signature=3de34be30941887f56ec6b58617ac265,Integrated genomic analyses of

Clinical materials

Institutional Review Board approvals for genetic studies, along with written consent from all study subjects, were obtained at the participating institutions. The specific approval committees included the Human Investigation Committee at Yale University, Ethikkommission der Medizinischen Fakultät der Universität zu Köln and Ethikkommission an der Medizinischen Fakultät der Rheinischen Friedrich-Wilhelms-Universität Bonn.

Selective tissue dissection

For each frozen specimen submitted for whole-exome sequencing, sections were re-reviewed to confirm the diagnosis and assess the adequacy of the frozen tissue for experimental analysis. On H&E-stained sections from frozen tissue blocks, areas of interest were identified and microscopically dissected to ensure that each sample consisted of >70% tumour cells; unwanted regions such as inflammatory and necrotic areas were excluded. Tumours in the replication cohorts did not undergo selective tissue dissection. DNA/RNA was prepared using the Allprep DNA/RNA Mini Kit (Qiagen) with the assistance of a QIAcube.

Exome capture and sequencing

Nimblegen/Roche human solution-capture exome array (Roche Nimblegen, Inc.) was used to capture the exomes of blood and tumour samples according to the manufacturer’s protocol. Sequencing of the library was performed on Illumina HiSeq instruments using paired-end 74 basepair reads by multiplexing two tumour samples or three blood samples per lane. Image analysis and base calling was performed by Illumina Pipeline with default parameters, installed on Yale University's High Performance Computing Cluster.

Whole-exome sequence analysis

We first filtered the reads based on Illumina quality score. The low-quality 3′-end of the reads using FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/index.html) and PCR primer-contaminated sequences, that are considered to lead to alignment artifacts, were trimmed using cutadapt. The sequences were kept only if both reads in a pair had more than 35 bases remaining after the above trimming and filtering quality measures. The reads were aligned to the human reference genome (version GRCh37) using Stampy (version 1.0.21) in a hybrid mode with BWA (version 0.5.9-r16)http://picard.sourceforge.net/). Mean target coverage was 246 and 160 for tumour and blood respectively. The average percentage of reads with at least 20 × coverage was 93% and 90% for tumour and blood respectively. We performed multi-sequence local realignment around putative and known insertion/deletion sites. This was followed by the base quality score recalibration using the Genome Analysis Toolkit (GATK, version 2.5–2)http://evs.gs.washington.edu/EVS/) and 1000 Genome Database. In addition we used our internal database of 2216 exomes to compare the variant allele frequencies of each gene and excluded the variants in the genes that have greater than 150 variant alleles. Finally, we annotated variant alleles using Ensembl database (version 69) with the help of Variant Effect Predictor (v2.7) toolhttp://archive.broadinstitute.org/cancer/cga/indelocator) and Strelka50). Mutational signatures of 6 main categories, T>C/A>G, A>C/T>G, G>C/C>G, C>T/G>A, A>T/T>A, G>T/C>A) are calculated based on filtered somatic variants.

Clonality analysis

Clonality rate of each somatic mutation was calculated based on the variant allele frequency and ploidy at that site, taking into account the admixture rate of each tumour.

CNV identification from exome data

The log ratio of depth of coverage between tumour and blood was calculated using GATK-Depth Of Coverage tool. CNV segments were then called from the log ratio of depth of coverage using ExomeCNV R package

Custom molecular inversion probe sequencing and analysis

Targeted sequencing of exomic regions and exon-intron boundaries of NF2, SMARCB1, TRAF7, PIK3CA, PIK3R1, PRKAR1A, SMO and SUFU plus the recurrent variants AKT1 p.Glu17Lys and KLF4 p.Lys409Gln was performed using molecular inversion probes (MIPs). Recurrent mutations in POLR2A were assessed with Sanger screening.

Custom amplicon sequencing and analysis. Libraries consisting of the coding exons from TRAF7, NF2, SMO, and the recurrent mutations for AKT1 p.Glu17Lys and KLF4 p.Lys409Gln were created using the TargetRich custom amplicon kit (Kailos Genetics).

Sanger sequencing

Coding variants detected by whole-exome sequencing or targeted next-generation sequencing were confirmed by Sanger sequencing using standard protocols.

Whole-genome genotyping

The Illumina Platform was used for WGG and CNV analyses of the samples. Human OmniExpress-12v1.0 BeadChips that contain 733,202 markers were used according to the manufacturer’s protocol (Illumina, San Diego, CA, USA). CNVs were detected by comparing the normalized signal intensity between tumour and matched blood or tumour and the average of all blood samples. Segmentation was performed on log intensity (R) ratios using DNACopy algorithm

Gene expression data

We used Illumina HumanHT12.v4 chips on the gene expression data. Data was normalized using normal-exponential convolution model-based background correction and quantile normalization using the limma R packageP-value<0.05.

Random forest prediction model using gene expression data

We used the randomForest R package (https://cran.r-project.org/web/packages/randomForest/index.html) for building a model to predict the histological gradehttps://www.bioconductor.org/packages/devel/bioc/manuals/affy/man/affy.pdf).

miRNA sequencing analyses

Sequencing was performed on Illumina HiSeq instruments using 75 basepair, single-end reads and multiplexing 8 tumour samples per lane. Adaptor sequences were trimmed using cutadapt

The set of known human miRNA precursors were downloaded from miRBase version 20 (ref. 60). We reported total read counts for 5p and 3p strands. Read counts for each sample were normalized to reads per million reads (RPM) and log2-transformed. Batch effect was corrected using ComBat in sva R package

Six algorithms were used for miRNA target prediction: Miranda, Mirbase, Mirtarget2, Pictar, Tarbase and TargetScan using the RmiR.Hs.miRNA R packageP<0.05) and prediction support in at least three databases. GO term enrichment analysis was performed using the GOStats R package

DNA methylation data

We performed DNA methylation profiling on 57 tumour samples and 3 control samples using the Illumina Infinium HumanMethylation450 BeadChip, which assesses the level of methylation at over 450,000 CpG sites across the entire genome (covering 99% of RefSeq genes and 96% of CpG islands). We first processed raw intensity files (*.idat) and obtained ratio between Illumina methylated probe intensity and total probe intensities (beta-values) using the champ package

Quality control and preprocessing for DNA methylation data

We removed sites containing missing values. The probes targeting a CpG with a SNP were also removed from analysis. Probes targeting the X and Y chromosomes were excluded. The sites having at least 50% samples with detection P-value>0.05 are removed. The detection P-value is calculated using the reported background signal levels of both the methylated and unmethylated channels. After preprocessing the raw data, we performed beta mixture quantile normalization (BMIQ), a normalization correction for the technical differences between the Type I and Type II array designs

Unsupervised clustering for DNA methylation

We performed consensus clustering on methylation b-values with 80% subsampling over 1,000 iterations of hierarchical clustering based on a Pearson correlation distance metric and average linking

Identifying differentially methylated sites

Differentially methylated sites were calculated using the empirical bayes method called limmaP-value<0.05 and median b-value difference >0.1 or

Immunofluorescence staining

Meningioma frozen tissue sections were washed with phosphate buffer saline (PBS) for 5 min, then placed in 4% formaldehyde in PBS for 3 min (fixation), rinsed in PBS and permeabilized with 0.3% Triton-X100 in PBS for 40 s. The sections were rinsed with PBS 3 times for 5 min each and were incubated in BSDSGS blocking solution (PBS with 1% bovine serum albümin, 5% donkey serum, 5% goat serum, 0.1% glycine, 0.1% lysine) with 0.1% Tween 20. Afterwards, the sections were stained with 1:100 rabbit polyclonal anti-EZH2 antibody (5246P, Cell Signaling, Danvers, MA, USA). After washing, the slides were incubated 30 min with 1:200 donkey-anti rabbit Alexa Fluor 555 (A315772, Life Technologies, Grand Island, NY, USA). After washing, slides were mounted with Vectashield DAPI medium (H1200, Vector Labs, Burlingame, CA, USA). Images were analyzed under an inverted microscope (Axio Vert A1, Zeiss, Obekochen, Germany) with fluorescent light source (X-Cite 120Q, Lumen Dynamics).

H3K27ac and H3K27me3 ChIP-seq

Briefly, 10 to 15 frozen sections of each tumour block or dura sample were collected for ChIP-Seq experiments. Tissue was crosslinked with 1% formaldehyde, quenched with glycine, and washed with PBS. Nuclei were extracted by dounce homogenization and resuspended in nuclear lysis buffer containing 0.3% SDS. Chromatin was sheared by sonication with a Q800R2 Sonicator by QSonica (60 min total, amplitude 30, 10 s pulses, 10 s rest). Soluble chromatin was incubated with magnetic beads coated with either H3K27ac antibody (ab4729) or H3K27me3 antibody (ab6002) overnight at 4 °C. Chromatin was precipitated using a magnet, washed extensively, and eluted with TE+1% SDS. Crosslinks were reversed, purified, and subjected to standard Illumina paired-end multiplexed library construction. H3K27ac and H3K27me3 ChIP and input samples were sequenced for each tumour (1 × 75 bp, HiSeq 2000). H3K27ac and H3K27me3 reads were aligned uniquely with bowtie (0.12.7)P-value threshold=0.05) between meningioma subtypes

Identifying super-enhancers from ChIP-seq data

To identify super-enhancers, first regions enriched in ChIP-seq reads for H3K27ac were identified using MACS with input control. Super-enhancers were separated from typical enhancers using ROSE pipeline (https://bitbucket.org/young_computation/rose) with parameters -s (stitching) 12,500, -t (promoter exclusion zone) 2000. We used DiffBind to estimate significance of super-enhancer read change (adjusted P-value threshold=0.05) between meningioma subtypes

Data availability

All somatic mutations identified through exome sequencing of meningiomas were submitted to the COSMIC database previously