Introduction

BCOR, a member of the polycomb repressive complex that was originally discovered as an interacting corepressor of BCL6, plays a critical role in transcriptional repression [1, 2]. Alterations in BCOR have been reported in various tumors of soft tissue, bone, and visceral organs. Pierron et al. were the first to identify bone and soft tissue sarcomas with BCOR::CCNB3 fusion in adolescents and young adults [3]. Subsequent studies on sarcoma identified broad types of BCOR alterations, including BCOR fusions with alternative partners [4,5,6] and internal tandem duplications (ITDs) within exon 15 of BCOR [7]. BCOR fusion and ITD have also been reported in clear cell sarcoma of the kidney and high-grade endometrial stromal sarcoma [8,9,10,11,12]. Although these tumors with BCOR alterations exhibit different clinical characteristics, they are associated with overlapping histological features, including a dense proliferation of uniform oval cells in a variably myxoid and hypervascular stroma, and share immunoprofiles, such as overexpression of BCOR and SATB2 [5]. Furthermore, genome-wide DNA methylation analysis identified these tumors as distinct clusters that juxtaposed with one another [13], indicating that they are related but non-identical entities.

In the central nervous system (CNS), BCOR ITD was first identified through (epi)genetic analysis of tumors originally diagnosed as primitive neuroectodermal tumors (PNETs) of the CNS [14]. These tumors exhibit characteristic histological findings and DNA methylation profiles [14]. This has led to the proposal of the tumor entity “high-grade neuroepithelial tumor with BCOR alteration (HGNET-BCOR),” which has been recently renamed “CNS tumor with BCOR internal tandem duplication” in the fifth edition of the World Health Organization (WHO) tumor classification [15]. The tumor affects pediatric and young adult patients and mainly exhibits a solid growth pattern of spindle to oval cells with perivascular pseudorosettes and branching capillary networks [14, 16,17,18]. Subsequent studies have further described brain tumors with fusions involving BCOR or BCORL1 (a homolog of BCOR), such as EP300::BCORL1 [19], CREBBP::BCORL1 [20], and EP300::BCOR fusions [21,22,23], which has further expanded the disease concept of BCOR-altered brain tumors. However, the clinicopathological and DNA methylation profiles reported for these tumors were not identical [21,22,23], and whether CNS tumors with BCOR/BCORL1 fusion represent a single nosologic entity has been a topic of controversy. This study reported a CNS tumor with EP300::BCOR fusion and compared its clinicopathological and molecular characteristics with those of previously reported tumors.

Clinical summary

A 32-year-old woman presented to an outside institution with complaints of right-sided visual impairment. The patient was healthy, except for the occurrence of occasional right-sided scintillating scotoma in the past year. The patient was referred to the National Cancer Center Hospital where the visual acuity test revealed right-lower homonymous hemianopsia. Imaging studies revealed a 60-mm mass in the left occipital lobe. The mass was well-circumscribed and hyper-intense on the T2-weighted magnetic resonance (MR) image and hypo-intense to focally hyper-intense on the T1-weighted MR image with heterogeneous gadolinium enhancement (Fig. 1A–D). Calcification and intratumoral hemorrhage were observed. The patient underwent tumor excision, followed by postoperative local radiation (54 Gy) therapy, based on the suspected diagnosis of anaplastic ependymoma. However, the tumor recurred locally 18 months later and was resected. The patient was disease-free three months after the second surgery although the right-lower homonymous hemianopsia was not resolved.

Fig. 1
figure 1

Magnetic resonance images of a high-grade neuroepithelial tumor with EP300::BCOR fusion. A–D The magnetic resonance image findings of a primary preoperative tumor. A well-demarcated occipital mass exhibited a hypo-intense signal with a focally hyper-intense signal on a T1-weighted image (A) and a hyper-intense signal on a T2-weighted (B) and fluid attenuated inversion recovery (FLAIR) (C) images. The tumor exhibited heterogeneous enhancement on a gadolinium-enhanced T1-weighted image (D)

Full size image

Histopathological and genetic findings

The analysis of the specimen obtained at the first surgery revealed solid growth of uniform small oval cells with prominent perivascular pseudorosettes and branching capillary blood vessels (Fig. 2A–D). The tumor was well-circumscribed. The tumor cells had poorly defined fibrillary cytoplasm and round-to-oval nuclei with fine or granular chromatin and small or inconspicuous nucleoli. The mitotic activity was counted 10 in 2 mm2. Necrosis, calcification, and focal microvascular proliferation were also observed. Microcystic changes, Rosenthal fibers, and eosinophilic granular bodies were absent. Immunohistochemical analysis revealed that the tumor tested positive for GFAP, D2-40, and CD99, and focally for OLIG2 (Fig. 2E) but tested negative for IDH1-R132H, epithelial membrane antigen (EMA), BCOR (Fig. 2F), and p65. The Ki-67 labeling index was 40%. The primary antibodies used in the immunohistochemical analysis are listed in Supplementary Table 1.

Fig. 2
figure 2

Histological findings of a high-grade neuroepithelial tumor with EP300::BCOR fusion. The tumor comprised uniform small oval cells with prominent perivascular pseudorosettes (A, B). Mitoses were frequently observed (B, inset). The tumor margins were well-demarcated (C). Necrosis was observed (D). Immunohistochemically, the tumor cells were positive for OLIG2 (E) but negative for BCOR (F). Original magnification: × 100 (A, C, D), × 200 (B), or × 400 (B inset, E, F)

Full size image

DNA pyrosequencing, which was performed at the time of diagnosis using methods described previously [19], did not reveal mutations in the IDH1, IDH2, H3-3A, BRAF, and TERT promoters. Reverse transcription polymerase chain reaction analysis did not reveal ZFTA::RELA fusion. The overall histology strongly suggested the diagnosis of anaplastic ependymoma. However, the tumor exhibited unusual features, including focal OLIG2 expression, a lack of RELA fusion, and negative p65 expression. Therefore, the definitive diagnosis was deferred, and the case was diagnosed as “glioma, most consistent with anaplastic ependymoma” with a note on the unusual features.

The histological characteristics of the specimen obtained at the second surgery were similar to those of the specimen obtained at the first surgery, including a dense proliferation of uniform oval cells and abundant perivascular pseudorosettes. However, examination of the tumor periphery revealed infiltrating growth of tumor cells that entrapped neurons and neurofilament-positive axons. Immunohistochemical analysis revealed that the tumor tested positive for GFAP, NeuN, and SATB2, and focally for OLIG2 expression. ATRX staining was retained. The tumor was negative for CD34, EMA, synaptophysin, MDM2, and BCOR. This specimen was obtained after the publication by Tauziède-Espariat et al. [22], and we found the tumor shared with their cases some histological features (e.g., perivascular pseudorosettes and delicate branching capillaries) and BCOR-negative phenotype. Suspecting the presence of EP300::BCOR fusion, we performed BCOR break-apart fluorescence in situ hybridization (FISH) assay (RP11-77G22 + RP11-665O2 labeled in orange; RP11-91I16 + RP11-1082P20 labeled in green, Chromosome Science Labo, Hokkaido, Japan), which revealed BCOR rearrangement in most tumor cells (Fig. 3A). To further characterize BCOR fusion, total RNA was extracted from formalin-fixed paraffin-embedded tumor sections. An RNA sequencing library was prepared using a TruSight RNA Pan-Cancer library kit (Illumina, San Diego, CA, USA). The library was subjected to paired-end sequencing on a MiSeq DNA sequencer. The fusion of EP300 (exon 31, NM_001429.4) and BCOR (exon 6, NM_001123385) was identified using the RNA-Seq alignment application (Illumina) (Fig. 3B). Clinical FoundationOne CDx (Foundation Medicine, Cambridge, MA, USA) confirmed the EP300::BCOR fusion but no other alterations were detected in the target genes. DNA methylation analysis was performed using an Infinium Methylation EPIC array platform (Illumina). The Deutsches Krebsforschungszentrum (DKFZ) classifier (v11b4) predicted the case as “methylation class CNS high-grade neuroepithelial tumor with BCOR alteration” with a low calibrated score of 0.38. However, the newer version of the DKFZ classifier (v12.5) classified the case as “methylation class CNS tumor with BCOR/BCORL1 fusion” with a calibrated score of > 0.99. To perform unsupervised nonlinear dimension reduction, the 1000 most variable probes were selected from the reference samples of 2801 CNS tumors (GSE109381) [24] based on the standard deviation. t-distributed stochastic neighbor embedding (t-SNE) plots were constructed using the Rtsne package (version 0.15). In the t-SNE plots, the tumors were clustered near the HGNET-BCOR reference samples (Fig. 3C). A copy number plot derived from methylome data revealed an overall flat profile (Fig. 3D).

Fig. 3
figure 3

Molecular findings of a high-grade neuroepithelial tumor with EP300::BCOR fusion. A BCOR break-apart fluorescent in situ hybridization (FISH) analysis. Split green and orange signals (arrows) were observed in most tumor cells, indicating BCOR rearrangement. Additional isolated green signals were also observed. B Schematic presentation of the predicted chimeric EP300::BCOR protein. RBD, RORA-binding domain; ZF, zinc finger domain; KIX, kinase-inducible domain of CREB-interacting domain; Bromo, bromodomain; HAT, histone acetyltransferase domain; NBD, NCOA2-binding domain; BBD, BCL6-binding domain; ANK, ankyrin repeats; PUFD, PCGF Ub-like fold discriminator. C t-distributed stochastic neighbor embedding analysis of DNA methylation data. The present tumor (red square) was clustered near the high-grade neuroepithelial tumor with BCOR alteration reference samples (blue dots). D Copy number profiling using DNA methylation data demonstrated a relatively silent chromosomal copy number status

Full size image

The results of the present case are summarized in Table 1 along with data from previously reported CNS tumors with BCOR/BCORL1 fusion.

Table 1 Reported CNS tumors with BCOR/BCORL1 fusions
Full size table

Discussion

This report describes a new case of CNS neuroepithelial tumor with EP300::BCOR fusion. EP300::BCOR fusion was initially reported by two groups in five CNS tumors. However, three tumors described by Torre et al. [21] and two tumors reported by Tauziède-Espariat et al. [22] exhibited different clinicopathological characteristics even though the tumors shared the same fusion profile. The tumors reported by Torre et al. [21] occurred in one female and two males aged 10–18 years and involved the basal ganglia/thalamus, frontal lobe, or occipital lobe. The tumor exhibited an infiltrating growth pattern, a myxoid/microcystic background, and prominent chicken-wire vessels but lacked perivascular pseudorosettes. One case exhibited low-grade histological characteristics, while the other two exhibited high-grade histology (i.e., necrosis, microvascular proliferation, and mitotic activity) in addition to low-grade areas. Immunohistochemical analysis revealed BCOR expression in these tumors. The DKFZ methylation classifier (probably v11b4 based on the publication time) could not classify the tumors. In the t-SNE plots, the three cases clustered together but away from CNS HGNET-BCOR [21]. In contrast, Tauziède-Espariat et al. [22] reported two tumors in the temporal or frontal cerebral lobe involving a 13-year-old man and a 27-year-old man. The tumor formed well-circumscribed masses with minimal peripheral infiltration. Perivascular pseudorosettes, microcyst formation, and chicken-wire vessels were also observed. The tumor exhibited high-grade histological features, such as necrosis, microvascular proliferation, and mitotic activity. The tumors were immunohistochemically negative for BCOR [22]. The DKFZ methylation classifier (probably v11b4 based on the publication time) classified the tumors as HGNET-BCOR [22]. The present tumor resembles the cases reported by Tauziède-Espariat et al. [22] as it exhibited overall circumscription, high-grade histological features, perivascular pseudorosettes, chicken-wire vessels, and BCOR-negative immunophenotype although microcysts were absent. The DKFZ classifier (v11b4) predicted a low-confidence match with HGNET-BCOR. In the t-SNE plot, the tumor was clustered near HGNET-BCOR. Whether these two groups represent separate entities or form the phenotypic spectrum of a single disease is unknown.

The present tumor may also be related to CNS tumors with a fusion involving BCORL1 (a BCOR homolog). Two CNS tumors have been reported to harbor BCORL1 fusion with related partners. Fukuoka et al. [19] reported a tumor harboring EP300::BCORL1 fusion in the occipital lobe of a male patient aged 72 years. Similar to the tumor described in this study, the tumor exhibited histological features of anaplastic ependymoma and was classified as HGNET-BCOR by the DKFZ methylation classifier (v11b4) with a low calibrated score of 0.44. Yamazaki et al. [20] described a glioma with CREBBP::BCORL1 fusion (CREBBP is an EP300 paralog) that exhibited infiltrating growth, microcysts, and a lack of perivascular pseudorosette [20]. The DKFZ classifier (v11b4) could not classify the tumor (score < 0.3). Similar to the present tumor, both cases were classified as “methylation class CNS tumor with BCOR/BCORL1 fusion” by a new version of the classifier (v12.5) [20]. Nonetheless, BCOR expression was positive in the tumor with CREBBP::BCORL1 and was not reported in the tumor with EP300::BCORL1. The role and utility of BCOR expression in tumor classification must be determined in future studies.

The correlation between BCOR expression and BCOR fusion is complex. Immunohistochemical analysis revealed that the present tumor exhibited a BCOR-negative phenotype despite BCOR fusion. This can be attributed to the use of anti-BCOR antibody (clone C-10), which recognize 300 amino acid residues at the N-terminal (exons 1, 2, and 3 and a part of exon 4), and the predicted fusion protein lacking an antibody recognition site with a BCOR breakpoint in exon 6. Negative BCOR expression of the tumors reported by Tauziède-Espariat et al. [22] can be explained similarly as the tumors had a BCOR breakpoint in exon 4. However, the tumors reported by Torre et al. [21] were immunopositive for BCOR even though the BCOR breakpoint was in exon 2 or exon 7, contributing to the loss of some or all of the first 300 amino acid residues of BCOR. Furthermore, a glioma with CREBBP::BCORL1 exhibited upregulated levels of BCORL1 and BCOR mRNAs and was immunopositive for BCOR even though it lacked BCOR genetic aberrations [20]. Similar inconsistent BCOR expression has been reported in high-grade endometrial stromal sarcoma with ZC3H7B::BCOR fusion [12]. In these tumors, BCOR immunoexpression was reported in 3 cases with BCOR breakpoints in exon 7 or 14 [12]. The difference in BCOR expression may involve the breakpoint of BCOR fusion and the expression of BCOR in the other allele on the X chromosome in women.

Pisapia et al. reported a CNS tumor with a reciprocal fusion combination BCOR::CREBBP [25]. The phenotypes of this tumor were completely different from those of the tumors discussed above [21, 22]. The tumor initially manifested as a “gliomatosis cerebri” in a 15-year-old boy [25] with a predominantly low-grade diffuse astrocytoma histology with ATRX loss. BCOR expression was absent, and the tumor harbored TERT promoter mutation (c.-124C > T) with no mutations in IDH1/IDH2 and H3-3A [25]. The recurrent tumor after chemoradiotherapy progressed to glioblastoma histology [25]. This reciprocal BCOR::CREBBP fusion does not share most exons with EP300::BCOR [21, 22] and was predicted to be an out-of-frame fusion resulting in premature stop codon in CREBBP [25], which along with a TERT promoter mutation may have influenced phenotypic differences.

Very recently, Wu et al. [23] reported 21 cases of CNS tumors that formed a coherent DNA methylation group and harbored EP300::BCOR fusion (n = 11), CREBBP::BCOR fusion (n = 1), MEAF6::CXXC5 fusion (n = 1), or BCOR stop-gain mutations (n = 2) or exhibited undetermined BCOR status (n = 6). These cases exhibited variable histological features but likely encompass the characteristics described in this and previous reports [21, 22]. The tumors with EP300::BCOR fusion occurred mainly in children and young adults (six males and five females) with the age of occurrence in the range of 5–72 years (median, 21 years), predominantly involving the cerebral hemisphere [23]. Most tumors were indeterminate for peripheral infiltration. Perivascular pseudorosettes were observed in half of the cases [23]. BCOR immunoexpression were observed in 4 of 8 tested cases, which included one case with EP300::BCOR fusion that was predicted not to maintain a BCOR antibody recognition site [23]. Most tumors exhibited high mitotic activity and necrosis [23]. The DKFZ classifier (v12b6) classified most tumors with EP300::BCOR fusion as “CNS tumor with EP300:BCOR(L1) fusion,” while two cases were classified as “neuroepithelial tumor with BCOR internal tandem duplication.” [23] One tumor with MEAF6::CXXC5 fusion was classified as “CNS tumor with EP300:BCOR(L1) fusion” even though it did not exhibit BCOR alterations. The imperfect concordance between methylation class, phenotype, and genetic aberrations suggests a complex (epi)genetic/phenotypic relationship and indicates that EP300::BCOR fusion is not the sole determinant of tumor characteristics.

In conclusion, we reported a new case of high-grade CNS tumor with EP300::BCOR fusion, which exhibited well-circumscribed ependymoma-like histological features and negative BCOR immunoexpression. BCOR/BCORL1-altered tumors should be considered in the differential diagnosis of supratentorial CNS tumors with ependymoma-like histological features, especially when they lack ZFTA fusion or express OLIG2 even in the absence of BCOR expression. CNS tumors with BCOR/BCORL1 fusions appear to share partly overlapping, but non-identical phenotypes, and further studies of additional cases are required to establish their classification.