The authors have declared that no competing interests exist.
Conceived and designed the experiments: WJM KW. Performed the experiments: FG LS LZ DJW. Analyzed the data: FG. Contributed reagents/materials/analysis tools: JR XC SH TCC SLG DJW GZ. Wrote the paper: FG DJW WJM KW.
Meningiomas are central nervous system tumors that originate from the meningeal coverings of the brain and spinal cord. Most meningiomas are pathologically benign or atypical, but 3–5% display malignant features. Despite previous studies on benign and atypical meningiomas, the key molecular pathways involved in malignant transformation remain to be determined, as does the extent of epigenetic alteration in malignant meningiomas. In this study, we explored the landscape of DNA methylation in ten benign, five atypical and four malignant meningiomas. Compared to the benign tumors, the atypical and malignant meningiomas demonstrate increased global DNA hypomethylation. Clustering analysis readily separates malignant from atypical and benign tumors, implicating that DNA methylation patterns may serve as diagnostic biomarkers for malignancy. Genes with hypermethylated CpG islands in malignant meningiomas (such as
Meningiomas are central nervous system tumors that originate from the meningeal coverings of the brain and spinal cord. They are the most frequently diagnosed primary brain tumor, accounting for 33.8% of all primary brain tumors in the United States, with an annual incidence of 40 to 60 cases per million persons
Several genetic studies have been conducted over the past a few years to investigate the genetic basis of meningioma pathogenesis. A genome-wide association study including 859 meningioma patients and 704 control subjects identified the
In addition to alteration in the DNA sequences, epigenetic modifications are also closely linked to cancer genesis and progression. Aberrant DNA methylation is one of the major types of epigenetic modifications in cancer
In the current study, we use meningiomas as a model system to investigate the role of DNA methylation in tumor malignancy. To our knowledge, this is the first study to explore genome-scale DNA methylation in malignant, atypical and benign meningiomas. Additionally, unlike most previous cancer genetics studies that compare DNA methylation patterns between tumor and normal tissue, our goal was to investigate whether benign and malignant tumors differ in DNA methylation patterns, and if these differences have biological and clinical significance. We assayed genome-scale DNA methylation and mRNA expression levels of meningiomas categorized as malignant, atypical and benign according to the WHO criteria. Our analysis provides a mechanistic view to uncover the regulatory role of DNA methylation in meningioma malignancy. We identified a list of differentially methylated genes that could serve as diagnostic biomarkers or as candidate genes for further investigation.
The study was approved by the Institutional Review Board at the University of Southern California. Informed written consent was obtained for all participants (either from patients directly or a surrogate with the highest priority for decision making if the patient is unable to provide consent) prior to undergoing surgical tumor resection. All patients presenting to Los Angeles County – USC Medical Center or Keck Hospital of USC with benign or malignant neoplasms of the brain or spine were eligible participants. Demographic information was obtained from patient medical records. Tumor samples were procured by the clinical pathology service at the time of surgery. Specimens were brought directly to the pathology laboratory from the operating room for rapid processing. After sufficient tissue was utilized for diagnostic purposes and released by pathology, the remaining tissue was stored in sterile saline and snap frozen in a −80°C freezer.
Genomic DNA samples were prepared by Qiagen DNeasy Blood & Tissue Kit using the manufacturer's recommended protocol. Genomic DNA was successfully extracted from all except one of the tissue samples collected. Genomic DNA extracted from the excluded sample had a very low DNA concentration and didn't pass the quality control (QC) test for the DNA methylation array. This sample was retrieved from the USC tissue bank, and no additional tissue sample was available for this subject. Genome-scale DNA methylation profiles of 482,421 methylation sites were assessed using the Illumina Infinium HumanMethylation450 (HM450) BeadChips. The array has 99% coverage of RefSeq genes, at gene promoter, 5′UTR, first exon, gene body, and 3′UTR. For DNA methylation data obtained after background correction, data points with a detection
Total RNA samples were extracted from meningiomas tissue by Qiagen RNeasy Plus Universal Kit following manufacture's recommended protocol. The liquid nitrogen grinding method was used to homogenize tissue samples. Tissue samples were first cut into small fragments with no more than 100 mg material in a 1.5 mL tube. Liquid nitrogen was then added into the tube, and the tissue was grinded with grinding rod, before adding 900 µL QIAzol Lysis Reagent. The tubes were vortexed for 60 sec and left at room temperature for 5 minutes before following the standard RNeasy Plus protocol for RNA extraction. The quality of total RNA was assessed using Experion™ RNA StdSens Chip on a Bio-Rad Experion system (bioanalyzer). Genome-scale gene expression profiles of approximately 47,000 transcripts were quantified using an Illumina HumanHT-12 v4 Expression Beadchip. Raw data were processed and normalized without background subtraction using the Illumina GenomeStudio software suite.
Clustering analysis was performed using MultiExperiment Viewer (MeV) software suite v.4.7
In the benign group, TSS200 DNA methylation probes having mean β values between 0.80 and 1 (6,834 probes) were selected as genes with heavily methylated promoters. To further identify genes significantly hypomethylated within this pool, differential mean β values between benign and malignant groups for each selected probe were calculated. The probes with differential mean β values greater than 0.6 were chosen for analysis.
Differential mean β values calculated for all the probes (genomic locations) together with the ChIP-Seq signals of EZH2, RING1B and H3K27me3 in human embryonic stem cells (GSE13084) were visualized using Integrative Genomics Viewer (IGV)
Publicly available gene expression data for meningiomas were downloaded from Gene Expression Omnibus database (GSE16153), including expression array data for forty-three benign and six malignant tumor samples. Both DNA methylation and expression data collected from 27 brain glioma samples were downloaded from the TCGA Data Portal (
Our study assayed DNA methylation in 19 primary brain tumor samples (ten benign, five atypical and four malignant meningiomas) from patients who underwent surgical tumor resection at the Keck Hospital of USC and the Los Angeles County – USC Medical Center. Patient characteristics are provided in
Analysis of sample DNA methylation mean β values for all probes, TSS200 probes, TSS1500 probes, 1st Exon probes, Gene Body probes, 3′UTR probes, and 5′UTR probes. Atypical and malignant meningiomas tend to have decreased levels of global DNA methylation in comparison to benign meningiomas.
Patient | Sample ID | Age | Hospital | Gender | Race | Location | Chemo/Radiation | WHOGrade | Methylation | Gene expression |
1 | 249 | 56 | LAC | M | H | LEFT FRONTAL | NONE | III | Y | Y |
2 | 362 | 68 | LAC | F | A | LEFT FRONTAL | NONE | II | Y | N |
2 | 489 | 44 | LAC | F | H | OLFACTORY GROOVE | NONE | II | Y | N |
4 | 479 | 60 | LAC | F | H | LEFT SPHENOID WING | NONE | I | Y | Y |
5 | 347 | 51 | LAC | F | H | LEFT POSTERIOR CLINOID | NONE | I | Y | Y |
6 | 392 | 53 | LAC | F | H | LEFT SPHENOID WING | NONE | I | Y | Y |
7 | 403 | 57 | LAC | F | C | LEFT TEMPORAL | NONE | I | Y | N |
8 | 010 | 58 | KH | M | C | RIGHT OCCIPITAL | NONE | III | Y | Y |
9 | 254 | 47 | KH | F | C | LEFT PARIETAL, RIGHT PARIETAL | NONE | III | Y | Y |
10 | 678 | 68 | KH | F | H | LEFT FRONTAL/TEMPORAL | NONE | II | Y | N |
11 | 238 | 61 | KH | F | H | LEFT FRONTAL | NONE | I | Y | N |
12 | 048 | 63 | KH | F | C | RIGHT TEMPORAL | NONE | I | Y | Y |
13 | 404 | 48 | KH | F | C | LEFT FRONTAL | NONE | I | Y | Y |
14 | 255 | 32 | KH | F | A | OLFACTORY GROOVE | Chemo/Radiation | III | Y | Y |
15 | 176 | 50 | LAC | F | H | LEFT TEMPORAL | NONE | 1 | Y | N |
16 | 221 | 38 | LAC | M | H | LEFT FRONTAL | NONE | III | N | Y |
17 | 274 | 35 | LAC | M | H | LEFT PETROUS | NONE | I | Y | N |
18 | 311 | 50 | LAC | F | A | LEFT PARIETAL | NONE | I | Y | N |
19 | 351 | 24 | LAC | M | C | RIGHT FRONTAL | NONE | II | Y | N |
20 | 320 | 55 | LAC | M | H | LEFT FRONTAL | NONE | II | Y | N |
Sample ID: Refers to sample number for analysis.
Age: Age at the time of surgery when the sample was obtained.
Hospital: LAC = LA County; KH = Keck Hospital of USC.
Race: C = Caucasian; H = Hispanic, A = Asian.
Chemo/Radiation: Refers to whether the patient received chemotherapy or radiation prior to surgery to obtain the specimen.
Methylation: Included in methylation analysis.
Gene Expression: Included in gene expression analysis.
To investigate whether global DNA methylation patterns can be predictive of WHO tumor grades, we performed hierarchical clustering analysis, using the top 2% of probes with highest variability (standard deviation) of DNA methylation levels across 19 meningiomas. Additionally, we included 4 pituitary adenoma samples in the same analysis to assess any similarity with different subtypes of meningiomas. Hierarchical clustering readily divides these samples into three distinct subgroups (
Malignant, atypical, benign meningiomas and pituitary adenoma samples were labeled with M, A, B and P, respectively. Malignant meningiomas can be readily separated from benign/atypical meningiomas or pituitary adenomas, but benign and atypical meningiomas are not separable based on methylation patterns alone.
To investigate whether genes affected by alterations of DNA methylation in malignant meningiomas tend to fall within specific molecular pathways, we next performed pathway analysis using the Gene Set Enrichment Analysis (GSEA) software suite
Rank | Gene sets in MSigDB | Size | ES | NES | NOM p-val | FDR q-val | FWER p-val |
1 | HSA05214_GLIOMA | 32 | −0.51 | −1.79 | 0.003 | 0.119 | 0.072 |
2 | HSA00561_GLYCEROLIPID_METABOLISM | 30 | −0.58 | −1.74 | 0.004 | 0.141 | 0.16 |
3 | HSA04540_GAP_JUNCTION | 53 | −0.55 | −1.74 | 0.000 | 0.095 | 0.161 |
4 | GLUCONEOGENESIS | 25 | −0.56 | −1.72 | 0.006 | 0.080 | 0.177 |
5 | GLYCOLYSIS | 25 | −0.56 | −1.72 | 0.006 | 0.064 | 0.177 |
6 | HSA00010_GLYCOLYSIS_AND_GLUCONEOGENESIS | 33 | −0.56 | −1.72 | 0.002 | 0.058 | 0.189 |
7 | GLYCEROLIPID_METABOLISM | 21 | −0.65 | −1.70 | 0.004 | 0.058 | 0.218 |
8 | HSA04060_CYTOKINE_CYTOKINE_RECEPTOR_INTERACTION | 121 | −0.52 | −1.68 | 0.000 | 0.070 | 0.278 |
9 | HSA04020_CALCIUM_SIGNALING_PATHWAY | 74 | −0.51 | −1.64 | 0.000 | 0.093 | 0.383 |
10 | ARGININE_AND_PROLINE_METABOLISM | 20 | −0.60 | −1.56 | 0.009 | 0.185 | 0.611 |
Analysis was performed to compare malignant and benign sample groups. Size – number of genes in a set; ES – enrichment score, represents the degree to which a gene set is overrepresented at the extreme of the ranked list; NES – normalized enrichment score, represents the corrected ES based on the size of the set; NOM p-val – nominal
Levels of gene expression can be regulated by multiple types of epigenetic control mechanisms that include but are not limited to DNA methylation, histone modification, transcription factor binding and higher-order chromosomal structure. To understand the relationship of DNA methylation and gene expression, genome-scale gene expression profiles in 12 meningioma samples (five benign, two atypical and five malignant) were measured using the Illumina HumanHT-12 v4 Expression BeadChip. All but one sample was previously subjected to the Illumina Infinium HM450 DNA methylation BeadChip for DNA methylation analysis (
Scatter plots with smoothed density showing the correlation of DNA methylation levels at different genic regions with gene expression. The plots reveal the relationship of DNA methylation and expression in benign and malignant tumors.
We further investigated the 15 most hypomethylated genes and asked whether these genes have altered expression in malignant tumors. Most of these genes were lowly expressed in both benign and malignant meningiomas (
In addition to global DNA hypomethylation, focal DNA hypermethylation at CpG islands is also typical in human cancers
To explore DNA hypermethylation of CpG islands, CpG island probes with increased mean β values in malignant meningiomas compared to benign tumors (total 56,677) were chosen for statistical analysis to identify significantly hypermethylated CpG islands. The probes with decreased mean β values in malignant meningiomas (total 85,342) were also tested to identify significantly hypomethylated CpG islands. Significance Analysis of Microarray (SAM) analysis (FDR = 1%, see Methods)
Aggregation plots of EZH2 (
Examination of expression data of genes with hypermethylated promoter CpG islands also identified one gene,
A), Venn diagram showing overlap of suppressed genes with hypermethylated genes; B), Plot showing
DNA hypermethylation is generally associated with epigenetic switching from transient suppression by PRC to permanent suppression by DNA methyltransferases. In malignant meningiomas, the
CpG island DNA hypermethylation of
To the best of our knowledge, the current study represents the first to interrogate whole-genome DNA methylation levels across three subtypes of meningiomas, and provides novel insights into the molecular pathophysiology of malignant transformation in meningioma. Globally decreased DNA methylation levels observed in the atypical and malignant meningiomas compared to their benign counterparts were in line with aggressiveness of this tumor type, and DNA hypomethylation was not restricted to specific gene regions. Malignant meningioma could be distinguished from atypical and benign ones according to DNA methylation patterns, suggesting the potential application of genome-wide DNA methylation profiling in specifying the diagnostic subtype of meningioma.
In our study, DNA methylation and gene transcription profiles were assessed in a group of patients with malignant, atypical, or benign meningiomas. Biological pathway analysis identified DNA hypomethylation events associated with the glioma pathway in malignant meningiomas. However, expression levels of the identified DNA hypomethylated genes was not significantly altered in the malignant group, suggesting change of DNA methylation does not directly contribute to altered gene expression. Consistent with that, we did not observe overlapping genes of differential DNA methylation at any genic regions with differential gene expression between malignant and benign meningiomas. In addition, we analyzed a publicly available dataset of glioma samples from the TCGA Data Portal. From a chosen group of twenty-seven glioma samples with both DNA methylation data and expression data available, we explored correlation of DNA methylation levels at different genic regions with gene expression. Similar to our findings in meningiomas, we did not observe strong correlation in the glioma dataset (
With genome-scale DNA methylation data, we also explored focal DNA hypermethylation in malignant meningiomas. Malignant tumors often demonstrate DNA hypermethylation at CpG island regions when compared to normal tissue
The majority of genes hypermethylated at promoter CpG islands are suppressed in both malignant and benign meningiomas. We did, however, identify one exception, the
The sample size in our study is relatively small, thus we are limited by statistical power from the analysis. For example, within identified DNA hypermethylated genes, only
In summary, we provided a genome-scale map of DNA methylation in malignant, atypical and benign meningiomas. We further discussed the functional relevance of DNA methylation in tumor malignancy. Genes harboring significantly different methylation patterns between malignant and benign tumors offer a promising avenue for diagnostic utility. However, these findings require further assessment and validation prior to potential clinical application.
DNA methylation levels of glioma-related genes in meningiomas. Heatmap illustrating DNA methylation β values of core enriched genes in glioma-related biological pathways. Genes that belong to the core enrichment group were highlighted. The color in each cell represents gene DNA methylation levels (red: higher methylation; blue: lower methylation).
(TIF)
Differential DNA methylation
(TIF)
DNA methylation
(TIF)
DNA methylation
(TIF)
Expression levels of DNMT1 gene in meningiomas. A) Comparison of
(TIF)
Expression data of genes hypomethylated at core promoter regions. Expression levels for genes severely hypomethylated at core promoters were compared in malignant and benign meningiomas. Group mean and standard error of mean (SEM) values were included in the table.
(DOCX)
Distribution of the CpG islands at different genomic regions. Distribution of the CpG islands at different genomic regions. Distribution at genomic regions such as gene promoter, gene downstream, 5′UTR, 3′UTR, coding exon, intron and distal intergenic regions was compared for the hypomethylated, hypermethylated and all CpG islands.
(DOCX)
Expression of genes with hypermethylated CpG islands. Expression of genes with hypermethylated CpG islands at promoter regions. Group mean expression values for those genes were calculated for malignant and benign meningiomas. Also the information of PRC targeting was included in the last column.
(DOCX)
We thank all of the members in both Wang and Mack laboratories for their comments and suggestions.