Single-cell RNAseq of childhood medulloblastoma

We explore cellular heterogeneity in 28 childhood medulloblastoma (MB) (1 WNT, 9 SHH, 7 GP3 and 11 GP4) using single-cell RNA sequencing (scRNA-seq), immunohistochemistry and deconvolution of bulk transcriptomic data. Neoplastic cells are broadly clustered according to subgroup, and within subgroups discrete sample clustering is associated with chromosomal copy number variance. Each subgroup contains subpopulations exhibiting mitotic , undifferentiated and neuronal differentiated transcript profiles , corroborating other recent medulloblastoma scRNA-seq studies and identifying new subpopulations. We identify a photoreceptor-differentiated subpopulation that is predominantly found in GP3 medulloblastoma, and in SHH, a subpopulation that constitutes differentiating nodules . Deconvolution of a large transcriptomic dataset shows that neoplastic subpopulations are associated with major and minor subgroup subdivisions, for example, photoreceptor subpopulation cells are more abundant in GP3-alpha. This scRNA-seq dataset also demonstrates medulloblastoma subgroup-specific differences in the tumor microenvironment and immune landscape, and reveals an SHH nodule -associated myeloid subpopulation. Additionally, we perform scRNA-seq on genetically engineered mouse (GEM) models of GP3 and SHH medulloblastoma. These models specifically matched the corresponding human subgroup-specific neoplastic subpopulations. We provide an interactive online resource that facilitates exploration of these MB single cell datasets. Collectively, our findings advance our understanding of the neoplastic and immune landscape of the main medulloblastoma subgroups in both humans and GEM models. Overall design: Characterization of childhood medulloblastoma using single-cell RNAseq and methylation profiling. ****Bam files produced by 10x Genomics cellranger were submitted to GEO. These files are available from the SRA database (SRP274357). If needed the bam files can be converted to the original paired-end FASTQs using bamtofastq from 10x Genomics****