ID #680 Survival prediction and multimodal characterization of tumor heterogeneity in posterior fossa ependymoma group B

Background Posterior fossa ependymoma can be classified into three distinct molecular types: PF-EPN-A, PF-EPN-B, and PF-EPN-SE. Characterization of PF-EPN-B has primarily focused on describing its molecular heterogeneity and identifying subtypes. PF-EPN-B is characterized by late relapses, without known risk markers to distinguish high- from low-risk cases. However, research focusing on identifying prognostic markers and characterizing this molecular type using high-resolution approaches such as single-nucleus RNA sequencing remains limited. Methods Using an updated cohort of 340 DNA methylation profiles, including 230 patients with available survival data and 16 single-nucleus RNA sequencing profiles, we performed an integrative analysis to identify prognostic markers and to discern the heterogeneity within PF-EPN-B at the single-cell level. Results Based on the DNA methylation data, we were able to identify the previously described subtypes 1-5, however, the clusters 1-3 exhibited less stable and less distinct features compared to clusters 4 and 5. Single-nucleus RNA sequencing of 82,000 cells revealed differences in immune infiltration and distinct immune cell-specific activated metaprograms across the subtypes. Training a linear support vector machine on the 10,000 most variable CpG sites allowed stratification into two risk groups with significantly different progression-free survival (median group 1: 56.6 months, median group 2: 165.0 months, p = 0.0026). Additionally, gain of chromosome 1q emerged as a potential negative prognostic marker for progression-free survival, with enrichment in subtype 1. Conclusion PF-EPN-B exhibits substantial heterogeneity in both copy number alteration profiles and gene expression patterns across its subtypes. The identification of chromosome 1q gain as a possible prognostic marker and differences in immune infiltration between the subtypes may enable improved risk stratification and support the development of more precise, patient-tailored therapeutic strategies.