Enrichment of Tumor-Infiltrating B Cells in Group 4 Medulloblastoma in Children

Abstract

Medulloblastoma (MB) is the most common malignant brain tumor in children. It is classified into core molecular subgroups (non-wingless activated (WNT), sonic hedgehog activated (SHH), Group 3 (G3), and Group 4 (G4)). In this study, we analyzed the tumor-infiltrating immune cells and cytokine profiles of 70 MB patients in Taiwan using transcriptome data. In parallel, immune cell composition in tumors from the SickKids cohort dataset was also analyzed to confirm the findings. The clinical cohort data showed the WNT and G4 MB patients had lower recurrence rates and better 5-year relapse-free survival (RFP) compared with the SHH and G3 MB patients, among the four subgroups of MB. We found tumor-infiltrating B cells (TIL-Bs) enriched in the G4 subgroups in the Taiwanese MB patients and the SickKids cohort dataset. In the G4 subgroups, the patients with a high level of TIL-Bs had better 5-year overall survival. Mast cells presented in G4 MB tumors were positively correlated with TIL-Bs. Higher levels of CXCL13, IL-36γ, and CCL27 were found compared to other subgroups or normal brains. These three cytokines, B cells and mast cells contributed to the unique immune microenvironment in G4 MB tumors. Therefore, B-cell enrichment is a G4-subgroup-specific immune signature and the presence of B cells may be an indicator of a better prognosis in G4 MB patients.

Clinical and Molecular Features in Medulloblastomas Subtypes in Children in a Cohort in Taiwan

“…Cell type deconvolutions were estimated as described in the previous study [ 21 ]. Briefly, the scores of 64 cell types in 5 major cell populations were computed with the gene expression data set normalized to TPM level of 489 cell population specific markers with xCell [ 22 ].…”

“…The DNA methylation array was performed as described in the previous study [ 21 ]. Raw data files were read and preprocessed using the capabilities of Minfi [ 23 ] and the ChAMP [ 24 ] package in the R environment.…”

“…The SNF method was performed in the cohort series as described in the previous study [ 21 ]. Briefly, subtype clustering was performed by the SNFtool package in the R environment based on the top 1% of the most differentially expressed common genes ( n = 216) and probes ( n = 3211) from a previous study [ 2 ].…”

“…Briefly, subtype clustering was performed by the SNFtool package in the R environment based on the top 1% of the most differentially expressed common genes ( n = 216) and probes ( n = 3211) from a previous study [ 2 ]. The parameters of SNF were referred to the previous study [ 21 ].…”

“…The subtyping of non-WNT/non-SHH MBs was based on a web-based classifier of MB G3/4 subgroups ( , accessed on 14 August 2021), as described in the previous study [ 21 ]. Briefly, Illumina Infinium MethylationEPIC array raw signal IDAT-files were uploaded and normalized by a two-factor linear model on log2 transition to the web-based classifier with adjustment for frozen derivatives and patient gender.…”