Novel insights into hypoxia-driven transcriptomic and epigenetic landscapes in grade 3 meningioma

Abstract

Background

Meningiomas are among the most prevalent central nervous system (CNS) tumors, with up to 20% of cases exhibiting recurrence or aggressive behavior. Hypoxia is a key driver of malignant transformation and therapeutic resistance, yet its molecular basis in meningioma remains poorly understood.

Methods

We conducted integrative transcriptomic and epigenomic profiling of IOMM-Lee cells (grade 3 meningioma) cultured under hypoxic (0.2% O₂) and normoxic conditions. RNA-sequencing and Illumina MethylationEPIC v2.0 data were analyzed in R using DESeq2 and minfi, respectively. Functional enrichment, transcription-factor binding analysis, and pathway mapping (clusterProfiler, enrichR) were performed. Findings were cross-validated in public meningioma datasets, in Indian meningioma patient cohort and cell line via RT-qPCR, and azacytidine-based demethylation assay. Functional role of the candidate gene was elucidated in vitro via cellular assays.

Results

Hypoxia triggered a canonical HIF1A-driven transcriptional program activating glycolytic and angiogenic pathways while downregulating genes associated with DNA repair and replication in meningioma. Several differentially expressed genes (DEGs) were identified as known oncogenes, tumor-suppressors, or associated with immune regulation and stemness. Promoter motif analysis identified HIF1, SP1, TP53, BRCA1, and E2F1 as enriched transcriptional regulators. We validated hypoxia and HIF1-mediated regulation of some of the top DEGs. DNA-methylation analysis revealed epigenetic silencing of RTN4IP1 and ZBTB7C under hypoxia, reversible upon azacytidine treatment. Integrative comparison with patient datasets highlighted SLITRK2, PDE4C, SGCD, and LRP1B as hypoxia-responsive genes associated with poor prognosis. Several hypoxia-regulated genes also showed significant correlation with known hypoxia biomarkers, VEGFA and CA9. IGFBP3 and NDRG1 were among the top hypoxia-associated upregulated genes, and IGFBP3 expression was linked to advanced meningioma grades. Knockdown of IGFBP3 via siRNA in hypoxia-treated IOMM-Lee cells was associated with reduced cell proliferation and migration.

Conclusions

This study presents the first integrated transcriptomic–epigenomic landscape of hypoxia in grade 3 meningioma, uncovering regulatory networks and candidate biomarkers with prognostic and therapeutic potential. These findings provide a foundation for future translational studies targeting hypoxia-driven tumor progression in meningioma.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07606-9.

Introduction

Meningiomas develop from the meninges surrounding the central nervous system (CNS) and are among the most prevalent primary brain tumors. These represent half of all primary benign tumors and one-third of all intracranial tumors [1]. According to the World Health Organization (WHO), approximately 80.1% of meningiomas are classified as benign (WHO grade 1), 18.3% as atypical (WHO grade 2), and 1.5% as malignant (WHO grade 3) [1, 2]. While predominantly benign, these tumors can manifest in various biological grades, associated with significant morbidity and mortality, especially when they are recurrent or located in critical intracranial regions [3]. For most patients diagnosed with benign meningiomas, a combination of surgical resection and stereotactic radiotherapy effectively controls the disease [4]. However, there is a lack of standardized treatment for progressive or recurrent meningiomas.

Tumor development relies on numerous factors, including the capacity to adapt to hypoxic conditions. Hypoxia is characterized by inadequate oxygenation and is a prevalent condition in solid tumors [5, 6]. The oxygen levels in normal tissues range between 2 and 9% (~ 40 mmHg on average) while hypoxic microenvironment varies between 0.02% and 2% O₂ (below 10 mmHg). Interestingly, despite the brain comprising of only 2% of total body weight, it accounts for 20% of the total oxygen usage within the body; the consumption and distribution of oxygen within the brain is region-dependent, with O₂ levels as low as ~ 0.5% in the midbrain, while ~ 8% in the pia region. Reduced oxygen levels within the tumor microenvironment (TME) bring about a variety of changes in tumor biology. Hypoxia-induced detachment of tumor cells triggers metastatic spread of tumors, often mediated by induction of epithelial to mesenchymal transition (EMT). Hypoxia is also associated with increased stemness in cancer cells including loss of differentiation, tumorigenesis and aggressiveness. Tumor hypoxia also triggers cytokine and chemokine secretion which in turn results in the recruitment of pro-tumor immune cells, suppressing anti-tumor response of a variety of immune cells. Tumor cells exploit a number of mechanisms to adapt to hypoxia which include extrusion of cytotoxic drugs by ABC-transporters, exhibition of quiescent state, metabolic adaptations and display of stemness features, which ultimately result in chemo- and radiotherapy failure [5]. Intra-tumoral hypoxia has been linked to poor disease-free survival in many cancers including prostate, cervical cancer, and head and neck squamous cell carcinoma, as well as brain tumors such as glioblastoma [7].

Extensive studies have been conducted on hypoxia-induced gene expression changes in various cancers; however, meningiomas are relatively underexplored. Some studies have highlighted the hypoxic microenvironment as a critical driver of the malignant phenotype, in meningiomas, which often correlates with WHO grade, tumor aggressiveness, therapeutic resistance, and higher recurrence rates, and is consequently linked to poor prognosis [8, 9]. Under hypoxic conditions, hypoxia-inducible factor 1-alpha (HIF-1A) acts as the primary transcriptional regulator. Elevated HIF-1A levels are associated with peritumoral edema, aggressive meningioma phenotypes, and unfavorable prognosis, establishing it as a potential prognostic biomarker for postoperative outcomes [10]. Spreckelsen et al. revealed that meningiomas with the KLF4^K409Q mutation exhibit enhanced hypoxia signaling due to increased HIF-1A activity, resulting from impaired HIF-1A degradation. In benign intracranial meningiomas, genes in the AhR (aryl hydrocarbon receptor) signaling pathway have been shown to mediate tumor adaptation to hypoxia in both HIF-1A-dependent and independent manner [11]. Initial studies have provided nascent insights on the association of hypoxia with meningioma biology through analysis of expression levels of key hypoxia-associated genes. Hence, it is critical to identify the molecular drivers of hypoxic response in meningioma and assess their impact on disease prognosis to reveal novel targets for the development of hypoxia-targeted therapeutics for meningioma.

Furthermore, epigenetic alterations are critical regulators of gene expression. Changes in DNA methylation patterns are the most extensively characterized epigenetic modifications associated with tumorigenesis [12]. Conserved cytosine-guanine (CpG) islands adjacent to gene regulatory regions often undergo abnormal methylation associated with increased tumor aggressiveness. Metastasis, severity, and recurrence in meningioma have all been associated with hypo- and hypermethylation of various genes [13, 14]. Despite the significance of hypoxia, comprehensive studies addressing hypoxia-induced alterations in gene expression and genome-wide DNA methylation profiles of meningiomas are lacking. To date, the specific mechanisms and molecular mediators of hypoxia that influence meningioma biology through physiological changes remain unclear.

In this study, we generated transcriptomic and DNA methylation profiles of IOMM-Lee cells (representative of grade 3 meningioma) exposed to chronic hypoxic conditions (~ 0.2% O₂ for 48 h). Differentially expressed genes (DEGs), probes with altered methylation patterns, and genes exhibiting differential methylation were identified. The Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses revealed biological roles of DEGs. Protein-protein interaction (PPI) networks unveiled densely interconnected components and hub genes. Further, many of these hypoxia-regulated genes were shown to correlate with disease prognosis (recurrence-free survival) in meningioma. Integration of gene expression and methylation data identified differentially expressed methylation-regulated genes (MeDEGs). To the best of our knowledge, this is the first study involving a comprehensive analysis of hypoxia-associated genes and methylation signatures in grade 3 meningioma and their association with disease prognosis, which reveals potent biomarkers and unlocks a new perspective towards the development of therapeutic alternatives targeting hypoxia in meningioma.

Materials and methods

Cell culture

IOMM-Lee cell line (grade 3 meningioma) was sourced from ATCC and cultured in DMEM (high glucose) medium supplemented with 10% Fetal Bovine Serum, 100 U/ml Penicillin, and 100 µg/ml Streptomycin. The cells were maintained at 37 °C and 5% CO₂ levels.

Hypoxia treatment

IOMM-Lee cells (2 × 10⁵ cells/well) were seeded in six-well plates in triplicate, 24 h prior to hypoxic treatment. The cells were cultured for 48 h in a hypoxia workstation (In vivo 200, Ruskinn Technology Ltd., UK) at 37 °C with 0.2% and 5% levels to mimic hypoxic conditions, while cells maintained under normoxic conditions for the same duration were used as controls. Cells were lysed for RNA extraction post 48 h of hypoxia treatment.

RNA extraction and quality assessment

Total RNA was isolated using the RNeasy mini kit (QIAGEN) following the manufacturer’s protocol. RNA concentration and integrity assessment were conducted via Qubit 4.0 Fluorometer and TapeStation (High Sensitivity D1000 ScreenTape assay), respectively. RNA samples with RNA integrity number (RIN) ≥ 9.0 and 28 S/18S ratio ≥ 1.9 were used for transcriptome profiling.

Complementary DNA (cDNA) library preparation and transcriptome profiling

The cDNA libraries were synthesized using NEBNext Ultra II RNA Library Prep Kit. These were then pooled and sequenced on an Illumina NovaSeq platform at approximately 40 million paired end reads (2 × 150 bp) per sample. Quality Control (QC) and pre-processing of FASTQ files was performed using the Fastp (v 0.20.1) [15] tool for adapter trimming and filtering of low quality reads. The Phred score cut-off was 30 & the minimum read length to retain was 50. Filtered reads were aligned to the human reference genome GRCh38 patch 13 using splice aware aligners like HISAT2 (v 2.1.0) [16] to quantify reads mapped to each transcript. Total number of uniquely mapped reads were quantified using featureCounts (v 2.0.1) [17].The uniquely mapped reads for protein-coding genes were filtered using Ensembl Biomart [18]. Total RNA transcriptomic data generated is available on the NCBI Gene Expression Omnibus (GEO) platform under the GEO Series accession ID- GSE282352.

Differential gene expression analysis and unsupervised clustering

To identify DEGs in hypoxia (w.r.t. normoxia), DESeq2 (v1.37.5) [19] package was used. DESeq2 estimates variance-mean dependence in count data from high-throughput sequencing and tests for differential expression based on a negative binomial distribution-based model. The |log₂FoldChange| and adjusted p-value thresholds of ≥ 1.0 and < 0.05, respectively, were used to filter out DEGs. Top 25 up- and downregulated DEGs were used to perform unsupervised clustering of hypoxic and normoxic samples using Euclidean distance and complete linkage methods, represented as a heat map using SRplot (http://www.bioinformatics.com.cn) [20]. Volcano plots were generated using GraphPad Prism (v8.0.2).

Indian meningioma patient cohort: sample collection and RNA extraction

Our validation cohort comprised patients (N = 21) treated surgically for meningioma at the Department of Neurosurgery at All India Institute of Medical Sciences (AIIMS) Hospital between 2018 and 2022 and healthy individuals (N = 2) treated surgically for epilepsy. Sample collection was carried out according to institutional ethical guidelines. Tumor samples collected at the time of surgery were snap frozen in liquid nitrogen for storage until use. The remaining portions were formalin fixed and paraffin embedded for routine histopathology. All meningiomas were evaluated by a senior neuropathologist (VS) according to the WHO CNS5 (2021) criteria for grading. Clinical features such as age at operation, sex, tumor location, histological subtype, and the MIB-1 labelling index were also recorded and are summarized in Table 1, while detailed clinical information is available in Supplementary File 1. Total RNA was extracted from freshly frozen tumor tissues using the MagMAX™ mirVana™ Total RNA Isolation kit (Invitrogen). Samples with RIN > 6.0 were considered for cDNA synthesis.

Table 1.

Clinicopathological characteristics of the meningioma patient cohort used for validation

Validation Cohort (N = 21 tumors + 2 controls)
Sr. No.	Characteristics	Samples
1	WHO Grade (2021)
	G1	9
	G2	7
	G3	5
2	Sex
	Female	9
	Male	12
3	Age
	> = 40 years	11
	< 40 years	10
4	Subtypes
	Meningothelial	2
	Transitional	6
	Fibrous	1
	Atypical	7
	Anaplastic	5
5	Tumor location
	Basal	8
	Convexity	11
	Intraventricular	2
6	Recurrence Status
	Recurrent	8
	Non-recurrent	13

cDNA synthesis and reverse transcription quantitative polymerase chain reaction (RT-qPCR) gene validation

IOMM-Lee cells (2 × 10⁵ cells/well) were seeded in six-well plates and exposed to hypoxic conditions as previously described, while control cells were maintained under normoxic conditions. Cells were collected at different time points- 6 h, 24 h, 48 h for total RNA extraction using the Qiagen RNeasy Mini Kit according to the manufacturer’s protocol. cDNA was synthesized from 1000 ng of total RNA from hypoxic and normoxic IOMM-Lee cells and patient tumor samples in a 10 µl reaction volume via iScript cDNA synthesis kit (Bio-Rad). For quantification of top DEGs, specific primers were designed (Supplementary File 2) and SYBR Green PCR Master mix (Bio-Rad) was used to set up the reaction in the CFX96TM real-time system (Bio-Rad). To avoid false reads from any genomic DNA contamination in the cDNA, the detection primers were designed to span exon‒exon junctions. The ΔCT values for candidate genes in each sample were obtained after normalisation to housekeeping gene- β-actin (ACTB) and beta-2 microglobulin (B2M) for tumor samples and IOMM-Lee cells, respectively.

Expression correlation analysis in Indian meningioma patient cohort

Expression correlation analysis was performed using ggplot2 in R (v4.3.0) [21]. The normalized Cq values (from RT-qPCR validation) of candidate genes in each sample were used. Pearson’s (r) and Spearman’s (ρ) correlation coefficients were computed to assess linear and monotonic relationships. A linear regression model (Y ~ X) estimated the slope, 95% confidence interval, and corresponding p-value for the relationship between genes. The coefficient of determination (R²) was derived from the model summary to quantify the proportion of variance explained.

Differential gene expression and correlation analysis in publicly available datasets with hypoxia hallmark genes

Differentially expressed genes within different meningioma grades were identified in four meningioma datasets from NCBI GEO database (GSE183653 [22], GSE252291 [23], GSE136661 [24], GSE173825 [25]) using the GEO2R tool (https://www.ncbi.nlm.nih.gov/geo/geo2r/). DEGs were also identified between individual meningioma grades versus healthy controls using the GSE43290 [26] dataset. The |log₂FoldChange| and adjusted p-value thresholds of ≥ 1.0 and < 0.05, respectively were used. The clinical parameters for each dataset are summarized in Table 2. The “Hypoxia_Hallmark” gene set comprising 200 genes upregulated in response to low oxygen levels was extracted from Molecular Signatures Database (MSigDB) (human) [27]. The multiple list comparator tool was used to identify “meningioma-associated hypoxia markers” defined as genes differentially expressed across different comparisons in at least 3 GEO datasets and also overlap with the MSigDB Hypoxia Hallmark list. The expression of the top 100 upregulated DEGs was correlated with meningioma-associated hypoxia markers as well as characteristic hypoxia markers. Pearson correlation coefficients (r) were calculated using GraphPad Prism (version 8.0.2).

Table 2.

Clinical characteristics of meningioma patient cohorts across GEO datasets included in the study

Illumina EPIC v2.0 DNA methylation array

Genomic DNA was extracted using the QIAamp DNA Mini Kit (QIAGEN). The DNA concentration and quality was assessed via a NanoDrop 2000 (ThermoFisher Scientific) spectrophotometer. Genomic DNA was sodium bisulfite converted using the EZ DNA Methylation kit (Zymo Research) and subsequently processed using the Illumina Infinium MethylationEPIC v2.0 array (Illumina), as per the manufacturer’s instructions. Raw IDAT files were processed using the minfi cross-package workflow in R studio (version 4.2.1) [28]. The probes in the array were annotated using the Illumina Human Methylation EPIC v2 anno.20a1.hg38(https://github.com/jokergoo/IlluminaHumanMethylationEPICv2anno.20a1.hg38) and Illumina Human Methylation EPIC v2 manifest (https://github.com/jokergoo/IlluminaHumanMethylationEPICv2manifest) R packages.

Identification of differentially methylated CpGs (DMCs), differentially methylated genes (DMGs), and differentially methylated regions (DMRs)

Normalization of probes was performed using the functional normalization method [29]. CpG probes with a detection p-value > 0.05, those located on sex chromosomes, as well as probes containing SNPs and cross-hybridizing probes obtained from [30], were excluded from downstream analysis. Post filtering, the resulting dataset included 888,670 CpG probes. Both β-values and M-values were computed for analysis, and DMCs between the groups were identified using the limma R package [31] with a significance threshold of p-value < 0.01 and methylation difference (Δβ) of ≥ 0.1 (representing a 10% change). Furthermore, we conducted Spearman’s correlation tests to investigate the relationships between replicates derived from the same sample based on the β-values.

The DMC sites were annotated with gene names and biotypes using the R-package EnsDb.Hsapiens.v86 (version 2.99.0) [32], biomaRt R package (version 2.54.1) [33], and C-It-Loci database (http://c-it-loci.uni-frankfurt.de/, accessed on April 28, 2024) [34]. Gene Ontology (GO) Biological Process (BP) analysis was conducted on the genes using the enrichR R package (version 3.2) [35]. Genomic regions, including promoters, exons, introns, and intergenic regions, as well as regulatory elements such as enhancers, were annotated using the annotatr R package (version 1.24.0) [36]. The GenomicRanges R package (version 1.50.2) [37] was utilized to overlap the genomic coordinates of human super-enhancer elements obtained from the SEdb 2.0 database [38] (http://www.licpathway.net/sedb/download.php, accessed on April 3, 2024) onto the locations of DMCs. The hg38 CTCF motifs were sourced from the JASPAR 2022 database through the CTCF R package (version 0.99.11) [39].

DMR analysis was performed using the DMRcate R package (version 3.2.1) [40] using default settings. The lambda parameter was set to 1000 bp which represent the maximum distance between two significant CpG probes to be considered part of the same DMR. This setting affects both the number and size of the DMRs generated. A bandwidth scaling factor (C) of 2 and a significance cutoff (p-cutoff) of 0.05 were also applied. Filtering was then applied to retain DMRs with HMFDR < 0.05, containing more than one CpG probe and a mean methylation difference of < − 0.05 or > 0.05.

Multi-omics integration of DNA methylation and transcriptome data

We applied unsupervised sparse Partial Least Squares (sPLS) using the MixOmics R package (version 6.32.0) [41] to integrate DNA methylation and transcriptome data. First, highly variable CpG probes and genes were selected based on thresholds of Δβ <-0.08 or > 0.08 between the mean methylation levels in hypoxia and normoxia samples for CpGs, and log₂(TPM + 1) < -1 or > 1 for gene expression. Subsequently, the number of dimensions and variables was determined using the default instructions in the MixOmics analysis pipeline. Finally, the spls function was used to perform sPLS in canonical mode and a heatmap was generated using the pheatmap function from the pheatmap package (version 1.0.13) to visualise the correlations between the CpG probes and genes identified by the sPLS analysis. GO and pathway analysis was conducted on the 488 genes identified from the integration analysis using the Enrichr web tool (https://maayanlab.cloud/Enrichr/).

5-Azacytidine treatment

5-Azacytidine (A2385, Sigma-Aldrich), supplied as a crystalline solid, was dissolved in dimethyl sulfoxide (DMSO) and reconstituted in complete DMEM media to reach working concentrations of 5 µM and 10 µM each, for cell treatment. Six-well plates were seeded with cells (2 × 10⁵ cells/well), 24 h prior to treatment. Azacytidine was administered in 24-hour intervals, and cells were lysed for RNA extraction after 48 h under hypoxic conditions.

Pathway enrichment, gene ontology (GO), and gene set enrichment analysis (GSEA)

KEGG pathway and GO analysis was performed for DEGs and genes identified post multi-omic integration to examine their roles in key pathways and underlying biological processes (BP), cellular components (CC), and molecular functions (MF) using the Enrichr web tool (https://maayanlab.cloud/Enrichr/) [35]; a p-value threshold of < 0.05 was considered statistically significant. Key DEGs were also mapped to the HIF-1 signalling pathway using the KEGG Mapper (https://www.genome.jp/kegg/mapper/) tool [42]. It is a pathway visualization tool that maps the user-provided gene list onto reference KEGG pathway diagrams. GSEA was also performed to identify biological pathways enriched in hypoxic cells using the GSEA software (4.4.0) [43] and the Hallmark gene sets (h.all.v2024.1.Hs.symbols.gmt) from MSigDB. Significant pathways were defined as those with FDR q value < 0.05.

Transcription factor enrichment analysis

Metascape (https://metascape.org/gp/) [43] was used to identify the enriched transcription factors (TFs) among the DEGs, which identified enrichment in the TRRUST (Transcriptional Regulatory Relationships Unraveled by Sentence-based Text mining) ontology category. Criteria for terms selection included a p-value below 0.01, an enrichment factor exceeding 1.5, and a minimum count of three. Selected terms were then grouped into clusters based on their degree of similarity.

Prediction of hypoxia response elements (HREs) in the promoter of DEGs

DEGs under hypoxic conditions were analysed for the presence of hypoxia response elements (HREs-consensus binding sites for HIF-1A/ARNT transcription factor complexes) in the promoter region, specifically within 5-kb upstream of the 5’-ends of the DEGs. The upstream region of DEGs were obtained via the Eukaryotic Promoter Database [44]. HREs were predicted using PROMO (version 3.0.2) [45, 46], which uses the TRANSFAC database (version 8.3).

Protein-protein interaction (PPI) network construction and molecular complex detection (MCODE) analyses to identify hub genes and functional modules

The “Search Tool for the Retrieval of Interacting Genes/Proteins” (STRING v12.0) database [47] was used to anticipate the functional and physical interactions of proteins. Network relationships with the highest confidence scores (≥ 0.90) were screened and imported into Cytoscape 3.10.1 [48] for PPI network visualization. Subsequently, depending on the degree of nodes (total number of edges/connections a node has within the network), the top 30 hub genes were determined using the CytoHubba plug-in [49].To recognize densely interconnected components in the PPI network, the MCODE plug-in [50] was used with specific parameter settings (Degree Cutoff = 2, Node Score Cutoff = 0.2, K-Core = 2, and Max-Depth = 100). Clusters were retained based on stringency-adjusted cutoff thresholds: upregulated gene clusters were selected with MCODE score > 4 and node density > 4, while downregulated gene clusters required higher stringency with MCODE score > 10 and node density > 10. GO analysis of top clusters was done using Enrichr.

Identification of prognostically relevant DEGs

To identify prognostically relevant genes, top 100 DEGs (based on log₂FoldChange) were examined using the PrognoScan database (http://dna00.bio.kyutech.ac.jp/PrognoScan/) [51]. It employs a minimum p-value approach to identify optimal expression cut-off thresholds for patient stratification and assesses the association between gene expression and prognosis (here, Recurrence-free survival (RFS)) using log-rank test statistics. A corrected p-value < 0.05 was considered statistically significant. Kaplan-Meier survival plots were constructed using GraphPad Prism (version 8.0.2).

Association of hypoxia-responsive gene signature with stemness

StemChecker (http://stemchecker.sysbiolab.eu/) [52] tool was used to investigate the role of DEGs in stemness. Statistical significance was determined using a hypergeometric test, which assesses the enrichment significance of genes within composite gene sets associated with various stemness signatures, stem cell types, and transcription factors among the input genes. Bonferroni adjustment for multiple testing was used to obtain adjusted p-values.

Plasmid construct and siRNA

IGFBP3 specific siRNA was purchased from MedChemExpress (Cat. No. HY-RS06626, pre-designed siRNA). A universal scrambled siRNA was used as a negative control (si-NC). For overexpression studies, the HIF-1A coding sequence (full match with transcript variants NM_181054.3 and NM_001530.4) was cloned into the pcDNA3.1(+) vector between HindIII restriction sites.

Transient transfection of HIF-1A overexpression clone and IGFBP3 siRNA

IOMM-Lee cells were seeded in six-well and twelve-well plates at seeding densities 2 × 10⁵ cells/well and 1.5 × 10⁵ cells/well, respectively, 24 h prior to transfection. The cells were transfected with HIF-1A overexpression plasmid (pcDNA3.1(+)-HIF-1A) (3g/well) or IGFBP3 siRNA (50 nM working concentration) with respective controls using Lipofectamine 2000 (Invitrogen) as per manufacturer’s protocol. The transfection mix was prepared in Opti-MEM (reduced-serum medium). Cells were lysed for subsequent total RNA and protein extraction at 48 h post-transfection for RT-qPCR and western blotting analysis for quantification of candidate genes (w.r.t. respective controls) at transcriptional and translational levels.

Cellular assays

The effect of siRNA-mediated IGFBP3 knockdown was assessed in vitro in IOMM-Lee cells on cellular proliferation (MTT cell viability assay) and migration (Wound healing assay). The detailed methodology for both assays is available in Supplementary File 2.

Western blotting

Forty-eight hours post-transfection, cells were harvested and lysed using RIPA buffer to obtain whole protein extracts and quantified via BCA protein assay (Real Gene, #501101). Equal amounts of protein, mixed with Laemmli buffer, were separated on 12% SDS–PAGE gels and transferred onto 0.2 μm nitrocellulose membranes at 90 V for 1.5 h at 4 °C in 1× Towbin transfer buffer. Membranes were blocked with 5% BSA for 1 h at room temperature with gentle agitation, followed by overnight incubation at 4 °C with primary antibodies (1:1000 dilution of IGFBP3 antibody (Cell Signaling Technology #64143), 1:1000 dilution of NDRG1 antibody (Cell Signaling Technology #5196), 1:1000 dilution of HIF-1A antibody (Cell Signaling Technology #36169). After three washes with 1× TBST (10–15 min each), membranes were incubated with secondary antibodies (1:10000 dilution of anti-mouse A16072 and anti-rabbit A16110 (Invitrogen) for 1 h at room temperature. Following three TBST washes (10 min each), protein bands were visualized using an enhanced chemiluminescence (ECL) substrate (Bio-Rad) and imaged with a ChemiDoc system.

Statistical analysis

All the experiments were carried out in triplicates, unless stated otherwise. Shapiro-Wilk test was performed to assess the normality of the data. For normally distributed data, a two-tailed unpaired Student’s t-test was used to evaluate statistical significance. P-value < 0.05 was considered statistically significant. The Welsh t-test (assuming unequal variance between comparison groups) was used to evaluate statistical significance for Indian meningioma patient data. The statistical comparison among groups for DNA methylation data for all genomic and regulatory elements was conducted using the unpaired Wilcoxon signed-rank test. All graphs and statistical analyses were conducted using GraphPad Prism.

Results

Identification of hypoxia-associated gene signature in malignant meningioma cells

To identify the DEGs in response to hypoxia (0.2% O₂ levels), we performed differential gene expression analysis using DESeq2. Significantly dysregulated genes were identified using (|log₂FoldChange| cut off ≥ 1.0 and adjusted p-value < 0.05). We identified 3984 genes that displayed differential expression in response to low oxygen levels (Supplementary File 3), of which 1863 were upregulated and 2121 were downregulated in hypoxia-treated IOMM-Lee cells compared to non-treated cells. The volcano plot and heat map of the DEGs highlighting top up- and downregulated genes are represented in Fig. 1A, B.

Fig. 1.

Hypoxia-responsive gene signatures identified in grade 3 meningioma. (A) Significant differentially expressed genes in hypoxic versus normoxic IOMM-Lee cells; volcano plot depicting the upregulated and downregulated genes under Hypoxic versus Normoxic conditions in IOMM-Lee (Grade 3 Meningioma cell line). (B) Cluster heatmap of DEGs showing top 25 upregulated and downregulated genes. The color transition from blue to red indicates a change from downregulation to upregulation. (C, D) RT-qPCR validation of gene expression changes at 6, 24, 48 h under hypoxic conditions (0.2% O₂). (*p-value < 0.05; **p-value < 0.01; ***p-value < 0.001). (E, F) Upregulation of IGFBP3 and NDRG1 at protein level under hypoxia (0.2% O₂) at 48 h

The expression patterns indicated in the RNA sequencing data were further validated for the top DEGs in hypoxic vs. normoxic conditions in IOMM-Lee cells via RT-qPCR at multiple time points under hypoxia (6 h, 24 h, and 48 h). GPM6A, TRAM1L1 and SGCD transcripts were downregulated while IGFBP3, NDRG1 and UBA52 were upregulated under hypoxia. The time-dependent variation in expression of these genes is depicted in Fig. 1C, D. Furthermore, the protein expression levels of IGFBP3 and NDRG1 after 48 h of hypoxic exposure were also validated by western blot (Fig. 1E, F). The results were consistent with RNA sequencing-derived expression patterns of said genes.

Gene function annotation and pathway enrichment analysis

To examine the functions of DEGs induced by hypoxia, GO and KEGG pathway analyses were performed. The top ten enriched KEGG and GO terms belonging to each category are shown in Fig. 2. For upregulated genes, the most enriched biological processes were negative regulation of DNA-templated transcription, cellular response to hypoxia, and glycolytic process. The most enriched CC terms included cell-substrate junction, glutamatergic synapse, and the most enriched MF terms were 2-oxoglutarate-dependent dioxygenase activity and histone demethylase activity (Fig. 2A). In contrast, for downregulated genes, BP analysis showed enrichment in DNA metabolic processes and double-strand break repair. CC was enriched in the mitochondrial matrix, CMG complex, nucleolus, and heterochromatin. MF analysis revealed enrichment of single-stranded DNA helicase, nuclease, and RNA endonuclease activities, producing 5’-phosphomonoesters (Fig. 2B).

Fig. 2.

Functional enrichment of DEGs and Gene set enrichment analysis. (A) Gene ontology analysis of upregulated genes under hypoxia showing the top10 enriched terms associated with BP: biological process; CC: cell component; MF: molecular function. (B) Gene ontology analysis of downregulated genes under hypoxia. (C) KEGG pathway analysis (upregulated genes). (D) KEGG pathway analysis (downregulated genes). The dot’s size reflects the number of genes that are enriched, while its color indicates the associated p-values. (E) GSEA showing top enriched gene sets under hypoxic conditions

KEGG pathway analysis showed enrichment of upregulated genes in the HIF-1 signaling pathway, MAPK and p53 signaling pathways (Fig. 2C). In contrast, downregulated genes were enriched in systemic lupus erythematosus, neutrophil trap formation, DNA replication and DNA repair (Fig. 2D).

Gene set enrichment analysis

GSEA was performed to identify curated Hallmark gene sets from MSigDB that exhibit statistically significant differences between hypoxic and normoxic conditions. The threshold for significance included a FDR q-value < 0.05 and a nominal p-value < 0.05. GSEA identified 20 gene sets out of 50 different ‘Hallmark gene sets’ to be significantly enriched with “hypoxia” phenotype having the highest enrichment score (Supplementary File 4). Other predominant enriched pathways were TNF-α signaling via NFkB, p53 pathway, glycolysis, hedgehog, TGF-β, mTORC1, epithelial-mesenchymal transition, and Wnt/β-catenin signalling pathways (Fig. 2E). Additionally, DEGs in our data that were also predicted to be involved in HIF-1 signaling pathway were highlighted in KEGG pathway (pink) (Fig.S1A) to understand their position in the HIF-1 signaling cascade. The enriched genes were also visualized in the form of a network, where the dark color of the nodes indicates a higher degree (Fig.S1B). Their respective fold change values in the RNA sequencing data were also annotated (Fig.S1C).

Identification of key transcriptional regulators of hypoxia-associated DEGs

To understand the mechanism of regulation of gene expression in response to hypoxia, upstream regulators such as transcription factors (TFs) predicted to bind the promoter regions of hypoxia-associated DEGs were identified. Evidently, HIF-1A was the most enriched TF for upregulated hypoxia-responsive genes. SP1, BRCA1, TP53, AR, and STAT3 were among other enriched TFs (Fig. 3A). For downregulated genes, E2F1 was the most enriched TF, along with RB1, SP1, STAT1, TP53, and MYC (Fig. 3B). The DEGs were also intersected with the list of all known human TFs from AnimalTFDB4.0, revealing 373 TFs dysregulated under hypoxia. Among the top altered TFs were TLX2, RORA, ESR2, FOXO6, ZBTB7C, TP73, MGA, DACH1, ZNF660, HOXC13, E2F2, along with several other members of the ZNF and HOX families.

Fig. 3.

Transcription Factor Enrichment and HRE Analysis in DEGs. (A) Transcription factor enrichment in upregulated genes. (B) Transcription factor enrichment in downregulated genes. (C) Illustration showing the position of hypoxia response elements (HREs) within the promoters of hypoxia-induced genes. (D) RT-qPCR validation of genes containing HREs upon HIF-1A overexpression. (*p < 0.05; **p < 0.01; ***p < 0.001)

Furthermore, HIF-1A binding sites in the promoter region of DEGs were identified, using the PROMO prediction program, HREs were predicted in the 5-kb region upstream of the top 100 up- and downregulated genes. A total of 22 genes were found to have one or more HREs (Fig. 3C) of which GRM4, SLC38A5, NDRG1, HILPDA, SLAMF9, CNTFR, CELF5, GPR146, LCN1, SLC22A8, ENO2, LRRC15, PPP1R3C, EEF1A2, and BNIP3 were upregulated, whereas CHTF8, RAPGEF5, ADCK1, SERPINB2, RNASEH1, AK7, and KCNMB1 were downregulated under hypoxic conditions (Supplementary File 5). To experimentally validate the regulatory role of HIF-1A, it was overexpressed in IOMM-Lee cells, and the expression levels of target genes with predicted HREs was assessed. NDRG1 expression levels were elevated following HIF-1A overexpression w.r.t. control, whereas RNASEH1 expression was reduced, consistent with our predictions. Expression of known HIF-1A targets, including VEGFA and CA9, was also increased under these conditions (Fig. 3D).

PPI network construction and identification of hypoxia-associated hub genes

To enhance our understanding of the interactions among hypoxia-associated dysregulated genes, PPIs were identified using the STRING tool (http://string-db.org) and networks were visualised using Cytoscape. Dense networks of upregulated genes comprising 1844 nodes (proteins) and 1345 edges (interactions) and of downregulated genes comprising 2101 nodes and 2993 edges, respectively were constructed. All singletons were eliminated from the network. Degree centrality of each node (protein) was used to identify top 30 up- and downregulated hub genes. Among these, UBA52, MAPK3, EP300, FAU, JUN, FOS, STAT3, and various members of the ribosomal subunits (small and large) (Fig.S2A) were upregulated, while several members of the histone H3, H4 and H2A family, MCM family, BRCA1, CDC45 and RAD51 were downregulated (Fig.S2B).

Identification of functional modules of hypoxia-associated DEGs in meningioma

The MCODE clustering algorithm was used to identify prominent PPI modules within the dense PPI networks, which were then assessed for their involvement in diverse biological processes. Thirty subnetworks (modules) with different MCODE scores were obtained from the primary PPI network of upregulated genes. The top six modules with scores > 4 and nodes > 4 are shown in Fig. 4A. Fifty subnetworks with different MCODE scores were obtained for downregulated hypoxia-responsive genes. Clusters of upregulated genes were predominantly involved in ribosome and cytoplasmic translation, glycolysis, and chromatin remodeling, whereas the downregulated modules were primarily enriched in nucleosome assembly and DNA replication processes. The top three most significant modules with scores > 10 and nodes > 10 are shown in Fig. 4B.

Fig. 4.

MCODE analysis highlighting the top clusters containing seed proteins in rectangular nodes responsible for cluster formation. (A) Clusters of upregulated genes under hypoxia with a score > 4. (B) Clusters of downregulated genes under hypoxia with a score > 10.(Each subnetwork contained one seed gene. Rectangular nodes represent the seed node, whereas the rest are clustered nodes.)

Key oncogenes (ONGs), tumor suppressor genes (TSGs) and Immune-related genes (IRGs) identified among hypoxic signatures in meningioma

Oncogenes and TSGs represent two significant classes of genes involved in tumor development and progression. We compared the DEGs with known TSGs and ONGs obtained from OncoKB (https://www.oncokb.org/) [53] relevant across different cancer types based on various sequencing panels, Vogelstein et al. [54] or Sanger Cancer Gene Census (Fig.S3A, B). We identified 51 ONGs among the hypoxia-upregulated genes (including HRAS, EGFR, NOTCH1, STAT3, JUN, EZH1) and 54 TSGs among the downregulated genes (BRCA1/BRCA2, TP53, NF2, PARP1, ATR/ATM, RAD51). Further, immune-related genes differently expressed under hypoxia were identified upon intersection of our list of DEGs with the human IRG list available in the ImmPort database (https://www.immport.org/home) [55]. A total of 191 differentially expressed immune-related genes (DE-IRGs) were identified, of which 103 were upregulated (TGFB1, CXCL8, IL11, NGFR, CD4) and 88 were downregulated (IL1B, TNF superfamily members) (Fig.S3C).

Hypoxia-responsive DEGs are associated with stemness in meningioma

Hypoxia drives transcriptional programs associated with stemness and is recognized as a critical driver of cancer stem cell maintenance. Therefore, to investigate the significance of DEGs in stemness, we used StemChecker which provides curated lists of stemness-associated gene signatures and transcription factors. Upregulated genes under hypoxia significantly overlapped with mammary stem cells, embryonal carcinoma, and embryonic stem cell signature genes (Fig.S4A) with the ‘mammary stem cells’ being the most significant with 45 overlapping DEGs, including KRT16, KRT17, GNG2, MTX1, YEATS2 and ZBTB1. Additionally, the upregulated genes also overlapped with transcription factors involved in stemness maintenance and pluripotency, with SUZ12 identified as the most significant (FDR < 0.05) (Fig.S4B). The downregulated genes were also enriched for stemness-related signatures, including hematopoietic (CALCRL, CRHBP, SERF1B, C11orf7), embryonic (POSTN, SERPINB2, ELAVL2) and neural stem cell genes (GPM6A, MAP2) (Fig.S4C). Furthermore, downregulated genes overlapped with E2F4, NANOG, and SOX2 transcription factors associated gene signature (Fig.S4D). All the corresponding genes that overlapped with each category are reported (Supplementary File 6).

Identification of prognostically relevant hypoxia-associated DEGs

As hypoxia is often associated with aggressive tumor behavior, we further investigated the effect of the top 100 hypoxia-associated DEGs on patient prognosis using the PrognoScan tool via Kaplan-Meier survival analysis. Higher expression of SLITRK2, KRT17, DIRAS1, ABCB6, PDE4C, PTGS1 and ZP1 was associated with poor prognosis. Low expression of 5 downregulated genes including SGCD, LRP1B, CALCRL, MGA and AK7 significantly correlated with poor patient survival (Fig. 5).

Fig. 5.

Kaplan-Meier plots depicting the correlation between progression-free survival in meningioma patients as obtained from PrognoScan

Hypoxia-associated gene signatures in meningioma correlate with known hypoxia-markers and are dysregulated across tumor grades in different datasets

We wanted to evaluate the extent of overlap between our hypoxic signature in advanced grade meningioma cell line with gene expression patterns of aggressive meningioma tumors. To begin with, the expression values of genes from the “Hypoxia_Hallmark gene set” from MSigDB, comprising 200 genes upregulated in response to hypoxia, were assessed in grade 3 meningioma samples across four different GEO datasets (GSE183653, GSE252291, GSE136661, GSE173825). Of these three genes SLC2A1, COL5A1, and PPFIA4 were consistently overexpressed in higher-grade meningiomas across multiple datasets and may serve as meningioma-associated markers of hypoxia. The expression of top upregulated hypoxia-responsive DEGs in our data was correlated with these three genes and two well-established hypoxia markers (VEGFA and CA9) using Pearson’s correlation analysis. The correlation coefficient matrix (r-matrix) and p-value matrix (p-matrix) for each gene with hypoxia markers were generated (Supplementary File 7). We reported genes with significant correlation with at least two of the five hypoxia markers for each dataset. Twenty-one genes in GSE136661, 11 genes in GSE173825, 30 genes in GSE183653, and 11 genes in GSE252291 exhibited a positive correlation with the hypoxia markers (Fig.S5A-D), including ENO2, HK2, PFKFB4, BNIP3, TMEM45A, ZP1, ACKR3, DIRAS1, and IGFBP3.

Our hypoxia-associated DEGs overlapped with DEGs across tumor grades (G1, G2, G3) w.r.t. normal meninges after differential gene analysis of data from a publicly available dataset (GSE43290) using the GEO2R tool with |log₂FoldChange| cut-off ≥ 1.0 and adjusted p-value < 0.05. Seven genes (ISYNA1, PTPRF, PPDPF, RAB11B, BAX, TRIO, and PLPPR2) were upregulated in all meningioma grades w.r.t. controls, as well as in hypoxia (Fig. 6). In contrast, CRABP2, TMEM45A, BMP7, and COL5A1 were upregulated in grade 3 meningioma and hypoxia. Furthermore, 147 genes that were downregulated under hypoxia were also downregulated in all grades, including GPM6A, SGCD, CALCRL, CD22, RAPGEF5, HGF, PZP, EPHA3, MAG, SERPINB2, and KCNMB1. Across different intergrade comparisons (using GSE183653, GSE252291, GSE136661) 32 genes were upregulated and 32 genes were consistently downregulated under hypoxia, as well as in grade 3 vs. grade 1 comparisons across all datasets. In grade 3 vs. grade 2 comparisons, 8 genes were commonly upregulated, while only one gene, ARHGEF28, was consistently downregulated (Supplementary File 8). For grade 2 vs. grade 1 comparisons, only 4 genes were common across all datasets, of which HJURP and NEFL were upregulated, whereas ENPP1 and SLC16A9 were downregulated. Expression of all common genes are depicted in Fig. 6B.

Fig. 6.

Overlap of DEGs under hypoxia with various meningioma grades and intergrade comparisons (A) Venn diagram illustrating the shared DEGs across different comparisons. (B) Heatmap depicting the expression patterns of these common genes across multiple datasets, with the color gradient representing log₂FoldChange, transitioning from blue (downregulation) to red (upregulation)

Hypoxia-associated DEGs positively correlate with VEGFA and are dysregulated across tumor grades in Indian meningioma patient samples

To assess the correlation of key hypoxia-induced genes (IGFBP3 and NDRG1) with well-known hypoxia marker (VEGFA) in Indian meningioma patient samples, we performed expression correlation analysis. Both IGFBP3 and NDRG1 were shown to positively correlate with VEGFA expression, with Pearson’s (r) coefficients 0.712 (p = 0.000293) and 0.612 (p = 0.00317), respectively, and Spearman’s (ρ) coefficients 0.713 (p = 0.000289) and 0.464 (p = 0.0342), respectively and R² values of 0.507 (IGFBP3-VEGFA) and 0.374 (NDRG1-VEGFA) (Fig. 7A, B). IGFBP3 was thus more significantly correlated with VEGFA than NDRG1.

Fig. 7.

IGFBP3 expression positively correlates with VEGFA and regulates meningioma cell viability and migration. (A–B) Correlation of IGFBP3 and NDRG1 expression with VEGFA in meningioma patients. (C) RT-qPCR validation of IGFBP3 across tumor grades and recurrent vs. non-recurrent cases. (D) Confirmation of IGFBP3 downregulation at the protein level upon IGFBP3 siRNA transfection via western blotting (n = 2) (E) IGFBP3 depletion reduces meningioma cell viability. (F-G) Wound-healing assay showing reduced migration upon IGFBP3 silencing

To assess the reproducibility of our hypoxia-associated gene signatures, we quantified the levels of IGFBP3 in twenty-one Indian meningioma tumor samples (G1 = 9, G2 = 7, G3 = 5; Recurrent = 8, Non-recurrent = 13) via RT-qPCR. Interestingly, IGFBP3 was significantly upregulated in advanced grades (Grade 2 and Grade 3) w.r.t. controls (p = 0.0403 and 0.0172, respectively) as well as in grade 3 meningioma w.r.t. grade 1 (p = 0.0144) (Fig. 7C). Upregulation of IGFBP3 in grade 2 meningioma w.r.t. grade 1 was close to statistical significance (p = 0.0516). Thus, a sharp rise in IGFBP3 levels is associated with advanced grade in Indian meningioma patients. IGFBP3 levels appear to diminish in recurrent meningiomas w.r.t. non-recurrent cases, however, not statistically significant.

Effect of IGFBP3 knockdown on cell viability under hypoxic conditions

IGFBP3 was found to be highly upregulated at the transcript level in hypoxia-treated IOMM-Lee cells compared to normoxia and was also significantly elevated in higher-grade meningiomas compared to controls, as well as in Grade 3 tumors relative to Grade 1. To investigate its functional role, siRNA-mediated knockdown of IGFBP3 was performed using 50 nM of siRNA (si-IGFBP3). The knockdown efficiency was confirmed at protein level by western blotting using cell lysates collected 48 h post-transfection with si-IGFBP3 or the respective control (si-NC) (Fig. 7D). Cell viability was assessed using the MTT assay at 24- and 48-hours post-transfection under hypoxic conditions (0.2% O₂). Knockdown of IGFBP3 resulted in a significant decrease in cell viability compared to control siRNA-treated cells, with viability reduced to 76.1% at 24 h and 61.4% at 48 h (Fig. 7E). The reduction in viability was more pronounced at 48 h post-transfection, indicating a time-dependent reduction in cell survival upon IGFBP3 depletion. These findings suggest that IGFBP3 contributes to the survival and proliferative capacity of IOMM-Lee cells in a hypoxic microenvironment.

IGFBP3 silencing reduces cell migration

To evaluate the role of IGFBP3 in meningioma cell migration, a wound healing (scratch) assay was performed under hypoxic conditions (0.2% O₂). Cells transfected with IGFBP3 siRNA exhibited a significant reduction in migration compared to control siRNA-treated cells. Quantification of relative migration revealed a 51% reduction in the migratory ability of IGFBP3-silenced cells compared to controls, indicating impaired motility upon knockdown (Fig. 7F, G); suggesting that IGFBP3 may function as a positive regulator of hypoxia-induced cell migration.

Distribution of DNA methylation pattern in the genomic regions and regulatory elements

First, we analyzed the methylation profile of normoxia and hypoxia (each condition included a biological replicate) at a global genomic scale. The global analysis indicated that genome-wide methylation levels were similar between normoxia and hypoxia (median of 58% and 57% respectively, Fig. 8A). As expected, gene promoters showed hypomethylation and gene body (exonic and intronic regions) showed hypermethylated state in both normoxia and hypoxia condition (Fig. 8A). Interestingly, intergenic regions showed intermediate methylation status (median methylation of 49% and 48%). However, the median methylation differences were small (Supplementary File 9) in global-scale and in genomic regions (promoter, intron, intergenic), hypoxic cells were significantly hypomethylated (p-values ranging from 1.1e-08 to < 2e-16). However, exonic regions had no significant difference between normoxia and hypoxia (p-value = 0.093). We also investigated methylation patterns in the genome regulatory elements (enhancers, super-enhancers and CTCF) (Fig. 8B). The enhancer and CTCF regions had similar methylation levels in normoxia and hypoxia (median methylation of 49% in normoxia vs. 48% in hypoxia for enhancers and 51% vs. 51% for CTCF). Whereas the super-enhancer showed higher methylation levels compared to enhancer and CTCF (median methylation of 62% in normoxia vs. 61% in hypoxia for super-enhancer). Consistent with the global pattern, hypoxic cells were significantly hypomethylated in CTCF and enhancer elements (p-value < 0.01).

Fig. 8.

Overview of DNA methylation patterns across genomic regions and regulatory elements. (A) Boxplot illustrating the methylation level of CpG probes at various genomic elements. (B) Box plots showing the distribution of methylation levels at different regulatory elements. The y-axis shows DNA methylation level (b value) ranged from 0 (unmethylated) to 1 (fully methylated). Boxplot shows the median as black thick line. The asterisk (*) indicates statistical significance. Wilcoxon test was used to compute p values between normoxia and hypoxia. Differential methylation and functional enrichment analysis in normoxia and hypoxia samples. (C) Percentage of hypermethylated and hypomethylated DMCs in the genomic regions. (D) Distribution of hypermethylated and hypomethylated DMCs across CpG island regions. (E) Top 10 GO biological processes using 332 unique genes. The number of genes enriched to the corresponding GO term has been represented by the x-axis, while the GO biological process term by the y-axis. (F) Heatmap of 40 DMCs with greater than 17.5% methylation variation in hypoxia cells compared to normoxia cells. Annotated with associated genes, gene biotype, and CpG island regions. The row represents CpG probes, and the column represents samples. Red indicates low methylation (hypomethylation) and yellow indicates high methylation (hypermethylation) beta values. Methylation-regulated DEGs (ME-DEGs). (G) Intersection of DEGs with hypermethylated (top) and hypomethylated genes (bottom). (H) RT-qPCR validation of expression of genes upon azacytidine treatment. (*p-value < 0.05; **p-value < 0.01; ***p-value < 0.001)

Identification of differentially methylated CpGs in normoxia and hypoxia cells

Differential methylation analysis was conducted between normoxia and hypoxia samples to identify specific CpG sites with significant methylation differences. A total of 581 DMCs were identified to have a 10% difference in methylation (p-value < 0.01). Majority of the DMCs (72%) were hypermethylated in hypoxia compared to normoxia cells. More than 60% of the DMCs mapped to intron and other genomic regions (such as 3’UTR, 5’UTR, CDS and exon-intron boundaries) and < 40% of the DMCs belonged to exon, promoter, and intergenic regions (Fig. 8C). For CpG features, ~ 70% of the DMCs were in the open sea regions and very small proportion of DMCs belonged to CpG islands (Fig. 8D). This result is expected since 60% of the HumanMethylationEPIC v2.0 CpG probes are in open sea region [56]. Furthermore, 581 DMCs were mapped to 332 unique genes and 70% of those genes were identified as protein coding. The gene ontology analysis of the 332 genes showed enrichment of several key biological processes (p-value < 0.01, (Supplementary File 10). The top pathways were nervous development and synapse assembly both of which are relevant in the context of meningioma as it’s a cancer of central nervous system (Fig. 8E). In addition, pathways related to stem cell function and metabolic activities were also highly significant (Fig. 8E). We further filtered the DMCs with a |Δβ| ≥ 0.175 (17.5% difference) to identify DMCs with more substantial DNA methylation differences between the two conditions. The results revealed 40 DMCs, including 22 hypermethylated probes and 18 hypomethylated probes (Fig. 8F).

DMR analysis

We have performed DMR analysis between hypoxia and normoxia conditions. Using DMRcate approach and after applying stringent thresholds (adjusted p-value < 0.05 and a mean methylation difference of < − 0.05 or > 0.05), we found 298 DMRs. Out of this 70 DMRs were hypo methylated and 220 were hypermethylated in hypoxia compared to normoxia. Majority of the DMRs reside in intronic regions (124 DMRs) and promoters (100 DMRs). The DMRs mapped a total 270 unique genes (Supplementary File 11). The DMR associated genes contained several direct HIF targets (NDRG1, TFRC, FLT1, IFI16, ENOX1) and HIF induced long non-coding RNA FILNC. In terms of pathway enrichment, we found DMR associated genes were enriched in ERBB4-ERBB4 signalling pathway and regulation of growth.

Integrative analysis of DNA methylation and gene expression

Integration of DNA methylation and transcriptome data was performed to identify global methylome-transcriptome relationship. MixOmics integration analysis identified 480 CpGs associated with 488 unique genes (in the first component analysis that explained majority of the variation) with high level of methylome-transcriptome correlation. These CpGs were either strongly positively (close to 1) or negatively correlated (close to − 1) with expression (Supplementary File 12 and Fig. S6A). We then asked the possible functional role of the 488 methylation regulated genes. Pathway analysis revealed that these genes were primarily associated with DNA replication and repair, including double-strand break repair, DNA elongation (Fig. S6B). Reactome pathway analysis showed that these genes are involved in RNA processing and mitochondrial function, such as rRNA and tRNA processing, mitochondrial RNA degradation, and regulation of mitochondrial RNA stability (Fig. S6C). Together, these findings suggest that DNA methylation is likely to control many genes and potentially involved in the brain’s response to hypoxia that includes DNA replication and repair, cellular migration, metabolism, RNA processing, and mitochondrial function.

Overlap of differentially methylated and expressed genes

To identify CpGs that were differentially methylated between the hypoxic and normoxic groups, the limma R package was utilized with a significance threshold of p-value < 0.01. Probes with > 10% methylation difference (|Δβ| ≥ 0.1) and p-value < 0.01 were considered significant for reporting results. We detected 165 hypermethylated and 416 hypomethylated CpG probes with associated genes. Not all CpG probes have corresponding annotated genes as CpG sites are distributed across the genome including non-coding and regulatory regions. Therefore, CpGs with annotated differentially methylated genes (DMGs) were intersected with the DEGs. Upon the intersection, we identified 24 methylation-regulated DEGs, of which 12 genes were upregulated and hypomethylated, and 12 were downregulated and hypermethylated under hypoxia (Fig. 8G).

Expression of genes upon Azacytidine treatment alters gene activity

Azacytidine, a known inhibitor of DNA methyltransferases, leads to DNA hypomethylation and the reactivation of silenced genes. Azacytidine was administered to IOMM-Lee cells at concentrations of 5 µM and 10 µM under hypoxic conditions, with DMSO serving as a control. We selected two genes, RTN4IP1 and ZBTB7C, from top methylation-regulated DEGs (ME-DEGs) to validate their expression upon treatment. As per RNA sequencing data RTN4IP1 was downregulated and hypermethylated whereas ZBTB7C was upregulated and hypomethylated under hypoxic conditions. Upon treatment with azacytidine, a rise in the expression of both RTN4IP1 and ZBTB7C was observed under both hypoxic and normoxic conditions (Fig. 8H). Changes in gene expression confirmed successful hypomethylation and validated the sequencing results.

Expression of DNA methylation regulators under hypoxia

To investigate the impact of hypoxia on DNA methylation regulation, we examined the normalized expression of key DNA methylation regulators in our hypoxic samples. Specifically, the expression of DNA methyltransferase 1 (DNMT1) was significantly downregulated (log₂FoldChange: -1.74), whereas the expression of Ten-Eleven Translocation methylcytosine dioxygenase 1 (TET1) and Ten-Eleven Translocation methylcytosine dioxygenase 2 (TET2) was significantly upregulated, with log₂FoldChange values of 1.62 and 1.54, respectively.

Discussion

Meningiomas account for ~ 41% of all CNS tumors. Despite high global incidence, surgery and radiation remain the mainstays of treatment, often inefficient in the management of advanced-grade and aggressive tumors at complex anatomical locations. Arising mainly from a lack of vascularity, hypoxia is a hallmark of solid tumors and has a wide influence on tumorigenesis, cell proliferation, elevated glucose metabolism, immunosuppression, increased metastasis, and resistance to chemo- and radiotherapies [57]. A substantial correlation has been established between HIF-1A expression and advanced WHO grades, as well as recurrent tumors in meningioma. Elevated expression of AhR pathway components (such as AhR and ARNT (HIF-1β)) has been positively correlated with advanced tumor grades in meningioma [58]. Interestingly, in a study including 40 meningiomas (majority grade 1), a direct correlation between microvessel density and VEGF expression (p = 0.0005) was established, with higher grades (grade 2 and 3) having numerous microvessels and elevated expression of soluble VEGF isoforms (121 and 165), while benign meningiomas were associated with fewer larger vessels and elevated levels of 189 VEGF isoform. A significant proportion of intracranial meningioma morbidity is linked to the degree of tumor vascularity and extent of peritumoral vasogenic edema [59], both of which may be regulated by hypoxia (through regulation of angiogenesis) within the TME. Heon Yoo et al. showed that carbonic anhydrase 9 (CA9), an endogenous hypoxia marker was strongly expressed in 50% (29 out of 62) of meningioma patients [60]. The expression of PD-L1 programmed death ligand-1), a prominent immunosuppressive factor in cancer, has been correlated with poor prognosis and advanced grade meningioma. Interestingly, PD-L1expression has also been associated with upregulation of genes involved in NFKB2 and hypoxia signaling in meningioma. A direct link between CA9 and PD-L1 peri-necrotic positivity in meningioma was established via IHC and associated with poor disease outcomes [61]. Thus, hypoxic TME in advanced-grade and aggressive meningiomas may contribute to immunosuppression via elevated levels of PD-L1. The distinct biomarker response pattern in malignant meningioma following radiotherapy is characterized by decreased HIF-1A, VEGF, and LDH [62]. HIF-1A is known to enhance glycolytic capacity through the induction of glycolytic regulators such as PFKL and ENO1, both of which were found to be upregulated under hypoxic conditions in the IOMM-Lee cells. Notably, ENO1 has been reported to correlate with immune cell infiltration, including B cells, CD8⁺ and CD4⁺ T cells, macrophages, neutrophils, and dendritic cells, as well as with tumor purity, suggesting a potential link between hypoxia-driven metabolic reprogramming and immune modulation [63]. However, their precise functional roles in meningioma remain poorly defined. While research till now has primarily focused on isolated aspects of hypoxia’s impact on meningiomas, such as the significance of HIFs in gene regulation and the association between hypoxia and aggressive tumor phenotypes. The specific mechanisms and molecular mediators through which hypoxia influences meningioma cells remain unclear.

Our study, for the first time, identified transcriptomic and epigenetic changes in IOMM-lee cells (representative of grade 3 meningioma) in response to hypoxic conditions (~ 0.2% O₂ levels). The identified DEGs were enriched in pathways related to tumor progression and survival under low-oxygen conditions. As predicted, genes involved in the HIF-1 signaling pathway (a master regulator of the hypoxic response) were prominently upregulated. Activation of this pathway facilitates tumor survival and adaptation through angiogenesis, metabolic reprogramming, and resistance to apoptosis, aligning with the aggressive phenotype observed in higher-grade meningiomas. Conversely, downregulated genes were enriched in pathways related to DNA replication and repair, neutrophil extracellular trap formation, and homologous recombination. This downregulation suggests hypoxia-induced impairment of DNA repair mechanisms, which could contribute to increased genomic instability and tumor progression.

The hypoxia-associated gene signatures identified in our study show significant overlap with well-established hypoxia signatures; 107 of 200 “hypoxia hallmark” genes from MSigDB were found to be upregulated under hypoxia in our data. NDRG1, IGFBP3, PPFIA4, ACKR, TMEM45A, SLC2A3, COL5A1, HK2 and LOX are among the top dysregulated genes. Furthermore, IGFBP3, COL5A1, ENO2, and LOX were identified as common genes with those defining epithelial-mesenchymal transition, as per MsigDB. Interestingly, NDRG1 and IGFBP3 have been previously reported to be elevated in atypical and anaplastic meningiomas, and linked to the development of fibrotic tumor vasculature in benign meningiomas [64]. PPFIA4 levels have been reported to be significantly rise in grade 3 meningioma [65]. Elevated levels of ACKR3 (or CXCR7) have been associated with tumor aggressiveness and proliferation in meningioma [66]. Transcriptomic analysis of KLF4^K409Q-mutated meningiomas (prone to PTBE-associated complications) revealed upregulation of SLC2A3 and HK2 [67]. Thus, the key hypoxia biomarkers identified in this study may have significant roles in predicting disease outcomes.

Our study also aimed to elucidate the key players in the upstream regulation of identified hypoxia gene signatures in meningioma. HIF-1A, a subunit of the HIF-1 transcription factor, is a major mediator of the cellular response to hypoxia. Regulatory motif analysis revealed HIF-1A to be the most enriched TF for the upregulated genes. In addition to HIF-1A, other TFs such as STAT3, ATF4, and NF-κB1 were also enriched, which are known for their roles in cellular hypoxic responses. STAT3 is activated under low-oxygen conditions and interacts with HIF-1A to modulate genes involved in immune evasion and tumor progression [68]. In most studies investigating cancer cells, activation of NF-κB in response to hypoxia leads to reduced apoptosis, enhanced angiogenesis, and promotion of cell motility (EMT transition) [69]. Elevated levels of ATF4 in tumor hypoxic regions promote cancer cell survival by regulating metabolic homeostasis, antioxidant biosynthesis, and autophagy [70]. Furthermore, our analysis using StemChecker identified SUZ12 as the most significant TF and 163 upregulated genes were found to be overlapped with it. Previous studies have demonstrated upregulation of SUZ12 in recurrent tumors [71]. Consequently, these transcription factors may be crucial in regulating gene expression under oxygen-deprived conditions.

Of the well-known genes reported to be associated with meningiomas, KLF4 and PTCH1 were upregulated under hypoxia, whereas NF2 was downregulated under hypoxia. NF2 (neurofibromatosis type 2) mutations represent the most common genetic alteration in meningiomas and loss of NF2 protein causes loss of contact-dependent growth inhibition and enhanced anchorage-independent growth [72]. This may be an additional mechanism for low NF2 levels in meningioma patients.

The hypoxia-associated gene signatures in this study have been derived from hypoxic exposure of IOMM-Lee cells, which represent a single grade 3 tumor (in vitro). To estimate the reproducibility of identified signatures in clinical context, we overlapped our DEGs with those identified from differential analysis of publicly available GEO datasets. A total of 335 upregulated genes in hypoxia were also differentially expressed in grade 3 meningiomas; 241 genes in grade 2; and 254 genes in grade 1. This finding suggests a strong association of hypoxia with meningioma grades, particularly advanced grades, thereby reinstating the potential of hypoxia-targeted therapies in meningioma management. CRABP2, TMEM45A, BMP7, and COL5A1 were among top upregulated genes while SGCD, LRP1B, and CALCRL were top downregulated genes in grade 3 meningioma patients and under hypoxia. Interestingly, lower expression of SGCD, LRP1B and CALCRL was associated with poor prognosis in a progression-free survival analysis using Prognoscan. SGCD, CALCRL, GPM6A, CD22, and RAPGEF5 were among the genes downregulated in all 3 meningioma grades w.r.t. controls (G1 v C; G2 v C; G3 v C). GPM6A was also found to overlap with neural stem cells and it showed highest enrichment in the brain as per the Human Protein Atlas Database (https://www.proteinatlas.org/) [73] analysis. Although it is a relatively less explored gene, one study highlights its role as a tumor suppressor gene that inhibits the proliferation of lung adenocarcinoma [74]. In glioblastoma stem cells (GBSC), GPM6A overexpression was associated with radioresistance [75]. In addition, higher expression of upregulated genes such as SLITRK2, KRT17, DIRAS1, ABCB6, PDE4C, PTGS1, andZP1 was associated with a poor patient prognosis, suggesting their potential utility as prognostic biomarkers in meningioma. Additionally, we validated the expression of top hypoxia-responsive genes- IGFBP3 and NDRG1 along with VEGFA in twenty-one Indian meningioma patient samples. Statistically significant upregulation was observed in the expression of IGFBP3 between Grade 2 and Grade 3 w.r.t. controls (p = 0.0403 and 0.0172, respectively) and in Grade 3 w.r.t. Grade 1 (p = 0.0144). Elevated IGFBP3 levels are thus associated with advanced grades in Indian meningioma patients. No significant changes in the levels of NDRG1 were observed, however, the levels seemed to rise from low to high grade tumors w.r.t. controls. Interestingly, both IGFBP3 and NDRG1 were shown to positively correlate with VEGFA expression. IGFBP3 expression more significantly correlated with VEGFA than NDRG1.

To better understand interactions of proteins at the molecular level in response to hypoxia, we constructed PPI networks using the top hypoxia-responsive DEGs from our data. UBA52, MAPK3, EP300, RPS19, and UBC emerged as the top five hub genes. Recent investigations have highlighted the involvement of UBA52 in the progression and development of colorectal cancer and hepatocellular carcinoma [76]. MAPK3 is a widely recognized prognostic factor in various tumors, such as breast carcinoma and glioma [77]. A case study supports the involvement of MAPK3 in recurrent meningioma [78]. In another study, it was found to be aberrated in a formalin-fixed paraffin-embedded (FFPE) meningioma cohort [79]. EP300 levels have been reported as reliable markers for predicting meningioma recurrence, independent of the WHO grade [80]. Members of histone family were among the downregulated hub genes in our hypoxic signature. In giant cell tumors of bone stromal cells reduced levels of canonical histones (H3C6, H2AC8, H4C6 and H3C7) increase genomic instability [81]. Consequently, it is expected that histone deregulation will have a dynamic function that involves multiple processes, which has not been explored extensively with regard to tumor progression. Few studies have highlighted the alterations in the expression of histone genes in cancers of TCGA (The Cancer Genome Atlas) project [82]; no information in the context of histone-family genes in meningioma is currently available.

Hypoxia can simultaneously activate oncogenic signaling pathways that promote proliferation and survival, repress TSGs involved in growth inhibition and apoptosis, and modulate immune-related genes that contribute to immune evasion, highlighting the multifaceted role of hypoxia in tumor progression. Comparison of hypoxia-responsive DEGs with curated cancer gene databases (OncoKB, Vogelstein et al., Sanger Cancer Gene Census) revealed significant enrichment for known oncogenes and tumor suppressors: 51 oncogenes were upregulated (including HRAS, EGFR, NOTCH1, STAT3, JUN, EZH2), while 54 tumor suppressors were downregulated (BRCA1, BRCA2, TP53, NF2, PARP1, ATR, ATM, RAD51). Additionally, intersection with the ImmPort immune gene database identified 191 DE-IRGs: 103 upregulated (TGFB1, CXCL8, IL11, CSF1R, CD274) and 88 downregulated (IL1B, TNF superfamily members).

Furthermore, through integrative analysis of gene expression as well as methylation data, hypermethylated-downregulated (SLCO2A1, RTN4IP1, IFI6) and hypomethylated-upregulated (CACNA1H, MLLT3, ZBTB7C) gene status were reported, thereby hinting at methylation-mediated transcriptomic regulation of gene expression. CACNA1H has previously been reported to be hypomethylated in high-grade (2, 3) meningiomas [83]. T-type calcium channels exhibit high levels across different gliomas and can potentially stimulate tumor cell proliferation. Inhibition of CACNA1H triggers endoplasmic reticulum stress (ERS), leading to apoptosis in glioma cells [84]. The expression profiles and functional mechanisms of ZBTB7C vary significantly among different types of tumor cells [85]; with proto-oncogenic roles in renal clear cell carcinoma [86]. ADGRB3 is an adhesion GPCR transmembrane protein abundantly expressed in the brain, and its hypermethylation has been reported in medulloblastoma [87]; it is also among the most significantly mutated genes in ovarian, breast, lung, and prostate cancer [88]. Changes in methylation patterns and the expression of these genes may contribute to hypoxia-mediated aggressiveness of meningiomas. To summarize, this study provided a comprehensive and integrative evaluation of the hypoxic signature of meningiomas by incorporating gene expression and methylation data. We provide novel insights into key molecular entities involved in hypoxia response in meningioma, their predicted upstream regulators (TFs), promoter methylation patterns, hypoxia-responsive ‘hub genes’, and hypoxia-inducible DEGs with prognostic relevance in meningioma. Key genes highlighted in this study may thus further be explored for their roles as promising biomarkers and/or targets for hypoxia-focused meningioma therapy.

Despite the strengths of a comprehensive integrative transcriptomic and methylation analysis in the context of hypoxia, a well-known hallmark of solid tumors, certain limitations of the study mandate careful and proportionate interpretation of the results. Hypoxia-induced signatures identified in the study were derived from transcriptomic and methylation analysis of cells from a single meningioma cell line, IOMM-Lee. While it is a well-established model for grade 3 meningiomas with reproducible growth and high tumor take rates (87–94% in preclinical models) [89], it represents a single genetic background of malignant meningiomas. Furthermore, established cell lines can accumulate culture-specific alterations over time; IOMM-Lee exhibits a complex karyotype that has likely evolved due to long-term culturing and thus may not fully represent the genetic patterns of primary tumors and heterogeneity [90]. To strengthen the translational relevance of our findings, future studies should incorporate additional meningioma cell lines, as well as primary patient-derived samples across different tumor grades and preclinical models, to validate whether the identified hypoxia-induced genes are conserved more broadly. Also, the methylation patterns identified in the study are associated with a key technical limitation. The Illumina EPIC array cannot discriminate between 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC). As 5hmC plays distinct regulatory roles and is particularly enriched and biologically relevant in tumors of the CNS, this limitation should be considered when interpreting methylation signals derived from array-based data.

Hypoxia is detectable in a substantial subset of meningiomas through tissue-based biomarkers (CA9, HIF-1A, GLUT1) and functional imaging; however, significant heterogeneity in detection methods to detect tissue hypoxia patterns complicates reliable patient selection and treatment endpoint definition [91, 92]. Translational progress will require standardized IHC panels with locked thresholds, convergence with hypoxia PET, biomarker-enriched and serial-imaging adaptive trials, and a clear demonstration that hypoxia biomarker–guided treatment meaningfully improves outcomes over current clinicopathologic models. Key hypoxia-associated genes identified in our study may be evaluated for their expression in additional patient cohorts. A cumulative “hypoxia score” may be derived from weighted expression of key hypoxia markers from this study to further categorize patients into “high hypoxia” and “low hypoxia” groups; this score may further be assessed for correlation with disease outcomes such as progression, recurrence, and aggressiveness, thereby aiding clinical management. Further, the functional validation of promising hypoxia-induced targets in this study via modulation of expression using CRISPR/Cas9-mediated knockout, siRNA-mediated silencing or overexpression is imperative to elucidate their mechanistic roles in meningioma pathogenesis and effects on disease hallmarks such as cell proliferation, migration, invasion, DNA damage and repair, glucose metabolism, and stemness under hypoxia. Thus, the clinical translation of key hypoxia-inducible markers of meningioma will require rigorous validation, advanced in vitro and in vivo functional studies, and multi-cohort testing.

Conclusion

The present study comprehensively analyzed hypoxia-induced gene expression and DNA methylation changes in meningioma. This research addresses the critical need to understand the molecular processes that underlie the adaptation of meningiomas to hypoxia, a prevalent feature of solid tumors that contributes to aggressive behavior, therapeutic resistance, and poor prognosis. Identifying key pathways, transcription factors, and prognostic markers offers valuable insights for developing targeted therapies and diagnostic tools for meningiomas.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CNS

Central nervous system

WHO

World Health Organization

TME

Tumor microenvironment

EMT

Epithelial to mesenchymal transition

HIF-1A

Hypoxia-inducible factor 1-alpha

CpG

Cytosine-guanine

DEGs

Differentially expressed genes

DMCs

Differentially Methylated CpGs

DMGs

Differentially methylated genes

DMRs

Differentially Methylated Regions

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

Me-DEGs

Methylation-regulated genes

MSigDB

Molecular Signatures Database

cDNA

Complementary DNA

RT-qPCR

Reverse Transcription quantitative Polymerase Chain Reaction

TFs

Transcription factors

HREs

Hypoxia response elements

PPI

Protein-Protein interaction

MCODE

Molecular complex detection

STRING

Search Tool for the Retrieval of Interacting Genes/Proteins

RFS

Recurrence-free survival

ONGs

Oncogenes

TSGs

Tumor suppressor genes

IRGs

Immune-related genes

Author contributions

Conceptualization & Supervision: RK & AC; Funding acquisition: RK & AC; Sample preparation: RJ; Data analysis & interpretation: MD & PA; Experiments & Validation: MD, RJ; Validation, Resources; JS, VS; Writing-Original draft preparation: MD, PA; Writing-Review and editing: RK, AC, MD, RJ, PA. All authors read and approved the final manuscript.

Funding

The work was supported by MFIRP/Proposal-367 with funding from IIT Delhi and University of Auckland collaboration fund award to RK and AC.

Data availability

The datasets supporting the conclusions of this article are available in the NCBI’s Gene Expression Omnibus (Edgar et al., 2002) and are accessible through GEO Series accession number GSE282352.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee (Ref. No. IEC264/05.04.2019) of the All India Institute of Medical Sciences, New Delhi, and informed consent was obtained from the patients. All procedures were performed in compliance with relevant laws and institutional guidelines and have been approved by the appropriate institutional committee (Ref. No. IEC-264/05.04.2019) of the All India Institute of Medical Sciences, New Delhi, and informed consent was obtained from the patients.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.