Single-cell multi-omics uncovers coordinated epigenetic and transcriptomic evolution in IDH-mutant glioma

IDH-mutant gliomas (IDH-G) typically present as low-grade slow-growing tumors at diagnosis, but invariably progress to high-grade and incurable tumors despite maximal treatment. While previous bulk sequencing studies have shown that IDH-mutant gliomas undergo unique changes in DNA methylation and transcriptional programs during the tumor progression, how these jointly facilitate IDH-G evolution at cellular level has been poorly understood. To address this question, we profiled a longitudinal cohort of IDH-G (36 samples from 19 patients) by multi-omics single-nucleus sequencing, co-capturing full-length transcriptional (by Smart-Seq2) and DNA methylation (by extended-representation bisulfite sequencing (XRBS)) from the same single-nuclei. Compared to reduced-representation bisulfite sequencing (RRBS) used for single-cell DNA methylation profiling in our prior multi-omic single-cell study (Chaligne, et al, Nature Genetics, 2021), the XRBS adopted in this study provided higher coverage of CpG islands (mean of 378,888 vs 198,345, P = 2.2 x 10 -16 ), especially in non-promoter regions. Single-nucleus RNA-sequencing analysis showed a longitudinal increase in the fraction of stem-like state and a reciprocal decrease in that of differentiated state after recurrence. Single-nucleus DNA methylation analysis revealed that in both IDH-G subsets (astrocytoma and oligodendroglioma), IDH-G progression was associated with methylation loss, which marks tumors with worse clinical outcome. This methylation loss was uniformly observed across malignant cells within the same individual tumors and was correlated with the increase in the fraction of stem-like population. This suggests that lower DNA methylation is not due to a change in cell state composition, but rather that it may underlie. Differentially methylated and expressed gene analyses of malignant cells with versus without methylation loss identified hypomethylation of PRC2 targets and increased expression of glioma stem-cell genes as potential mechanisms underlying the expansion of stem-like states. Leveraging our quantitative framework that directly measure cell state heritability and transition dynamics based on high-resolution lineage trees build by DNA methylation information suggested that decreased methylation reshapes cellular transitions to increased heritability of stem-like states and decreased differentiation. This study integrating single-cell methylation and transcription within an evolutionary framework shows how DNA methylation loss drives stem-like transcriptional states and altered cell-state dynamics in IDH-G, offering a blueprint for dissecting evolution in human cancers.