Search

Novel treatment approaches for pediatric central nervous system (CNS) tumors are urgently needed. A lack of patient-derived tumor cells impedes progress towards developing such new therapies. We intended to overcome this challenge by establishing methods to create a biorepository of viable single cell suspensions of pediatric brain tumor surgical specimens. Quantitative and qualitative comparisons of tissue processing strategies were performed to preserve viability of heterogeneous tumor and immune cells. Novel drug targets were identified by analyzing pathways affected by RNA transcripts that are highly expressed (outliers) in a patients’ tumor; outliers for each patient were determined by RNA-seq data from individual patients’ tumors compared with a compendium of 12,747 pediatric and adult samples harmonized by the Treehouse Childhood Cancer Initiative at the UCSC Genomics Institute. The predicted anti-tumor efficacy of small molecule inhibitors of the outlier pathways was tested in cell viability assays against short-term cultured cells from matched patients. Successful tissue collection required obtaining informed consent, standard operating procedures, and sample recording using a Laboratory Inventory Management Software (LIMS). Since 2020 we have banked 67 pediatric CNS brain tumor specimens at Stanford. Amongst those, 51 cases yielded sufficient tissue for RNA-seq and cryoprotection. The most common tumor histology was low-grade glioma (LGG, 26 of 67), the majority of which were pilocytic astrocytoma (18 of 26). The second and third most common tumor types are embryonal tumors (6 medulloblastoma, 3 AT/RT) and ependymoma (4), respectively. We identified significant differences in cell viability with different preservation media. An outlier pathway previously not implicated in LGG was identified and sensitivity to a small molecule inhibitor of this outlier pathway was demonstrated. Taken together, we established feasibility for validating therapeutic vulnerabilities identified by a genome-guided approach in short-term cultures from surgical specimens. This works facilitates the rapid development of personalized CNS tumor treatment.

Library preparation protocols for RNA sequencing (RNA-Seq) libraries vary and affect the downstream analysis of RNA-Seq data. Available RNA sequencing datasets often lack library construction protocol information in the metadata. This information is necessary to be able to compare RNA sequencing datasets appropriately. For example, non-polyadenylated transcripts are measured when a ribosomal RNA-depletion method is used (riboD library preparation), but not when transcripts are selected by poly(A) tails (polyA library preparation). Without knowing the library preparation method, it would appear that the non-polyadenylated transcripts were expressed at much higher levels in the samples prepared via riboD. In order to tackle this issue, we have developed a Random Forest classifier that can delineate between riboD and polyA RNA-Seq datasets. A grid search method was applied to the number of trees, maximum depth, and the minimum datasets included in each leaf node, which determined the best parameters to be 100, 8, and 1 respectively. However, applying the model on the Pediatric Brain Tumor Atlas (PBTA) showed strong overfitting for polyA samples. We examined the performance of the model investigating each maximum depth increment from 1 to 8, and we determined that the best performance on our validation set was achieved with maximum depth of 1. We subsequently proceeded to train our classifier on our own curated compendiums of pediatric cancer polyA and riboD samples including 188 and 264 samples respectively, after balancing the two datasets in terms of disease prevalence, and selecting the top 5,000 most variable genes as the default input dimensionality of the model. For samples whose genes do not exactly match the predetermined 5,000 genes, we substitute the expression of the missing gene with the mean expression of the gene observed in the training data. We show that it achieves 100% classification accuracy of samples to their respective library preparation protocols in GTEX (all polyA), CCLE (all polyA), and for 7 of 9 SRA projects. Five of the SRA projects contained only riboD datasets, 3 contained only all polyA datasets, and one was 50% riboD, and 50% polyA. Notably the SRA datasets are not all cancer related datasets, showing the power of our model to distinguish between library preparation protocols in vastly different settings. The model will become available on Docker so that it is readily, and easily accessible for application on new samples. Our model serves as an important step towards robust library preparation identification. Including samples in the training procedure from diverse contexts would make our model more widely applicable. Need summary statement here and/or future work statement Citation Format: Ioannis Anastopoulos, Holly Beale, Geoff Lyle, Allison Cheney, Olena M. Vaske, Joshua M. Stuart. Detection of RNA-Seq library preparation type via random forest [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2287.

Diffuse pediatric gliomas harboring a Histone H3 K27M mutation are significantly more aggressive than H3-wild type gliomas and demonstrate global hypomethylation at the K27 residue(1). As a result, these tumors show global aberrant gene expression, resulting in a primitive, stem-like proliferative cell population(2). Posterior fossa (PF) ependymomas, on the other hand, harbor few significantly recurrent somatic mutations, but PF-A and PF-B subgroups have been defined on the basis of epigenetic differences(3). Compared to PF-B, the PF-A subgroup demonstrates H3K27 hypomethylation, aberrant gene expression and aggressive tumor growth(4,5). We recently identified a set of long non-coding RNA (lncRNA) that are transiently expressed in early brain development(6), and hypothesized that H3K27M gliomas and PF-A ependymomas may share methylation-related dysregulation of long non-coding RNA (lncRNA) networks responsible for maintaining normal differentiation programs. Here we describe a network of regulatory lncRNA with increased expression in both H3K27M gliomas and PF-A ependymomas, as compared to H3-WT gliomas and PF-B ependymomas. We demonstrate that increased expression of this lncRNA network correlates with the over-expression of signaling pathways involved in maintaining a non-differentiated, proliferative phenotype and driving tumorigenesis. We hypothesize that in both H3K27M gliomas and PF-A ependymomas, aberrant global methylation may be driving lncRNA to activate and maintain stem-like states in early neural development, suggesting similarities in epigenetically driven, developmental origins for both tumor types. References: 1. Chan KM, Fang D, Gan H, et al. Genes Dev. 2013;27(9):985–90; 2. Filbin MG, Tirosh I, Hovestadt V, et al. Science. 2018;360(6386):331–5; 3. Witt H, Mack SC, Ryzhova M, et al. Cancer cell. 2011;20(2):143–57; 4. Bayliss J, Mukherjee P, Lu C, et al. Sci. Transl. Med. 2016;8(366):366ra161; 5. Mack SC, Witt H, Piro RM, et al. Nature. 2014;506(7489):445; 6. Field AR, Jacobs FM, Fiddes IT, et al. bioRxiv. 2017:232553.

Genomic characterization is often used for the identification of therapeutic targets in tumors. Recently, comparative transcriptomics has begun to be utilized for this purpose. In this pilot, we compare the transcriptome of a patient with recurrent high grade glioma (HGG) to our cohort to identify potential therapies. We reviewed transcriptomic profiles from patients who had resection of HGG at our institution over the past year as well as the UCSC cancer compendium. Briefly, tumor RNA was extracted from embedded tumor tissue sections with tumor cellularity higher than 20%. RNA libraries were sequenced to obtain approximately 65 million reads on an Illumina HiSeq 4000 System utilizing patterned flow cell technology. The RNA profile of a 24 male with Li-Fraumeni syndrome and recurrent HGG with leptomeningeal spread underwent comparative transcriptomics to identify targets. A Bayesian statistical framework for gene expression outlier detection was used. These comparisons allowed for the identification of genes and pathways that are significantly overexpressed. Our internal HGG cohort consisted of 44 adult patients and was evenly distributed among the 4 HGG Verhaak subtypes. Our patient of interest had druggable outlier expression in HDAC1, STAT1 and STAT2 in comparison to our internal cohort indicating vorinostat and ruxolitinib as potential therapies, respectively. We then compared our patient of interest to 12,747 patients in the cancer compendium and STAT2 expression was high but not an outlier. In comparison to 738 glioma samples, STAT1 and STAT2 were outliers but not HDAC1 again indicating ruxolitinib as a potential targeted therapy. The patient did not have outlier expression in notch transcriptional targets or immune checkpoint biomarkers when compared to all cohorts. In conclusion, comparative Transcriptomics can identify therapeutic targets in a patient with recurrent HGG even in small cohorts. In our pilot, we identified ruxolitinib as a potential candidate to treat leptomeningeal recurrence.

Gene expression analysis of single samples shows increasing promise for clinical applications. However, obtaining high quality RNA from a human tumor sample can be challenging because medical, surgical, and pathological requirements often lead to sparse or degraded RNA. The variability in RNA quality presents challenges for defining input sample requirements, which are required to calculate sensitivity, specificity and reference ranges as required for a Clinical Laboratory Improvement Amendments (CLIA)-approved test. Clinical analysis of a single RNA-Seq dataset for the purpose of gene expression profiling involves not only the patient's sample, but a comparison cohort. We use 12,236 total tumor samples and require at least 20 samples for within-disease comparisons. Many of these samples do not have associated metadata about the quality of the sample, and so we have prioritized quality measures that can be derived from the sequence data alone. In order to characterize variability present in RNA-Seq datasets, we analyzed paired-end Illumina RNA sequencing (RNA-Seq) data from 1088 tumor samples from 29 data providers. We categorized reads based on where and how well they map to the genome, as well as by their PCR duplicate status. We defined reference ranges for five types of reads found in sequencing data: unmapped (0-13%); multi-mapped (2-15%); mapped duplicate (2-66%); mapped non exonic (0-26%) and mapped, exonic, non-duplicate (MEND, 27-76%). Only 64% of the 1088 tumor samples had read type fractions within the reference ranges. Of the remainder, most exceeded the reference ranges of more than one type of read. We then measured the relationship of sensitivity and specificity to input MEND read depth. We subsampled 5 deeply sequenced samples. With each subsample, we identified exceptionally highly expressed genes and samples with similar gene expression profiles. With subsampling to 20 million MEND reads, we detected over-expressed genes (“up-outlier” genes) with a median sensitivity of 96.1% and specificity of 99.8%; sample similarity had 96.6% sensitivity and 100.0% specificity. We estimate that a sample sequenced to a depth of 70 million total reads will typically have sufficient data for the up-outlier and sample-similarity gene expression analysis assays described here. With this analysis, we have identified a conservative approach to measuring the quality of RNA-Seq read data, which can then be used to define the sensitivity and specificity of single-sample assays to support their ultimate clinical adoption. Citation Format: Holly C. Beale, Jacquelyn M. Roger, Matthew A. Cattle, Liam T. McKay, Katrina Learned, Geoff Lyle, Ellen T. Kephart, Rob Currie, Du Linh Lam, Lauren Sanders, Jacob Pfeil, John Vivian, Isabel Bjork, Sofie R. Salama, David Haussler, Olena M. Vaske. Determining accuracy of RNA sequencing data for gene expression profiling of single samples [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5464.