Amplification of the PLAG-family genes—PLAGL1 and PLAGL2—is a key feature of the novel tumor type CNS embryonal tumor with PLAGL amplification

Methylation analysis

Unsupervised visualization of genome-wide DNA methylation data using t-distributed stochastic neighbor embedding (t-SNE) of > 90,000 pediatric and adult tumor samples of numerous types revealed a subset of 46 tumor samples clustering closely together, but away from established DNA methylation reference classes. Investigation of copy number alterations in each sample indicated amplification of the genetic loci corresponding to one of two different PLAG-family genes (PLAGL1 at 6q24.2 or PLAGL2 at 20q11.21) in the majority of tumors—a genetic aberration not known to be a characteristic feature in any of the currently defined CNS tumor types. We, therefore, provisionally named this novel DNA methylation class “CNS embryonal tumor with PLAG-family gene amplification”—ET, PLAGL (Fig. 1a).

Fig. 1figure 1

DNA methylation clustering identifies a novel epigenetically distinct subtype of CNS embryonal tumor characterized by focal PLAG-family gene amplification. a Left: DNA methylation-based t-SNE analysis of > 90,000 pediatric and adult tumor samples. Circled are different medulloblastoma (MB) and embryonal tumor with multilayered rosettes (ETMR) subtypes, the ET, PLAGL type, and various low grade and high grade glioma subtypes—pilocytic astrocytoma (PA), pleomorphic Xanthoastrocytoma (PXA), H3 G34-mutant diffuse hemispheric glioma (G34), H3 K27-altered diffuse midline glioma (K27), diffuse pediatric-type high grade glioma, RTK subtype (pedRTK). Right: enlarged depiction of samples belonging to the ET, PLAGL type. The arrows mark two slightly outlying samples based on t-SNE. Methylation classes are color-coded as described in [12], grey color means the sample could not be matched to any of the existing methylation classes. b DNA methylation-based analysis using t-SNE dimensionality reduction on 33 ET, PLAGL tumors and a reference cohort of 910 different CNS tumors including 780 gliomas/glioneuronal tumors and 130 medulloblastomas. Methylation classes are color-coded and labeled using the respective group abbreviations. ET, PLAGL tumors are differentially colored according to their amplified PLAG-family gene. Two outlying ET, PLAGL samples are circled and marked with an arrow. Samples belonging to the ET, PLAGL type are depicted enlarged on the right. Full group names are: adult-type diffuse high grade glioma, IDH-wild type, subtype E (HGG_E), diffuse pediatric-type high grade glioma, RTK1 and 2 subtype (pedRTK1, pedRTK2), HGG-IDH wild type-subclass midline (HGG_MID), diffuse hemispheric glioma, H3 G34-mutant (G34), diffuse midline glioma, H3 K27-altered, subtype EGFR-altered (EGFR), diffuse midline glioma, H3 K27-altered (K27), glioblastoma, IDH-wild type, subtype posterior fossa (CBM), Glioblastoma, IDH-wild type, RTK1 and 2 subtype (RTK1, RTK2), Glioblastoma, IDH-wild type, mesenchymal subtype (MES), diffuse pediatric-type high grade glioma, MYCN subtype (pedMYCN), embryonal tumor, not otherwise specified (EMB), high-grade astrocytoma with piloid features (HGAP), Pleomorphic Xanthoastrocytoma (PXA), diffuse leptomeningeal glioneuronal tumor, subtype 1 and 2 (DLGNT_1, DLGNT_2), Medulloblastoma, SHH-activated (MB_SHH), Medulloblastoma, WNT-activated (MB_WNT), Medulloblastoma, non-WNT/non-SHH, Group 3 and 4 subtype (MBg34), Inflammatory microenvironment (LYMPH_HI), neuroepithelial tumor with PATZ1 fusion (PATZ), embryonal tumor with PLAG-family gene amplification (ET, PLAGL)

Out of the 46 samples initially belonging to the ET, PLAGL cluster, 11 samples were found to be duplicate or relapse samples based on genotype matches. One additional sample was excluded based on quality control indicating array hybridization issues. Another tumor with primary extracranial location was also excluded. This resulted in a set of 33 individual tumors classified as ET, PLAGL that were subjected to further analysis. Including information about the PLAGL1/PLAGL2 amplification status of each sample, we repeated t-SNE analysis using a select subset of 910 reference tumors of various types—including HGGs, medulloblastomas, and a set of the recently published neuroepithelial tumors with PATZ1 fusions [6]—together with the 33 ET, PLAGL tumors (Fig. 1b). All ET, PLAGL tumors formed one distinct cluster regardless of their PLAG gene amplification status, which confirmed their group affiliation and epigenetic similarity. The ET, PLAGL cluster was not located in proximity to any of the HGG, medulloblastoma, or other embryonal tumor clusters (Fig. 1a, b) underlining its epigenetic divergence from those tumors—an important point to stress since apart from HGG, medulloblastoma or other embryonal tumors were frequently among the initial histopathological diagnoses for the PLAGL-amplified cases, especially when occurring in the cerebellum. Two samples were found to be outliers that clustered close to ET, PLAGL, but slightly apart from the core group (Fig. 1a) as well as further apart in the refined t-SNE analysis (Fig. 1b). Both outlying samples were PLAGL1-amplified tumors, one of which was from an adult patient (age 59 years) and one with unknown age. These two samples were subsequently excluded and the remaining analyses were focused on the core cluster of 31 samples (Fig. 1b). When investigating possible further substructure within this cluster, there was some evidence that the ET, PLAGL cluster could potentially be subdivided into two different sub-clusters based on their location on the t-SNE plot, separating the PLAGL1-amplified from the PLAGL2-amplified samples. Three samples without apparent PLAG-family gene amplification were also part of the core group based on their DNA methylation pattern, with two seemingly PLAGL1-like and one PLAGL2-like. In a further t-SNE analysis, which also included a set of the recently published supratentorial ependymoma-like tumors with PLAGL1 fusions [54], as well as 1,124 sarcomas in addition to the previous reference cohort of 910 tumors, the PLAGL-amplified samples maintained its own unique cluster (Supplementary Fig. 1).

Copy number analysis

We derived copy number (CN) plots and assessed CN status for all 31 samples based on the raw intensities of the DNA methylation array probes, which revealed focal amplification of PLAGL1 or PLAGL2 in 28 of the 31 core samples (90.3%) with 11 samples being PLAGL1-amplified (35.5%) and 17 samples being PLAGL2-amplified (54.8%). Three samples showed no amplification of any PLAG-family gene (9.7%). CN summary plots were derived for PLAGL1- and PLAGL2-amplified samples separately to visualize broad chromosomal copy number changes in each subtype (Fig. 2a). As the segmentation algorithm used to produce the summary plots does not always recognize amplicons of very small size as a segment, only a subset of the PLAGL2 amplifications were detected automatically, but manual screening of the PLAGL-regions confirmed focal amplification of PLAGL1 or PLAGL2 as described above (Fig. 2b, c). Differential comparative analysis was performed using GISTIC2.0 to compare PLAGL1-amplified versus PLAGL2-amplified samples and detect significantly altered regions across all samples and per subtype (Fig. 2d, Supplementary Fig. 3).

Fig. 2figure 2

Copy number analysis of CNS embryonal tumors with PLAGL gene amplification. a Copy number summary plots were derived per subgroup for PLAGL1-amplified and PLAGL2-amplified tumors. b, c Chromosome 6 and chromosome 20 amplifications in ET, PLAGL tumors were verified using IGV. Samples are grouped according to their amplification status. b Shown are focal amplifications on chromosome 6 and chromosome 20 for PLAGL1 and PLAGL2, respectively. c Zooming in on the amplified regions around PLAGL1 and PLAGL2 (with co-amplification). d GISTIC amplification plot of all 31 samples belonging to the ET, PLAGL type. The genome is displayed vertically on the y-axis and genomic positions of amplified regions are indicated. Normalized amplification signals (G-score) and q values (log scale) are indicated on the X-axis on the top and bottom, respectively. The green line represents the significance cutoff (q value = 0.25)

Ten of the 17 PLAGL2-amplified samples (58.8%) showed co-amplification of a region immediately downstream of PLAGL2 on chromosome 20, which mainly affected the gene CBFA2T2 (Fig. 2b, c). GISTIC2.0 analysis confirmed the region containing PLAGL1 (6q24.2; q value 4.83*10–23), PLAGL2 (20q11.21; q value 2.20*10–22), and the downstream region of co-amplification on 20q11.21 (q value 2.98*10–16) (Fig. 2d; Supplementary Tables S2, S3) as significantly amplified segments in the PLAGL1- and PLAGL2-amplified tumors, respectively. Multiple ET, PLAGL group-wide and subgroup-specific deletions were identified, including a common deletion of region 11p15.4 (q value 2.07*10–11) (Supplementary Fig. 3, Supplementary Tables S2, S4), but this region encompassing various olfactory receptors as well as further affected genomic regions are likely to represent copy number polymorphisms and/or technical artifacts rather than functional somatic alterations.

Patient and sample characteristics

Patient characteristics were summarized (n = 31, Table 1) and visualized (Supplementary Fig. 2). The median age as well as the age range differed between PLAGL1- and PLAGL2-amplified tumors (p = 0.005497, Wilcoxon rank sum test). PLAGL1-amplified cases occurred mainly in school age children and teenagers, with only a few younger patients (1–19 years, median age of 10.5 years), while PLAGL2-amplified cases were mostly prevalent in infants and toddlers, with an age range from 1 to 5 years (with the exception of one adult case of 36 years; median age of 2 years). The three ET, PLAGL tumors without PLAGL1/2 amplification occurred in very young patients (0–2 years, median age of 1). The incidence of PLAGL1 tumors was higher in girls than in boys (M:F 3:8), while the opposite trend was seen for the incidence of PLAGL2 tumors (M:F 10:7), but this difference in sex distribution was not statistically significant (p = 0.1021, Chi-square test). Tumors occurred at several anatomic sites, mainly the cerebral hemispheres (35.5%) and the cerebellum (25.8%), but were also found in the brainstem (6.5%), other midline structures (9.7%), or growing into the ventricles (6.5%). Nine out of ten hemispheric PLAGL-amplified tumors occurred in female patients, compared to only three out of seven cerebellar PLAGL-amplified tumors and one tumor that was growing into the ventricles (Supplementary Fig. 2). Initially rendered histopathological diagnoses based on morphological features comprised various tumor types—medulloblastoma (16.1%), other embryonal tumors (22.6%), HGGs (19.4%), indeterminate neuroepithelial tumors (9.7%), and sarcoma (3.2%) (Table 1).

Table 1 Patient characteristics for the cohort of CNS embryonal tumors with PLAG-family gene amplification (ET, PLAGL) samples (n = 31)Targeted next-generation DNA sequencing analysis

A subset of the tumors (n = 14) were further analyzed by targeted next-generation DNA sequencing to interrogate the genomic landscape beyond PLAGL amplification (Supplementary Table S1). One tumor with PLAGL2 amplification (#A113—excluded from the core DNA methylation cohort due to QC issues, but with a clear signal for ET, PLAGL), harbored additional focal high-level amplifications of the MDM4 oncogene on chromosome 1q32.1 and the MYCN oncogene on chromosome 2p24.3. In one tumor with PLAGL1 amplification (#A388), we found separate focal high-level amplification of the GLI2 oncogene on chromosome 2q14.2, and in another tumor with PLAGL1 amplification (#A93), we found focal amplification of MYB on chromosome 6q23.3. One PLAGL2-amplified tumor (#A105) and one PLAGL1-amplified tumor (#A93) harbored deleterious missense mutations in the TP53 tumor suppressor gene. The remaining seven tumors with PLAGL2 amplification and three tumors with PLAGL1 amplification harbored no additional likely oncogenic amplifications, homozygous deletions, mutations, insertions/deletions, or gene fusions among any of the evaluated genes. Specifically, all 14 evaluated tumors were wild type for the IDH1 and IDH2 genes, as well as the histone H3 genes (H3F3A, H3F3B, HIST1H3B, and HIST1H3C). None harbored amplifications, fusions, or mutations of receptor tyrosine kinase genes including EGFR, PDGFRA, FGFR1, MET, ALK, ROS1, or NTRK2 that are common in pediatric HGG. None harbored alterations within genes of the MAP kinase signaling pathway (e.g., BRAF, KRAS, NF1, and PTPN11) that are also common in pediatric gliomas. None harbored mutation or deletion of the SMARCB1 or SMARCA4 genes, thereby distinguishing these tumors from atypical teratoid/rhabdoid tumors. None harbored DICER1 mutation or amplification of the chromosome 19 microRNA cluster (C19MC), thereby distinguishing these tumors from embryonal tumor with multilayered rosettes. None harbored BCOR fusions or internal tandem duplication, and none of the evaluated tumors (n = 6) harbored MN1 or BEND2 fusions. Genetic alterations known to contribute to telomere maintenance (TERT promoter mutation or ATRX mutation/deletion) were also not identified in any of the tumors.

Histopathological characterization

Histopathological review was performed on a subset of the tumors, including 6 with PLAGL2 amplification, 8 with PLAGL1 amplification, and 1 with no PLAG gene family amplification. The predominant morphological pattern was a densely cellular neoplasm with solid growth composed of primitive, embryonal-like cells with brisk mitotic activity (Fig. 3, Supplementary Fig. 4, and Supplementary Table S5). Less common patterns included spindled and more uniform/monotonous round cells, but tumors with these patterns always had other areas with more primitive, embryonal-like cells. While most tumors demonstrated a solid growth pattern with a paucity of entrapped neuropil and a sharply circumscribed border with adjacent brain parenchyma, a couple of tumors displayed focal infiltrative growth. Many tumors had regions of necrosis, usually without palisading of tumor cells at the periphery. No well-developed microvascular proliferation was observed in any of the reviewed tumors. Ependymal canals or perivascular pseudorosettes (characteristic histological features of ependymoma) were not observed.

Fig. 3figure 3

Imaging and histologic features of CNS embryonal tumors with PLAGL gene amplification. Shown are pre-operative T2-weighted MR images and low/high resolution H&E-stained histology images of a a PLAGL2-amplified tumor in a 2-year-old female patient (#A110) and b a PLAGL1-amplified tumor in a 13-year-old female patient (#A387)

Immunostaining for markers of glial differentiation (GFAP and OLIG2) was mostly negative, with only a few PLAGL2-amplified tumors showing labeling of rare scattered tumor cells (Fig. 4, Supplementary Table S5). Several tumors demonstrated patchy weak staining for synaptophysin, while others were negative. Neurofilament expression was often seen in scattered tumor cells, but otherwise revealed an absence of entrapped neuropil, confirming the solid growth pattern of these tumors. Two PLAGL2-amplified tumors demonstrated focal collections of tumor cells with paranuclear dot-like positivity for EMA staining, while the majority of tumors lacked EMA expression. All evaluated tumors had intact/retained expression of INI1/SMARCB1 and BRG1/SMARCA4. All evaluated tumors had minimal to absent immunostaining for LIN28A, BCOR, and CD99. A subset of tumors demonstrated positivity for YAP1 and GAB1, while no tumors had nuclear beta-catenin staining. Desmin expression was present in the majority of evaluated tumors (9/12, 75%), which ranged from rare scattered cells to diffuse strong labeling of all tumor cells in a small number of the PLAGL2-amplified cases. Other markers of myogenic differentiation (myogenin, smooth muscle actin, and MyoD1) were negative in all evaluated tumors. Ki-67 labeling indices ranged from 30 to 70%.

Fig. 4figure 4

Immunohistochemical features of CNS embryonal tumors with PLAGL gene amplification. Shown are representative immunostains of a a PLAGL2-amplified tumor in a 1-year-old female patient and b a PLAGL1-amplified tumor in a 13-year-old female patient. c Summary of IHC results in PLAGL1/2-amplified tumor samples

Gene expression analysis

Differential gene expression between tumors with PLAG-family gene amplification and a selection of other CNS tumor types was examined using the R2 Genomics Analysis and Visualization Platform (http://r2.amc.nl). In concordance with the observed gene amplification, ET, PLAGL tumors showed overexpression of the respective amplified PLAG-family gene as assessed by RNA-seq (Fig. 5a, c), while both PLAGL1 and PLAGL2 are downregulated postnatally in normal brain and cerebellar tissues (https://apps.kaessmannlab.org/evodevoapp/) [13] (Supplementary Fig. 5). Leveraging our expression data set of 11 PLAGL1- and PLAGL2-amplified tumor samples and 279 samples from other CNS tumor and normal tissue types (HGGs with H3 G34R/V or K27M mutation and GBM_pedRTK1 or 2 (n = 76), PA with BRAF fusion (n = 25), PXA (n = 25), normal brain tissue (n = 36), embryonal tumors such as ATRT, ETMR, or medulloblastomas (n = 117)), we first compared gene expression of the PLAGL-amplified tumors to our subset of embryonal tumors. We derived a gene set specific to the ET, PLAGL type (Fig. 5a) as well as a PLAGL-specific gene-signature consisting of the top 250 differentially expressed genes (Supplementary Table S6). In addition to PLAGL1/2 overexpression, we found differential expression of several genes involved in developmental and differentiation processes such as CDX1, NR5A1, TLX1, TBX1, FGF19, and DLK1 (Fig. 5a, Supplementary Table S6); known direct PLAGL target genes such as IGF2, H19, CDKN1C and DLK1 [64] (Fig. 5a, Supplementary Fig. 6), as well as CYP2W1 and the kinase RET, both putative treatment targets (Fig. 5a, c). We screened expression of 86 human IGs in the PLAGL1/2-amplified samples (Fig. 5a). A subset of 13 IGs (Meg3, Ndn, Grb10, Dlk1, Igf2, Cdkn1c, Plagl1, Peg3, Mest, Nnat, Asb4, H19, and Ppp1r9a) described as having high connectivity with other IGs [5] were differentially expressed in the PLAGL1/2-amplified tumors (Fig. 5a, Supplementary Figs. 6, 7). We also ran the same differential expression analyses comparing ET, PLAGL versus glial tumors as well as versus normal fetal and adult brain tissues. This analysis yielded similar results regarding the overrepresentation of imprinted genes as well as developmental and differentiation-related genes (Supplementary Fig. 8). Expression of classical pan-neuronal, glial, sarcoma/mesenchymal, neural stem cell, and proliferation marker genes was also examined in the ET, PLAGL tumors versus our subset of CNS embryonal tumors, gliomas, and normal tissues, but was inconclusive in terms of possible cell/lineage of origin, as there was no set of marker genes that was clearly differentially expressed in the ET, PLAGL tumor type—with the exception of high Desmin expression in the PLAGL-amplified tumors in all three comparisons (Supplementary Fig. 9). Overexpression of the myogenic marker Desmin was more pronounced in the PLAGL2-amplified samples (Supplementary Fig. 9d). Furthermore, this analysis showed a lack of glial marker expression in the PLAGL tumors. We compared bulk RNA-seq data of the ET, PLAGL tumors to a single-nucleus sequencing atlas containing transcriptomes from different cell types, differentiation states, and subtypes of the developing human cerebellum to map the cellular origins of the ET, PLAGL tumors (Supplementary Fig. 10) as described in Okonechnikov et al. [39]. None of the lineages that were used as reference could be identified as the origins of ET, PLAGL tumors. Consequently, we analyzed the expression of genes representing different developmental states and locations as markers for pluripotency, germ layers (ectoderm, mesoderm, endoderm), neuroectoderm, forebrain and pallium, subpallium (including the ganglionic eminence), midbrain, hindbrain, spinal cord, as well as various additional pan-neuronal and glial markers [56] (Fig. 5b, Supplementary Fig. 11). The early neural genes OTX2, TLX1, SIX3, MSI1, and DACH1 were overexpressed in the PLAGL-amplified tumors, as were some subpallial neural markers such as DLX5, DLX6, the lateral ganglionic eminence (LGE) marker ISL1, and the germ layer markers KRT18 and GATA4 pointing to a cell of origin at an early and rather undifferentiated developmental stage.

Fig. 5figure 5

Gene expression profiles of CNS embryonal tumors with PLAGL gene amplification. a, b Volcano plots showing fold-change and p-value for the comparison of differential gene expression of 11 PLAGL1/2-amplified tumors versus 117 embryonal tumors from different types and subtypes. Highlighted are a 86 human IGs (ocher) and 13 IGs with high connectivity (lilac) as described in reference [5]. Shown in black: selection of genes with large magnitude fold-changes (x axis) and high statistical significance (− log10 of p-value, y-axis). b Genes with differential expression in different brain regions and during different developmental states as described in reference [56] c Boxplots comparing gene expression between CNS tumor types for a select set of genes. The subset of 117 embryonal tumor samples (atrt, etmr, med) is identical to a and b. plagl, ET,PLAGL; pa, pilocytic astrocytoma; pxa, pleomorphic xanthoastrocytoma; hgg, high-grade gliomas (G34R/V, K27M, pedRTK1/2); norm, normal brain tissues; atrt, atypical teratoid rhabdoid tumor; etmr, embryonal tumor with multilayered rosettes; med, medulloblastomas (WNT, SHH, group 3, group 4); red: samples with PLAGL1 amplification, blue: samples with PLAGL2 amplification. Significance bars indicate groups whose differences in gene expression are statistically significant when compared to samples with PLAGL1/2 amplification (t-test, Bonferroni-corrected p-value = 0.00714286). PLAGL1/2 upregulation is statistically significant compared to all other groups when looking at PLAGL1 or PLAGL2 tumors separately

ChIP-seq analysis

Chromatin immunoprecipitation studies using antibodies against PLAGL1 and PLAGL2 proteins (n = 5) was performed to identify gene loci bound by these two transcription factors in PLAGL-amplified tumors. This ChIP-seq data confirmed the known Plagl1 targets IGF2, CDKN1C, and DLK1, as well as most of the other IGs with high connectivity, as being direct targets of PLAGL1 and PLAGL2 TF binding in the PLAGL-amplified tumors (Supplementary Fig. 6, Supplementary Fig. 12). The receptor tyrosine kinase RET and the cytochrome P450 family member CYP2W1—both potential drug targets—were also revealed as further direct PLAGL1/2 targets (Supplementary Fig. 13). In addition, we identified components of the Wnt/β-Catenin signaling pathway, FZD2 and FZD9, to be targets of PLAGL1/2 TF binding that are also differentially expressed in the PLAGL-amplified tumors (Supplementary Fig. 14).

Survival analysis

Clinical outcome data were available for 21 patients with PLAGL1/2-amplified tumors. Five-year and 10-year OS for patients with PLAGL1- and PLAGL2-amplified tumors as well as for male and female patients was determined. Survival rates across the cohort remained constant after 5 years, hence both 5- and 10-year OS was 66% for patients with PLAGL1-amplified tumors, 25% for patients with PLAGL2-amplified tumors, 18% for male patients, and 82% for female patients, respectively. Although a trend towards a worse prognosis for patients with PLAGL2-amplified tumors was noticeable—with 5 out of 12 patients with a PLAGL2-amplified tumor being deceased compared to 2 out of 9 patients with a PLAGL1-amplified tumor (Fig. 6b)—PFS and OS did not differ significantly between the two different groups (Fig. 6a, p value = 0.096 and 0.44, respectively). Patient sex was also not a significant predictor for PFS or OS (Fig. 6a, p value = 0.12 and 0.2), but more deaths in male patients were recorded irrespective of the subgroup. With respect to different treatment regimens, the inclusion of chemotherapy agents beyond temozolomide (TMZ) early on in treatment showed a potential benefit for patient survival (Fig. 6b, Supplementary Table S7) while the inclusion of radiotherapy as part of the initial treatment seemed to have limited effect (Supplementary Fig. 15), but this should be judged with caution given the overall low numbers.

Fig. 6figure 6

Clinical outcomes of patients with CNS embryonal tumor with PLAGL gene amplification. a Kaplan–Meier plots showing OS and PFS stratified by subgroup and sex. The log-rank test was used to show differences between the curves, p-values of the log-rank test are shown in each graph. b Swimmer plot showing available OS and PFS times per patient, including treatment information and clinical response/relapse. Samples are stratified by sex, PLAGL1/2 amplification status is indicated. Information about surgical resection (SUR) and presence of metastasis (MET) at the time point of primary diagnosis is displayed in the squares on the left where available (resections or metastases at later time points are not displayed), GTR, gross total resection; STR, subtotal resection; RES, resection (unknown, if GTR or STR). Information about chemotherapy (CT) and radiotherapy (RT) treatment regarding the entire follow-up time is displayed in the squares on the left where available

留言 (0)

沒有登入
gif