Astroblastoma is a glial neoplasm previously diagnosed solely on the basis of histologic features including tumor cells with stout processes arranged in perivascular pseudorosettes (so-called “astroblastomatous” rosettes), often embedded within a sclerotic stroma [2, 3]. However, refinement in molecular-based tumor classification schemes has led to the realization that this histologic pattern is not specific to astroblastoma and can also be seen in other genetically defined tumor entities, including BRAF-mutant pleomorphic xanthoastrocytoma, ZFTA-fused supratentorial ependymoma, and IDH-wildtype glioblastoma [1, 7, 13]. A recent genomics study demonstrated that a large subset of tumors with astroblastoma histologic appearance harbor MN1 gene fusion (typically with BEND2 as the fusion partner) and resolve into a distinct epigenetic cluster, which was initially termed “high-grade neuroepithelial tumor, MN1-altered” [11]. Multiple subsequent studies have confirmed that MN1 fusions define a group of circumscribed glial neoplasms with astroblastoma-like morphology (with or without a primitive embryonal-like component), female predominance, and a propensity for local recurrence, but with mostly favorable clinical outcomes [1, 7, 13]. The identification of MN1 fusion by next-generation sequencing or MN1 rearrangement by break-apart fluorescent in situ hybridization (FISH) can be used to support the diagnosis of “astroblastoma, MN1-altered” per recent cIMPACT-NOW recommendations [8], and this molecularly-defined tumor type has been adopted in the 5th edition of the WHO Classification of Tumors of the Central Nervous System [4]. Of note, however, are a few recent case reports of gliomas with astroblastoma-like histology lacking MN1 alterations and instead harboring EWSR1-BEND2 fusion [9, 10, 12, 14]. The biologic relationship of these gliomas with EWSR1-BEND2 fusion compared to the newly defined tumor type “astroblastoma, MN1-altered” is currently unresolved.

We performed targeted next-generation sequencing and genome-wide DNA methylation profiling on a cohort of four patients with EWSR1-BEND2 fused gliomas. The one male and three females had a median age at time of initial diagnosis of 13 years (range 6–26 years) (Fig. 1a). Two patients had tumors located in the frontal lobe and presented with seizures, whereas the other two patients had tumors located in the brainstem or cervical spinal cord and presented with limb weakness. Imaging revealed enhancing mass lesions, sometimes associated with a cystic cavity (Fig. 1b, c; Supplementary Fig. 1 [Online Resource 1]). Extent of resection, adjuvant therapy, and clinical outcomes are listed in Supplementary Table 1 [Online Resource 2]. Histologically, these were circumscribed gliomas composed of epithelioid tumor cells embedded within a densely sclerotic stroma, with variably prominent astroblastomatous rosettes (Fig. 1d, e; Supplementary Fig. 2 [Online Resource 1]; Supplementary Table 2 [Online Resource 2]). A more primitive, embryonal-like component was prominent in two tumors and focally present in the other two. Mitotic activity was conspicuous (often > 5 mitoses per 2 mm2), and foci of necrosis were identified in all cases, but were not associated with peripheral palisading of tumor cells. No microvascular proliferation, Rosenthal fibers, or eosinophilic granular bodies were identified. Immunohistochemical evaluation revealed variable patchy expression of glial markers (GFAP and OLIG2), minimal to absent expression of neuronal markers (synaptophysin and neurofilament), and often diffuse membranous expression of EMA (Supplementary Fig. 3 [Online Resource 1]; Supplementary Table 3 [Online Resource 2]). Notably, immunostaining for BCOR demonstrated diffuse strong nuclear positivity in two tumors, despite an absence of BCOR mutation, tandem duplication, or other gene rearrangement. Meta-analysis of patient outcomes and other clinical data together with previous case reports revealed occurrence during childhood or early adulthood, location most frequently in midline structures (5/7 in the brainstem or spinal cord), and a propensity for local recurrence, with CSF dissemination occurring in two patients leading to mortality (Fig. 1f–h) [9, 10, 12, 14].

Fig. 1
figure 1

Clinicopathologic features and epigenomic profiling of astroblastoma-like gliomas with EWSR1-BEND2 fusion. a Summary table of the eight patients with gliomas harboring EWSR1-BEND2 fusion, four from this study and four previously reported. b MR imaging from patient #2 demonstrating an enhancing mass within the cervical spinal cord. c MR imaging from patient #4 demonstrating a solid and cystic mass with an enhancing mural nodule in the left frontal lobe of the brain. d Histology from patient #1 showing a glial neoplasm with dense pericellular and perivascular sclerosis. e Histology from patient #3 showing astroblastomatous rosettes. f Dot plot of patient age at diagnosis for astroblastoma-like gliomas with EWSR1-BEND2 fusion. g Dot plot of tumor anatomic location for astroblastoma-like gliomas with EWSR1-BEND2 fusion. h Kaplan–Meier plot of overall survival and recurrence-free survival for astroblastoma-like gliomas with EWSR1-BEND2 fusion. i tSNE dimensionality reduction plot of genome-wide DNA methylation profiles from the 4 EWSR1-BEND2 fused gliomas alongside 366 reference tumors spanning 11 CNS tumor entities. j Unsupervised hierarchical clustering of DNA methylation data from the 4 astroblastoma-like gliomas with EWSR1-BEND2 fusion and 3 astroblastomas with confirmed MN1-BEND2 fusion, with a heatmap of the 20,000 most differentially CpG sites shown

Full size image

By targeted next-generation DNA sequencing, all four tumors demonstrated EWSR1 fusion with BEND2 as the 3′ partner (Supplementary Fig. 4 [Online Resource 1]; Supplementary Table 4 [Online Resource 2]). In three of the tumors, the gene fusion linked exons 1–8 (codons 1–270) of EWSR1 with exons 2–14 (codons 9–799) of BEND2, whereas the remaining tumor linked exons 1–7 (codons 1–199) of EWSR1 with exons 4–14 (codons 126–799) of BEND2. The EWSR1-BEND2 fusion was the solitary oncogenic alteration identified in all tumors, with an absence of accompanying alterations involving IDH1/2, histone H3 genes, BRAF, NF1, PRKCA, FGFR1/2/3, NTRK1/2/3, EGFR, PDGFRA, MET, PIK3CA, PIK3R1, PTEN, CDKN2A, TP53, TERT (including promoter region), ATRX, CIC, FUBP1, MN1, ZFTA (c11orf95), RELA, YAP1, MYC, MYCN, MYB, and MYBL1 [6]. The quantity of chromosomal copy number aberrations was variable, but no tumors harbored whole arm co-deletion of chromosomes 1p and 19q, focal amplifications, or homozygous deletions (Supplementary Fig. 5 [Online Resource 1]; Supplementary Table 5 [Online Resource 2]). Genome-wide DNA methylation profiling was performed on the four tumors using Infinium EPIC 850k Beadchips (Illumina) following the manufacturer’s recommended protocols (see Supplementary Methods [Online Resource 3]). tSNE clustering of the DNA methylation data alongside a reference cohort of CNS tumors revealed that the EWSR1-BEND2 fused gliomas resolved into a single epigenetic group that was divergent from all established reference methylation classes of CNS tumors, but was in close proximity to “High-grade neuroepithelial tumor, MN1-altered” (Fig. 1i; Supplementary Fig. 6 [Online Resource 1]; sample manifest in Supplementary Table 6 [Online Resource 2]). Random forest classification using the online DKFZ Brain Tumor Classifier (www.molecularneuropathology.org) demonstrated high-confidence match to the “HGNET, MN1” methylation class for all four tumors (Supplementary Table 7 [Online Resource 2]) [5]. Unsupervised hierarchical clustering successfully segregated the 4 EWSR1-BEND2 fused gliomas from 3 astroblastomas with confirmed MN1-BEND2 fusion but revealed highly similar DNA methylation patterns overall (Fig. 1j).

Here we show that gliomas with EWSR1-BEND2 fusion exhibit astroblastoma-like histologic features and resolve into a single epigenetic subgroup most similar to the methylation class “HGNET, MN1” that encompasses the new tumor type astroblastoma, MN1-altered. It remains unclear why MN1-BEND2 and EWSR1-BEND2 fusions are selected for in astroblastoma, and why EWSR1 fusion appears to be more common in brainstem and spinal cord astroblastomas relative to MN1 fusion in hemispheric tumors, but we speculate that these represent two biologically distinct subtypes of astroblastoma which awaits confirmation in future studies. In epigenetically-defined astroblastomas with available sequencing data delineating the fusion partners, only a single tumor with MN1 fused to a partner other than BEND2 has been reported to date (dkfz_CNS-PNET_15-0184, GSM1881234 with MN1-CXXC5 fusion) [11]. Including this case series, there are now eight reported astroblastomas with EWSR1-BEND2 fusion (seven with available epigenetic data), indicating that BEND2 may be a more critical gene for molecular oncogenesis and relevant for tumor classification compared to MN1. This is reminiscent of how supratentorial ependymomas were initially defined as being RELA fusion positive, but the recognition of a group of supratentorial ependymomas with ZFTA (C11orf95) fused to other partners besides RELA has led to a revised nosology based on ZFTA as the defining fusion partner. Thus, these findings support consideration of BEND2 fusion as the defining molecular alteration for astroblastoma, rather than MN1 rearrangement as is currently recognized. Future studies are required to define the full biologic and clinicopathologic spectrum of astroblastomas, including establishment of evidence-based grading criteria and determination of optimal treatment regimens.