cIMPACT-NOW update 8: Clarifications on molecular risk parameters and recommendations for WHO grading of meningiomas

Abstract

Meningiomas represent the most common of all primary CNS tumors. The recent fifth edition of the WHO classification of central nervous systems tumors (CNS5) introduced some significant changes. For the first time, molecular biomarkers (oncogenic variants in the TERT promoter and homozygous deletion of CDKN2A/B) were introduced as independent diagnostic criteria for meningioma CNS WHO grade 3 (Figure 1A). Additionally, the grading of rhabdoid and papillary meningiomas was amended from assignment to CNS WHO grade 3 on the basis of rhabdoid or papillary histology alone towards stratification into CNS WHO grade 1, 2, or 3 based on the presence or absence of histological criteria for CNS WHO grade 2 (atypical) or CNS WHO grade 3 (anaplastic) meningiomas. There were also substantial insights into meningioma biology, yielding alterations or “molecular subtypes” that may be further leveraged into clinically useful diagnostic categories. Such data were not mature enough yet to be incorporated into CNS5. Since its publication, however, the body of literature on novel diagnostic and prognostic markers in meningioma has substantially increased, and predictive markers have emerged. In addition, common questions related to the practical application of CNS5 criteria necessitate clarification to ensure standardized application in routine diagnosis.

Figure 1.

(A) CNS5 grading scheme with the molecular markers CDKN2A/CDKN2B and oncogenic variants in the TERT promoter as independent criteria for CNS WHO grade 3 independent of morphology. Select cases with CNS WHO grade 1 or 2, morphology, may benefit from additional molecular analyses for more precise risk assessment (Figure 2). (B) Overview of improved risk assessment approaches to meningioma with CNS WHO grade 1 or 2, morphology. * for 1p, deletions of >5 % have shown prognostic significance; for other chromosomal arms, insufficient data is available on whether a smaller than whole-arm deletion is also indicative of prognosis. **Specifically for brain-invasive otherwise benign (BIOB) meningioma, this update proposes to assign CNS WHO grade based on molecular markers.

Open in new tabDownload slide

The consortium to inform molecular and practical approaches to CNS tumor taxonomy-not official WHO (cIMPACT-NOW) was formed in late 2016 by a group of neuropathology and neuro-oncology experts to provide practical recommendations (published as cIMPACT-NOW updates) to improve the diagnosis and classification of CNS tumors, in advance of the publication of a new WHO Classification of CNS tumors. Between the revised fourth (2016) and the fifth (2021) edition of this classification, 7 cIMPACT-NOW updates were published.¹ In 2022, the cIMPACT-NOW steering committee resumed its activities and directed a working group of neuropathologists and clinical advisors to address ambiguities and advances related to the classification and grading of meningiomas. The following summarizes our current understanding and provides practice recommendations, which could potentially be considered for adoption by the WHO in its next edition (CNS6).

Clarification on Mitotic Count

Mitotic activity is one of the key criteria used for assigning meningioma grade. As described in 1997 and 1999 and adopted by the WHO 2000 (third edition) CNS tumor classification, a mitotic count of ≥4 (up to 19) and ≥20 mitoses in 10 high power fields (HPFs) of 0.16 mm² each has long been used for labeling the tumor as atypical (CNS WHO grade 2, in the third edition “II”) or anaplastic/malignant (CNS WHO grade 3/III) meningioma, respectively.^2,3 Mitoses should be identified by careful review of good quality H&E-stained sections (3–5 microns in thickness), with counts obtained in consecutive HPFs (×400, a combination of ×40 objective lens with a ×10 eyepiece lens) in the area with the highest mitotic activity (“hotspot”). Importantly, as the objective lens of most current microscopes results in larger HPF areas (generally 0.23–0.24 mm²), adjustments need to be made. As a practical guideline, 10 HPFs of 0.16 mm² correspond approximately to 7 HPFs of 0.23–0.24 mm². Meanwhile, in order to standardize counts and in preparation for the transition to digital pathology, mitotic count thresholds used in the WHO CNS5 classification are now expressed as number of mitoses per mm² rather than per HPF: ≥2.5 and ≥12.5 mitoses/mm² for CNS WHO grades 2 and 3 meningiomas, respectively.⁴ Of note, the evaluation of 1.6 mm² is still required and extrapolation from smaller areas should not be the basis for grading. Although counts using digital microscopy may differ from the manual approach on glass slides, some pathologists advocate circling mitoses on a virtual slide and then outlining the 1.6-mm² region containing the highest number of mitoses (personal communication from several colleagues). Computational algorithms to assess mitotic count in digitized slides are becoming increasingly reliable and may allow for improved, more reproducible assessments in the not-too-distant future: They provide the basis to generate large, standardized, and shareable datasets that are not subject to variations between microscopes. In addition, these algorithms may aid in reducing observer bias by providing a “second opinion,” which, however, may introduce new biases of their own. Hence, comparative studies will need to be performed to avoid potential grade migration.⁴

Clarification on Brain Invasion

Brain invasion is defined in CNS5⁵ as the infiltration of tumor cells through the leptomeninges into the subjacent brain parenchyma. As such, extension along the Virchow–Robin spaces does not qualify for brain invasion because of an intact pia mater in this scenario. Brain invasion has been shown in some studies to portend aggressive behavior with shorter recurrence-free survival,^3,6 but other studies do not support this,^5,7 as recently summarized.⁸ Importantly, these tumors do not necessarily have similar cytogenetic alterations to those typically found in other histologically defined atypical meningiomas.⁹ This raises the unresolved question of whether these tumors have a unique epigenetic signature that could help with their integrated diagnosis in a more robust way. Additional challenges and questions exist, including the prognostic significance of the extent of invasion and the variable interpretations of this feature, given that it has been reported in anywhere from 4% to a third of all meningiomas.⁸ As in CNS5, we recommend strict criteria to avoid overcalling this feature. The association with anatomic location also needs to be clarified—specifically determining whether CNS parenchymal invasion in spinal cord meningiomas has the same prognostic and biological significance as in their intracranial counterparts. A similar question can be raised about the pituitary gland or peripheral/cranial nerve invasion. Additionally, tumor sampling is critical in assessing the presence and degree of invasion.^10,11 Optimal histologic assessment in cauterized/fragmented (eg, obtained via CUSA/SONOPET) specimens, which constitute a fair number of specimens, remains a challenge. Similarly, interpretation of GFAP immunostaining can be challenging as both false-positive and false-negative instances can occur, particularly since it is not always obvious exactly where the thin pial surface lies. Addressing the clinical significance (and impact on grading) of brain invasion will require careful analysis from existing tumor repositories that have thorough clinical annotation. However, many of these longitudinal studies lack consistency and standardization in patient management and are often limited to a single sample or multiple samples without orientation. Newer neurosurgical technologies such as intraoperative fluorescence tools and confocal endomicroscopy may improve the quality of the surgical specimen, thereby enabling a more accurate assessment of brain invasion.¹² Generally, the neuropathological evaluation is subject to sampling bias of the tissue submitted, highlighting the importance of providing a comprehensive specimen. Finally, little is known about the impact of brain invasion on the prognosis of meningiomas in the pediatric age group where the developing brain and distinct molecular characteristics impact tumor biology and potential prognosis. In those cases, perivascular spread along Virchow–Robin spaces is also more common, representing a diagnostic pitfall that is often misinterpreted as true brain invasion. Collectively, suspected brain invasion needs to be thoroughly evaluated to avoid overcalling. Figure 2 illustrates the challenges in the interpretation of possible brain invasion with GFAP stains clarifying a case of brain-invasive otherwise benign (BIOB) meningioma with true invasion (left) versus perivascular spread along Virchow–Robin spaces without true invasion beyond the pia (right).

Figure 2.

GFAP immunostains in a brain-invasive otherwise benign meningioma with entrapped reactive CNS tissue within the main tumor mass (left) and a meningioma with perivascular spread along Virchow–Robin spaces but no true parenchymal invasion with tumor extending beyond the pial barrier (right). Scale bar app. 200 µm.

Open in new tabDownload slide

While a definite correlation between brain invasion and specific molecular alterations remains elusive, evidence suggests that molecular data (eg, copy-number variations [CNVs], DNA methylation profiling [DNAMP]) can identify lower- or higher-risk cases among morphologically ambiguous samples. Hence, in cases with possible but not definite brain invasion, additional molecular testing is advised and should also be incorporated into the grading of a brain-invasive meningioma otherwise benign (sometimes referred to as a BIOB) meningioma. According to this novel recommendation, a BIOB meningioma would be assigned CNS WHO grade 2 in case of 1p deletion, in line with the new guidance to assign all cases with 1p deletion (in the presence of 22q deletion/NF2 alteration) to CNS WHO grade 2 (Figure 1B). Specifically, for BIOB meningiomas, other molecular markers of high-risk (DNA methylation, RNA subgroups) may also qualify as a criterion for CNS WHO grade 2 designation (or CNS WHO grade 3 in case of CDKN2A/CDKN2B homozygous deletion, TERT promoter oncogenic variant, or possible upcoming other molecular criteria for CNS WHO grade 3). A case with the same morphological findings but lacking molecular high-risk features is graded as CNS WHO grade 1. Consequently, a BIOB meningioma in which no testing could be performed should be termed “meningioma, not otherwise specified” with the CNS WHO grade left off.

TERT Promoter and CDKN2A/B

Oncogenic variants in the TERT promoter and CDKN2A/CDKN2B deletion are genetic alterations observed in a small (<5%) subset of meningiomas, typically those with oncogenic variants in NF2 and/or loss of chromosome 22q. These alterations are associated with a high risk of recurrence and dramatically shorter progression-free intervals, regardless of other histological features (TERT promoter^13–15; CDKN2A/B^16–20). Thus, they are considered significant independent prognostic factors in meningiomas, warranting a CNS WHO grade 3 designation in the latest scheme. However, no guidelines have been established before in which cases these markers should be tested.

Oncogenic variants in the TERT and CDKN2A/CDKN2B deletions should be tested for in CNS WHO grade 2 (atypical) and grade 3 (anaplastic) meningiomas to identify patients at increased risk of recurrence and progression. TERT promoter oncogenic variants have also been detected in rare CNS WHO grade 1 meningiomas,¹⁴ suggesting the potential value of comprehensive genomic profiling for all meningiomas regardless of histologic grade. However, given that the yield is low overall, each center must decide for itself which cases will be routinely tested. One possible approach for when to add genetic/epigenetic testing in general is offered in Table 1. Testing methods for the TERT promoter include targeted DNA sequencing covering the 2 hotspots, C228T and C250T, as well as more comprehensive sequencing approaches that adequately cover this region. For the association of CDKN2A/B deletion with prognosis, older studies emphasized the need for homozygous deletion—hence, the requirement for homozygous deletion in the WHO 2021 criteria—whereas meningiomas with either homozygous or hemizygous loss exhibit similar recurrence rates in more recent studies.^21,22 The case numbers with hemizygous CDKN2A/B losses in the recent studies are still too small to draw firm conclusions on the role of hemizygous deletion for meningioma grading. The same applies to other oncogenic variants in CDKN2A and CDKN2C detected in rare cases of higher-grade meningiomas.

DNA fluorescence or chromogenic in situ hybridization (FISH/CISH) is a widely available method to assess the copy number of this genomic locus. CISH has the advantage of allowing for concurrent assessment of morphology during the evaluation of the probe binding. Due to the large size of CISH/FISH probes, however, small deletions may occasionally go undetected. Additional methods to assess copy number include droplet digital PCR and computation from data acquired using various next-generation DNA sequencing techniques, which offer more comprehensive genomic profiling of the meningioma. Additionally, array-based assays for copy number or DNA methylation arrays can be used to evaluate deletions at this locus. An immunostain for p16 may also be a reasonable surrogate marker when there is a complete loss of expression, though this would still need further molecular confirmation, since some CNS WHO grade 1 meningiomas, in particular, express very little protein which may thus mimic loss of expression secondary to gene loss.²³ Loss of cytoplasmic MTAP (methylthioadenosine phosphorylase) immunohistochemical staining may also serve as a potentially promising surrogate due to the close proximity of the respective gene to CDKN2A/B and an initial immunohistochemical study suggesting its utility.²⁴ Additional studies on MTAP and other surrogate markers may reveal new immunohistomestry-based approaches to assess CDKN2A/B status. This genomically closely located marker and proteins involved in the respective pathways may reveal new immunohistochemistry-based approaches. In certain instances, rare events like TERT activation through gene rearrangements and amplifications can be inferred from DNA and RNA sequencing assays and high-resolution copy-number analysis.¹⁵ Considering that oncogenic variants in the TERT promoter and CDKN2A/CDKN2B deletions can arise during progression, it is advisable to prioritize the selection of tissue blocks or regions for DNA extraction based on the most malignant-appearing and proliferative regions.^13,25 Scenarios in which testing for TERT and CDKN2A/CDKN2B is advised are given in Table 1.

H3K27me3

Methylation and acetylation of histone proteins are primary epigenetic modifications that substantially impact transcriptional regulation.²⁶ In the central and peripheral nervous system, H3K27me3 loss is associated with clinically aggressive behavior in posterior fossa ependymomas, peripheral nerve sheath tumors, and a subset of diffuse gliomas.^27–29

Immunohistochemistry (IHC) for H3K27me3 has been investigated within specific cohorts of meningioma to determine if lost expression might serve as a prognostic marker. All studies have concluded that loss of H3K27me3 is more frequent with increasing grade, with the largest cohort showing loss in 3% of grade 1, 10% of grade 2, and 18% of grade 3 meningiomas.^30–33 Multiple investigations have demonstrated that loss of H3K27me3 is associated with a greater risk of meningioma recurrence on multivariable analysis and that its loss is especially predictive of recurrence in grade 2 tumors.^30–33 Most have concluded that loss of H3K27me3 does not stratify risk among grade 3 tumors. However, one investigation that included a relatively large cohort (n = 66) of CNS WHO grade 3 meningiomas concluded that loss of H3K27me3 was independently associated with shorter overall survival.³⁴ Others have focused exclusively on recurrent meningiomas and demonstrated that loss of H3K27me3 is more frequent in recurrent than in primary meningiomas for both grades 1 and 2 tumors and is also predictive of shorter time to re-recurrence.³⁵

Overall, the literature indicates that loss of H3K27me3 increases in frequency with higher grade and at recurrence. Lost expression is predictive of recurrence in grade 2 meningiomas, while further study is needed to determine clinical relevance in grades 1 and 3 meningiomas. The retrospective studies performed thus far have had variable inclusion criteria, especially related to tumor grade, making direct comparison of the frequency of H3K27me3 loss among the cohorts challenging. However, the reported overall frequency of H3K27me3 loss varied widely from 5% to 34%.³⁶ Interpretative challenges also arise from ambiguous or patchy immunohistochemistry staining, which has been noted in 5%–7% of cases.^30,31 The relevance of weak staining or focal loss is currently not clear. Hence, an interpretation of lost expression should only be rendered when tumor cells are clearly devoid of staining and internal controls (eg, endothelial cells) are positive. Beyond artificial staining patterns, true regional intratumoral staining heterogeneity correlating with proliferation and other morphological features has been reported. The relevance of such regionally heterogeneous staining of tumor cells is not yet clearly established, although it may indicate a more aggressive subclone. Overall, the literature supports the use of H3K27me3 as a prognostic marker of recurrence among CNS WHO grade 2 meningiomas. However, these two subgroups of CNS WHO grade 2 meningiomas based on H3K27me3 stratification remain distinct from CNS WHO grade 1 and 3 meningiomas.³⁷ Hence, H3K27me3 status in CNS WHO grade 2 meningioma offers prognostic information but does not currently inform grading.

BAP1, PBRM1, and SMARCE1 Alterations for Meningioma Subtyping and Grading

The existing literature highlights the connection between BAP1, PBRM1, and SMARCE1 inactivation and specific meningioma subtypes, as well as their impact on tumor behavior and prognosis. BAP1 inactivation is frequently observed in meningiomas with rhabdoid and/or papillary morphology.^38–40 PBRM1 inactivation has been linked to meningiomas displaying papillary features.⁴⁰ Notably, BAP1 and PBRM1 are located closely on chromosome 3p21.1, within a distance of 150 kb. Although usually inactivated independently, concurrent inactivation of BAP1 and PBRM1 can occur in the same tumor,⁴¹ likely contributing to the histologic diversity observed in these tumors. Rhabdoid and/or papillary features can be variable ranging from highly prominent to only focal or absent. SMARCE1 inactivation is strongly associated with clear cell meningioma, which primarily occurs in the cerebellopontine angle and spine, predominantly affecting younger individuals, including children and young adults.^42–45 Additionally, these tumors cluster away from other meningioma subtypes on DNA methylation profiling, providing further evidence that they are unique.⁴⁶ While BAP1-mutant meningiomas and SMARCE1-mutant meningiomas can occur sporadically, they can also arise in patients with pathogenic germline variants. Therefore, a relevant family history and/or young age of onset should prompt consideration of genetic counseling and germline testing.

Losses of BAP1, PBRM1, and SMARCE1 can be detected through IHC. Genetic evaluation requires both DNA sequencing to identify somatic (or germline) variants and copy-number analysis to identify deletions, including focal and intragenic deletions. SMARCE1-mutant clear cell meningiomas demonstrate aggressive behavior, including recurrence and occasional cerebrospinal fluid seeding, leading to their designation as CNS WHO grade 2 tumors. Since this appears to be a unique meningioma subtype, SMARCE1 inactivation is highly useful in establishing the diagnosis and grade. Whether or not other genetic alterations may be encountered in rare examples remains uncertain, though a designation of “meningioma with focal clear cell features” should not be rendered merely because of focal clear cytoplasm or increased glycogen within an otherwise classic meningioma subtype. The assignment of a specific CNS WHO grade for meningiomas with BAP1 and/or PBRM1 alteration remains uncertain. A study reported that over 80% of BAP1-mutant meningiomas and nearly 90% of PBRM1-mutant meningiomas were initially classified as CNS WHO grades 2 and 3 meningiomas before genetic testing.⁴¹ Additionally, BAP1 loss has been linked to a shorter time to recurrence in a small cohort.³⁸ These findings suggest that tumors with inactivation in these genes may be innately higher grade. However, further comprehensive studies with larger sample sizes and long-term patient follow-up are necessary to confirm whether elevated recurrence rates are associated with tumors that otherwise look benign.

TRAF7, AKT1, KLF4, SMO, PIK3CA, and POLR2A Alterations for Meningioma Grading

While NF2 alterations are the most frequent oncogenic variants in meningioma, another set of recurrent somatic oncogenic variants that are mutually exclusive with NF2, including TRAF7, AKT1, KLF4, SMO, PIK3CA, and POLR2A, is associated with tumor location, morphology, and clinical outcomes.^47–52 Oncogenic KLF4 variants typically co-occur with oncogenic TRAF7 variants (mostly in the secretory variant), while a lower fraction of AKT1-mutant cases have accompanying TRAF7 alterations, and rarely oncogenic TRAF7 variants occur as isolated events.^{47,50,53–55}

Based on the prominent biological differences between NF2 and non-NF2 driver alterations, some have suggested dividing meningiomas into NF2 (harboring NF2 variants), TRAKLS (with TRAF7, AKT1, KLF4, and SMO alterations), and not otherwise classified (NOS) molecular tumor types.⁵⁶AKT1, SMO, and PIK3CA alterations are enriched in the skull base,^47,52,57 with SMO being predominant in the olfactory groove location.⁵⁸ The so-called TRAKLS genotype is highly associated with histologic grade 1 meningiomas. However, SMO and PIK3CA oncogenic variants are also noted in grade 2, albeit uncommonly.^52,56,59 On univariable analysis, individual gene alterations of TRAF7 and KLF4, as well as the TRAKLS genotype, have been shown to be statistically associated with improved progression-free survival (PFS) compared to NF2-altered meningiomas.^56,60 In outcome analyses, PFS was considered superior to the “average” (including NF2-mutant) CNS WHO grade 1 tumors. However, none of the individual alterations or the TRAKLS genotype remained statistically significant on multivariable analysis that corrected for age, biological sex, methylation class, and grade.⁶⁰ Another study confirmed the favorable outcome of KLF4-mutant cases (regardless of TRAF7 status) but showed a significantly higher risk of recurrence within 60 months for cases with isolated TRAF7 alterations.⁵³

In an integrated, multiplatform analysis of meningiomas that included CNVs, RNA expression and DNAMP, TRAF7, AKT1, KLF4, SMO, and POLR2A cases were found exclusively in one integrated molecular cluster (MG2) that was similarly associated with a favorable prognosis.⁶¹ The meningiomas with these alterations accounted for fewer than half of the tumors in MG2 and the independent prognostic significance of individual or grouped oncogenic variants was not distinguished from other factors that may influence molecular clustering.

Another methylation-based analysis of meningiomas found that one specific methylation class (MC ben-2) had very few copy-number alterations and was enriched for oncogenic variants in TRAF7, AKT1, KLF4, and SMO. Alterations of this group were also seen in MC ben-3 but at a much lower frequency. PFS times for both of these methylation classes were superior to those of “average” (irrespective of molecular markers) CNS WHO grade 1 meningioma in this analysis. Additionally, the methylation class, but not individual or grouped oncogenic variants, was the strongest prognostic factor.⁶²

The prognostic significance of oncogenic variants in these genes occurring individually or in combination is evident on univariable analysis but has not been demonstrated on multivariable analysis to be prognostically independent of grade, molecular class, or methylation class. Therefore, individual oncogenic variants in this class cannot be used as strict grading criteria but can be used to inform clinical decision-making in challenging cases. Clinical trials to target a subset of these alterations with precision medicine approaches for patients with recurrent or progressive meningiomas are ongoing (NCT02523014); if effective, these events may direct treatment decisions in the future.⁶³

Copy-Number Alterations Other Than CDKN2A/CDKN2B

CNVs are well documented in meningiomas, especially in higher-grade tumors.^64–67 Comprehensive genomic analyses have demonstrated that, depending on the cohort and threshold for defining CNV size, approximately one third of all meningiomas have no observable CNVs, while more than 50% harbor at least monosomy of chromosome 22.^22,68,69 In line with the monosomy of chromosome 22, those meningiomas with CNVs are mostly NF2-mutant.^{22,47,50,61,62,69} In higher-grade meningiomas, monosomy 22 occurs in combination with other recurrent CNVs, including losses on 1p, 6p/q, 10q, 14q, and 18p/q, and less frequently on 2p/q, 3p, 4p/q, 7p/q, 8p/q, and 13p/q.^22,69–73 CNVs can accumulate with the progression of NF2-mutant cases, and NF2-mutant meningiomas tend to harbor the greatest abundance of genomic instability.⁷⁰ Especially higher-grade meningiomas may also exhibit hemizygous or homozygous deletions of CDKN2A and/or CDKN2B, which can be focal or occur in conjunction with arm-level loss on chromosome 9p. Gains of chromosomal arms are less frequent and may not always indicate aggressive behavior: High-grade cases may show a gain of 1q, whereas gain of whole chromosomes (mostly 5 and 7) is a feature observed in biologically indolent angiomatous, metaplastic, and microcystic meningiomas.^{22,69,72,74,75}

Although the specific CNVs can vary substantially among individual tumors, loss of 1p is the most frequently observed in the context of adverse clinical outcomes. Loss of 1p occurs up to twice as often as losses on 14q, 6p/q, and 18p/q, and nearly 4 times as often as losses on 10p/q.^22,69–73 Thus, oncogenic phylogeny models propose 1p loss as the first CNV after monosomy 22q in NF2-mutant, high-grade meningiomas,^69,76,77 with rare cases that show 1p loss in the absence of 22q loss (and hence no reliable data on this subset).

Given the strong association between CNVs and higher-grade meningiomas and the increased risk of recurrence, the assessment of CNVs presents an appealing approach to predicting clinical behavior. Several studies have proposed grading based on CNVs alone, or integrating CNVs with other findings, such as morphology, clinical features, radiologic data, sequence variant status, and/or DNA methylation, with the first attempts to do so reaching back more than 2 decades.^{22,69–73,78} Although the exact hierarchy of relevant CNVs differs among studies and may depend on cohort composition and statistical approach, 1p deletion consistently confers an increased risk of recurrence across studies. 1p loss alone conferred a significantly higher risk of recurrence compared to tumors with intact 1p in a recent comprehensive study.²² Monosomy of 1p remained an independent marker when adjusting a large cohort for methylation subgroups, and also in a multi-center prospective trial with standardized radiation treatment.⁷⁹ Importantly, the outcome of 1p-deleted cases did not differ from the average of CNS WHO grade 2 cases in another study.⁶⁹ The prognostic relevance of 1p alterations occurred in the presence of any segmental loss above 5% of the chromosomal arm in one study,⁸⁰ which is important since 1p is most prone to subtotal alterations⁶⁹ among all chromosomes in meningiomas. Upcoming studies may further refine this threshold.

One of the appealing aspects of using CNVs in clinical practice is the flexibility of evaluating these events through various analytical approaches. These include widely available methods in clinical laboratories like targeted FISH/CISH which has limitations depending on probe binding site, as well as the utilization of copy-number arrays, methylation profiling arrays, and the inference of CNVs from targeted next-generation sequencing data. Thus, similar information can be obtained using widely available techniques, without a reliance on proprietary technologies and proprietary algorithms. Of note, 1p and other CNVs are subject to intratumoral heterogeneity, hence sampling the morphologically most aggressive area (eg, fragment with the highest mitotic count or proliferation index) is advisable.⁸¹

In summary, based on the available literature, we propose that meningioma with complete or segmental 1p deletion and concurrent monosomy of 22q and/or NF2 oncogenic variants should be graded as at least CNS WHO grade 2, even in histologically otherwise benign appearing meningiomas.

DNA Methylation (DNAMP) Profiling and RNA-Based Analyses

The DNAMP of tissues provides information relating to cell type and differentiation status of its major constituents. Meningiomas are highly suited to methylation-based tumor typing due to their solid growth pattern and high tumor fractions. However, an exception is evident for the rare lymphoplasmacyte-rich meningioma. In contrast, the angiomatous meningioma with usually hyalinized blood vessels as the dominant component rarely poses a problem because of the low number of nuclei in the vascular compartment. Sizable cohorts of meningiomas have been subjected to DNAMP and multiple classifiers have been developed.^70,82–85 Further developments have led to the integration of histological features, CNVs, gene expression profiles, and oncogenic DNA variants together with methylation data^61,69 into multi-layered, integrated prognostic scores. In general, DNAMP and integrated molecular classification systems for meningioma grouping provide similar results across unsupervised systems, which appear to provide different group assignments than supervised systems that are trained on (or incorporate) clinical endpoints for tumor classification.⁸⁶ Regardless, all molecular grouping systems incorporating DNAMP provide valuable prognostic information for meningioma outcomes.

All types of methylation data acquisition are compatible with the principles of these classifiers. In practice, ready-to-use solutions are available for methylation data generated on the Illumina platform, while employing data from methylation sequencing or the Nanopore platforms requires adaptation of these scripts. Gene expression-based meningioma classification also faces technical challenges associated with RNA preservation in frozen versus formalin-fixed paraffin-embedded tissues, but when optimized, classification using RNA sequencing closely approximates the results of DNA methylation grouping.^86,87 RNA hybridization techniques that do not rely on transcript polymerization for amplification and sequencing are robust across tissue preservation conditions,⁸⁸ and may provide more generalizable prognostic or predictive systems for meningiomas. More recent work also suggests an RNA-based marker signature to stratify more accurately for risk of recurrence and even to predict response to radiotherapy.⁸⁹

All major DNAMP-based proposed classification schemes share a largely benign subgroup with NF2-wildtype meningiomas (“Merlin-intact,” “benign-2”), a mostly benign subgroup with isolated NF2/22q-alteration lacking any other alterations (“immune-enriched/activated,” or in systems with higher granularity “benign-1”) and a highly aggressive subgroup with NF2/22q-alteration and several other changes on CNV (1p, 10, 14, CDKN2A/B) or sequence (TERT promoter) level (“hypermitotic,” “proliferative,” “malignant”). In addition, some schemes propose finer granularity for the NF2/22q-altered cases (“intermediate-A/B,” “hypermetabolic”). A potentially large benefit in prediction is for patients with CNS WHO grade 2 meningiomas and those at the grade 1/2 interface, where there is known to be great clinical variability. The integrative evaluation of CNVs and specific methylation patterns is more likely to assist in distinguishing between patients experiencing early and later tumor recurrence.^22,89 While promising, caution is suggested as most studies investigating molecular classification of meningiomas have been retrospective. Nonetheless, these classifiers are being incorporated into clinical trials under development, and early data are encouraging for prospective validation.^69,79,90 As such, DNAMP may further inform patient management for CNS WHO grade 2 and borderline grade 1/2 meningiomas. Importantly, however, no definite criteria are currently available for adjusting meningioma grade up or down when DNAMP is discordant with histopathologic criteria. Based on the available evidence that methylation and CNV patterns outperform current grading practices on morphological criteria in several studies, these data should nonetheless be leveraged for grading when available. Since limited independent prospective studies on DNAMP alone are available, we propose to rely only on a concordant DNAMP and CNV result (eg, both indicating a higher grade in a morphological CNS WHO grade 1 case or vice versa). If both DNAMP and CNVs unequivocally indicate a clinical behavior different from the morphological criteria, the meningioma should for now be graded as “CNS WHO grade 1, with molecular data suggestive of higher risk of recurrence” or “CNS WHO grade 2, molecular data suggestive of lower risk of recurrence.” In either instance, this should be accompanied by a comment clarifying the basis of this interpretation for the specific case and with regard to the evidence summarized here. Of note, this should not be interpreted as a requirement to perform molecular analyses on every meningioma case at this time, but rather as guidance on how such information could be useful. Subsequent updates or volumes of the WHO classification may translate this into definite criteria. Recommendations for when molecular analyses are advisable are summarized in Table 1.

A practical approach to the decision of whether to employ DNAMP or RNA-based analyses for meningioma patients depends on downstream consequences. In all instances where postsurgical therapy strategies remain uncertain, additional diagnostic studies can further aid this critical decision. The costs of radio- or chemotherapy are far higher than molecular analyses and the adverse effects of therapy to patients at low risk of recurrence may be severe. Similarly, patients with potentially underdiagnosed high-grade meningiomas should still have the opportunity of beneficial postsurgical therapy. Hence, the actual role of these markers in the treatment decision needs to be further investigated. For centers where it is not feasible to run DNAMP (or RNA-based, once validated classifications are available) analyses on each meningioma, we suggest the tabulated indications for DNAMP (and potentially upcoming RNA-based approaches) in meningiomas (Table 1). An overview of how to incorporate these data into a diagnostic report is depicted in Figure 1B. Classifiers for typing and grading meningiomas based on DNAMP can be accessed at http://www.molecularneuropathology.org, https://william-c-chen.shinyapps.io/MeninMethylClassApp/, https://william-c-chen.shinyapps.io/MeningiomaGeneExpressionBiomarker/. However, none of them has been certified as a diagnostic tool and all thus need local independent validation.

Summary of Recommendations

Recent advances in our understanding of meningioma on a molecular level can now inform diagnostics and especially grading. The morphological criteria for meningioma build on the experience of the past century and still represent the central pillar for specimen work-up. Intriguingly, many subtypes initially described with morphological criteria decades later have proved to have distinct molecular features. On the basis of morphology assessment and the current WHO classification, this cIMPACT-NOW update adds the following recommendations:

• (1) Brain-invasive but otherwise morphologically benign (BIOB) meningiomas should not be graded before molecular data are obtained that further inform grading;

• (2) Cases with borderline morphological grading criteria should undergo additional molecular testing; in addition, this update suggests possible molecular testing in a variety of other scenarios to increase diagnostic precision (Table 1);

• (3) Histologically low-grade or borderline meningiomas with chromosome 1p deletion and concurrent 22q deletion/NF2 oncogenic variant should be assigned to “CNS WHO grade 2,” unless further molecular markers indicating CNS WHO grade 3 are present. Testing for 1p status is currently not recommended for all otherwise low-grade cases, but those scenarios are stated in Table 1.

• (4) Some areas of inquiry do not have available data that is currently sufficient for recommendations, most notably for H3K27me3, single gene variants often but not consistently associated with CNS WHO grade 1, and hemizygous CDKN2A/B deletion status.

It is important to note that these recommendations only pertain to sporadic meningiomas in adults since pediatric, radiation-associated, and neurofibromatosis type 2-related meningiomas are poorly represented in currently available studies. Hence, more comprehensive studies focusing on these cases are warranted. Likewise, while some morphological subtypes are meanwhile matched with largely overlapping molecular groups (eg, clear cell/SMARCE1, secretory/KLF4/TRAF7), the relevance of subtypes neither showing consistent alterations nor a clear-cut outcome (eg, chordoid) will require critical reappraisal. These recommendations pertain to tissue-based prognostic criteria, but the extent of resection is also a well-established prognostic factor that naturally informs clinical management as well. Attempts to combine histopathological, molecular, and extent of resection data have been performed in the past and warrant further studies.

The field of meningioma prognostication is advancing quickly, and it is hoped that these recommendations provide useful interim guidance for neuropathologists between CNS5 and CNS6. It is further hoped that such cIMPACT-NOW recommendations may lead to more uniform implementation of emerging findings to guide the care of patients with these common tumors.

Conflict of interest statement

A.v.D. reports being co-founder and shareholder of Heidelberg Epignostix GmbH, royalties for IDH1 R132H mutation-specific antibody from DIANOVA GmbH, and royalties for BRAF V600E mutation-specific antibody from Roche Ventana. D.C. reports being co-founder and shareholder of Heidelberg Epignostix GmbH, royalties for IDH1 R132H mutation-specific antibody from DIANOVA GmbH, and royalties for BRAF V600E mutation-specific antibody from Roche Ventana, and research funding from NOVOCURE. D.R.R. reports honoraria for consulting or lectures from Gamma Tile, Tipping Point, Nerviano Medical Science, Boston Scientific, Duke University, Memorial Sloan Kettering Cancer Center, St Jude, and serves as SNO Radiation Track Co-Chair. F.S. reports being co-founder and shareholder of Heidelberg Epignostix GmbH, and patents or patent applications for classification of cancer and methylation-based classification, honoraria from Bayer, Roche and Illumina, speaker support from Agilent. G.R. serves on the data safety board of HIT HGG trial. KDA reports patents or patent applications for DNA methylation-based cancer diagnostics for tumors of the central nervous system, kidney and hematopoietic system, and for predicting DNA methylation and tumor types from histopathology. M.R.G. reports honoraria from George Washington University and serves on the Advisory Board for the ARISTOCRAT Trial, and as grant reviewer and advisor for the Brain Tumour Charity. P.K.B. reports grants for institution from Breast Cancer Research Foundation, Eli Lilly, Merck, Kinnate, MGH Research Scholar Award, American Association for Cancer Research, Melanoma Research Alliance, Krantz, Mirati, and consulting fees from Advise Connect Inspire Atavistik Bio, Axiom, CraniUS, Genentech, InCephalo, Kazia, Medscape, MPM Capital Advisors, Sintetica, honoraria from Genentech, MPM Capital Advisors, serves on the monitoring or advisory boards of Kazia, N2M2, CraniUS, and as ASCO co-chair multisite guidelines committee, SNO Medical Oncology Representative, and received other support from Amgen, Eli Lilly, AstraZeneca, Merck, Bristol Myers Squibb, Genentech-Roche, GSK, Kazia, PfizerAlliance Neuro-Oncology Committee Co-Chair. P.W. reports lecture honoraria from Chimerix. S.D. serves as Siteman Cancer Center Protocol Review and Monitoring Committee member and Chair of the Constitution Committee, American Association of Neuropathologists. S.S., D.N.L., G.Z., A.P., D.J.B., and C.S. report no disclosures.

Funding

None declared.

Acknowledgments

This paper has been reviewed by the Steering Committee and Clinical Advisory Panel of cIMPACT-NOW and by the International Society of Neuropathology Executive.

The full cIMPACT Steering Committee comprises Drs. Ken Aldape, Dan Brat, David Capper, Andreas von Deimling, Dominique Figarella-Branger, Cynthia Hawkins, Tom Jacques, Takashi Komori, Arie Perry, Guido Reifenberger, Brent Orr, Felix Sahm, Chitra Sarkar, Pascale Varlet, and Pieter Wesseling, with David Louis as cIMPACT-NOW Advisor.

The cIMPACT Clinical Advisory Panel comprises Drs. Martin van den Bent, Mark Gilbert, Sabine Mueller, Stefan Pfister, Uri Tabori, and Michael Weller. The authors also thank Abigail Suwala, Niek Maas, and Luca Bertero for critically reviewing the manuscript independently of committee assignments. All authors conducted literature review, participated in consensus discussions, contributed to writing, and approved the manuscript.

References