2.1 Study design, patients, data collection

This study included patients diagnosed with MAM, provided by the consultation archive database (1993–2022) from the department of neuropathology at GHU-Paris Psychiatry and Neurosciences, Sainte-Anne Hospital. Epidemiological data (sex and age at diagnosis) and lesion and treatment-related data (location of lesion, preoperative magnetic resonance imaging (MRI) features, duration of symptoms, extent of resection, relapses, and complementary treatments) were retrospectively analyzed. The extent of the initial resection was assessed by early (within 48 h) postoperative MRI. All patients' parents or legal guardians signed informed consent forms before treatment began. We obtained human subjects approval for genetic analyses.

2.2 Central histopathological review

The central pathology review was performed conjointly by two neuropathologists (ATE and PV). Samples were stained with hematoxylin-phloxin-saffron (HPS) according to standard protocol. For each case of MAM, the following pathological features were researched: cell density, pattern (nodules of lesional cells or only infiltration of Virchow Robin spaces), and psammoma bodies. Mitotic count was monitored using 5 high-power fields (HPF), which corresponded to 1.6 mm² on our microscope, and was counted jointly by two neuropathologists in the hotspot area. Meningioma subclassification was performed in accordance with the current WHO 2021 classification. For each MAM + M case, each component was macrodissected and molecularly analyzed (fluorescence in situ hybridization—FISH, DNA sequencing, and DNA-methylation profiling).

2.3 Immunohistochemistry

Unstained 3-μm-thick slides of formalin-fixed, paraffin-embedded (FFPE) tissues were obtained and submitted for immunostaining with an automated stainer (Dako Omnis, Glostrup, Denmark). The following primary antibodies were used: a marker of meningioma SSTR2a (1:200, clone UMB1, Abcam, Cambridge, UK), GFAP (1:200, clone 6F2, Dako, Glostrup, Denmark), INI1 (1:50, clone 25/BAF47, BD-Biosciences, Erembodegem, Belgium), BAP1 (1:100, clone C4, Diagomics, Blagnac, France) which may be lost in a subset of meningiomas and meningiomatosis, and Ki-67 (1:200, clone MIB-1, Dako, Glostrup, Denmark). External positive and negative controls were used for all antibodies and staining.

2.4 FISH analyses

A FISH study was performed on interphase nuclei according to the standard procedures and the manufacturer's instructions. The SMARCB1 and NF2 gene copy numbers were assessed using the following centromeric and locus-specific probes: Z-2178-50 (Zytovision, Bremerhaven, Germany) and FG0003 (Abnova, Taipei, Taiwan). A deletion was defined if more than 30% of nuclei presented no signal for CDKN2A locus (at least 100 nuclei counted for each case). Results were recorded using a DM600 fluorescence imaging microscope (Leica Biosystems, Richmond, IL) fitted with appropriate filters, a charge-coupled device camera, and digital imaging software from Leica (Cytovision, v7.4).

2.5 DNA sequencing

The design of a custom next-generation sequencing (NGS) panel called DRAGON (for the Detection of Relevant Alterations in Genes involved in Oncogenetics by NGS) and marketed by Agilent under the name of SureSelect CD Curie CGP has been developed specifically for the molecular analysis of lesions. It is composed of 571 genes of interest in oncology from diagnostic, prognostic, and molecular therapy points of view, including, the full sequence of NF2, SMARCB1, AKT1 SMO, and PIK3CA, and the hotspot region of interest at the codon 409 of KLF4. The nucleotide sequence (variant calling is performed using Varscan2) as well as the number of copies (deletion and focal amplification) were explored. 50 ng of DNA input was extracted from FFPE lesions and used to prepare the library with the Agilent SureSelect XT-HS preparation kit in accordance with the manufacturer's protocol. The design uses 571 genes and an additional backbone of probes across the whole genome with an average resolution of one probe every 200 Kb. This allowed us to determine a ploidy and an estimated cellularity, together with a genomic profile spanning every chromosome. The copy number profile for each case was estimated using a combination of homemade R scripts and a facets package (v0.6.0) with a sex-specific unmatched-germline control, previously sequenced using the same panel for normalization. Thirty-two DNA were sequenced per 2 × 100 Sp flowcell of the NovaSeq Sequencer (Illumina) to reach an average depth of 1500X and a minimum depth of 100X on the region of interest.

2.6 DNA methylation profiling

DNA was extracted from FFPE tissue samples (a macrodissection of FFPE blocks was performed in cases of low cellular density) using the Qiagen DNeasy Blood & Tissue Kit (Cat NO./ID 69504), according to the manufacturer's instructions. 500 ng of DNA was extracted from each tissue sample. The DNA was sent to the Genotyping facility of the German Cancer Research Center (Heidelberg, Germany). All patient samples were analyzed using either Illumina Infinium Methylation EPIC or HumanMethylation450 BeadChip arrays according to the manufacturer's instructions. Affiliation predictions were obtained from a DNA methylation-based classification web platform for CNS tumors (www.molecularneuropathology.org, version 11b4). Next, a t-distributed stochastic neighbor embedding (t-SNE) analysis was performed and compared with the genome-wide DNA methylation profiles from the brain tumor reference cohort [4], as well as with a series of 30 pediatric meningiomas from our center. Data was generated by the DKFZ Genomics and Proteomics Core Facility (Heidelberg, Germany) as previously described [4]. Dimensionality reduction was then performed using the uniform manifold approximation and projection method (uwot R package) with the following non-default parameters: n_neighbors = 10, spread = 2, min-dist = 0.2.

3.1 Clinical characteristics

The main results are summarized in Table 1. All patients included in this series were pediatric, except one (#2). The median age at diagnosis was 10.0 years (patients' ages ranged from 9 to 28 years) for patients with pure MAM and 9.0 years (ranging from 1 to 12 years) for patients with MAM + M. The male/female sex ratio was 1.3 (4 males and 3 females) and 3.0 (3 males and 1 female), respectively in pure MAM and MAM + M. Lesion locations varied, with the frontal lobes being the most common location (6/11 cases, 55%). All cases, except one (case #7), were revealed by seizures. The remaining patient was followed for a medulloblastoma, non-WNT/non-SHH, treated by surgery, chemotherapy, and craniospinal radiation therapy, and presented during the radiological follow-up (6 years after) an asymptomatic supratentorial MAM. There was no context of neurofibromatosis type 2 in the whole cohort. Another patient (#11) presented in the same surgical time, an atypical rhabdoid and teratoid tumor (AT/RT) component, SMARCB1-deficient, a meningioma, and a MAM. All three lesions were located in the left parietal lobe. There was no known SMARCB1 germline alteration. All patients, except one (case #3) underwent total resection. No patient received adjuvant treatment. Outcome data was available for all patients included in the cohort. Only one (9%) patient (case #3) had progression (after subtotal resection), with a mean progression-free survival of 124.1 months. All patients were alive at the end of follow-up (median overall survival: 193 months, ranging from 41 to 690 months).

3.2 Radiological characteristics

Preoperative imaging was available for 10/11 cases (see Supplementary Table 1 and Figure 1). All lesions affected the cortex and 8/10 cases (80%) were associated with subcortical white matter abnormalities. MRI showed thickening of the cortex, with a T2-weighted low signal in 7/10 cases (70%) and an intermediate signal in 1/10 cases. The T1-weighted signals were slightly higher than the normal cortex in 6/10 cases, and intermediate in 3/10 cases. An existence of cortical calcifications in all the cases with available CT (7/7) may explain this phenomenon. Only one case (#7) exhibited a high cortical signal on T2-weighted images and low signal on T1-weighted images. This case was detected during postoperative medulloblastoma follow-up, without any symptoms. The unique imaging characteristics may be attributed to the early detection of the radiological diagnosis before MAM calcification. Contrast enhancement varied from none to high (similar to choroid plexus). When available, diffusion was not restricted (high or intermediate apparent diffusion coefficient, 8/10 cases). Cerebral blood flow was low on arterial spin labeling perfusion MRI (7/10 cases).

In 80% of cases, white matter involvement manifested as a subcortical T2-weighted high signal; however, in case #11, the T2-weighted low signal and calcifications extended into the white matter toward the lateral ventricles. Three cases presented with enlarged Virchow-Robin spaces. These spaces varied in size and location, with some being subcortical with minor extensions (#11), while others were very large extending from the frontal cortex to the lateral ventricle (#3) or found within the basal ganglia, associated with an infiltration around the proximal middle cerebral artery (#1).

MRI was available for 3/4 MAM + M cases. The cortical MAM presented similarities to others but was connected to an extra-axial nodular mass in the immediate vicinity, without dural thickening. These masses showed imaging features typical of meningioma, with strong contrast enhancement. In one instance (#9, Figure 2), the meningioma was not detected during the radiological diagnosis, it appeared after a follow-up of 3 years with subsequent growth after 2 years, and a gradual radiological progression of the MAM.

3.3 Histopathological and Immunohistochemical characterization

In the MAM cohort, two morphological patterns were observed. First, three lesions (cases #1–3) presented low cellular density, composed of a fibrous infiltration of the Virchow-Robin spaces (MAM with a fibroblastic-like pattern) (Figure 3A–D). The four other cases (cases #4–7) presented an association of fibrous patterns and small nodules of meningothelial cells (MAM with cellular areas) (Figure 4A–D). There were whorls and psammoma bodies, and the meningothelial cells did not present cytological atypia or mitoses. In these cases, we observed a transition between a fibrous component at the surface of the leptomeninges with a fibrous spread in the Virchow-Robin spaces and the formation of meningothelial nodules (Supplementary Figure 1). Necrosis was absent in all MAM cases. Using IHC, INI1, and BAP1 immunoexpressions were retained in all MAM cases. The MIB1 labeling index was weak (less than 1% in all cases) and all lesions expressed SSTR2a.

In the MAM + M cohort, all cases presented a morphological spectrum from a paucicellar fibrous infiltration of the Virchow-Robin spaces to the formation of a circumscribed meningothelial proliferation, corresponding to a meningioma. All meningiomas were of a meningothelial subtype. Three of them presented 4, 5, and 6 mitoses per 1.6 mm² (grade 2), whereas the remaining case was grade 1. There was no brain parenchyma invasion (no irregular infiltration or protrusion of tumor cells into the underlying GFAP-positive parenchyma). Using IHC, INI1, and BAP1 immunoexpressions were retained in all MAM + M cases. The MIB1 labeling index was weak in the MAM component (less than 1% in all cases) whereas it varied from 1% to 8% in the meningioma component. In all lesions, both components (MAM and meningiomas) expressed SSTR2a.

3.4 Molecular results

DNA-sequencing analyses failed to reveal any mutation NF2, AKT1, KLF4, SMO, or PIK3CA genes in the whole cohort (pure MAM, and in both components of MAM + M). A hemizygous deletion of chromosome 22q was detected in 2/7 pure MAM (cases #5 and 7) and in all MAM + M (presence of the deletion in both components) (Figure 4E, F). An inactivating nonsense KMT2C mutation with an allele frequency of 4.38% was evidenced in one pure MAM (case #1). Concerning the case #11, the hemizygous deletion of chromosome 22q included SMARCB1 gene, and no alteration, no nucleotidic mutation or structural variant such as deletion, of SMARCB1 gene, was detected in the MAM and the meningioma (whereas the AT/RT component harbored a homozygous deletion of SMARCB1 gene). FISH analyses for NF2 and SMARCB1 failed to reveal any deletion in MAM cases which did not present a deletion of chromosome 22q on CNV (Figure 3E, F). MAM with a cellular pattern presented additional CNV alterations such as losses of chromosomal material on chromosome 1q (case #7) and chromosome 13q (case #5). MAM associated to meningiomas presented additional CNV alterations such as losses of chromosomal material on chromosome 1q (case #9) and chromosome 2 (case #11), or gain on chromosome 3p (case #8) and chromosome 5 (case #10). These alterations were also found in the meningioma component except for the gain of chromosome 5 of the case #10 but, interestingly, a focal amplification of EGFR appeared in the meningioma component in this case (confirmed by DNA-methylation profiling). The size of the loss of chromosome 22q differ from one case to another. For one case (case #7), the entire chromosome 22 was lost, for some cases, the chromosome 22q loss start from chromosome band 22q11.22 (case #9) or 22q11.23 (case #10) to the end of the chromosome 22q (hence encompassing both SMARCB1 and NF2), and finally for one interesting case (case #8), the deletion started from chromosome band 22q12.1 to the end of chromosome 22q hence encompassing NF2 but not SMARCB1. This last result hence defined a minimal deleted region that include NF2 gene that seems the most important for MAM development but not SMARCB1.

According to the DNA methylation-based classification and the DKFZ Classifier (version 12.5), none of the lesions were classifiable (calibrated scores for DNA methylation class <0.9), except one meningioma (case #11, which classified as a benign meningioma, with a calibrated score of 0.99). A t-SNE analysis was performed to compare the genome-wide DNA methylation profiles of pure MAM, MAM + M, pediatric meningiomas, and the different methylation classes (MC) of adult meningiomas. Most cases (three pure MAM and the four pairs of MAM + M) clustered within close proximity to the pediatric meningioma methylation class (Figure 5 and Supplementary Figure 2). Interestingly, meningiomas associated with MAM clustered within the MC of pediatric meningiomas. Moreover, MAM (pure or associated with meningiomas) formed a distinct cluster from other meningioma MC. The four remaining cases (#1, 2, 3, and 6) clustered in close vicinity with control hemispheric tissue.