We graciously thank the patients and families who have donated their tumors to the Children's Brain Tumor Network and/or the Pacific Pediatric Neuro-oncology Consortium, without which, this research would not be possible. This work was funded through the Alex’s Lemonade Stand Foundation (ALSF) Childhood Cancer Data Lab (JNT, CSG, JAS, CLS, CJB, SJS), ALSF Young Investigator Award (JLR), ALSF Catalyst Award (JLR, ACR, PBS), ALSF Catalyst Award (SJS), ALSF CCDL Postdoctoral Training Grant (SMF), Children’s Hospital of Philadelphia Division of Neurosurgery (PBS and ACR), the Australian Government, Department of Education (APH), and NIH Grants 3P30 CA016520-44S5 (ACR), U2C HL138346-03 (ACR, APH), U24 CA220457-03 (ACR), K12GM081259 (SMF), R03-CA23036 (SJD), and NIH Contract No. HHSN261200800001E (SJD). This project has been funded in part with Federal funds from the National Cancer Institute, National Institutes of Health, under Contract No. 75N91019D00024, Task Order No. 75N91020F00003 (JLR, ACR, APH). Additionally, this work was supported by the Intramural Research Program of the Division of Cancer Epidemiology and Genetics of the National Cancer Institute. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products or organizations imply endorsement by the U.S. Government.
The authors would like to thank the following collaborators who contributed or supervised analyses present in the analysis repository that were not included in the manuscript: William Amadio, Holly C. Beale, Ellen T. Kephart, A. Geoffrey Lyle, and Olena M. Vaske. Finally, we would like to thank Yuanchao Zhang and Eric Wafula for adding to the project codebase and Jessica B. Foster and Kristina A. Cole for helpful discussions while drafting the manuscript.
|Author|Contributions| |---|---|{% for author in manubot.authors %} |{{author.name}}|{% if author.contributions is string %}{{author.contributions}}{% else %}{% for contribution in author.contributions %}{{ contribution }}{% if not loop.last %}, {% endif %}{% endfor %}{% endif %}|{% endfor %}
CSG's spouse was an employee of Alex's Lemonade Stand Foundation, which was a sponsor of this research. JAS, CLS, CJB, SJS, and JNT are or were employees of Alex's Lemonade Stand Foundation, a sponsor of this research.
Figure 1. Overview of the OpenPBTA Project. A, The Children's Brain Tumor Network and the Pacific Pediatric Neuro-Oncology Consortium collected tumor samples from 943 patients. To date, 22 cell lines were created from tumor tissue, and over 2000 specimens were sequenced (N = 1035 RNA-Seq, N = 940 WGS, and N = 32 WXS or targeted panel). Data was harmonized by the Kids First Data Resource Center using an Amazon S3 framework within CAVATICA. B, Stacked bar plot summary of the number of biospecimens per phase of therapy per broad histology (Abbreviations: GNG = ganglioglioma, Other LGG = other low-grade glioma, PA = pilocytic astrocytoma, PXA = pleomorphic xanthoastrocytoma, SEGA = subependymal giant cell astrocytoma, DIPG = diffuse intrinsic pontine glioma, DMG = diffuse midline glioma, Other HGG = other high-grade glioma, ATRT = atypical teratoid rhabdoid tumor, MB = medulloblastoma, Other ET = other embryonal tumor, EPN = ependymoma, PNF = plexiform neurofibroma, DNET = dysembryoplastic neuroepithelial tumor, CRANIO = craniopharyngioma, EWS = Ewing sarcoma, CPP = choroid plexus papilloma). Only samples with available descriptors were included. C, Overview of the open analysis and manuscript contribution model. In the analysis GitHub repository, a contributor would propose an analysis that other participants can comment on. Contributors would then implement the analysis and file a request to add their changes to the analysis repository ("pull request"). Pull requests underwent review for scientific rigor and correctness of implementation. Pull requests were additionally checked to ensure that all software dependencies were included and the code was not sensitive to underlying data changes using container and continuous integration technologies. Finally, a contributor would file a pull request documenting their methods and results to the Manubot-powered manuscript repository. Pull requests in the manuscript repository were also subject to review. D, A potential path for an analytical pull request. Arrows indicate revisions to a pull request. Prior to review, a pull request was tested for dependency installation and whether or not the code would execute. Pull requests also required approval by organizers and/or other contributors, who checked for scientific correctness. Panel A created with BioRender.com.
Figure 2. Mutational landscape of PBTA tumors.
Shown are frequencies of canonical somatic gene mutations, CNVs, fusions, and TMB (top bar plot) for the top 20 genes mutated across primary tumors within the OpenPBTA dataset. A, Low-grade astrocytic tumors (N = 227): pilocytic astrocytoma (N = 104), other low-grade glioma (N = 69), ganglioglioma (N = 35), pleomorphic xanthoastrocytoma (N = 9), subependymal giant cell astrocytoma (N = 10); B, Embryonal tumors (N = 128): medulloblastomas (N = 95), atypical teratoid rhabdoid tumors (N = 24), other embryonal tumors (N = 9); C, Diffuse astrocytic and oligodendroglial tumors (N = 61): diffuse midline gliomas (N = 34) and other high-grade gliomas (N = 27); D, Other CNS tumors (N = 194): ependymomas (N = 60), craniopharyngiomas (N = 31), meningiomas (N = 17), dysembryoplastic neuroepithelial tumors (N = 19), Ewing sarcomas (N = 7), schwannomas (N = 11), and neurofibroma plexiforms (N = 7). Additional, rare CNS tumors are displayed in Figure {@fig:S3}A. Tumor histology (Cancer Group) and patient sex (Germline sex estimate) are displayed as annotations at the bottom of each plot. Only samples with mutations in the listed genes are shown. Multiple CNVs are denoted as a complex event.
Figure 3. Mutational co-occurrence and signatures highlight key oncogenic drivers.
A, Bar plot of occurrence and co-occurrence of nonsynonymous mutations for the 50 most commonly mutated genes across all tumor types (annotated from cancer_group if N >= 10 or Other if N < 10); B, Co-occurrence and mutual exclusivity of nonsynonymous mutations between genes; The co-occurrence score is defined as cancer_group with N >= 3. E, Sina plots of RefSig signature weights for signatures 1, 3, 8, 11, 18, 19, N6, MMR2, and Other across cancer groups. Box plot lines represent the first quartile, median, and third quartile.
Figure 4. TP53 and telomerase activity
A, Receiver Operating Characteristic for TP53 classifier run on FPKM of stranded RNA-Seq samples. B, Violin and box plots of TP53 scores plotted by TP53 alteration type (Nactivated = 11, Nlost = 100, Nother = 866). C, Violin and box plots of TP53 RNA expression plotted by TP53 activation status (Nactivated = 11, Nlost = 100, Nother = 866). D, Box plots of TP53 and telomerase (EXTEND) scores grouped by cancer_group. Mutation status is highlighted in orange (hypermutant) or red (ultra-hypermutant). E, Heatmap of RefSig mutational signatures for patients who have least one tumor or cell line with a TMB >= 10 Mut/Mb. F, Forest plot depicting the prognostic effects of TP53 and telomerase scores on overall survival, controlling for extent of tumor resection, LGG group, and HGG group. G, Forest plot depicting the effect of molecular subtype on overall survival of HGGs. For F and G, hazard ratios (HR) with 95% confidence intervals and p-values are listed. Significant p-values are denoted with black diamonds. Reference groups are denoted by grey diamonds. H, Kaplan-Meier curve of HGG tumors by molecular subtype.
Figure 5. Transcriptomic and immune landscape of pediatric brain tumors A, First two dimensions from UMAP of sample transcriptome data. Points are colored by the broad histology of the samples they represent. B, Heatmap of GSVA scores for Hallmark gene sets with significant differences, with samples ordered by cancer group. C, Box plots of quanTIseq estimates of immune cell proportions in select cancer groups with N > 15 samples. Note: Other HGGs and other LGGs have immune cell proportions similar to DMG and pilocytic astrocytoma, respectively, and are not shown. D, Forest plot depicting the additive effects of CD274 expression, immune cell proportion, and extent of tumor resection on overall survival of medulloblastoma patients. Hazard ratios (HR) with 95% confidence intervals and p-values are listed. Significant p-values are denoted with black diamonds. Reference groups are denoted by grey diamonds. Of note, the Macrophage M1 HR was 0 (coefficient = -9.90e+4) with infinite upper and lower CIs, and thus it was not included in the figure. E, Box plot of CD274 expression (log2 FPKM) for medulloblastoma samples grouped by molecular subtype. Bonferroni-corrected p-values from Wilcoxon tests are shown.
Table 1. Molecular subtypes generated through the OpenPBTA project. Listed are broad tumor histologies, molecular subtypes generated, and number of specimens subtyped within the OpenPBTA project.
Table 2. Patients with hypermutant tumors. Listed are patients with at least one hypermutant or ultra-hypermutant tumor or cell line. Pathogenic (P) or likely pathogenic (LP) germline variants, coding region TMB, phase of therapy, therapeutic interventions, cancer predisposition (CMMRD = Constitutional mismatch repair deficiency), and molecular subtypes are included.