Posterior fossa ependymomas can be subdivided into two types based on molecular profiling: A (PFA) and B (PFB) [4]. Immunophenotypically, they differ in the expression of H3K27me3: PFA show loss of this mark while PFB retain its expression. Furthermore, they also differ in the age of onset (pediatric vs. adult in PFA and PFB, respectively) [5]. However, because the maintainance of H3K27me3 immunoexpression is not pathognomonic, the most recent World Health Organization (WHO) classification has defined that a diagnosis of PFB may be rendered using DNA-methylation profiling [4]. Moreover, rare posterior fossa ependymomas with ZFTA fusion, exhibiting a preserved expression of H3K27me3, have been reported in pediatric cases [2]. To date, no PFB pediatric series has been reported and only epigenetic data are available in the literature [1]. The aim of this study was to clinically, radiologically, and molecularly (including DNA-methylation profiling) characterize a retrospective series of pediatric PFB (diagnosed solely by histopathology and immunohistochemistry) to detect potential differential diagnoses.

This pediatric series included 15 tumors initially diagnosed as PFB before the 2021 WHO guidelines, based on retained immuno-expression of H3K27me3. Subependymomas were excluded. The children were aged 1 to 17 years-old (median age: 11) with a female predominance (sex ratio: 1.5). A central neuroradiological review confirmed that all tumors were located in the posterior fossa. The DNA-methylation profiling classified tumors into four groups: PFB (n = 9), PFA (with low calibrated scores) (n = 2), supratentorial ependymomas, ZFTA-RELA fusion positive (n = 2), and spinal ependymomas (n = 2) (Table 1 and Additional files 1, 2, 3, 4 for details). We also performed a t-SNE (t-distributed stochastic neighbor embedding) analysis to better classify tumors with low calibrated scores (< 0.9) (Fig. 1). Using t-SNE, one of the two cases (#10) classified as PFA was reclassified as PFB whereas the other (#11), which presented a heterogeneous staining for H3K27me3 (without any positivity for EZHIP and H3K27M) was in close vicinity with PFA. The two cases classified as ependymomas, ZFTA-RELA fusion positive were confirmed by immunohistochemistry (L1CAM and NFκB immunopositivities), and RNA sequencing analysis (ZFTA::RELA fusion). Central neuroradiological review confirmed that both tumors were located in the posterior fossa without a supratentorial component. Interestingly, both cases presented clinical and radiological similarities and were distinct from PFB: they concerned the youngest patients (aged 1 and 3 years-old) and were revealed by a solid tumor located in the upper part of the fourth ventricle towards the aqueduct (compared to other PFB from the cohort which were located in the lower part of the fourth ventricle and crossed the Magendie foramen). The two cases classified as spinal ependymomas were located in the medulla oblongata or at the bulbo-medullar junction in a context of neurofibromatosis type 2 (NF2).

Fig. 1
figure 1

t-distributed stochastic neighbor embedding analysis of DNA methylation profiles of the investigated tumors alongside selected reference samples. Reference DNA methylation classes: diffuse midline glioma H3 K27M mutant (DMG_K27), diffuse midline glioma EGFR_altered (DMG_EGFR), diffuse midline glioma EZHIP_overexpressed (DMG_EZHIP), low-grade glioma, ganglioglioma (GG), ependymoma, myxopapillary (EPN_MPE), ependymoma, posterior fossa group A (EPN_PFA), ependymoma, posterior fossa group B (EPN_PFB), ependymoma, ZFTA fusion (EPN_ZFTA), ependymoma, spinal (EPN_SPINE), ependymoma, YAP fusion (EPN_YAP), subependymoma, posterior fossa (EPN_PF_SE), subependymoma, spinal (EPN_SPINE_SE), subependymoma, supratentorial (EPN_ST_SE)

Full size image
Table 1 Summary of clinicopathological features of the cohort
Full size table

Herein, the integrative histopathological, genetic and epigenetic analyses, including t-SNE (Fig. 1) segregated tumors into: 10 PFB (66.7%, with an enrichment of subclass 4 as previously reported [1]), two ependymomas, ZFTA-fusion positive (13.3%), and two NF2-associated spinal ependymomas (13.3%). The last case (#11) remained not elsewhere classified (NEC) posterior fossa ependymoma (6.7%) because of the discrepancy between immunohistochemistry and DNA-methylation profiling (performed two times). Further similar EPN, NEC cases are needed to clarify their classification and the existence of other potential diagnostic biomarkers. Previously, infratentorial (one cervicomedullary and two cerebellar) ependymomas, ZFTA-fusion positive (one ZFTA::MAML2, one ZFTA::NCOA2, and one ZFTA::RELA fusion) were reported in children aged 3, 4 and 11 years [2]. Like our cases, the DNA-methylation profiling confirmed that they clustered with their supratentorial counterpart ZFTA-fused [2]. Our results reinforce that ZFTA fusion can occur in the posterior fossa, and therefore, constitutes a potential diagnostic pitfall. In these conditions, L1CAM and NFκB immunostainings may represent useful diagnostic biomarkers in the detection of ZFTA-fused cases. Interestingly, ZFTA-fusion positive posterior fossa ependymomas seem to occur more frequently in younger patients than pediatric PFB, which are very rare before adolescence [1]. Similarly to a previously reported case, our series highlights an epigenetic distinction between spinal ependymomas centered in the medulla and PFB ependymomas, suggesting distinct cellular origins [3]. This precise anatomical site is well described in NF2 in which almost all ependymomas occurr in the cervicomedullary junction [3].

In conclusion, the results of this pediatric series are in line with the newly established essential diagnostic criteria for PFB in the WHO classification. A diagnosis of pediatric PFB ependymoma cannot be proposed solely based on the retention of H3K27me3 immunoexpression, as they encompass at least three different histomolecular tumoral types. Our study highligths also the importance of the integration of clinical, radiological, and neuropathological data to achieve an accurate diagnosis. According to our results, it may be recommended to perform L1CAM and NFκB immunohistochemistry and to search for clinical and radiological criteria for NF2 before performing DNA-methylation profiling for the differential diagnosis of pediatric PFB ependymoma.