The Intersection of Epigenetic Alterations and Developmental State in Pediatric Ependymomas Open Access

Ependymomas are tumors that occur throughout the central nervous system. Until recently, ependymomas were divided based on histopathology into three grades predicted to correlate with patient outcome (I, II, III) [1]. Due to difficulty in distinguishing grades II and III and well as lack of consistency in treatment outcomes within grades, concerted effort has been placed into reclassifying these tumors based on molecular characteristics such as DNA methylation, gene expression, and major molecular drivers [2‒4]. As of 2016, ependymomas are divided into 9 groups based on location in the central nervous system and major genomic and epigenomic characteristics [4]. Generally, pediatric ependymomas along with other pediatric cancers are devoid of large mutational burdens or copy number variations compared to adult cancers [5]. This directly challenges typical models of cancer where cumulative mutations progressively allow cells to gain the hallmarks of cancer such as immune evasion and replicative immortality, finally resulting in neoplastic transformation [6]. Rather, we posit that ependymomas and other pediatric brain cancers may be more accurately described as a set of developmental diseases, where one main driver takes advantage of the epigenetic, transcriptomic, and even metabolic states seen in the cells of origin to initiate neoplastic transformation. This concept is seen in other cancers such as neuroblastoma, where spontaneous cure is thought to happen due to differentiation mechanisms and epigenetic changes in transformed cells [7]. Disease drivers that have been identified in ependymoma, such as fusion oncoproteins or overexpression of certain genes of interest, may interact with cell of origin programs to cause neoplastic transformation. Because of this, it is of vital importance to better understand the cell types where these cancers originate as well as how these cell programs are changed as the tumor develops. Investigating the intersection between oncogenic and developmental programs, could lead to therapeutic insights for ependymoma and its specific subtypes.

Childhood brain tumors are thought to have origins in development distinct from adult tumors [5, 8]. Several brain tumor types mirror the transcriptional and epigenetic programs of distinct progenitor cell lineages, such as granule progenitor cells with Sonic Hedgehog medulloblastoma, unipolar brush cells with group 4 medulloblastoma, and gliogenic precursor cells with astrocytoma [8‒10]. One major hypothesis is that oncogenic transformation leads to a block in differentiation of rapidly dividing progenitor cells. Pediatric ependymomas are currently thought to originate from radial glial cells (RGCs), a cell type seen in early brain development that gives rise to many other cell types such as neurons, astrocytes, and ependymal cells [11]. The most common driver of cortex-derived ependymoma (i.e., supratentorial, ST) is a gene fusion between zinc finger translocation-associated protein (ZFTA; previously called C11ORF95, see ref 2) and reticuloendotheliosis viral oncogene homolog A (RELA). The fusion protein ZFTA-RELA, while sufficient to drive one subtype of ependymoma [2], is not seen in any other cancer type. This leads to a hypothesis that the developmental and epigenetic context of RGCs may be susceptible to oncogenic changes induced by the fusion, whereas more differentiated cell types are not. This may occur, for example, via ZFTA-RELA binding sites being accessible in RGCs but less so or absent from other cell types (Fig. 1a). Arabzade et al. [17] 2021 found that ZFTA-RELA binds to DNA motifs that are shared with PLAGL/1, transcription factors that are highly expressed in embryonic development [12, 13], and correlate with poorer outcomes in gliomas and other cancers [14‒16]. Plagl1 is an imprinted gene and a subset of overexpressed genes in ependymoma is also imprinted, such as Axl, Dlk1, Igf2. This evidence supports the concept that stalled developmental programs are at the root pediatric brain tumor pathogenesis [8]; in ependymoma through hijacking of developmental epigenetic programs.

The most common and aggressive forms of ependymoma are those that arise in the posterior fossa (PF), a region comprising the cerebellum and brain stem. PF ependymoma are thought to originate in the cerebellum early in embryonic development from prenatal gliogenic progenitor cells [10]. PF ependymoma are broadly divided into two major subgroups, labelled type-A (PFA) or type-B (PFB), and have distinct clinical outcomes. PFAs and PFBs, while similar in histopathology, have different molecular characteristics and outcomes. PFAs are more common in younger children and infants than PFBs and have a worse prognosis, especially when accompanied by 1q gain [4, 18‒20]. Interestingly, while PFBs have a better prognostic outcome, they are also characteristic of more genetic instability with greater copy number alterations and gain/loss of chromosomal arms. PFAs are defined by elevated EZHIP expression, a protein that phenocopies oncohistone mutations observed in subsets of pediatric high-grade glioma [21‒24] EZHIP binds to and inhibits the PRC2 complex, which contains the methylating enzyme EZH2, leading to global loss of the repressive mark H3K27me3 [22‒24]. There is also corresponding increase of active mark H3K27ac, including at key glycolytic and TCA cycle-related genes [25]. This leads to global alterations of metabolic pathways, which could include future therapeutic targets, and may be linked to PF ependymoma growth in specific hypoxic-microenvironments [26]. How EZHIP overexpression and H3K27me3 loss is able to initiate PFA ependymoma is unclear, and not yet established in animal models. This is in contrast with midline PHGG that have accompanying mutations in P53, ATRX, ACVR1B, and amplification of PDGFRA that can be modeled in mice using a variety of approaches. One hypothesis is that the cell of origin for PFA ependymoma are particularly vulnerable to EZHIP overexpression and that specific cell-type targeting in development is needed, assuming those cells in humans are also observed in mice.

In ST ependymomas, epigenetic changes are often driven by fusion oncoproteins thought to act as transcription factors [2, 4, 27, 28]. Each oncogenic fusion protein works to drive disease by activating different downstream oncogenic transcriptional and epigenetic programs. Pediatric ST-EPNs are divided into subtypes driven by ZFTA-fusions and YAP1-gene fusions. These fusions arise by different means, activate different downstream pathways, and have different clinical outcomes, outlining the potential needs for distinct treatments.

The ZFTA-RELA protein is a fusion between ZFTA and v-rel avian RELA. Although ZFTA is known to have important roles in binding and remodeling of chromatin and translocation into the nucleus, it remains mostly uncharacterized, especially outside of a cancer context [29]. RELA, however, is well studied. It is a heterodimer in the NF-κB signaling pathway, which plays roles in several cellular processes such as inflammation, metabolism, and chemotaxis [30, 31]. This fusion arises through chromothripsis on chromosome 11, where genomic instability leads to catastrophic local chromosomal rearrangements. Chromothripsis plays a role in several other cancers, such as leukemia, myeloma, and myxofibrosarcoma [32‒34]. The ZFTA-RELA transcriptional program involves inflammation and activation of NFkB, consistent with the role of RELA, but also fusion-specific programs such as inhibition of differentiation and development programs [17]. Induction of fusion-specific programs may be aided by recruitment of chromatin remodelers such as BRD4 and EP300 that interact specifically with the fusion but not with ZFTA or RELA alone [29] and importantly are “druggable.” Furthermore, specific transcription factors are vital to fusion-mediated oncogenesis such as Sox9 [35]. Sox9 is a transcription factor that has roles in glial development and is elevated in ependymoma [36, 37], and its loss leads to increased survival in ependymoma [35].

Yes1 associated transcriptional regulator (YAP1) is a component of the Hippo signaling pathway. It is a potent oncogene that increases proliferation and downregulates apoptosis via association with the TEAD family of transcription factors [38, 39]. In ependymoma, YAP1’s most common fusion partner is mastermind-like domain containing protein 1 (MAMLD1), another transcriptional coactivator. Interestingly, while ZFTA-RELA is seen solely in ST ependymoma, YAP1-MAMLD1 is seen in some other cancers, such as schwannoma and thymoma [40, 41], implying that YAP1 fusions are not as restricted in their cell of origin as ZFTA-RELA fusions. Work from the Kawauchi laboratory illustrates the importance of DNA binding locations and partner transcription factors in the etiology of this ependymoma. The DNA binding motifs of nuclear factor 1 (NFI) and TEAD1-4 are highly enriched at YAP1-bound regulatory elements within this subtype, and these motifs are within genes important for oncogenesis and tumor progression, such as those for cell migration and proliferation. Mutation of the TEAD interaction site within YAP1-MAMLD1 prevents it from inducing oncogenesis in mice. In addition to these important interactions with TEAD genes, YAP1-MAMLD1 is known to physically interacts with NFI proteins NFIA and NFIB and knockdown of these proteins leads to decreased proliferation and fusion target expression [42]. Importance of this interaction and further study of the NFI family in ependymoma is also warranted by its involvement in other cancers, such as its promotion of metastasis in small cell lung cancer by increasing chromatin accessibility [43].

While ZFTA-RELA and YAP1-MAMLD1 are the most common fusion proteins in ependymoma, there are other less common disease driving fusion proteins (see Table 1, ref [16, 27, 28, 42]). ZFTA-RELA has 7 different varieties seen in ependymoma [2]. Additionally, Yap1 has also been seen to partner with other proteins such as FAM118B (uncharacterized), TFE3 (transcription factor), and SS18 (SWI/SNF) [4, 27], highlighting the importance of epigenetic regulation in progression of this cancer. Patjler et al. [42] also found expression of other less common fusions. For example, the PTEN-TAS2R1 fusion leading to a frame shift and subsequent disruption of PTEN, a tumor suppressor gene commonly mutated in cancer. Loss of PTEN activity has previously been demonstrated to induce NF-κB activity through activation of Akt/mTOR [44, 45], highlighting the importance of this pathway in ST ependymoma development [4]. Importantly, Kupp et al. showed that an artificial ZFTA-EP300 fusion is not able to induce tumor formation, demonstrating that transcriptional activators simply brought to the correct area in the genome are not sufficient to induce oncogenic transformation. ZFTA-EP300 also failed to upregulate the transcription factor GLI2 as ZFTA-RELA does, suggesting aberrant noncanonical hedgehog signaling is required for fusion mediated tumorigenesis. Not only is GLI2 required for RELA fusion driven tumor formation in mouse models, but GLI2 binding sites are also highly enriched in active enhancers and super enhancers of human ST-EPN-RELAs [46, 47]. Targeting of the hedgehog signaling pathway may be a beneficial treatment in the future, as arsenic trioxide targeting of GLI2 has been in mouse models [46]. Additionally, this pathway is crucial in development of the nervous system and is highly active in embryonic and neural precursor cells, further illustrating the interaction of fusion mediated transcriptional changes with the current developmental landscape is vital when initiating neoplastic transformation.

ZFTA-RELA is a potent oncogenic fusion protein, with fusion transformed neural stem cells able to form tumors in mice at 100% penetrance within 30 days [2]. Surprisingly, ZFTA-RELA is seen only in ST ependymoma and not other cancer types. This has led the field to ask how such a potent oncogene be so restricted. Evidence suggests the fusion interacts with the developmental landscape of the cell of origin to initiate cancer, preventing transcriptional programs rich in genes for proliferation and stem cell function from turning off at the correct points in development. ZFTA-RELA shares a binding motif with the transcription factor Pleiomorphic Adenoma Gene-Like 1 (PLAGL1) [17]. PLAGL1 has an important role in development of RGCs and is only present in this cell type in embryonic development, with altered expression of the gene leaving radial glia in an undifferentiated state unable to move up and populate the cortex [12]. Normally, PLAGL1 regulates an imprinted gene network and is silenced early in development [48], but ZFTA-RELA may prevent silencing of programs activated by PLAGL1. Several direct targets of ZFTA-RELA are shared with PLAGL1, such as Igf2 and Dlk1 [17]. ZFTA-RELA deletion also leads to decreased expression of genes involved in differentiation and development [17]. ZFTA-RELA may also be working with PLAGL1 or other cell type-specific genes to cooperatively change gene expression. For example, the growth of ZFTA-RELA ependymoma but not high-grade glioma is dependent on the expression of Sox9 [35], another developmental transcription factor. This principle of hijacking developmental programs is also present in PFA ependymomas. PFA is characterized by overexpression of EZHIP and corresponding global decrease in H3K27me3 [22‒24]. In normal development, H3K27me3 is absent within the embryo brain at key developmental genes but is added by the PRC2 complex. Correspondingly, components of the PRC2 complex decrease over time until the switch from neurogenesis to gliogenesis. Early knockdown of these components, which is functionally the same as increased EZHIP expression, leads to lengthened periods of neurogenesis and lasting high expression of genes for proliferation and self-renewal [49, 50]. This series of evidence supports the importance of the early developmental state of ependymoma as necessary for cellular transformation.

Traditional considerations of adult cancer where mutations sequentially accumulate to transform cells are not observed in pediatric malignancies, including ependymomas. Accordingly, different avenues of treatment compared to adult cancers need to be explored. As discussed in this review, pediatric ependymomas frequently utilize epigenetic changes for tumor initiation and growth, some of which may be able to be targeted directly in new forms of treatment. For example, PFA tumors are characterized by EZHIP overexpression, PRC2 complex inhibition, and corresponding loss of H3K27me3. Hyperactivity of PRC2 complex members such as EZH2 has been implicated in the initiation, growth, and metastasis of other cancers [51‒54], and several PRC2 complex inhibitors are in clinical trials, with tazemetostat being FDA approved for treatment of epithelioid sarcoma and follicular lymphoma [53, 55]. Drugs to increase H3K27me3 levels or inhibit EZHIP (decrease H3K27me3 levels) are an exciting avenue of future research for PFA ependymomas. Histone modifications in general are frequently used as drug targets [56‒58] and further investigation of histone methyltransferase inhibitors and demethylating agents as treatment for PFA may be important. Drugs of these classes show promise in other cancers such as colorectal [59] and lymphoma [54].

While some subtypes of ependymoma may be promising candidates for epigenetic targeting, other subtypes may benefit more from looking at effects of epigenetic and cell identity changes. ZFTA-RELA is an oncogenic transcription factor and transcription factors are notoriously difficult to drug, requiring targets to be found elsewhere. Work in Mack et al. 2017 [47] used epigenetics methods to find new druggable targets in ependymoma. They performed oncogenic enhancer profiling and compared a list of super enhancer regulated genes with the Washington University Drug Gene interaction database, finding several druggable targets including ion channels. This is interesting because recent data has highlighted the importance of the tumor environment and normal brain activity in brain cancer development. Neural activity has been shown to aid in the progression of tumors in the central [60‒62] and peripheral nervous system [63], but not yet directly in ependymomas. More work into better understanding this relationship in ependymomas specifically could lead to better therapeutic approaches, as has already been shown in other brain cancers. For example, use of the antiepileptic drug lamotrigine has been shown to rescue electrophysiological properties in several models of neurofibromatosis type 1 [64]. This work highlights the importance of considering epigenetics not just as a therapeutic target but as a tool to find new avenues of treatment.

Additionally, questions about basic mechanisms behind pediatric ependymoma development still need to be addressed. Pediatric ependymomas closely mirror cellular development and depend on these programs for initiation and possibly maintenance. If ependymoma cells depend upon specific transcriptional and epigenetic programs observed in cells of origin, a better understanding of the developmental cell state of ependymoma is needed. Ependymomas may arise from RGCs or similar progenitor cells, cell types prevalent in early brain development that dimmish at birth. Further study of the cell of origin could lead to more avenues for treatment that would not be apparent when studying tumors at end stage when cells have transformed and new mutations have accumulated, such as drugging certain transcriptional programs present in embryonic development. Due to the role of developmental cell states in ependymoma, directed differentiation where RGCs are encouraged down a path to more mature cell states could also be an encouraging form of therapy. This principle has been using in the treatment of APL [65] and has shown promise for the treatment of other cancers such as breast cancer [66]. If pediatric brain tumors like ependymoma are fundamentally distinct from adult counterparts (as supported by multiple datasets), new approaches to identifying targets and treatments are needed.

Ependymomas are a diverse set of brain tumors. Extensive characterization of these tumors has been performed, leading to more clinically relevant divisions compared to histopathological grading. Investigation concerning the epigenetic mechanisms behind these diseases has also occurred, implicating histone modifications and transcriptional changes in the etiology of several subtypes of ependymoma. Developmental context will be an important area of future study in order to better understand how this tumor type originates and interacts with unique developmental states to support its own growth. Further research into the basic science behind ependymoma will be necessary in order to develop better treatments, such as those involving recruitment of the immune system and surrounding healthy tissue. This will be especially important for subtypes with no targeted therapies and poor survival rates. Basic mechanistic principles underlying ependymoma can also be brought to other pediatric cancers, such as those driven by other fusion proteins.

The authors have no conflicts of interest to declare.

No funding was received in the preparation of this manuscript.

A.S.K. and S.M. conceptualized and outlined the review article. A.S.K. and S.M. performed material preparation and literature search. A.S.K. wrote the first draft, and S.M. commented and edited. All authors approved the final draft.