Supratentorial non-RELA, ZFTA-fused ependymomas: a comprehensive phenotype genotype correlation highlighting the number of zinc fingers in ZFTA-NCOA1/2 fusions

The cIMPACT-NOW Update 7 has replaced the WHO nosology of “ependymoma, RELA fusion positive” by “Supratentorial-ependymoma, C11orf95-fusion positive”. This modification reinforces the idea that supratentorial-ependymomas exhibiting fusion that implicates the C11orf95 (now called ZFTA) gene with or without the RELA gene, represent the same histomolecular entity. A hot off the press molecular study has identified distinct clusters of the DNA methylation class of ZFTA fusion-positive tumors. Interestingly, clusters 2 and 4 comprised tumors of different morphologies, with various ZFTA fusions without involvement of RELA. In this paper, we present a detailed series of thirteen cases of non-RELA ZFTA-fused supratentorial tumors with extensive clinical, radiological, histopathological, immunohistochemical, genetic and epigenetic (DNA methylation profiling) characterization. Contrary to the age of onset and MRI aspects similar to RELA fusion-positive EPN, we noted significant histopathological heterogeneity (pleomorphic xanthoastrocytoma-like, astroblastoma-like, ependymoma-like, and even sarcoma-like patterns) in this cohort. Immunophenotypically, these NFκB immunonegative tumors expressed GFAP variably, but EMA constantly and L1CAM frequently. Different gene partners were fused with ZFTA: NCOA1/2, MAML2 and for the first time MN1. These tumors had epigenetic homologies within the DNA methylation class of ependymomas-RELA and were classified as satellite clusters 2 and 4. Cluster 2 (n = 9) corresponded to tumors with classic ependymal histological features (n = 4) but also had astroblastic features (n = 5). Various types of ZFTA fusions were associated with cluster 2, but as in the original report, ZFTA:MAML2 fusion was frequent. Cluster 4 was enriched with sarcoma-like tumors. Moreover, we reported a novel anatomy of three ZFTA:NCOA1/2 fusions with only 1 ZFTA zinc finger domain in the putative fusion protein, whereas all previously reported non-RELA ZFTA fusions have 4 ZFTA zinc fingers. All three cases presented a sarcoma-like morphology. This genotype/phenotype association requires further studies for confirmation. Our series is the first to extensively characterize this new subset of supratentorial ZFTA-fused ependymomas and highlights the usefulness of ZFTA FISH analysis to confirm the existence of a rearrangement without RELA abnormality.

Background Ependymomas (EPN) are glial neoplasms that affect mainly children and young adults. New insights in the genomic and epigenetic landscape of EPN has led to the identification of different groups accordance to their anatomic location (supratentorial, posterior fossa and spinal) [1]. Three subgroups have been identified among supratentorial tumors (ST-EPN): subependymomas; EPN, YAP1 fusion-positive; and EPN, RELA fusion-positive (according to the World Health Organization-WHO-2016 classification) [1][2][3][4]. Infrequently (6.5% of cases in one series) [4], C11orf95 or RELA genes have fused with other genes as a result of chromothripsis. The Consortium to Inform Molecular and Practical Approaches to CNS Tumor Taxonomy (c-IMPACT NOW) Update 7 recently proposed the nosology "ST-EPN, C11orf95-fusion positive" instead of "ST-EPN, RELA-fusion positive" [5]. This modification reinforces the idea that when ST-EPN exhibits fusion that implicates the C11orf95 gene with or without the RELA gene, it represents the same histomolecular entity [4,[6][7][8]. In recent papers, the methylation classifier based on Forest plot random classification highighted that cases with C11orf95fusion without RELA presented epigenetic vicinity with tumors of the EPN-RELA methylation class (MC) and subdivided them into two satellite clusters (2 and 4) by multidimensional reductionality (more specifically t-Distributed Stochastic Neighbor Embedding (t-SNE) analysis) [8,9]. However, these alternative partners to RELA seem to produce original morphological patterns which challenge the histopathological diagnosis.
In fact, recent studies have reported a large spectrum of morphologies, including glial, glioneuronal, embryonal and even mesenchymal and epithelial patterns in tumors harboring C11orf95-fusions without RELA [6,7,9]. In this study, we performed a clinico-pathological and molecular analysis (including DNA-methylation profiling and the identification of the new clusters of methylation) of 13 new cases of ST-EPN with C11orf95 (now called ZFTA for Zing Finger Translocation Associated by the new HUGO gene Nomenclature Committee) fusion without the RELA gene to more suitably characterize these tumors and compare them with their counterparts which have classical ZFTA:RELA fusion.

Study design, patients, data collection
This study included patients diagnosed with ST EPN or glial ST tumors with ZFTA rearrangement but no RELA rearrangement during ependymal cell differentiation, determined by FISH analyses (techniques previously described [3]). Epidemiological data (gender and age at diagnosis) and tumor-and treatment-related data (location of tumor and extension, extent of resection, relapses and complementary treatments) were retrospectively analyzed. The extent of the initial resection was assessed by magnetic resonance imaging (MRI) or computed tomography performed after surgery. All the patients' parents or legal guardians signed informed consent forms before treatment was started. We obtained human subjects approval from our institutional review board.

Central radiological review
The central radiological review was performed by two neuroradiologists (NB and VDR). Preoperative MRIs were read and the following features were analyzed: location, tumor size, signal in a T1-weighted sequence and a T2-weighted sequence, susceptibility imaging, the diffusion and apparent diffusion coefficient map (ADC), fusion was frequent. Cluster 4 was enriched with sarcoma-like tumors. Moreover, we reported a novel anatomy of three ZFTA:NCOA1/2 fusions with only 1 ZFTA zinc finger domain in the putative fusion protein, whereas all previously reported non-RELA ZFTA fusions have 4 ZFTA zinc fingers. All three cases presented a sarcoma-like morphology. This genotype/phenotype association requires further studies for confirmation. Our series is the first to extensively characterize this new subset of supratentorial ZFTA-fused ependymomas and highlights the usefulness of ZFTA FISH analysis to confirm the existence of a rearrangement without RELA abnormality.

Central histopathological review
The central pathology review was performed conjointly by two neuropathologists (ATE and PV).

Immunohistochemistry
Unstained 3-μm-thick slides of formalin-fixed, paraffin-embedded tissues were obtained and submitted for immunostaining with an automated stainer (Dako Omnis, Glostrup, Denmark

FISH analyses
A FISH study was performed on interphase nuclei according to the standard procedures and the manufacturer's instructions. The CDKN2A gene copy number was assessed using the following centromeric and locus specific probes: Vysis CDKN2A/CEP9 FISH Probe Kit (Abbott Molecular, USA). Deletion was considered if they were detected in more than 30% of nuclei respectively. Results were recorded using a DM600 imaging fluorescence microscope (Leica Biosystems, Richmond, IL) fitted with appropriate filters, a CCD camera, and digital imaging software from Leica (Cytovision, v7.4).

DNA sequencing
Mutations for the hTERT promoter was developed using Massarray iPlex technology and Massarray online design tools (Agena Bioscience) as previously described [23].

RNA sequencing
RNA was isolated from FFPE (Formalin-fixed paraffinembedded) tissues with sufficient tumoral density. RNA was extracted using the High Pure FFPET RNA Isolation Kit (catalogue # 06650775001 Roche diagnostics GmbH) according to the manufacturer's instructions. The RNA concentrations were measured on a Qubit 4 Fluorometer (# Q33238, Thermo Fisher Scientific) with the Invitrogen Qubit RNA BR Kit (# Q10210, Thermo Fisher Scientific). The percentage of RNA fragments > 200 nt (fragment distribution value; DV200) was evaluated by capillary electrophoresis (Agilent 2100 Bioanalyzer). DV200 > 30% was required to process the next steps in the analysis. NGSbased RNA sequencing was performed using the Illumina TruSight RNA Fusion Panel on a Nextseq550 instrument according to the manufacturer's instructions (Illumina, San Diego, CA, USA). This targeted RNA sequencing panel covers 507 fusion-associated genes, to assess the most recognized cancer-related fusions. The TruSight RNA fusion panel gene list is available at https:// www. illum ina. com/ conte nt/ dam/ illum ina-marke ting/ docum ents/ produ cts/ gene_ lists/ gene_ list_ trusi ght_ rna_ fusion_ panel. xlsx. 7690 exonic regions are targeted with 21,283 probes. Libraries were prepared according to the Illumina instructions for the TruSight RNA fusion Panel kit. STAR_v2.78a and Bowtie software were used to produce aligned readings in relation to the Homo Sapiens Reference Genome (UCSC hg19). Manta v1.4.0, Tophat2 and Arriba v2.1.0 tools were used for fusion calling.

DNA methylation profiling
Tumor DNA was extracted from freshly frozen tissue samples using the Qiagen DNeasy Blood & Tissue Kit (Cat NO./ID 69504) according to the manufacturer's instructions. 500 ng of DNA were extracted from each tissue sample. DNA was sent to the Genotyping facility at the German Cancer Research Center (Heidelberg, Germany). All patient samples were analyzed using either Illumina Infinium Methylation EPIC or HumanMethyla-tion450 BeadChip arrays according to the manufacturer's instructions. Affiliation predictions were obtained from a DNA methylation-based classification web platform for central nervous system tumors (www. molec ularn europ athol ogy. org, version 11b4). Next, a t-SNE analysis was performed and compared with the genome-wide DNA methylation profiles from the brain tumor reference cohort [24] as well as with a previous series of ZFTA:RELA-fused EPN [3] and with the series of ZFTAfused ependymomas reported by Zheng et al. [9]. Data was generated at the DKFZ Genomics and Proteomics Core Facility (Heidelberg, Germany) as previously described [24].

Clinical and radiological characteristics
Relevant clinical data are summarized in Table 1. The median age at diagnosis was 6.7 years (patients' ages ranged from 9 months to 41 years). The male/female sex ratio was 1.6 (8 males and 5 females

Molecular results
FISH analyses for CDKN2A failed to reveal any deletion in any of the cases tested (n = 13). No mutation of hTERT was evidenced in any of the cases tested (n = 13). We found a new MN1:ZFTA fusion which was verified by RT-PCR and Sanger sequencing for case #2 (Additional file 3). Other ZFTA partners have been previously described [4,[6][7][8][9]. The anatomy of the 11 in  According to the DNA methylation-based classification and the DKFZ Classifier (version 11b4), none of the tumors were classifiable (calibrated scores for DNA methylation class < 0.9). Although none of the cases received a calibrated score ≥ 0.9 in the current version (11b4) of the CNS tumor classifier, most of the tumors obtained the highest score for ependymal subclasses (EPN-RELA) with valid quality controls for all samples. A t-SNE analysis was performed to compare the genome-wide DNA methylation profiles of our previous EPN-RELA cohort with proven RELA:ZFTA fusion (n = 80) [3], EPN-YAP (n = 26), HGNET-BCOR (n = 23) and HGNET-MN1 (n = 21) in the CNS reference cohort [19]. All cases clustered in close proximity to EPN-RELA (Fig. 7). Copy number profiles are detailed in Additional files 5-17. In a more focused t-SNE analysis of DNA methylation data of these samples alongside the recently described satellite clusters of ZFTA-fusion positive EPN (cluster 1, n = 9; cluster 2, n = 40; cluster 3, n = 17, and cluster 4, n = 27) [9], four of the cases grouped with cluster 4 and nine with cluster 2 (Fig. 7).

Discussion
Like ST ZFTA:RELA-fused EPN, ST non-RELA ZFTAfused EPN affected mainly children [4,[6][7][8]. The sex ratio was 1.3 (13 males and 10 females) [4,[6][7][8]. Radiologically, non-RELA ZFTA-fused EPN presented some similarities with their classical counterparts with ZFTA:RELA fusion [3]. In fact, they were mainly characterized by well-demarcated solid and cystic lesions with peripheral enhancement of the cystic content [3]. However, contrary to ZFTA:RELA-fused EPN, peripheral edema was significant in our cases and for the most part the cystic component was not hyperintense on the FLAIR sequence [3]. ST non-RELA ZFTA-fused EPN presented high morphological heterogeneity with only rare cases having histopathological and immunohistochemical features of ZFTA:RELA-fused EPN [8,9]. In the literature, their histological appearance was sarcoma-like, PXA-like, highgrade glioma-like, malignant teratoma-like, embryonal tumor-like, or had neuronal differentiation and a granular cell component [6,7,9]. We also identified four cases with astroblastoma-like features. Despite this phenotypical heterogeneity, all tumors were in close epigenetic proximity to the MC EPN-RELA. As expected, our cases with ZFTA fusion without RELA were subclassified in clusters 2 and 4 [9]. In the original report, the cluster 2 corresponded almost exclusively to tumors with ependymal morphology [9]. In our series, 4/9 tumors in cluster 2 showed ependymal features and 5 presented astroblastoma-like features, noted for the first time. In the original report [9], the tumors in cluster 4 corresponded to highly malignant poorly differentiated tumors including one malignant small-cell sarcomatoid carcinoma and one undifferentiated sarcoma [9]. In our series, all three tumors presenting with sarcoma histology were classified in cluster 4. None of the cases in our series or from those in the literature exhibited significant nuclear expression of NFκB [7,8], which supports previous studies showing that p65 immunoexpression is highly correlated to the presence of RELA fusion [3,25,26]. However, all except two cases [6,7,9] showed L1CAM immunoexpression to varying degrees and intensities, confirmed by the RNA expression data [4]. Consequently, L1CAM may represent a diagnostic tool for non-RELA, ZFTA-fused EPN. Further immunohistochemical series including different   [6][7][8][9]. These fusions alone are sufficient to drive tumorigenesis in vivo [6,7,9]. In the original report, the main cases of ZFTA:MAML2 fusion were in cluster 2 and showed ependymal features [9]. Our data are in line with this report as our three cases with ZFTA:MAML2 were classified as cluster 2 and showed a histological phenotype of EPN but also of astroblastoma. A MN1:ZFTA fusion with ZFTA as a 3' partner was noted for the first time in another of our cases, as was previously reported in one case with LTBP3:ZFTA fusion [4]. Interestingly, this case of MN1:ZFTA fusion presented astroblastoma-like features, but was in close vicinity of the MC EPN-RELA (cluster 2) and not HGNET-MN1. The MN1 breakpoint is similar to that of the MN1:BEND2 fusion, which could constitute a diagnostic pitfall if only the MN1 breakapart FISH is used. We found a ZFTA fusion with a non-coding region which probably leads to a truncated ZFTA protein at the C terminal end. At the mRNA level, the truncation of 3'UTR of ZFTA eliminates three miRNA binding sites (hsa-miR-424-3p URS00002BCF86_9606) involved in regulating ZFTA expression. This loss of regulation could lead to a nuclear accumulation of ZFTA sustaining oncogenicity with a cis-acting mechanism instead of the trans-activating mechanism that could not take place in the absence of a coactivating partner. This cis-acting mechanism has been suggested by Zhu et al. because only half of the top-scoring ChiP-seq peaks of ZFTA-RELA chimeric proteins containing one or more ZFTA DNA binding motifs [27]. They hypothesized that the no-motif peaks might be bound by a ZFTA-RELA-containing protein complex that uses another pioneer subunit to initiate chromatin binding, necessitating a cis-acting mechanism. In our case, this ZFTA truncation led to an astroblastic phenotype that could correspond to the purely oncogenic cis-effect of ZFTA. This cis-acting hypothesis should be tested in mechanistic studies that are beyond the scope of our descriptive study. It is interesting to note that in our series with detailed histological typing, none of the ZFTA:NCOA1/2 fusions showed any astroblastoma phenotype, which highlights the potential role of the ZFTA partner in the histological phenotype. However, the number of ZFTA zinc fingers is also important. Previous studies have shown that the number of ZFTA zinc fingers in ZFTA:RELA fusions impact oncogenicity as well as the chromatin binding sites [28]. In ZFTA:RELA fusions, the 5'ZFTA part of the chimeric protein has one (RELAfus1) or two (RELAfus2) zinc finger domains while RELA is consistently almost full length. Parker et al. showed that neural stem cells (NSCs) transduced with RELAfus 1 generated tumors after intra-cerebral implantation in nude mice [4]. RELAfus2 NSCs also generated tumors albeit with clearly lower lethality. Using ChiP sequencing, Zhu et al. showed that chimeric protein RELAfus1 binds to 32,135 binding sites, RELAfus2 to 13,954 with only 5338 common sites with fus1 [27]. Therefore, the number of zinc finger domains in the ZFTA fusion necessarily impact the tumoral biology together with the fusion partner. Four ZFTA zinc finger domains is the rule in cases of non-RELA ZFTA-fused EPN published so far [6][7][8][9] and this was present in ten of our cases. However, whereas four ZFTA:NCOA1/2 with four zinc fingers corresponded to three classical ependymal histomorphologies and one PXA-like morphology, all three of our ZFTA:NCOA1/2 fusions with only one zinc finger corresponded to a sarcoma-like phenotype. We are the first to . The cases in this study, indicated as black dots, were in close proximity to the EPN_RELA subgroup.
In a more focused t-SNE analysis of the samples alongside the recently described satellite clusters of ZFTA-fusion positive ependymoma and YAP1-altered ependymoma, four of the cases grouped with cluster 4 and nine with cluster 2