CNS tumors with YWHAE:NUTM2 and KDM2B-fusions present molecular similarities to extra-CNS tumors having BCOR internal tandem duplication or alternative fusions.

BCOR (BCL6 Corepressor) internal tandem duplication (ITD) has been implicated in a wide variety of tumors of different organs [1]: clear cell sarcomas of the kidney (CCSK), high-grade endometrial stromal sarcomas (HGESS), undifferentiated round cell sarcomas (URCS) in the soft tissue, and tumors within the central nervous system (CNS). BCOR is part of the polycomb repressive complex 1.1 (PRC1.1), in association with the KDM2B (Lysine Demethylase 2B) protein, that mediates transcriptional repression of oncosuppressors through post-translational modifications of histones. A variable proportion of CCSK, HGESS and URCS present the YWHAE:NUTM2 fusion which is always found in mutual exclusion with the BCOR ITD. Tumors with YWHAE:NUTM2 fusions also exhibit BCOR up-regulation, reinforcing the hypothesis that these two alterations activate a common pathogenetic pathway [2, 3]. CNS tumors, isolated from a series of primitive neuroectodermal tumors by a distinct methylation profile, were initially named high-grade neuroepithelial tumors (HGNET) with BCOR alteration [4]. Because almost all HGNET-BCOR harbored BCOR ITD and of an unknown cellular origin, the cIMPACT-NOW update 6 recommends the terminology “CNS tumor with BCOR ITD” [4-6]. Here, we report four CNS tumors with YWHAE:NUTM2 or KDM2B fusions, which did not cluster with HGNET-BCOR by DNA-methylation analysis, but clustered with extra-CNS sarcomas with BCOR ITD or analog fusions.

These pediatric cases included a 1-year old boy (Case 1), a 7-year old boy (Case 2), a 10-year old girl (Case 3), and a 7-year old girl (Case 4). Tumors were located in the left temporal dura-mater (Case 1), the left temporal and parietal lobes (Case 2), the right parietal and occipital lobes (Case 3), and in the pons (Case 4). Case 1 was an extra-axial mass with an intense and homogeneous enhancement, and a restricted apparent diffusion coefficient (ADC), reflecting high cellularity (Fig. 1A). Case 2 was a large and well-circumscribed solid tumor with hemorrhaging and necrosis, and a slight enhancement after contrast injection (Fig. 1E). Case 3 was a well-circumscribed polycystic tumor with an intense enhancement after contrast injection and slight hypersignal on diffusion weighted images (Fig. 1I). Case #4 was a multinodular enhancing mass in the pons with non-enhancing additional tumoral infiltration on FLAIR sequence (Fig. 1M). Whole body imaging did not evidence a spinal, leptomeningeal or extra-cranial location. Three patients were alive at the end of the follow-up (12, 11, and 179 months, respectively for Cases 1, 2, and 3), only Case 4 died postoperatively. Histopathologically, these tumors were circumscribed from the parenchyma. There was an intra-tumoral hetereogeneity: oligodendroglial-like, undifferentiated (Fig. 1B and N), or ependymal features with pseudorosettes (Fig. 1F and J). Microcysts containing a myxoid substance and calcifications were respectively observed in Cases 2 and 3. Features of malignancy were obvious with necrosis, a high mitotic count and proliferation index, and microvascular proliferation in both cases. Using immunohistochemistry, there was a preserved expression of INI1 and BRG1 and no immunoexpression of LIN28A. The expression of GFAP and Olig2 was absent in three cases and focal in the last tumor. Expression of at least one neuronal marker (NeuN and neurofilaments) was present in three cases. A BCOR immunoexpression was absent or only focally present in 3/4 cases (Fig. 1D, H and L). Analysis of the RNA-seq data identified YWHAE:NUTM2A (Case 1), KDM2B:NUTM2B (Case 2), YAP1:KDM2B (Case 3), and CHST11:KDM2B (Case 4) fusions and confirmed by two of the five different methods of detection we use (Defuse V0.6.2, StarFusion v1.2.0 (STAR v 2.5.4a), Fusion Catcher v1.00, FusionMap (Oshell toolkit v10.0.1.50) and ARRIBA v1.2.0) (Fig. 2). Using the Heidelberg DNA methylation classifier, Cases 1 and 4 were not classified and Cases 2 and 3 were classified as HGNET-BCOR (with calibrated max-scores of 0.8 and 0.2). To better characterize the potential cellular origin of these cases, we performed a t-Distributed Stochastic Neighbor Embedding plot (t-SNE) analysis including HGNET-BCOR and sarcomas with BCOR alterations (HGESS and URCS, also referred to as Small Blue Round Cell Tumours—SBRCT—in the Heidelberg database) and four soft tissue sarcomas with YWHAE:NUTM2 or KDM2B fusions from our in-house database (Fig. 3). Interestingly, our four CNS cases clustered in close vicinity with sarcomas and not HGNET-BCOR (Fig. 3).

Fig. 1

Radiological and histopathological features. A Axial T1-weighted sequence with gadolinium showing an enhanced extra-axial left temporal mass. B Compact tumor with undifferentiated morphology (HPS, magnification × 400). C No immunoexpression of GFAP (magnification × 400). D Expression of BCOR in tumor cells (magnification × 400). E Axial T1-weighted sequence with gadolinium showing an enhancing polycystic mass in the right parietal lobe. F Proliferation composed of pseudorosettes (HPS, magnification × 400). G Focal expression of GFAP (magnification × 400). H BCOR immunoexpression in a part of tumor cells (magnification × 400). I Axial tomodensitometry with contrast showing a hemorrhagic lesion of the left parietal lobe with slight contrast enhancement. J Proliferation composed of pseudorosettes (HPS, magnification × 400). K Absence of immunoexpression of GFAP (magnification × 400). L Expression of BCOR (magnification × 400). M Sagittal T1-weighted sequence with gadolinium showing an enhancing multinodular lesion of the pons. N Compact tumor with nodular arrangement of undifferentiated cells (HPS, magnification × 400). O No immunoexpression of GFAP (magnification × 400). P No immunoexpression of BCOR (magnification × 400). Black scale bars represent 50 µm. Each line represents a case: Case 1 for A-D pictures; Case 2 for E–H pictures, Case 3 for I-L pictures, and Case 4 for M-P pictures. HPS: Hematoxylin Phloxin Saffron

Fig. 2

Genetic features. A RNAseq analysis highlights a fusion between YWHAE (pink) and NUTM2A (blue) genes, respectively located on chr17p13.3 and chr10q23.2. As the breakpoints are intra exonic (in exon 5 for YWHAE, and exon 2 for NUTM2A), the fusion point can easily been detected by split and span reads encompassing the rearrangement with a good coverage. Localized on minus strand (inverse orientation), the DNA sequence of NUTM2A is switched in frame with YWHAE (Circos plot). B RNAseq analysis highlights a fusion between KDM2B (pink) and NUTM2B (blue) genes, respectively located on chr10.q22.3 and chr12.q24.31. As the breakpoints are intra exonic (in exon 22 for KDM2B, and exon 5 for NUTM2B), the fusion point can easily been detected by split and span reads encompassing the rearrangement with a good coverage. Localized on minus strand (inverse orientation), the DNA sequence of NUTM2B is switched in frame with KDM2B (Circos plot). C RNAseq analysis highlights a fusion between YAP1 (pink) and KDM2B (blue) genes, respectively located on chr11q22.1 and chr12q24.31. As the breakpoints are intra exonic (in exon 5 for YAP1, and exon 13 for KDM2B), the fusion point can easily been detected by split and span reads encompassing the rearrangement with a good coverage. Localized on minus strand (inverse orientation), the DNA sequence of KDM2B is switched in frame with YAP1 (Circos plot). For this case, the quality of the method of detection was not perfect and we cannot affirm that it was not a reciprocal fusion. D RNAseq analysis highlights a fusion between CHST11 (pink) and KDM2B (blue) genes, respectively located on chr12q23.3 and chr12q24.31. As the breakpoints are intra exonic (in exon 2 for CHST11, and exon 12 for KDM2B), the fusion point can easily been detected by split and span reads encompassing the rearrangement with a good coverage. Localized on minus strand (inverse orientation), the DNA sequence of KDM2B is switched in frame with CHST11 (Circos plot). The representations of fusion transcripts were obtained by RNA-seq analysis and built by the ARRIBA bioinformatics fusion finder tool which sometimes highlights reciprocal fusions. The fusion transcripts between the 5' gene (pink) and the 3' gene (blue) are represented on the left. The sequences of the fusion points are written below the transcript pattern. The fusions are systematically exon-exon and preserve the reading frame. The right columns show the fusion proteins and the domains retained by the fusions. Protein domains legend: 14-3-3: 14-3-3 domain superfamily; JmjC: Jumonji family of transcription factors; Cxxc: Zinc finger CXXC-type; Fbox: F-box-like domain superfamily

Fig. 3

t-SNE distributions based on DNA Methylation and RNA-sequencing. The t-SNE is built with the 10 first principal components (PC) of the DNA methylation -values with a standard deviation superior to 20%. Our four tumors were compared with 154 reference samples from the Heidelberg CNS tumors and sarcomas cohorts, belonging to the HGNET-BCOR (n = 23), HGNET-MN1 (n = 21), EPN-RELA (n = 70), EPN-YAP (n = 10), EFT-CIC (n = 13), SBRCT-BCOR (n = 8) and ESS HG (n = 9) methylation classes. We added four in-house sarcomas with YWHAE:NUTM2 or KDM2B fusions. Our four tumors clustered with sarcomas and not HGNET-BCOR. HGNET-BCOR: high-grade neuroepithelial tumors with BCOR alteration; HGNET-MN1, high-grade neuroepithelial tumors with MN1 alteration; EPN-RELA, ependymomas with RELA fusion; EPN-YAP, ependymomas with YAP fusion; EFT-CIC, Ewing’s sarcoma family of tumors (with CIC alteration); SBRCT-BCOR, small blue round cell tumors with BCOR alteration; ESS HG, high grade endometrial stromal sarcomas

Contrary to HGNET-BCOR, CCSK and HGESS may present either BCOR ITD or YWHAE:NUTM2A/B fusions [3]. Here, we report four CNS tumors harboring YWHAE:NUTM2 or KDM2B fusions which differed from classical CNS tumors with BCOR ITD by clinical, radiological, immunophenotypical and molecular findings. Indeed, whereas the prognosis of HGNET-BCOR is poor [5], the outcome of our cases seems to be better, except for Case 4 most likely due to the tumoral pontine location. Imaging of our cases was different from those described in HGNET-BCOR such as large, solid and well-circumscribed intra-axial tumors that abut the overlying dura, with restricted diffusion and weak heterogeneous enhancement after contrast injection [6, 7]. Morphologically, all tumors were undifferentiated or arranged in small nodules of round clear cells, mimicking CCSK [7]. However, two cases presented ependymoma-like features with pseudorosettes, which have never been seen in extra-CNS tumors, contrary to CNS tumors with BCOR ITD. Immunohistochemically, only 1/4 of these tumors expressed Olig2, which is classically diffuse in CNS tumors with BCOR ITD [6, 7]. As previously described, BCOR immunoexpression was absent or focal in our cases without BCOR ITD [8, 9]. Lastly, DNA methylation clustering showed a close proximity of these cases to sarcomas with BCOR alterations. Because DNA methylation profiles are thought to represent a combination of both somatically acquired DNA methylation changes and a signature reflecting the cell of origin, and because no extra-CNS lesion was found in our cases, it is therefore reasonable to believe that they represent another tumor type than classical CNS tumors with BCOR ITD. Here, we report for the first time fusions implicating the KDM2B gene in the CNS; interestingly, only one case was previously described with a EPC1-KDM2B fusion in soft tissue [10]. KDM2D fusions were also found in HGESS and SBRCT [11].

To conclude, YWHAE:NUTM2 and KDM2B-fused CNS tumors aggregate within the category of sarcomas with BCOR alterations based on their DNA methylation signatures, despite some morphological features mimicking CNS tumors with BCOR ITD. Our results suggest that CNS tumors with these types of fusions represent a CNS location of mesenchymal tumors, but more cases (with DNA methylation and expression analyses) are needed for confirmation. Current diagnostic tools involving automated classification based on DNA methylation were proved inaccurate in two of our four cases. The diagnosis of CNS tumors, BCOR ITD must therefore be ascertained by precise clustering studies.