Tumor organoids are important tools for cancer research, but current models have drawbacks that limit their applications for predicting response to therapy. Here, we developed a fast, efficient, and complex culture system (IPTO, individualized patient tumor organoid) that accurately recapitulates the cellular and molecular pathology of human brain tumors. Patient-derived tumor explants were cultured in induced pluripotent stem cell (iPSC)-derived cerebral organoids, thus enabling culture of a wide range of human tumors in the central nervous system (CNS), including adult, pediatric, and metastatic brain cancers. Histopathological, genomic, epigenomic, and single-cell RNA sequencing (scRNA-seq) analyses demonstrated that the IPTO model recapitulates cellular heterogeneity and molecular features of original tumors. Crucially, we showed that the IPTO model predicts patient-specific drug responses, including resistance mechanisms, in a prospective patient cohort. Collectively, the IPTO model represents a major breakthrough in preclinical modeling of human cancers, which provides a path toward personalized cancer therapy.

Recent advances in the genomics of glioblastoma (GBM) led to the introduction of molecular neuropathology but failed to translate into treatment improvement. This is largely attributed to the genetic and phenotypic heterogeneity of GBM, which are considered the major obstacle to GBM therapy. Here, we use advanced human GBM-like organoid (LEGO: L aboratory E ngineered G lioblastoma-like O rganoid) models and provide an unprecedented comprehensive characterization of LEGO models using single-cell transcriptome, DNA methylome, metabolome, lipidome, proteome, and phospho-proteome analysis. We discovered that genetic heterogeneity dictates functional heterogeneity across molecular layers and demonstrates that NF1 mutation drives mesenchymal signature. Most importantly, we found that glycerol lipid reprogramming is a hallmark of GBM, and several targets and drugs were discovered along this line. We also provide a genotype-based drug reference map using LEGO-based drug screen. This study provides new human GBM models and a research path toward effective GBM therapy.

Supplementary Data 1. This supplementary data comprises a step-by-step description on how our gene-expression based classification models were generated and validated

Cox proportional hazards regression for overall survival (105 neuroblastomas)

Genes that significantly i) lost H3K27me3 coverage ii) lost DNA methylation iii) gained expression in Be(2)-C and IMR5-75 cells upon combination treatment with DAC and EPZ-6438

Legends to Supplementaries. This documents comprises the legends to the supplementary figures and tables

Supplementary Data 2. This supplementary data comprises the r-algorithm scripts required to perform the both model selection and validation as described in the step-by-step protocol

Gene Ontology (GO) terms significantly enriched among genes whose hypermethylation and downregulation are associated with high-risk disease (HyperDownHR).

Supplementary Table 3. This table summarizes the results of the Kaplan-Meier estimates for both EFS and OS of clinically relevant subgroups of neuroblastoma patients according to classification by the four remaining classifiers SVM_th22, SVM_th24, SVM_th26 and SVM_th44.

Dendrogram and probability values for hierarchical clustering of 105 neuroblastomas based on genome-wide DNA methylation (S1); Kaplan-Meier estimates of overall survival for DNA methylation subgroups (S2); Genes whose hypermethylation and downregulation are associated with neuroblastoma high-risk disease are induced during neuronal differentiation of neuroblastoma cells (S3); Examples for genes whose hypermethylation and downregulation are associated with neuroblastoma high-risk disease and that are marked by both H3K27me3 and DNA methylation at putative regulatory regions in Be(2)-C cells (S4); MYCN deregulation is associated with activation of PRC2 components (S5); Genes whose hypermethylation and downregulation are associated with high-risk disease are H3K27me3-marked in a MYCN-dependent fashion (S6); Example for genes whose hypermethylation and downregulation are associated with neuroblastoma high-risk disease that significantly lost H3K27me3 and DNA methylation at putative regulatory regions upon treatment with DAC and EPZ-6438 in Be(2)-C cells (S7); Genes whose hypermethylation and downregulation are associated with high-risk disease are preferentially induced upon DAC/TSA treatment (S8).

Description of additional methods and procedures used in the study. Also includes Supplementary References.

Genes significantly induced in Be(2)-C or IMR5-75 upon treatment with DAC, EPZ-6438 or a combination of both drugs (adjusted p-value<0.05, sorted by log2 fold change)

Supplementary Table 1. Clinical co-variates for the 709 patients who participated in the study. This supplementary table comprises detailed information on clinical co-variates for all 709 neuroblastoma patients who participated in the study.

Supplementary Figure 1. Highlights the EFS and OS for the cohort of neuroblastoma patients with MYCN-amplified disease (n=114, Fig 1a) and for th subcohort of patients >18 months of age with stage 4, MYCN non-amplified disease (n=102, fig. 1b). F, favorable; UF, unfavorable.

Supplementary Table 2. External performance validation of the top five classifiers. This supplementary table comprises the external performance metrics of the top five classifiers. Indicated are the values for classification accuracy, sensitivity , specificity and Matthew's correlation coefficient (MCC) in the prediction of 325 patients of the test set who fulfilled the criteria for classifier training (Favorable, n=187; Unfavorable, n=138).

Supplementary Table 4. Transcripts that contribute to the SVM_th10 classifier. This supplementary table highlights the transcripts that constitute the SVM_th10 classifier. Indicated for each feature are the Oligo-ID, the Gene symbols, the Entrez gene IDs, the RefSeq IDs, the Gene IDs and the Transcript IDs.

Legends for Supplementary Figures S1-S8.

Aberrant expression of MYC transcription factor family members predicts poor clinical outcome in many human cancers. Oncogenic MYC profoundly alters metabolism and mediates an antioxidant response to maintain redox balance. Here we show that MYCN induces massive lipid peroxidation on depletion of cysteine, the rate-limiting amino acid for glutathione (GSH) biosynthesis, and sensitizes cells to ferroptosis, an oxidative, non-apoptotic and iron-dependent type of cell death. The high cysteine demand of MYCN -amplified childhood neuroblastoma is met by uptake and transsulfuration. When uptake is limited, cysteine usage for protein synthesis is maintained at the expense of GSH triggering ferroptosis and potentially contributing to spontaneous tumor regression in low-risk neuroblastomas. Pharmacological inhibition of both cystine uptake and transsulfuration combined with GPX4 inactivation resulted in tumor remission in an orthotopic MYCN -amplified neuroblastoma model. These findings provide a proof of concept of combining multiple ferroptosis targets as a promising therapeutic strategy for aggressive MYCN -amplified tumors.

Background Mutation specific synthetic lethal partners (SLPs) offer significant insights in identifying novel targets and designing personalized treatments in cancer studies. Large scale genetic screens in cell lines and model organisms provide crucial resources for mining SLPs, yet those experiments are expensive and might be difficult to set up. Various computational methods have been proposed to predict the potential SLPs from different perspectives. However, those efforts are hampered by the low signal-to-noise ratio in simple correlation based approaches, or incomplete reliable training sets in supervised approaches. Results Here we present mslp, a comprehensive pipeline to identify potential SLPs via integrating genomic and transcriptomic datasets from both patient tumours and cancer cell lines. Leveraging cuttingedges algorithms, we identify a broad spectrum of primary SLPs for mutations presented in patient tumours. Further, for mutations detected in cell lines, we develop the idea of consensus SLPs which are also identified as screen hits, and show consistency impact on cell viability. Applied in real datasets, we successfully identified known synthetic lethal gene pairs. Remarkably, genetic screen results suggested that consensus SLPs have a significant impact on cell viability compared to common hits. Conclusions Mslp is a powerful and flexible pipeline to identify potential SLPs in a cancer context-specific manner, which might aid in drug developments and precise medicines in cancer treatments. The pipeline is implemented in R and freely available in github.

How overall tumor growth emerges from the properties of functionally heterogeneous tumor cell subpopulations is a fundamental question of cancer biology. Here we combined lineage tracing, continuous monitoring of tumor mass, proliferation assays and transcriptomics with mathematical modeling and statistical inference to dissect the growth of glioblastoma in mice. We found that tumors grow exponentially at the rate of symmetric divisions of brain tumor stem cells (BTSCs). Spatial modeling predicts, and data show, that BTSCs accumulate at the tumor rim rather than in the core. The physiological differentiation hierarchy downstream of BTSCs is preserved in mice and humans: transit amplifying progenitors give rise to terminally differentiated cells. Consistent with our quantification of the mechanisms underlying tumor growth, molecular data show elevated expression of cell cycle-and migration-related genes in BTSCs. Our systematic approach reveals fundamental properties of glioblastoma and may be transferable to the study of other animal models of cancer.