade glioma, which were associated with different clinical outcomes, burden of neoepitope, immune infiltration, and stemness (34). The emerging functions of m⁶A regulators in GSCs and immune infiltration have been confirmed, including roles in radio-chemotherapy resistance, tumorigenesis, promotion of the self-renewal of cancer stem cells, programmed proliferation of cancer cells, induction of apoptosis, and reduction of migration (8, 35, 36). Xu et al.’s work demonstrated the carcinogenic activity of FTO in promoting the invasion and migration of breast cancer cells via the FTO/miR-181b-3p/ARL5B signaling pathway, which highlighted the important role of FTO in tumor pathogenesis (37). Chang et al. uncovered an essential role of YTHDF3 in regulating the interaction between breast cancer cells and brain microenvironment by upregulating key brain metastatic proteins, thereby facilitating brain metastasis (38). Wang et al.’s research suggested that the upregulation of METTL14 could lead to the decrease of PERP mRNA levels via m⁶A modification, promoting the growth, invasion, and metastasis of pancreatic cancer cells (39). Our results were consistent with the former findings. The m⁶A regulators played an important role in meningiomas and segregated them into two distinct m⁶A clusters, which were correlated with different m⁶A regulator gene expression, interleukin gene expression, and immune cell infiltration. Besides, Mathoux et al. elaborated that m⁶A was enriched in the brain and emerged as a key regulator of neuronal activity and function in processes including neurodevelopment, learning and memory, synaptic plasticity, and stress response (40). Li et al. revealed that YTHDF2, an m⁶A regulator, functioned as a contributor to lung adenocarcinoma development through the upregulation of the AXIN1/Wnt/β-catenin signaling pathway (41). Another study revealed that m⁶A mRNA methylation contributed significantly to regulate the Wnt/β-catenin pathway (33). Similar to the research listed earlier, our present study also found that m⁶A modification patterns were related to several important signaling pathways in meningioma, such as Wnt signaling pathway, and development of organs and tissues in the nervous system.

With the rapid development of multi-omics and big data analysis, more research focuses on meningioma molecular subtypes. DNA methylation-based classification and grading system for meningioma had a higher power for tumor recurrence and progression prediction compared with the WHO classification (42). Another study demonstrated a highly distinct epigenetic signature of clear cell meningiomas, which was associated with frequent mutations within the SMARCE1 gene and/or loss of SMARCE1 protein expression (43). Zador et al. found that WHO grade II meningiomas could be further segregated into two distinct subgroups (a benign “grade I-like” and a malignant “grade III-like”) with different tumor recurrence rates (0 and 75%, respectively) (44). Williams et al. found that the patterns of genomic alterations in high-grade/progressive meningiomas were commonly grouped into three different categories. The NF2-associated canonical group frequently harbored CDKN2A/B alterations, which was potentially amenable to targeted therapies. An NF2-agnostic group harbored frequent TERTp and TP53 mutations. An NF2-exclusive group was partly characterized by BAP1/PBRM1 alterations (rhabdoid/papillary histology) or skull-base disease (45). Also, it was of great importance to identify some marker genes to help distinguish the molecular subtypes of meningioma. EPN1 was identified as one of the hub genes in pediatric medulloblastoma by multiple-microarray analysis (46). EXOSC4 functioned as a potential oncogene in the development and progression of colorectal cancer and was identified as a potential diagnostic molecular biomarker (47). To the best of our knowledge, EPN1, EXOSC4, H2AX, and MZT2B have not been identified as the marker genes of specific meningioma subtype.

This study had several limitations. First, the dataset we analyzed lacked survival information, and our results might have been affected by the small sample size. Therefore, improving the sample size, sequencing data, and clinical information of patients with meningioma is of great necessity in further studies. In addition, our results and conclusions are based on the bioinformatics analysis of datasets, which require further verification by basic biological experiments and clinical research.

In conclusion, this study suggested that the m⁶A regulators played an important role in meningiomas and segregated them into two distinct m⁶A clusters, which were correlated with different m⁶A regulator gene expression, interleukin gene expression, and immune cell infiltration. Also, m⁶A modification patterns were related to several important signaling pathways in meningioma and the development of organs and tissues in the nervous system. Among 32 key hub genes screened, EPN1, EXOSC4, H2AX, and MZT2B not only showed significant differences between meningioma molecular subtypes but also had the potential to be the marker genes of specific meningioma subtype.

Publicly available datasets were analyzed in this study. This data can be found here: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE136661 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE43290.

JC and YG designed the study. SS, LR, DW, and QX obtained and processed the data from GEO database. JC, SS, LH, and JD performed the data analysis. JC, HW, and MW prepared this manuscript. JD managed the research work in the subsequent revision. All authors contributed to the article and approved the submitted version.

This study was supported by grants from the National Key R&D Program of China (2018YFC1312600 and 2018YFC1312604 to YG), National Natural Science Foundation of China (82072788 and 81772674 to YG), Shanghai Sailing Program (20YF1403900 to LH).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.