Glioblastoma transcriptomic changes in migratory subpopulations at tumor-brain interface

Introduction: Glioblastoma (GBM) is a highly aggressive primary brain tumor with poor prognosis. Primary GBM tumors are highly heterogeneous whose cells are known to invade from the tumor edge diffusively into adjacent brain tissue, making these cells impossible to resect and difficult to treat. These poorly characterized invasive cells are thought to cause tumor recurrence seen in nearly all patients. The difference in transcriptomic expression between migratory and non-migratory cells is not well understood. Recent studies have shown a mechanical difference exists between the tumor bulk and peritumoral brain tissue; however, the extent to which mechanical changes can influence the migratory behavior of GBM cells is not clear. Utilizing genetically stable DNA barcodes and single-cell transcriptomics, we are investigating whether specific tumor cell subpopulations predominately migrate and how mechanics influence migratory behavior of distinct GBM cell subpopulations.

Materials and

Methods: Patient-derived primary GBM cells were transduced with a unique heritable DNA barcode which is expressed as mRNA. To investigate migration at the tumor-brain interface, we have developed an interface hydrogel platform to mimic the mechanical change from the tumor edge to peritumoral tissue. Hydrogels consist of thiolated-hyaluronic acid photocrosslinked with PEG-norbornene, which allows us to adjust the crosslinking density to match the stiffness of tumor and brain. HA concentration, RGD Concentration and diffusion rates were held constant, so that the only effect on migration was stiffness. Barcoded GBM cells were encapsulated into an interface hydrogel with tumor-like stiffness and allowed to migrate into a hydrogel with brain-like stiffness. Cells were cultured for two weeks then cells from each hydrogel were isolated and submitted for single-cell RNA sequencing. We investigated the transcriptomic differences between the migratory and non-migratory cells and lineages by running differential gene expression analysis using the Scanpy toolkit. Gene set enrichment analysis (GSEA) was conducted using pathway datasets from Reactome (PMCID: PMC7710148) and WikiPathways (PMCID: PMC2475545). We calculated the probability that cells from each particular lineage would migrate randomly using a binomial distribution to investigate whether pre-existing subpopulations associated with higher migratory activity were present.

Results, Conclusions, and Discussions: Evaluating differential gene expression between the migratory and non-migratory cells revealed that the most downregulated genes in the migrating cells were related to the matrisome and ECM production (Figure 1a). When GSEA was conducted, we found the most downregulated pathways were those related to collagen biosynthesis, ECM organization, and integrin signaling (Figure 1b). In contrast, many of the highly upregulated genes were associated with motility and poor survival, including 3 members of the metallothionein family (Figure 1a). Upregulated pathways were associated with the HIF1 transcription factor and glycolysis.

We identified 5 lineages with statistically significant migratory activities and 10 lineages with statistically significant non-migratory tendencies. We conducted differential gene expression on these pre-existing lineages using their expression in a control soft gel to investigate the genes and pathways which may indicate a higher tendency for a cell to be migratory. Enriched pathways in lineages more likely to migrate include transportation of secretory cargo to and from the Golgi apparatus and regulation of the microtubule and actin cytoskeletons. Interestingly, the pre-existing migratory subpopulations were enriched for the same metabolomic pathways that were observed as cells migrated across the interface. These results indicate that the pre-existing migratory populations may be better adapted for a migratory phenotype based on metabolism, cytoskeleton rearrangement, and protein secretion.

To investigate cellular composition, we scored each cell based on its expression of 39-50 marker genes from 6 classifications for GBM tumor cells described in Neftel, et al. (PMCID: PMC6703186). We found that clusters with mostly migratory cells had transcriptional signatures most similar to mesenchymal-like cells; however, clusters that were mixed or mostly non-migratory cells included similar numbers of cells with signatures for mesenchymal, astrocyte, OPC-like, and NPC-like tumor subpopulations, indicating the heterogeneity of the patient-derived GBM cell line.

While GBM invasion is influenced by many factors, our results indicate that actively migrating cells downregulate their ECM-integrin interactions and ECM deposition while upregulating pathways for glycolytic metabolism. Furthermore, we identified subpopulations with a pre-existing tendency for migratory behavior that display the same transcriptional signatures as actively migrating cells even prior to migration.

Acknowledgements (Optional):