Integrated genomic analyses of de novo pathways underlying atypical meningiomas

Abstract

Meningiomas are mostly benign brain tumours, with a potential for becoming atypical or malignant. On the basis of comprehensive genomic, transcriptomic and epigenomic analyses, we compared benign meningiomas to atypical ones. Here, we show that the majority of primary (de novo) atypical meningiomas display loss of NF2, which co-occurs either with genomic instability or recurrent SMARCB1 mutations. These tumours harbour increased H3K27me3 signal and a hypermethylated phenotype, mainly occupying the polycomb repressive complex 2 (PRC2) binding sites in human embryonic stem cells, thereby phenocopying a more primitive cellular state. Consistent with this observation, atypical meningiomas exhibit upregulation of EZH2, the catalytic subunit of the PRC2 complex, as well as the E2F2 and FOXM1 transcriptional networks. Importantly, these primary atypical meningiomas do not harbour TERT promoter mutations, which have been reported in atypical tumours that progressed from benign ones. Our results establish the genomic landscape of primary atypical meningiomas and potential therapeutic targets.

Figure 1. Somatic mutations and copy number variations in benign and atypical meningiomas.

(a) Oncoprint depicting the mutational profile of 75 exome-sequenced meningioma samples is shown. Histological grade, chromosome 22 loss status and recurrently mutated genes, which are clustered based on mutually exclusive meningioma subgroups (left) are summarized at the top panel. The distributions of somatic mutations according to their functional consequences are shown in the middle panel, whereas the mutational signatures are shown at the bottom. The colour codes are explained on the right. (b) Atypical versus benign meningiomas harbour similar number of damaging somatic mutations (atypical n=10, benign n=65). The lines above the bars indicate statistical analysis (Student’s t-test; ns: non-significant). Lines depict the median values; boxes plot 25th to 75th percentiles, whereas separately plotted dots show the outliers. (c) The percentage of genome alteration is statistically significantly different between atypical versus benign meningiomas (atypical n=55, benign n=153) (Student’s t-test). Lines depict the median values; boxes plot 25th to 75th percentiles, whereas separately plotted dots show the outliers. (d) Large-scale genomic events associated with atypical NF2 samples are shown (atypical n=43, benign n=57). Losses of chromosomes 14q, 10q, 10p, 1p and 6q significantly associate with atypical tumours (FDR adjusted Fisher’s exact test). (e) Differences in genomic alterations between atypical NF2 and benign NF2 mutant samples are shown (atypical n=43, benign n=57). Along the horizontal axis, losses are depicted in blue, whereas gains are shown in red. The vertical axis represents the genome. Significantly altered tumour suppressor consensus cancer genes in atypical NF2 samples are noted.

Figure 2. mRNA expression profile in atypical meningiomas.

(a) Unsupervised hierarchical clustering of 138 meningiomas by genome-wide expression profiling is shown. Atypical versus benign histology, underlying meningioma driver mutations, copy number variations, which are colour coded, are shown on the left. Although the expression profile accurately clusters meningioma samples based on driver mutations, it does not fully differentiate atypical versus benign tumours. Consensus clustering also separates non-NF2 meningiomas (Hedgehog, POLR2A, TRAF7/PI3K, TRAF7/KLF4) into different subgroups. (b) PC analysis using mRNA signature genes separates atypical and benign samples. (c) GO Term and pathway enrichment for genes associated with the atypical phenotype compared with benign are shown (atypical n=26, benign n=112). Black line indicates an FDR of 0.05. (d) EZH2 and E2F2 gene expressions, which are upregulated in atypical versus benign samples, are plotted (empirical Bayesian method). Lines depict the median values; boxes plot 25th to 75th percentiles, whereas separately plotted dots show the outliers.

Figure 3. miRNA expression profile in atypical meningiomas.

(a) Unsupervised hierarchical clustering of genome-wide miRNA expression of 32 meningiomas. Integrated view of the miRNA expression clustering, combined with the underlying meningioma driver gene mutation, histological grade and CNV-high or -low status, which are all colour coded, are shown. (b) Heatmap visualization of differentially expressed miRNAs between atypical versus benign samples. The colour scale for log₂ (fold change) is shown at the bottom. MiRNAs affected by CNVs are marked by an asterisk. (c) Differentially expressed miRNAs with either inhibitor or activator effects on cancer hallmark functions are shown. Red coloured miRNAs are upregulated whereas blue coloured miRNAs are downregulated in atypical samples. (d) The top gene ontology terms for the genes regulated by the 14q32 miRNA cluster. Black line indicates an FDR value of 0.05. (e) Expression of let-7 family, a known inhibitor of EZH2, negatively correlates with EZH2 mRNA expression (Spearman correlation test). Lines depict the median values; boxes plot 25th to 75th percentiles, whereas separately plotted dots show the outliers.

Figure 4. DNA-methylation patterns across benign versus atypical meningiomas.

(a) Integrated view of DNA methylation clustering combined with the underlying meningioma driver gene mutations, histological grades and genome copy numbers are shown (atypical n=11, benign n=46). Consensus clustering defines four main subgroups based on methylation status. The CNV-high or -low categories, as well as NF2 and SMARCB1 mutations are marked. (b) PC analysis of meningioma methylation data, which separate NF2 CNV-high, -low and non-NF2 samples into different groups, is shown (atypical n=11, benign n=46). (c) Percentages of hyper- or hypo-methylated sites among differentially methylated sites in various subgroups, as compared with benign non-NF2 samples are plotted. All atypical as well as benign CNV-high NF2 mutant meningiomas are hypermethylated when compared with benign non-NF2 samples (atypical n=11, benign n=46). (d) Percentage of altered genome significantly correlates with the extent of genome-wide DNA hypermethylation (atypical n=11, benign n=46). Along the horizontal axis, the meningioma samples are grouped according to the consensus cluster numbers as in a. The number of differentially hypermethylated sites divided by the total number of sites is shown as the green dashed line and quantified along the vertical axis on the right. Lines depict the median values; boxes plot 25th to 75th percentiles, whereas separately plotted dots show the outliers. (e) Comparisons for atypical versus benign as well as various other molecular subgroups, which are colour coded, are shown for Molecular Signatures Database (MSigDB) gene set enrichment summary plots. The most significant gene sets are plotted.

Figure 5. Histone modifications in meningiomas.

(a) Unsupervised hierarchical clustering of meningioma samples using H3K27me3 profiles is shown (Pearson correlation) (atypical n=3, benign n=3). Grades are colour coded. (b) Starburst plot for atypical versus benign H3K27me3 ChIP-seq signal fold change (horizontal axis) and FDR (vertical axis) is shown. Red and blue circles indicate regions that have increased binding in atypical versus benign meningiomas, respectively. PRC2-hESC targets are marked in red colour. These analyses reveal increased H3K27me3 binding in atypical samples, which shows particular enrichment of PRC2-hESC targets. (c) GO Term enrichment plot for genes with increased H3K27me3 binding and decreased expression is shown (atypical n=3, benign n=3). Black line indicates an FDR of 0.05 (d) Unsupervised hierarchical clustering of 18 meningiomas and 2 controls based on H3K27ac profile in super-enhancer sites is shown (Pearson correlation). Super-enhancer profiles are highly correlated among different meningioma subtypes. All samples are colour coded, which are shown on the right. (e) H3K27ac ChIP-seq results correlate with gene expression profiles (n=18). Atypical versus benign H3K27ac ChIP-seq signal fold change is plotted along the horizontal axis, whereas atypical versus benign gene expression fold change is shown along the vertical axis. Red dots indicate genes that have both activated super-enhancers and are overexpressed in atypical versus benign meningiomas, whereas blue points indicate genes that are associated with de-activated enhancers and are underexpressed. An FDR threshold of 0.05 is used both for gene expression and H3K27ac ChIP-seq data (empirical bayes and DiffBind methods). (f) H3K27ac ChIP-seq occupancy at a super-enhancer near ZIC1 (n=18). The horizontal axis shows genomic position whereas the vertical axis shows signal of ChIP-seq occupancy in units of reads per million (r.p.m.). Super-enhancer region that is differentially bound in atypical samples is depicted as a black line over the gene track and highlighted. (g) ZIC1 gene expression across atypical and benign samples is plotted (n=138) (empirical bayes method). Lines depict the median values; boxes plot 25th to 75th percentiles, whereas separately plotted dots show the outliers.

Figure 6. De novo pathway in the formation of atypical meningiomas.

Samples with NF2 loss were highly enriched among primary (de novo) atypical cases, which co-occur either with genomic instability or recurrent mutations in SMARCB1. These tumours have hypermethylation of the PRC2 binding sites in hESCs, thereby phenocopying a more primitive cellular state. Epigenetic changes acting on the PRC2 networks are associated with upregulation of the catalytic subunit EZH2, and down regulation of its regulator let-7. Primary atypical tumours harbour increased expression of cell-cycle related genes, including the E2F and FOXM1 transcription factor networks. TERT promoter mutations are not present in de novo tumours, and limited to recurrent samples only.