Hypermitotic meningiomas harbor DNA methylation subgroups with distinct biological and clinical features

Abstract

Background: Meningiomas, the most common primary intracranial tumors, can be separated into 3 DNA methylation groups with distinct biological drivers, clinical outcomes, and therapeutic vulnerabilities. Alternative meningioma grouping schemes using copy number variants, gene expression profiles, somatic short variants, or integrated molecular models have been proposed. These data suggest meningioma DNA methylation groups may harbor subgroups unifying contrasting theories of meningioma biology.

Methods: A total of 565 meningioma DNA methylation profiles from patients with comprehensive clinical follow-up at independent discovery (n = 200) or validation (n = 365) institutions were reanalyzed and classified into Merlin-intact, Immune-enriched, or Hypermitotic DNA methylation groups. RNA sequencing from the discovery (n = 200) or validation (n = 302) cohort were analyzed in the context of DNA methylation groups to identify subgroups. Biological features and clinical outcomes were analyzed across meningioma grouping schemes.

Results: RNA sequencing revealed differential enrichment of FOXM1 target genes across two subgroups of Hypermitotic meningiomas. Differential expression and ontology analyses showed the subgroup of Hypermitotic meningiomas without FOXM1 target gene enrichment was distinguished by gene expression programs driving macromolecular metabolism. Analysis of genetic, epigenetic, gene expression, or cellular features revealed Hypermitotic meningioma subgroups were concordant with Proliferative or Hypermetabolic meningiomas, which were previously reported alongside Merlin-intact and Immune-enriched tumors using an integrated molecular model. The addition of DNA methylation subgroups to clinical models refined the prediction of postoperative outcomes compared to the addition of DNA methylation groups.

Conclusions: Meningiomas can be separated into three DNA methylation groups and Hypermitotic meningiomas can be subdivided into Proliferative and Hypermetabolic subgroups, each with distinct biological and clinical features.

Keywords: DNA methylation; cancer; central nervous system; meningioma; tumor.

Figure 1.

Meningioma DNA methylation groups harbor biologically distinct subgroups. (A) Unsupervised hierarchical clustering of meningiomas from the University of California San Francisco (UCSF) discovery cohort (n = 200) using 2094 differentially methylated DNA probes. Dendrogram colors represent previously reported DNA methylation groups or novel DNA methylation subgroups. Heatmaps for DNA methylation probes and FOXM1 target genes from RNA sequencing are shown below the DNA methylation dendrogram. (B) Network of gene circuits distinguishing the subgroups of Hypermitotic meningiomas without enrichment of FOXM1 target genes (Hypermetabolic meningiomas) from the subgroup of Hypermitotic meningiomas with enrichment of FOXM1 target genes (Proliferative meningiomas) from RNA sequencing of the UCSF discovery cohort. Nodes represent pathways and edges represent shared genes between pathways. Nodes colored in red or blue are enriched or suppressed pathways in Hypermetabolic compared to Proliferative meningiomas, respectively. (C) Incidence of meningioma DNA methylation groups (Merlin-intact, Immune-enriched, or Hypermitotic) or subgroups (Hypermetabolic or Proliferative) after SVM classification of consecutive meningiomas from The University of Hong Kong (HKU) validation cohort. (D) Volcano plot from RNA sequencing differential expression analysis of Hypermetabolic versus Proliferative meningiomas from the validation cohort. (E) Heatmap for FOXM1 target genes from RNA sequencing of the validation cohort (n = 302) is shown below the DNA methylation group or subgroup.

Figure 2.

Meningioma clinical and histologic features and DNA methylation groups and subgroups. Sankey plot showing the number and percentage of meningiomas from the discovery and validation cohorts across clinical features, DNA methylation groups, and DNA methylation subgroups.

Figure 3.

Meningioma DNA methylation groups and subgroups predict postoperative outcomes. (A) Meningioma extent of resection (n = 565) across DNA methylation groups and subgroups (Chi-squared test). (B) Meningioma WHO grades (n = 565) across DNA methylation groups and subgroups (Chi-squared test). (C) Kaplan–Meier curves for meningioma local freedom from recurrence (LFFR) or overall survival (n = 565) across DNA methylation groups and subgroups (Log-rank tests). (D) Time-dependent area under the receiver operating characteristic curves (AUC) for meningioma LFFR or overall survival across histologic or DNA methylation classifications. (E) Brier error scores for meningioma LFFR or overall survival across histologic or DNA methylation classifications.

Figure 4.

Meningioma DNA methylation groups and subgroups across alternate molecular grouping schemes. Reclassification of the meningioma in this study according to alternate molecular grouping schemes based on gene expression profiles (gene expression types, n = 502), orthogonal DNA methylation analysis techniques (DNA methylation families, n = 565), copy number variants (CNVs) integrated with histologic features (integrated grade, n = 565), or DNA methylation profiling integrated with CNVs and histologic features (integrated score n = 565). Each of these molecular schemes proposes 3 groups of meningiomas, and concordance across schemes was calculated by dividing the observed distribution of low/intermediate/high-risk cases by the expected distribution from meningioma DNA methylation grouping..