Comparative analysis of 2D and 3D cell culture methods on DNA methylation patterns in medulloblastoma

Epigenetic processes including histone modifications, DNA methylation, chromatin remodeling, and regulatory non-coding RNAs such as microRNAs regulate gene expression without direct alterations to the genome. Among these, DNA methylation which involves the addition of a methyl group to the fifth carbon of cytosine (C), forming 5-methylcytosine (5mC) most frequently in CpG dinucleotides (CpGs), is one of the most studied epigenetic modifications. DNA methylation has been shown to influence tumor development, tumor progression and therapy response in various cancers, including medulloblastoma, the most common malignant brain tumor in children. Due to the extensive genome-wide profiling done by several international consortia and independent groups, extensive genomic databases, including DNA methylation profiles, are publicly available for medulloblastoma making it suitable as a model system to study whether different cell culture techniques influence DNA methylation patterns. Here we sought to investigate whether 3D cell culture systems, commonly referred to as more closely recapitulating in vivo tumor microenvironments, promote DNA methylation patterns in tumor cell lines that are more closely aligned with those of patient tumors. Most commonly used human medulloblastoma cell lines, traditionally cultured as 2D monolayers, were subjected to both anchorage-independent (spheroids) and scaffold-based (collagen, Matrigel, synthetic peptide hydrogel) 3D cell culture methods. Genome-wide methylation profiles were generated using the Infinium MethylationEPIC BeadChip Arrays and compared to the Molecular Neuropathology group/Deutsches Krebsforschungszentrum (DKFZ) brain classifier database. Surprisingly, none of the human medulloblastoma cell lines was considered a match (score equal to or higher than 0.9) and three out of the five cell lines initially tested had values of less than 0.3 and did not meet the study criterion. Moreover, there was considerable differences in scores between cell lines but not between culture condition (2D versus 3D) for a given cell line even when applying different normalization methods to reduce non-biological artifacts. Thus, spheroid- or scaffold-based 3D cell culture methods did not translate into better alignment with patient-derived methylation profiles and suggests limited potential for these models to replicate patient tumor profiles. Furthermore, our data reveal that even in conjunction with advanced cell culture technology, the choice of biological model is a crucial component in engineering accurate disease models.