Molecular and clinicopathologic characteristics of CNS embryonal tumors with BRD4::LEUTX fusion

Genomic DNA and RNA was extracted from 5 μm sections of formalin fixed paraffin embedded tissue (FFPE) sections mounted on slides using the QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany), RNeasy FFPE Tissue Kit (Qiagen), or AllPrep DNA/RNA FFPE Kit (Qiagen). DNA was set aside for methylation analysis as described below. Next-generation sequencing (NGS) was performed using commercial TruSeq RNA Exome panel (Illumina). Exome RNA sequencing libraries were prepared with 100 ng tumor RNA using the Illumina RNA Prep with Enrichment (L) Tagmentation kit (Illumina) with Exome Probe Panel (Illumina). Final enriched libraries are sequenced on NextSeq 550DX or NovaSeq 6000 (Illumina).

After sequencing, the FASTQ files were aligned to human reference genome hg19 (GRCh37) using STAR aligner [9] to generate BAM files. The resulting BAM files were used by Arriba tool [10] to predict fusion calls. The filtered fusions (VCF file) are uploaded to the QIAGEN Clinical Insight (QCI; Qiagen) for annotation, classification, and interpretation. All variants were manually reviewed by visualizing the raw sequencing read alignments using the Integrative Genomics Viewer [11].

Samples were processed as previously described [12]. Briefly, genomic DNA (250 ng as the standard) was bisulfite-converted (EZ DNA Methylation Kit, Zymo Research D5001). When the total amount of DNA available was less than 250 ng, the sample was run as well. Bisulfite-converted FFPE DNA was processed with the Infinium FFPE DNA Restore kit (Illumina, USA) and was assayed on Infinium MethylationEPIC kit (Illumina, USA), according to the Infinium HD FFPE Methylation Assay automated protocol (Illumina, USA). Methylation data was processed using versions 11.b6 and 12.b6 of the DKFZ-Heidelberg classifiers and NCI-Bethesda classifier [6]. Copy number variant (CNV) profiles were inferred using the R “conumee” package (http://​bioconductor.​org/​packages/​conumee/​) as implemented in the classifier package. Tumor purity was estimated using RF_purify package and LUMP score from methylation profile [13, 14].

Five previously published cases [4‐8] were identified in the following manner. One case presented a dedicated case report for a tumor matching to the methylation class “CNS embryonal tumor with BRD4::LEUTX fusion”, with demonstration of the fusion [4]. A second case had been discovered in a large series of pediatric cancers where the BRD4::LEUTX fusion had been identified in one case of the series [5]. A third case was found in a series of medulloblastoma, where re-evaluation using version 12 of the Heidelberg classifier showed a match to CNS embryonal tumor with BRD4::LEUTX fusion [7]. A fourth case was included as a low-score match to group 3 medulloblastoma in the initial paper describing the Heidelberg version 11 methylation classifier [6], but which in our analysis of the Heidelberg version 12 classifier, matched to CNS embryonal tumor with BRD4::LEUTX fusion. The fifth case was described in a series of FOXR2-activated neuroblastomas, as a potential diagnostic mimic [8]. In addition to these 5 cases, 4 cases were identified in the clinical diagnostic methylation service at the National Institutes for Health. Clinical and molecular features for these 9 cases are summarized in Table 1. All patients for whom age was known were young children (up to 4 years) with large masses; three female, two male, and four of unknown sex. The initial (pre-methylation profiling) diagnosis, when known, suggested a CNS embryonal tumor, either NOS or a CNS neuroblastoma, FOXR2-activated.

Histologic and immunohistochemical features from our cases (#1–4), along with data retrieved from two previously published cases are described here. Histologically, all six of the assessed cases were highly cellular, showing embryonal morphology with diffuse arrangement (Fig. 1 and Additional file 1: Table S1), with a cell-streaming pattern seen in two cases, and Homer–Wright rosettes in one case. In two cases, the borders with brain parenchyma could be histologically assessed, showing sharp tumor circumscription. Neoplastic cells were small with round to oval nuclei, with two cases presenting some more irregularly shaped nuclei. Cytoplasm was scant. Scattered large, bizarre nuclei were seen in three cases. Four cases displayed necrosis. Brisk mitotic activity and apoptosis were seen in all samples. All cases displayed microvascular proliferation, often with marked multilayering of vessel walls. Calcifications were detected in two tumors histologically, and suspected in a third tumor by imaging exams, but could not be confirmed on the small biopsy examined. All tumors expressed synaptophysin (diffusely), and five expressed OLIG2 variably (Fig. 1). GFAP and neurofilament were expressed in a subset of tumor cells in 2/5 cases examined. Interestingly, one tumor had complete loss of H3k27me3 expression in tumor cells while three tumors showed a markedly attenuated expression of H3k27me3 in tumor cells with areas showing clear loss (Fig. 1) and areas with weak expression in a mosaic configuration. In all three cases tested, immunohistochemistry (IHC) was negative for the mutant protein H3 p.K27M. INI1 and BRG1 proteins were expressed in 5/5 tumors and LIN28 was negative in 3/3 tested. EZHIP was negative for 1/1 cases tested. We note a number of CNS tumor types where H3K27me3 expression is described as lost in specific cases, including IDH-mutant gliomas [15]). Results for all IHC performed are listed in Additional file 1: Table S2.

The DKFZ-Heidelberg classifier version 12.5 (and its subsequent iterations) included a methylation class “CNS embryonal tumour with BRD4:LEUTX fusion” which allowed for identification of tumors that matched to this class. Methylation profiling of our 4 cases, as well as 4 cases culled from the literature matched to this class. Seven of these 8 cases reached the 0.84 confidence threshold in the match, with the 8th case showing a ‘suggestive’ score (0.79) for this class (Table 1). RNA sequencing was performed for the 4 NIH cases and 3/5 outside cases, identifying the presence of the BRD4::LEUTX fusion for each of the 7 cases interrogated. All fusion junctions involved exon 2 of LEUTX, retaining the homeobox domain, while the BRD4 component of the fusions involved exons 11, 13, or 14, maintaining the bromodomain in all cases (Table 1 and Fig. 2). Inspection of the DNA copy number changes did not show recurrent alterations (copy number profiles for the 4 NIH cases are shown in Additional file 2: Fig. S1), and in addition copy number breakpoints at either the BRD4 or LEUTX loci were not a feature of this neoplasm. Gene expression was available for 3 of the 4 cases profiled at the NIH, where the expression of LEUTX and BRD4 was compared to the clinical dataset of > 2700 samples of a variety of tumor types with gene expression data available. In all 3 cases, expression of BRD4, and especially that of LEUTX was high relative to expression of these genes in the broad dataset of tumor types (Fig. 3). The elevated expression of LEUTX seen in these tumors is in line with a prior study that reported a BRD4::LEUTX fusion in a pediatric sarcoma [16]. Inspection of methylation signatures using UMAP for the four NIH cases and 2 of the previously published cases fore which (IDAT) files were available showed that CNS tumors with BRD4::LEUTX fusion cluster together, near but separate from group 3 medulloblastoma (Fig. 4). Lastly, DNA sequencing did not identify any recurrent pathogenic or likely pathogenic mutations in the 3 cases (cases 1–3) that were sequenced on a large (~ 500 gene) panel.

A representative series of images are shown for case #3, showing a large tumor with heterogeneous contrast enhancement, cystic change and mass effect (Fig. 5). With respect to patient follow-up, cases #1 (15-month-old female) and #2 (3-year-old male) were both alive without evidence of recurrence or progression at with near total resection followed by chemotherapy. Case #1 was alive one year post-therapy. Case #3 was a 14-month-old male who underwent a stereotaxic biopsy followed by four cycles of chemotherapy, with marked reduction on the enhancing solid tumor component, and gross total resection four months after the initial diagnosis. Two weeks post-surgery, there were signs of local recurrence and salvage therapy was indicated. Case #4 was a 12-month-old female with hydrocephalus with a heterogeneously enhancing mass with evidence of leptomeningeal spread. The patient underwent a stereotactic biopsy and the tumor showed striking response to chemotherapy with tumor shrinkage and nearly complete resolution of the intracranial leptomeningeal enhancement after chemotherapy (Fig. 6). Of note, cases #1–4 were all described to have been treated in a manner similar to clinical trial NCT00336024, where the chemotherapy regimen included cisplatin, cyclophosphamide, vincristine, and etoposide followed by thiotepa. Case #5 was a 4-year female [4] presenting with a large left parietal mass with multiple smaller supra- and infra- tentorial tumor nodules, as well as signs of subependymal and leptomeningeal dissemination. The patient died of disease shortly after surgery.