Pediatric ependymoma: an overview of a complex disease

Pediatric ependymomas comprise biologically distinct tumor entities with different (epi)genetics, age distribution and localization, as well as a different prognosis. Regarding risk stratification within these biologically defined entities, histopathological features still seem to be relevant. The mainstay of treatment is gross total resection (GTR) if possible, achieved with intraoperative monitoring and neuronavigation—and if necessary second surgery—followed by adjuvant radiation therapy. However, there is growing evidence that some ependymal tumors may be cured by surgery alone, while others relapse despite adjuvant treatment. To date, the role of chemotherapy is not clear. Current therapy achieves reasonable survival rates for the majority of ependymoma patients. The next challenge is to go beyond initial tumor control and use risk-adapted therapy to reduce secondary effect and therapy-induced morbidity for low-risk patients and to intensify treatment for high-risk patients. With identification of specific alterations, targeted therapy may represent an option for individualized treatment modalities in the future.


Introduction
Ependymomas represent the third most common malignant intracranial neoplasm in children. Pediatric ependymomas can occur in all compartments of the CNS; the most frequent location is the posterior fossa followed by supratentorial sites, while spinal tumors are rather rare [1,2]. A slight overall male predominance is reported [1][2][3]. Based on their histological resemblance, ependymomas across all locations were historically considered as one entity but assigned different tumor grades. Assuming that histopathological indicators of anaplasia, namely mitotic activity, vascular proliferation, and necrosis, were adequate predictors for outcome independent of patient age and tumor location, WHO grading was used for risk stratification and hence for treatment decisions. However, the utility of grading for risk stratification has remained controversial, especially because of its lack of reproducibility due to inter-observer variability [4,5].
Taylor et al. were able to show that tumors of different locations in the CNS represent distinct tumor entities derived from spatially restricted progenitor cell populations of radial glia cells, indicating that spatial heterogeneity may be associated with distinct tumor biology [13]. These findings indicate that tumors in different locations constitute genetically and biologically distinct diseases despite their histopathological resemblance. Pajtler  In this review, the subdivision is partly orientated on the classification by Pajtler et al. However, pediatric ependymoma is distinct from adult tumors, as indicated in the section below; therefore, we aimed to focus on ependymomas occurring in childhood. Current treatment modalities will be discussed separately, since most trials published to date have used histopathological grade, age, and extent of resection for stratification rather than localization or epigenetically defined subgroups. Information regarding studies identifying prognostic parameters and results from prospective clinical trials provided in the text are summarized in Tables 1 and 2.

Distinction between pediatric and adult ependymoma
Approximately 30% of all ependymomas affect children and adolescent patients [1,2]. With several differences between pediatric and adult ependymoma, a differentiated approach is warranted. Regarding clinical parameters, intracranial location is frequent in children, while spinal tumors are rather rare compared to the adult population, where this location represents the most frequent manifestation site [1,2]. Regarding histological classification, anaplastic ependymoma is more frequent in pediatric patients, while myxopapillary ependymomas are mostly found in the adult population [1]; subependymomas are extremely rare in childhood and will therefore not be discussed in this review.
Regarding treatment, complete resection and radiation therapy are administered in a higher frequency in pediatric patients [1]; however, some bias may be owed to the fact that most controlled studies for ependymoma are exclusively including pediatric patients.
With respect to genetic differences, infants with posterior fossa tumors more frequently seem to have balanced genetic profiles compared to polyploid ones in older patients [7,15]. Furthermore, the gain of chromosome 1q seems to be only slightly more frequent in the pediatric population [16], while PF-EPN-A seem to be merely exclusive to the pediatric population [17]. RELA-fusion as well as CDKN2A deletion may be detected in pediatric as well as adult patients; however, differences in frequency have not been analyzed in detail so far [9,11,12,[18][19][20].

Supratentorial ependymoma
Supratentorial ependymomas are more frequently diagnosed in young patients and show a decreasing incidence with age [1].
On MRI supratentorial ependymomas appear as large, inhomogeneously contrast-enhancing tumors with frequent cystic areas as well as calcifications and more rarely hemorrhage and necrosis [21]. Supratentorial ependymomas of childhood are considered to have a better prognosis than those located in the posterior fossa, possibly because gross-total resection is more easily achieved in the former group. Supratentorial ependymal tumors can be divided into three epigenetic groups: ST-EPN-RELA, ST-EPN-YAP, and ST-SE, with supratentorial subependymoma being irrelevant in pediatric patients [11]. Furthermore, there is evidence of supratentorial ependymal tumors different from ST-EPN-RELA nor ST-EPN-YAP [22][23][24].

Supratentorial ependymoma with C11orf95-RELA fusions
Approximately 70 % of supratentorial ependymomas harbor one of several alternative C11orf95-RELA fusions resulting from chromosomal instability/microchromothripsis involving the chromosomal region 11q13. 1 [11, 12, 20]. C11orf95-RELA fusion proteins translocate into the nucleus, activating the canonical NF-κB signaling pathway. This pathway controls cell proliferation, apoptosis, and vascularization. In addition, it is involved in inflammatory response [12,20]. C11orf95-RELA fusion-positive ependymomas have a densely capillarized phenotype [20,25], presumably due to the vasogenic action of the NF-κB signaling pathway, as well as a clear cell morphology [26]. Regardless of the specific fusion gene, elevated nuclear protein expression of RelA protein can be identified by immunohistochemistry, offering an accessible cost-effective tool for detection [20,27]. Furthermore, compared to fusion detection by RT-PCR, immunohistochemistry can identify all different fusions at once using antibodies against p65/RELA (with higher specificity compared to L1CAM) [27], while RT-PCR-based analysis requires specific primers for each individual fusion transcript.
Regarding prognosis, lower survival rates for ST-EPN-RELA ependymoma compared to other ependymal tumors in this location were reported in a retrospective analysis, including adult and pediatric patients [11]. However, these findings are challenged by data derived from exclusively pediatric cohorts enrolled in three recent prospective trials, showing no prognostic disadvantage for ST_EPN-RELA when compared to other ependymoma subtypes [28][29][30]. In the revised WHO classification of CNS tumors of 2016, supratentorial ependymomas with C11orf95-RELA fusions were introduced as a novel specific tumor entity [31], initiating the era of a genetically based ependymoma classification.

Supratentorial ependymoma with YAP1-MAMLD1 fusion
A small subset of pediatric supratentorial ependymomas harbor a YAP1-MAMLD1 fusion. This fusion almost exclusively occurs in  young children and shows an excellent prognosis, possibly even without further therapy after resection [6,11,12,30]. The involved fusion partners YES-associated protein 1 (Yap1) and mastermind-like domain containing 1 (Mamld1) are involved in several pathways, e.g., Hippo signaling pathway and Wnt/βcatenin pathway, active in various cancers [6].

CDKN2A deletion in supratentorial ependymoma
Besides the two aforementioned gene fusions, a frequently reported genetic alteration associated with poorer prognosis in (pediatric) supratentorial ependymomas is homozygous deletion of CDKN2A [9,11,18,19]. The CDKN2A locus on chromosome 9p21 encodes the tumor suppressor proteins p14 ARF and p16 INK4A , which control the cell cycle via Mdm2/p53 or CDK4/6 and the retinoblastoma (RB) family of proteins [32]. Consequently, a deletion/inactivation of CDKN2A may result in uncontrolled cell growth. However, only a limited number of cases have been reported so far, and therefore general conclusions should be drawn with caution. In childhood ependymomas, CDKN2A deletion is restricted to ependymomas with RELA fusion [19].
Supratentorial ependymoma without C11orf95-RELA or YAP1-MAMLD1 fusions (non-RELA/non-YAP ependymomas) Finally, supratentorial ependymomas harboring neither C11orf95-RELA nor YAP1-MAMLD1 fusions have been reported. Some of these tumors may represent histological mimics of ependymoma misdiagnosed as ependymomas. However, the remaining ones may represent truly novel entities and need to be characterized in more detail [22][23][24]. These tumors may be sub-stratified, according to their predominant histological appearance and biology as RELA-like, tanycytic, and astroblastoma-like variants. Novel fusion transcripts may be encountered among these tumors as well [24].

Posterior fossa ependymoma
Pediatric ependymomas are frequently located in the posterior fossa [1,2]. In addition to histological grading distinguishing between classic (WHO grade II) and anaplastic ependymomas (WHO grade III), there is an association between midline and lateral tumor localization and prognosis. On MRI, posterior fossa ependymoma may appear as homogenous and well-demarcated tumors with hemorrhage and possible calcification spots showing variable contrast enhancement due to necrosis and cyst formation. Tumors may be located inside the fourth ventricle with possible lateral expansion through the foramina of Luschka or the foramen of Magendie.
With GTR being the strongest predictor of outcome [8,29,[33][34][35], tumors arising from the floor and the lateral aspect of the fourth ventricle have a worse prognosis than those arising from the roof. One explanation is that achieving GTR without postoperative deficits is more difficult [36] in the former group. Another explanation may be that the location is associated with different (epi-)genetically driven growth patterns [14,37].
Analysis of genomic copy number profiles revealed distinct cytogenetic patterns with different prognostic impact. Tumors displaying partial chromosomal alterations (structural alterations) had the worst prognosis, especially those harboring a gain of the q-arm of chromosome 1 [7][8][9]33]. Tumors with a balanced genetic profile without any chromosomal alterations were associated with a slightly better prognosis and occur predominantly in very young children [7,15,16]. In contrast, whole chromosomal alterations (numerical genetic profile) were found in tumors of older patients and adultmostly spinal-tumors and associated with a better prognosis [7,16]. Epigenetically, ependymomas of the posterior fossa can be divided into two defined groups [10,11,14], which are discussed in detail in the following paragraphs.

PF-A ependymoma
PF-A ependymomas occur predominantly in younger children with a decreasing frequency in adolescence and near absence in adulthood. There is a male predominance, and male sex is associated with poorer prognosis [11,14,37,38]. PF-A tumors show a predominantly balanced genetic profile without frequent recurrent somatic mutations, but gain of the long arm of chromosome 1 may be noted in 17-25 % [9,11,14,33,37]. However, in infants below the age of 18 months, chromosomal aberrations may be absent [15]. Furthermore, CpG-island hypermethylation (compared to PF-B tumors) is present in PF-A tumors [10,11,14], caused by overexpression of EZHIP (CXorf67; enhancer of Zeste inhibitory protein) [37]. By binding EZH2, EZHIP silences targets of the polycomb repressive complex 2 which represses the expression of differentiation genes through trimethylation of H3K27 [10]. Indeed, reduced or absent H3K27me3 expression is characteristic of PF-A ependymomas [39,40]. The loss of trimethylation of H3K27 can be evaluated by immunohistochemistry, which offers a cost-effective and readily accessible tool to characterize posterior fossa ependymomas. Eventually, DNA methylation analysis revealed further possible subdivision of posterior fossa ependymomas [37,41]. Pajtler et al. published a subdivision of PF-A ependymomas into 2 main and 9 epigenetically defined subgroups, characterized by different frequencies of H3K27M mutations, gain of chromosome 1q, and levels of CXorf67 or OTX2 expression [37].

PF-B ependymoma
Compared to PF-A, PF-B tumors are more frequently found in older children and adult patients and frequently show relatively hypomethylated genomes and a polyploid chromosomal profile with gains of complete chromosomes, retained H3K27me3, and slight female predominance [10,11,14,17,[37][38][39][40][41]. Cavalli et al. were able to identify five different epigenetically defined PF-B subgroups distinct from subependymomas, PF-A, and spinal ependymomas [41]. The relatively favorable prognosis of these tumors constitutes a rationale for reduction of adjuvant therapy after surgery for these tumors, avoiding the long-term sequelae of radiotherapy.

Spinal ependymoma
In contrast to the adult population, spinal ependymomas are rare in children [1,2]. Furthermore, pediatric patients suffering from ependymomas in this location seem to be generally older than those with supratentorial or infratentorial tumors [1,2]. Spinal ependymomas can be histologically classified into myxopapillary ependymoma WHO grade I, as well as classic (WHO grade II) and the less frequent anaplastic ependymoma (WHO grade III) [31,42,43]. Additionally, a rare and aggressive genetic variant with MYCN amplification has been described [44][45][46].

Classic and anaplastic spinal ependymoma
Classic and anaplastic ependymomas arise predominantly in the cervicothoracic cord and are located intramedullary [42]. On MRI, they appear as T1-hypo-and T2-hyperintense intramedullary, contrast-enhancing lesions with frequent cystic, hemorrhagic, and necrotic components as well as possible calcifications. About 60 % are associated with syringomyelia [47]. Classic and anaplastic spinal ependymomas have a predominantly numerical cytogenetic profile, similar to PF-B ependymomas and are associated with a better prognosis [7,16]. The most frequent chromosomal alteration in spinal ependymomas is the loss of the long arm of chromosome 22 [11], and the most frequent somatically mutated gene in classic and anaplastic spinal ependymoma is NF2, located on chromosome 22q [48], while NF2 is neither mutated in intracranial nor myxopapillary ependymomas. Epigenetically, based on a mixed and predominantly adult cohort, spinal ependymomas were subdivided into three groups SP-SE, SP-EPN, and SP-MPE [11]. Spinal ependymomas show a methylation signature distinct from subependymomas, myxopapillary ependymomas, and MYCN-amplified (anaplastic) ependymomas [11,46] (see below). Overall, patients with spinal ependymomas have a favorable prognosis regardless of age with progression-free and overall survival rates of 70-90 and 90-100 %, respectively [49] as well as limited evidence of a worse prognosis for anaplastic tumors.

Myxopapillary ependymoma
The name is derived from their histopathological features which include perivascular "sleeves" of myxoid ground substance, microcysts, and radial (pseudopapillary) arrangements of the tumor cells around vessels. In contrast to classic and anaplastic tumors, myxopapillary ependymomas are mostly found in the conus-cauda region in intradural but extramedullary location [42,50] and are relatively rare in pediatric patients [51]. Multifocal growth as well as locations other than the conus-cauda region [52,53] and extra spinal manifestation in the sacrococcygeal soft tissue [54] have been described. Furthermore, with a given tendency to leptomeningeal dissemination, craniospinal MRI is warranted [55].
The diagnosis of myxopapillary ependymoma is straightforward if typical histological features are present. However, even if these are absent, their distinct methylation profile will establish the diagnosis, since it differentiates them from classic/anaplastic ependymomas [11,17,26]. Of note, rare anaplastic variants with an increased mitotic index as well as vascular proliferation and necrosis have been described [54]; these features do not appear to influence survival, though [52]. Overall survival of myxopapillary ependymoma is favorable with >90 % at 10 years [51,56,57]. Progression-free survival is less favorable, with frequent persistent and recurrent disease as well as dissemination sometimes already present at diagnosis [56,57]. In incompletely resected cases, adjuvant radiotherapy improves the prognosis [58].

Spinal ependymoma with MYCN amplification
A rare and highly aggressive subtype of anaplastic spinal cord ependymoma with MYCN amplification predominantly affecting young adults but older children as well has been identified recently, further supporting the prognostic relevance of biological classification. Their site of predilection is the cervico-thoracic spine [44][45][46]59], and tumors are often large at diagnosis, involving multiple spinal segments and displaying an exophytic growth pattern as well as frequent dissemination [44,46,59]. Histologically, ependymomas with MYCN amplification present with characteristic ependymal features and signs of anaplasia [44,46,59]. MYCN protein overexpression can be demonstrated by immunohistochemistry [44,59]. Genetically, all cases show high-level MYCN amplification, which remains stable throughout the course of disease [44][45][46]59]. This subtype also shows a distinct methylation profile [44,59]. The cases reported so far showed unfavorable progression-free and overall survival. Therefore, their distinction from other spinal ependymomas is essential.

Published trial cohorts
The most important factor in the treatment of pediatric ependymoma, regardless of age and location, is maximal safe surgery [28-30, 33-35, 60, 61], ideally GTR, and may include second surgery, whenever needed and considered feasible [28,30,34]. The mainstay of approved adjuvant treatment has been postoperative radiation therapy regardless of location and extend of resection [34,35]. Weighing possible longterm sequelae against tumor recurrence, the cut-off age for radiation therapy is currently debated; it is performed as early as 12 months in recently published studies [29,30,34,35]. The most favorable single center results so far reported for a cohort of pediatric ependymoma with supra-and infratentorial location were achieved by Merchant and colleagues [35] published in 2009 with 5-year EFS and OS of 74 % and 85.0 %, respectively, while promising multi-center results were published by Massimino et al. in 2016 [34] with 5-year EFS and OS of 65.4 % and 81.1 % and consecutive 10-year EFS and OS rates of 58 % and 73 % [28,34], respectively; the largest cohort published to date (n=356) also achieved good 5-year EFS and OS rates of 62.7 % and 83.8 %, respectively [29]. Since local recurrence is reported more frequently than distant failure [34,62,63], Massimino et al. advocated local intensification of RT with an extra boost on the tumor region [34].

Genetic parameters, prognostic, and possible therapeutic implications
Regarding the different emerging biologically defined subtypes, Merchant et al. recommended an observation-only approach for completely resected WHO grade II supratentorial tumors with promising results [29]. A similar approach for PF-B-type tumors was suggested by Ramaswamy et al. [38]. The option of observation only for certain subentities is currently investigated in the SIOP-Ependymoma II trial and in the ACNS0831 trial. In contrast to PF-B tumors, PF-A-type ependymoma with gain of the q-arm of chromosome 1 showed a dismal prognosis and needed maximal safe surgery as well as radiation-and chemotherapy, especially since local and distant failure and the tendency for dissemination are more frequently reported compared to other ependymoma subtypes [28,29]. For supratentorial ependymoma, the negative prognostic impact of RELA-fusion-positive ependymoma that was initially reported [11] could not be validated by others [29,30]. On the other hand, recent data by Jünger et al. and Massimino et al. suggest that supratentorial ependymoma with CDKN2A deletion represents a more aggressive variant, possibly requiring intensified treatment [19,28]. Given Palbociclib, an oral inhibitor of CDK4/CDK6, is currently being tested for clinical efficacy in high-grade gliomas with amplification of CDK4/CDK6 or homozygous deletion of CDKN2A [64], CDKN2A deletion in pediatric supratentorial ependymomas should be the subject of further investigation since it might represent a therapeutic target. Based on the advancing characterization of the biological features of ependymoma, targeted therapies may be identified and proven in a clinical setting in the future [65]. However, for now, maximal safe surgery remains the upfront goal and postoperative radiation therapy the most effective treatment. Furthermore, despite ependymoma being divided into biologically distinct sub-entities, WHO grading was used for risk stratification in most trials and confirmed to be a significant predictor of prognosis [28,29,34,35]. Gender and age might be prognostically relevant as well [29,34,35]. However, the prognostic impact of age may be due to the fact that all treatment modalities are not necessarily used for very young children in all departments.

Radio-and chemotherapy
Regarding proton versus photon therapy, the two modalities are considered of equal efficacy. Regarding sequelae, proton beam therapy, however, may be more suitable for ependymoma typically arising at very young age in order to spare the highly vulnerable surrounding developing brain [62,63]. For clinical validation, further evaluation in prospective trial cohorts with longer follow-up is needed.
To date, chemotherapy according to different protocols was often administered in children with incompletely resected or anaplastic (WHO grade III) tumors in addition to radiotherapy [29,30,34]. In very young children, chemotherapy was applied with the aim to delay radiation therapy. However, efficacy of chemotherapy could not be proven [60,61]. The value of chemotherapy in ependymoma therapy is currently being re-evaluated in the ongoing SIOP Ependymoma II trial. Furthermore, the ACNS 0831 trial, also investigating the possible benefit of chemotherapy, has shown first promising results, indicating a possible benefit of maintenance chemotherapy regarding EFS (presented at ISPNO 2020).

Therapy of spinal ependymoma
Data on treatment protocols specific for pediatric patients with spinal ependymomas are sparse, since pediatric trials focused on intracranial ependymoma and the majority of larger retrospective analyses described mixed or adult cohorts. Data from the prospective HIT-trial on 29 spinal ependymomas have shown gross-total resection as the main predictor of favorable outcome [49]. However, uniform adjuvant treatment protocols for pediatric spinal ependymomas do not exist [49] and the relevance of postoperative radio-and chemotherapy is not proven. Nevertheless, the European Association of Neuro-Oncology (EANO) recommended postoperative radiotherapy for all anaplastic (WHO grade III) tumors and incompletely resected classic (WHO grade II) tumors [55]. For myxopapillary ependymomas, radiotherapy was recommended even after complete resection as well as in relapsed cases after-if feasible-second surgery in combination with chemotherapy [55]. For myxopapillary ependymomas which have a higher rate of local recurrence than classic grade II tumors, GTR seems to be the only predictor of progressionfree and overall survival. Relapse rates for pediatric patients seem to be generally higher when compared to adult patients [66].
Treatment of primarily metastatic ependymoma and recurrent ependymoma CNS metastases in ependymoma are rare at presentation [31]. Benesch et al. recently published data of ten such pediatric patients (2.4 % of 402) enrolled in the multi-center HIT-2000 ependymoma trial [67]. The four patients with ST-EPN-RELA were alive in first or second remission 7.5-12.3 years after initial diagnosis, while all 4 with PF-EPN-A had died, indicating that chemotherapy may be required for these patients and that genetic subgrouping may adequately predict prognosis [67]. With regard to radiotherapy, in disseminated ependymomas, the full craniospinal volume will be targeted by the radiotherapy, whereas in localized tumor, only the treatment of tumor bed is standard of care [55].
For recurrent ependymomas, the EANO suggests reoperation and reirradiation if possible as well as chemotherapy. If no local treatment is feasible, chemotherapy can be administered alone [55]. However, prognosis of children with recurrent ependymoma remains poor [68]. A recently published retrospective analysis reported an increased risk for recurrence in patients with tumors with gain of chromosome 1q and/or PFA-methylation profile, further demonstrating only little benefit from resurgery and reirradiation in case of relapse [69].

Conclusion
In conclusion, pediatric ependymomas comprise biologically distinct tumor entities with different (epi)genetics, age distribution and localization, as well as different prognosis. However, histological classification within these biologically defined entities is still relevant for risk stratification. Furthermore, molecular classification gains importance and will be addressed in the upcoming WHO classification of CNS tumors.
To date the mainstay of treatment remains maximal safe surgery (ideally GTR), achieved with intraoperative monitoring and neuronavigation and if necessary second surgery. Adjuvant radiation therapy as soon as justifiable, given potential harm to the surrounding nervous tissue, is recommended for most tumors; however, there is substantial evidence that it might be spared in defined subentities, while the role of chemotherapy still needs to be established.

Remaining open questions
Since pediatric ependymomas depict rare entities, all children should be treated in clinical trials, analyzing demographic, clinical, histological, molecular, and treatment-derived parameter in order to answer the questions remaining to date:

i)
Is a mere molecular risk stratification superior to classic histopathological classification to differentiate between low-and high-risk tumors? ii) Does surgery alone depict sufficient treatment for lowrisk tumors? iii) What time, dose, and extent are suitable for which type of tumor in terms of photon and proton radiation; and which modality may be more suitable? iv) What is the role of classic chemotherapy in pediatric ependymoma? v) With potential therapeutic targets identified, what will be the role of targeted therapy? vi) Which is the optimal treatment for recurrent and primarily metastatic pediatric ependymoma; do we need a more aggressive treatment regime?
Data availability statement The data that support the findings of this study are available on reasonable request from the corresponding author.
Funding Open Access funding enabled and organized by Projekt DEAL.

Declarations
Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.