(P 114) METHYLOME PROFILING OF CARTILAGE TUMOURS: A PROMISING NEW TOOL FOR BIOPSY MATERIAL?

Objective: DNA methylation-based tumor classification has been shown to be a reliable diagnostic tool for CNS tumours and have impacted also the new WHO classification in adding new tumours subtypes mainly defined by their methylome profile. There is a growing amount of evidence that this approach is also helpful in classifying soft tissue and bone tumours, particularly in those subtypes that lack specific markers and/or are histologically similar.
The histologic features of higher grade chondrosarcoma (CS 2-3) can be focal and therefore difficult or even impossible to distinguish from atypical cartilaginous tumour (ACT) / low-grade chondrosarcoma (CS 1) on core needle biopsy. Patient management differs between the two groups and new approaches to reach a reliable diagnosis are urgently needed. Enchondroma (EC) and ACT / CS1 can be challenging to distinguish as well but pose a clinically less important problem. Grading by an experienced bone tumour pathologist should always consider also the corresponding clinical and imaging findings. Aiming to elucidate if methylome profiling could aid in this diagnostic challeng, we studied a series of 118 cartilage tumours including EC, ACT and CS for their epigenomic profiles and compared them to a larger set of reference data.

Methods: Hundred and eighteen chondrogenic tumors were analysed in this study, including EC, ACT, conventional and dedifferentiated CS characterized based on histomorphology features as well as radiological imaging. In addition, IDH1 and IDH2 mutation status has been determined for all cases. Array-based genome-wide DNA methylation evaluation was conducted using Infinium Human MethylationEPIC Array (Illumina). Our series of methylomes were compared to own and publicly available data including the cases derived from the study of Koelsche et al. (2020). Copy number profiles were extrapolated directly from DNA methylation arrays. Finally, DNA methylation levels of selected gene promoters were interrogated, in particular pCDKN2A.

Results: DNA methylation-based clustering of chondrosarcomas did not result in the formation of a unique cluster but rather multiple subclusters. The main determinant of the distribution of samples among the seven distinct subgroups was the presence of IDH1/2 mutations. IDH-mutant CS arising in the head and neck region (mostly skull base) presented a unique signature as has been described before. Similarly, dedifferentiated CS were also found in a dedicated cluster and IDH wildtype CS 2-3 showed a different clustering based on grade. This trend could also be observed between IDH-mutant CS 2-3, although to a lesser extent. Surprisingly, most ACT (12/21) formed a distinction cluster. The remaining ACT (9/21) were displayed together with the EC and few CS 2 preventing a clear discrimination between the subtypes. As expected, the proportion of copy number alterations increased according to tumor grade, thus supporting increased gemomic complexity upon chondrosarcoma progression. Finally, clusters composed of either CS 3 or dedifferentiated CS were characterized by an impairment of the CDNK2A locus through copy losses or hypermethylation of its promoter.

Conclusion: DNA methylation-based classification of cartilage tumours appears to be an informative tool for discriminating the vast majority of low- and intermediate-/high-grade CS and also dedifferentiated CS, particularly if correlated with histology, imaging and clinical findings. Furthermore, our set of ACT splits into two groups carrying different genome-wide methylation signatures: the first one characterized by an ACT-specific methylation pattern whereas the second group being positioned within the cluster of enchondromas.