This study was intended for patients with intracranial meningioma. The study population was retrospectively selected from the institutional tumor registry. Patients who had agreed with biobanking and genetic testing of tumor tissue for a scientific purpose have enrolled under the approval of the institutional review board. Patients were not compensated for their participation, and each participant provided written informed consent. All experiments were performed following the relevant guidelines and regulations.

Formalin-fixed paraffin-embedded (FFPE) tissue specimens were obtained after surgical resection. Two senior pathologists (K.S.L. and G.C.) independently reviewed the hematoxylin and eosin (H&E)-stained slides and confirmed the diagnosis according to the fourth edition of the World Health Organization classification of tumors of the CNS (WHO CNS) to identify the histological subtypes of meningiomas.

Immunostaining was performed for a single representative sample block showing specific histomorphologic characteristics from each case. Sections were transferred to poly-L-lysine coated adhesive slides, dried, deparaffinized, and rehydrated. The slides were subsequently subjected to heat-induced antigen retrieval. The following antibodies were used according to the manufacturer’s instruction: c-MET (prediluted, rabbit monoclonal antibody, Ventana, Tucson, AZ, USA), EGFR (1:150, mouse monoclonal antibody, Dako, Camarillo, CA, USA), p16 (prediluted, mouse monoclonal antibody, Ventana, Tucson, AZ, USA), and Ki-67 (1:50, mouse monoclonal antibody, Dako, Carpinteria, CA, USA). The sections were incubated with appropriate reagents from the Dako REAL EnVision Detection System (Dako, Glostrup, Denmark) and were counterstained with Mayer’s hematoxylin. Each case was categorized as either positive — when it showed moderate-to-strong cytoplasmic and/or membranous positivity in tumor cells — or negative, as controls without antibodies. For p16 staining, loss of expression was recorded if the tumor cells showed complete negative, while partial expression was considered positive.

A sample FFPE block with the most representative morphologic characteristics was selected from each case, and tumor areas were manually microdissected from the unstained slides. DNA was extracted using the Maxwell 16 FFPE Purification Kit (Promega, Madison, WI, USA) according to the manufacturer’s instructions. The TERT promoter region covering the hotspot mutations C228T (g.1,295,228C>T in GRCh37) and C250T (g.1,295,250C>T) were amplified by polymerase chain reaction (PCR) using the HotStarTaq Plus Master Mix Kit (Qiagen, Valencia, CA, USA) and primers forward 5’-GTCCTGCCCCTTCACCTT-3’ and reverse 5’-CAGCGCTGCCTGAAACTC-3’, which was run on Verti 96-well Thermal Cycler (Applied Biosystems, Waltham, MA, USA) with the following conditions: 96°C for 5 min; 44 cycles with 96°C for 30 s, 64°C for 30 s and 72°C for 1 min; 72°C for 10 min and 4°C for hold. The quality of PCR products was confirmed by gel electrophoresis on 2% agarose gel. Pyrosequencing of the TERT promoter mutations C228T and C250T was carried out on the PyroMark Q24-MDX system (Qiagen, Valencia, CA, USA) with a sequencing primer 5’-ACCCCGCCCCGTCCCGACCCC-3’, based on the manufacturer’s instructions. TERT promoter mutation was considered positive if T allele frequency at the mutation sites was ≥10%.

DNA was extracted from the FFPE tissue sample using the QIAamp DNA FFPE Tissue Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. The extracted DNA was quantitated using a Quantus fluorometer (Promega, Madison, WI, USA) and Agilent 4200 TapeStation system (Agilent Technologies, Santa Clara, CA, USA). DNA was sheared into fragments with a mean peak size of approximately 180 bp to 200 bp, using Adaptive Focused Acoustics (Covaris, Woburn, MA, USA). The paired-end libraries were prepared with a SureSelectXT HS Target Enrichment System kit (Agilent Technologies, Santa Clara, CA, USA), based on the manufacturer’s instruction, using commercially available targeted panels that cover NF2, AKT1, TRAF7, KLF4, SMO, PIK3CA, SMARCB1, and other cancer-related genes ( Supplementary Table 1 ). The quality of the DNA library was evaluated using a Bioanalyzer 2100 (Agilent Technologies, San Francisco, CA, USA) and quantified using a Qubit dsDNA HS Assay kit and a Qubit 2.0 fluorometer (Life Technologies, Carlsbad, CA, USA). The libraries were paired-end sequenced (2 × 150 bp) using a NextSeq 550Dx (Illumina, San Diego, CA, USA) or a HiSeq 2500 (Illumina, San Diego, CA, USA).

Sequencing data quality assessment and trimming were performed using FastQC v0.11.8 (RRID : SCR_014583). The sequence reads were aligned to the reference genome GRCh37 (hg19) using BWA v0.7.15 (RRID : SCR_010910), Picard v2.16.0 (RRID : SCR_006525), and Samtools v1.3.1 (RRID : SCR_002105). Single nucleotide variants (SNVs) and indels were identified using GATK v4.1.4.0 (RRID : SCR_001876), where a threshold of at least 10 reads and an allelic frequency of ≥10% were used for variant calling. The sequencing variants were annotated using SnpEff & SnpSift v5.1 (RRID : SCR_005191/SCR_015624). The variants found in the population databases, the Exome Aggregation Consortium (RRID : SCR_004068), and the Genome Aggregation Database (RRID : SCR_014964), were filtered out to demonstrate somatic mutations minimizing germline contamination. All variants were interpreted based on the Association for Molecular Pathology (AMP) standards and guidelines (5).

All patients received regular follow-up after initial treatment. Magnetic resonance images (MRIs) were performed within 48 hours after surgery, and the extent of resection was evaluated by neurosurgeons. For patients with visible residual tumors, adjuvant stereotactic radiosurgery or radiotherapy was usually performed within three to six months from surgery. Patients with grade 2 or 3 meningiomas were referred to the institution’s multidisciplinary neuro-oncology outpatient clinic, where a team of neurosurgeon, neuroradiologist, pathologist, medical oncologist, and radiation oncologist participated in a discussion to select the most optimal therapeutic plan for each patient.

Patients were usually followed up every six months for the first two years and then annually thereafter. The recurrence-free survival (RFS) was calculated from the date of surgical resection to the first recurrence of meningioma requiring further therapeutic plans. Evidence of local tumor recurrence was provided by MRI showing meningioma in a location contiguous with the previous operation site.

For descriptive statistics, the frequency with percentages was provided for categorical data and median with interquartile ranges for continuous data. Differences between grades 1 and 2/3 meningiomas were compared using the Chi-square test for categorical variables and the t-test for continuous variables. The statistical correlations between genomic features and clinicopathological parameters were evaluated using the Fisher’s exact test. Differences in RFS between distinct subgroups were assessed by the log-rank test. The prevalence of known recurrent variants in this study was compared to the pooled data from external cohorts using the Chi-squared test. Each result from statistical tests was considered statistically significant if a two-tailed p-value was less than 0.05. All statistical analyses were performed using R v3.6.3 (RRID : SCR_001905).

A total of 713 cases with intracranial meningiomas who received surgical treatment between 2003 and 2020 were reviewed ( Figure 1 ). The median age of the study population was 56 (47-65) years, and females made up 71.6% of the population. Grade 2/3 meningiomas were identified in 143 (20.1%) cases. Patients with grade 2/3 meningiomas tended to have a lower female predominance compared with those with grade 1 tumors ( Table 1 ): male, 44.8% vs. 24.3% (p < 0.001). There was no significant association between grades and the location of tumors.

A sample of 40 patients with grade 2/3 meningiomas participated in the targeted NGS panel study, whose tissue specimens were acquired between 2012 and 2019. The baseline characteristics of the sample and the study population were evenly distributed. Most candidates provided the tissue specimens from the initial surgery, except three patients with recurrent meningiomas who had a history of prior surgery and/or stereotactic radiosurgery. The candidates had no past medical history of underlying malignancies at the time of tissue acquisition, except one patient who had extracranial metastasis and received systemic chemotherapy with hydroxyurea before tumor sampling.

The study population was followed up for 59.1 (30.0-98.0) months. During the follow-up period, 59 cases of grade 1 meningiomas and 40 cases of grade 2/3 meningiomas had an event of recurrence, showing distinct outcomes (p < 0.001) ( Figure 2A ). Five-year recurrence-free survival (RFS) estimates were 86.0 [95% CI, 82.7-89.5%] and 66.8% [95% CI, 58.5-76.2%] for grade 1 and 2/3 meningiomas, respectively.

For grade 2/3 meningiomas, Simpson grade I resection was achieved in 33 (23.1%) cases, followed by 44 (30.8%) cases for grade II, 19 (13.3%) cases for grade III, 43 (30.1%) cases for grade IV, and 4 (2.8%) cases for grade V resection. While the survival difference between grades 2 and 3 was insignificant, the achievement of Simpson grade I resection significantly influenced the clinical outcome in grade 2 meningiomas (p = 0.005) ( Figure 2B ). Other treatment factors regarding adjuvant stereotactic radiosurgery or radiotherapy did not significantly change the outcome.

In the sample population, local recurrence has been developed in 10 patients during the follow-up period of 68.5 (55.8-97.3) months. Simpson grade I resection was achieved in 10 (25.0%) cases. Those patients with recurrence received additional treatment with stereotactic radiosurgery, and two of them underwent reoperation for surgical resection of the recurred tumor. The five-year RFS was 65.4% for a subset of participants whose long-term follow-up data were available, representing the grade 2/3 meningiomas in the study cohort.

To evaluate the clinicopathologic relevance regarding the molecular diagnostic criteria in WHO CNS 5, we assessed the TERT promoter and CDKN2A/B status of meningioma. The TERT promoter hotspot mutation PCR was available for 52 cases (40 samples and additional 12 cases) with grade 2/3 meningiomas. Positive results were identified in three cases; these were conventionally diagnosed as atypical meningiomas with high mitotic activity, and all recurred during the follow-up period and highly refractory to the selected treatments ( Figures 3 – 5 ). Interestingly, two of them had distant metastasis with rapid progression. Those formerly diagnosed as grade 3 meningiomas (7 cases) did not present a positive TERT promoter mutation. The molecularly integrated classifying group with conventional grade 3 and positive TERT promoter mutation showed a poorer outcome than the others ( Figure 2C ).

We also assessed the immunohistochemistry of CDKN2A (p16) that homozygotic deletion of the corresponding gene would be represented by loss of expression. Among 40 sample participants, 17 patients with grade 2 meningiomas were reclassified into a higher grade, as they presented negative p16 staining. However, the immunohistochemical status of CDKN2A did not show a significant difference in RFS ( Figure 2D ).

All 40 samples achieved the institutional quality standard of the NGS study. The tumor purity of FFPE samples ranged from 50 to 90%. The sequencing achieved the 100× coverage of 98.0% (96.2-98.5%) for the target regions with a mean depth of 750 (517-901) ( Supplementary Table 2 ). A total of 160 somatic variants from 66 different genes were identified ( Supplementary Table 3 ). The mutational profiles were highly heterogeneous, and no statistical association between genomic alterations and clinicopathologic features was observed ( Figure 6 ).

SNVs and indels were the most frequent in NF2 (42.5%), which consists of eight frameshift, three stop-gain and five splice site variants. Other mutations in previously documented genes associated with meningiomas, including AKT1 (one missense and one splice site variants) and TRAF (five missense and one splice site variants), were also found. In particular, AKT1 c.49G>A (p.E17K) mutation was recurrent in five cases, most of which concurrently harbored a missense variant in TRAF7. However, there was no clinical relevance in exploratory survival analyses ( Figures 2E, F ).

The prevalence of recurrent variants in meningioma was investigated by comparing the results between pooled data from previous literature and this study ( Table 2 ). Interestingly, the prevalence of mutations in AKT1 and TRAF7 were significantly higher than that in grade 2/3 cases previously reported in external cohorts [AKT1 p.E17K mutation in samples vs. external cohorts (#1-6), p < 0.001; TRAF7 mutations, p = 0.017]. Other known recurrent mutations, such as KLF4 c.1225A>C (p.K409Q) or SMO c.1234C>T (p.L412F), were not detected. A few individual cases presented possible pathogenic variants, PIK3CA c.1633G>A (p.E545K) or SMARCB1 (also known as BAF47, INI1, or SNF5L1) c.1130G>A (p.R377H), while others were tiered as benign or had uncertain significance.