Recurrent fusions in PLAGL1 define a distinct subset of pediatric-type supratentorial neuroepithelial tumors

Tumor samples and retrospective clinical data from 40 patients were obtained from multiple national and international collaborating centers and collected at the Department of Neuropathology of the University Hospital Heidelberg (Germany). Sample selection was based on unsupervised visualization of genome-wide DNA methylation data that revealed a molecularly distinct group of tumors (n = 40) forming a cluster separate from all established entities. Due to the aspect of a multicenter cohort (23 different centers) including DNA methylation data that have been uploaded via the web platform https://​www.​molecularneuropa​thology.​org, availability of tissue and/or clinical data was restricted for some of the cases. A proportion of data was generated in the context of the Molecular Neuropathology 2.0 study. Analysis of tissue and clinical data was performed in accordance with local ethics regulations. Clinical details of the patients are listed in Supplementary Table 1 (online resource).

For all cases with sufficient material (n = 16), histological review of an H&E-stained slide was performed according to the World Health Organization (WHO) 2016 classification of tumors of the CNS [17]. Immunohistochemical staining was performed on a Ventana BenchMark ULTRA Immunostainer using the ultraView Universal DAB Detection Kit (Ventana Medical Systems, Tucson, AZ, USA). Antibodies were directed against: glial fibrillary acid protein (GFAP; Z0334, rabbit polyclonal, 1:1000 dilution, Dako Agilent, Santa Clara, CA, USA), epithelial membrane antigen (EMA; clone GP1.4, mouse monoclonal, dilution 1:1000, Thermo Fisher Scientific, Fremont, CA, USA), Sry-related HMG-BOX gene 10 (SOX10; clone EP268, rabbit monoclonal, dilution 1:100, Cell Marque Corp., Rocklin, CA, USA) and oligodendrocyte lineage transcription factor 2 (OLIG2; clone EPR2673, rabbit monoclonal, dilution 1:50, Abcam, Cambridge, UK).

Genome-wide DNA methylation profiling of all samples was performed using the Infinium MethylationEPIC (EPIC) BeadChip (Illumina, San Diego, CA, USA) or Infinium HumanMethylation450 (450 k) BeadChip array (Illumina) according to the manufacturer’s instructions and as previously described [5]. Raw data were generated at the Department of Neuropathology of the University Hospital Heidelberg, the Genomics and Proteomics Core Facility of the German Cancer Research Center (DKFZ) or at respective international collaborator institutes, using both fresh-frozen and formalin-fixed paraffin-embedded (FFPE) tissue samples. All computational analyses were performed in R version 3.6.0 (R Development Core Team, 2016; https://​www.​R-project.​org). Copy-number variation analysis from 450 k and EPIC methylation array data was performed using the conumee Bioconductor package version 1.12.0 [4]. Raw signal intensities were obtained from IDAT-files using the minfi Bioconductor package version 1.21.4. Illumina EPIC and 450 k samples were merged to a combined data set by selecting the intersection of probes present on both arrays (combineArrays function, minfi). Each sample was individually normalized by performing a background correction (shifting of the 5% percentile of negative control probe intensities to 0) and a dye-bias correction (scaling of the mean of normalization control probe intensities to 10,000) for both color channels. Subsequently, a correction for the array type (450 k/EPIC) was performed by fitting univariable, linear models to the log2-transformed intensity values (removeBatchEffect function, limma package version 3.30.11). The methylated and unmethylated signals were corrected individually. Beta-values were calculated from the retransformed intensities using an offset of 100 (as recommended by Illumina). All samples were checked for duplicates by pairwise correlation of the genotyping probes on the 450 k/EPIC array. To perform unsupervised non-linear dimension reduction, the remaining probes after standard filtering [5] were used to calculate the 1-variance weighted Pearson correlation between samples. The resulting distance matrix was used as input for t-SNE analysis (t-distributed stochastic neighbor embedding; Rtsne package version 0.13). The following non-default parameters were applied: is_distance = T, theta = 0, pca = F, max_iter = 10,000 perplexity = 20. DNA methylation sites of the PLAGL1 imprinted region (received via http://​www.​humanimprints.​net/​#data) were visualized in a heatmap using the R-package ‘pheatmap’. Control tissue DNA methylation samples (n = 119) as previously described [5] were used for comparison.

RNA was extracted from FFPE tissue samples using the automated Maxwell system with the Maxwell 16 LEV RNA FFPE Kit (Promega, Madison, WI, USA), according to the manufacturer’s instructions. Transcriptome analysis using messenger RNA (mRNA) sequencing of samples for which RNA of sufficient quality and quantity was available was performed on a NextSeq 500 instrument (Illumina) as previously described [31]. This was possible for 20 tumors within the novel group and 14 ZFTA:RELA-fused ependymomas. In addition, a reference cohort of other glioma and glioneuronal subtypes were used for differential gene expression analysis (YAP1:MAMLD1-fused ependymoma (n = 3), central neurocytoma (n = 9), extraventricular neurocytoma (n = 8), dysembryoplastic neuroepithelial tumor (n = 11), papillary glioneuronal tumor (n = 9), KIAA1549:BRAF-fused pilocytic astrocytoma (n = 14), diffuse midline glioma H3 K27M mutant (n = 14) and glioblastoma IDH wild-type (n = 9)). Fastq files from transcriptome sequencing were used for de novo annotation of fusion transcripts using the deFuse [20] and Arriba (v1.2.0) [36] algorithms with standard parameters. All further analysis was performed in R (version 3.6.0; R Core Team, 2019) using the DESeq2 package (v1.28.1) [19]. Principal Component Analysis (PCA) was performed after variance stabilizing transformation of the count data and normalization with respect to library size, based on the selection of the top 1,000 most variable genes with relative log expression normalization. Similarities between samples were determined by computing Manhattan distances on the variance stabilized data followed by unsupervised hierarchical clustering. Differential expression testing was performed on raw count data after fitting a negative binomial model. P-values were adjusted for multiplicity by applying the Benjamini–Hochberg correction.

Genomic DNA was extracted from FFPE tumor tissue samples of 18 patients within the cohort using the automated Maxwell system with the Maxwell 16 FFPE Plus LEV DNA Purification Kit (Promega, Madison, WI, USA), according to the manufacturer’s instructions. Capture-based next-generation DNA sequencing was performed on a NextSeq 500 instrument (Illumina) as previously described [29] using a custom brain tumor panel (Agilent Technologies, Santa Clara, CA, USA) covering the entire coding and selected intronic and promoter regions of 130 genes of particular relevance in CNS tumors (Supplementary Table 2, online resource).

Statistical analysis was performed using GraphPad Prism 9 (GraphPad Software, La Jolla, CA, USA). Data on survival could be retrospectively retrieved for eleven patients. Distribution of time to progression or recurrence (TTP) after surgery was estimated by the Kaplan–Meier method. Patients lost to follow-up are censored at date of last contact in analysis of TTP.

DNA methylation profiling has emerged as a powerful approach for robust classification of CNS neoplasms [5]. Using a screening approach based on unsupervised visualization of a large cohort of genome-wide DNA methylation data, we identified a highly distinct group of tumors (n = 40) forming a cluster separate from all established entities of which a high proportion of tumors (19/32, 59%) were histopathologically diagnosed as ependymoma. A more focused t-SNE analysis of DNA methylation patterns of these samples alongside 1100 other well-characterized glial and glioneuronal neoplasms (reference samples included in the current version of the Heidelberg DNA methylation classifier with a calibrated score > 0.9) confirmed the distinct nature of this novel group (Fig. 1). Analysis of copy-number variations (CNVs) derived from DNA methylation array data revealed a relatively balanced profile in most of the cases, with structural aberrations on chromosome 22q (21/40, 52.5%) and 6q (19/40, 47.5%) most frequently observed (Supplementary Fig. 1a, online resource). A chromothripsis-like pattern affecting chromosomes 6 and 13 was seen in one of the samples (Supplementary Fig. 1b, online resource). In one case, a homozygous deletion of CDKN2A/B was detected. An integrated plot of CNVs identified in all samples is given in Supplementary Fig. 1c (online resource).

Since a high proportion of supratentorial ependymomas are driven by gene fusions involving ZFTA (C11orf95, most frequently fused to RELA) or YAP1, we performed mRNA sequencing of all samples with sufficient material (n = 20). In 19/20 of the cases, a gene fusion involving PLAGL1 was detected, conserving the zinc finger structure (C2H2 type) as part of the fusion product, with either EWSR1 as 5’ partner or FOXO1 or EP300 as a 3’ partner (Fig. 2a–c). In the most common EWSR1:PLAGL1 fusions (n = 13), exons 1–9 or 1–8 of EWSR1 (NM_013986), which is located on chromosome 22q12.2, were fused to exon 5 of PLAGL1 (NM_001289039), which is found on chromosome 6q24.2. Five out of 20 cases with exons 1–5 of PLAGL1 fused to FOXO1 upstream of exons 2–3 (NM_0017612) were also observed. In one case, exons 1–5 of PLAGL1 are fused to exons 15–31 of EP300 (NM_001429). In all rearrangements, the DNA binding domain (zinc finger structure) of PLAGL1 was retained and fused to the respective transactivation domain (TAD) of the partner gene (Fig. 2a–c). We next performed an exploratory differential gene expression analysis of tumor samples (n = 20) within the novel group in comparison to ZFTA:RELA-fused ependymomas (n = 14). Unsupervised hierarchical clustering demonstrated a clear segregation of tumor samples in comparison to ZFTA:RELA-fused ependymoma (Fig. 2d). These results were recapitulated by PCA of normalized transcript counts (Fig. 2e). Quantification of mRNA expression revealed that the PLAGL1 gene itself was more highly expressed in tumors within the novel group than in ZFTA:RELA-fused ependymoma (adjusted p = 1.22e − 14; Fig. 2f, g). Additionally, upregulated genes of potential interest included H19 and IGF2 (adjusted p = 1.31e − 83, adjusted p = 5.04e − 08; Fig. 2h, i), both regulated by PLAGL1 and with known functions in the tumorigenesis of different cancers [38]. RELA and ZFTA transcript levels were upregulated in ZFTA:RELA-fused ependymomas (adjusted p = 1.03e − 61 and adjusted p = 1.10e − 19, respectively; Fig. 2j, k). Differential gene expression analysis between tumors within the novel group and a reference cohort of other glial and glioneuronal subtypes confirmed high transcript levels of PLAGL1 (adjusted p = 2.35e − 18), H19 (adjusted p = 9.12e − 15), IGF2 (adjusted p = 7.91e − 06) and DLK1 (adjusted p = 1.12e − 10) in the PLAGL1-fused cohort (Fig. 3a–c and Supplementary Fig. 2, online resource). Expression of particular markers differentially expressed in astrocytic and in ependymal neoplasms [10, 14, 21] revealed low OLIG2 and SOX10 expression (adjusted p = 3.89e − 26 and adjusted p = 7.75e − 65) within the novel group, with similar expression of GFAP (Fig. 3d–f and Supplementary Fig. 2, online resource). Moreover, the nearby imprinting control region (ICR) of PLAGL1 showed evidence for loss of imprinting in the corresponding DNA methylation profiles (Supplementary Fig. 3, online resource). Analysis of the mutational landscape of 19/40 tumors in the novel group using targeted next-generation sequencing revealed TERT promoter mutations (C228T) in two of the cases (Supplementary Table 1, online resource), with no other relevant events involving putative brain tumor genes.

Analysis of available clinical data demonstrated that median age of the patients at the time of diagnosis was 6.2 years (n = 25; range 0–30; with 92% of the tumors occurring in patients < 17 years of age, Fig. 4a) and the sex distribution was relatively balanced (F/M = 1:1.2, Fig. 4b). All tumors in our series were located supratentorially (Fig. 4c). The proportion of PLAGL1-fused tumors from all supratentorial tumors cannot yet be accurately determined. However, within the pediatric Molecular Neuropathology 2.0 study, PLAGL1-fused tumors account for approximately 0.7% of all supratentorial neoplasms included in the study. Outcome data were available for 11 patients. Median progression-free survival was 35 months (range 10–85 months; Fig. 4d). The initial histopathological diagnoses of the tumors within the cohort were relatively wide, although a high proportion of cases were designated as ependymoma (19/32, 59%). Other recurrent diagnoses included ‘embryonal tumor’ and different low- and high-grade gliomas (Supplementary Table 1, online resource). More detailed descriptions of the cases are given in Supplementary Table 1. A histopathological review of samples with available material (n = 16) confirmed a relatively wide morphological spectrum of tumors with ependymoma-like features (Fig. 5a–i). Histologically, all reviewed tumors shared a moderate to high increase in cellular density in a mostly fine neurofibrillary matrix with prominent microcystic changes (Fig. 5a–d). The tumor cells typically had monomorphic, round to oval nuclei with finely dispersed chromatin and prominent nucleoli. Single cases presented more pleomorphic cells. In many cases, perivascular pseudorosettes were observed, at least focally. Two of the cases showed focal oligodendroglial morphology with perinuclear halos due to cytoplasmatic clearing (Fig. 5e). Extensive calcification was seen in a small number of tumors (n = 3). Necrosis was not observed. Mitotic activity was generally low, with exception of two cases. Immunoreactivity for GFAP was present in all cases (Fig. 5f). The tumor cells neither expressed OLIG2 nor SOX10 (Fig. 5g, h). In 3/16 of the cases, a dot-like positivity for EMA was detected (Fig. 5i).