ID #1114 Integrated bulk and single cell profiling of diffuse midline glioma, H3K27-altered tumours and models
Yura Grabovska, Alan Mackay, Laura Bevington, Anna Burford, Drenusha Sejdiu, Sara Temelso, Rita Pereira, Haider Tari, Diana Carvalho, Molina Das, Shauna Crampsie, Valeria Molinari, Rebecca Rogers, Ketty Kessler, Lynn Bjerke, Jiin Song, Emma Lamden, Iza Jasyck, Joselin de Faye, Leslie Bridges, Zita Reisz, Safa Al-Sarraj, Navneet Singh, Simon Stapleton, Cristina Bleil, Samantha Hettige, Bassel Zebian, Julia Cockle, Fernando Carceller, Matthew Clarke, Chris JonesAbstract
Background
Diffuse midline glioma, H3K27-altered (DMG-H3K27) are recognised as a distinct type of paediatric diffuse high-grade glioma with dismal clinical outcome in the WHO2021 CNS classification, defined by an integrated diagnosis including the defining oncohistone H3 alterations. Increasing evidence suggests the presence of distinct subtypes, but these are not yet well-defined.
Methods
Published and unpublished DNA sequencing from n = 760 DMG-H3K27 cases were integrated with n = 808 cases with methylation array profiling, along with bulk (n = 379) and single-cell (n = 35) RNAseq data. We additionally prospectively generated 28 patient-derived models both in vitro and in vivo that were molecularly credentialled in parallel.
Results
Clustering of methylation data by tSNE/UMAP showed a clear separation largely driven by anatomical location, with a cluster enriched for pontine lesions (DIPG, n = 291) separating from those enriched in non-brainstem lesions such as the thalamus (n = 414), tightly linked to H3 variant status, and highly enriched for gain of chromosome 1q or loss of 5q, respectively. The two major methylation-based clusters were readily discriminated on the basis of differentially methylated regions driving differential expression and scRNA-seq, highlighting the previously reported developmental origins associated with location, but also novel markers associated with differences in tumour immune microenvironment, and distinct subpopulations of cells driving tumour infiltration. Co-segregating mutations in the MAPK pathway, incorporating variously BRAF_V600E, NF1, FGFR1, and PTPN11 may define a clinically distinct group of tumours in the thalamus, though PIK3R1 InDels and non-canonical BRAF mutations in the pontine DMGs highlighted aggressive tumours for which in vitro modelling reveals initial sensitivity but also complex resistance to targeted MAPK inhibition. These tumours were clearly distinct from those of the DMG-EGFR subgroup, which clustered by methylation with diffusely infiltrating H3-WT hemispheric glioma, harboured frequent chromosome 7 gains absent from other DMG subtypes, hypomethylation of the EGFR/IGF1R loci, and EZHIP overexpression.
Conclusion
DMG-H3K27 comprise multiple subgroups with clinically relevant phenotypes driven by molecular features which arise within the context of distinct anatomical locations. A novel panel of patient-derived models which reflect these subgroups will aid in mechanistic and preclinical studies to develop biologically-rational treatments.