Glioblastoma (GBM) is a highly aggressive brain tumor that exhibits remarkable heterogeneity at the transcriptional level, which cannot be solely explained by its genetic mutations. This diversity is evident not only across different tumor specimens but also within the same tumor. GBM cells, which originate in the adult brain parenchyma, display a wide range of transcriptional profiles reminiscent of cell states emerging during brain development (Couturier et al., 2020; Phillips et al., 2006); such exceptional heterogeneity is connected to poor therapeutic outcomes and relapses. New results published in the preprint by Mossi Albiach, Janusauskas and co-authors describe the spatial cellular organization in a wide range of glioblastoma (Mossi Albiach et al., 2023 preprint). They find that GBM spatial organization resembles processes of brain development, hypoxia response, tissue damage and wound healing that are independent of the tumor mutational profile. Based on these findings, the authors propose a model describing the process leading to such spatial organization. This may provide new avenues for novel therapeutic interventions targeting the process of construction of distinct spatial organization.
The diversity of cellular states in GBM was addressed at the level of single-cell transcriptomes with astrocyte-like (AC-like), oligodendrocyte progenitor cell-like (OPC-like), neural progenitor cell-like (NPC-like) and an additional mesenchymal state (MES-like) being the core cell states present in tumors in various proportions (Neftel et al., 2019). Moreover, the plasticity of cell states assumes transitions between core modules, which was implicated as being the primary reason for therapy resistance.
The presence of several cell states within individual tumors led to the idea that these states may be involved in different functions and consequently need to be placed in a spatial context. Spatial transcriptomic approaches necessary to address the organization of cellular states within the tissues fall into two categories: (1) unbiased profiling of mRNA species within the tissue captured by beads (Slide-Seq) or spots (10X Genomics Visium), which are usually larger than a cell (thus precluding cellular resolution); and (2) single-molecule in situ hybridization or antibody-based approaches that profile a pre-selected set of genes, transcripts or proteins (thus being biased) within the tissues at a high resolution approaching the level of individual cells or sub-cellular localization. Therefore, there is a trade-off between the cellular resolution and the number of molecules that can be profiled within the tissue. Glioblastoma spatial resolution is now being profiled by both approaches in a complementary way.
Unbiased spatial profiling of GBM samples from 20 patients by 10X Genomics Visium has revealed five distinct transcriptional programs, including radial glia, neurodevelopment, spatial OPC, reactive immune and reactive hypoxia, pointing to the recapitulation of developmental processes and the response to a hypoxic environment at the tumor core being the main determinants of the spatial organization of the tumor (Ravi et al., 2022). In the new preprint by Mossi Albiach, Janusauskas and colleagues, the authors further enhanced the cellular resolution for the spatial characterization of GBM by profiling 888 RNA species by single-molecule in situ hybridization of 57 sections from 41 surgical samples obtained from 27 patients, including two rare samples of oligodendrogliomas. The design of the study allowed them to address inter-patient and intra-tumoral heterogeneity at near-single-cell resolution (called m-cells) across nearly 6.5 million m-cells, building a valuable resource that the authors used to derive several important insights into GBM composition and spatial organization.
First, cross-comparison of GBM spatial profiles with single-cell transcriptomic profiles of the developing brain resulted in a classification of undifferentiated GBM cell types since they matched to glioblasts, pre-oligodendrocyte progenitor cells, outer radial glial cells and neural intermediate progenitor cells. The fine classification of these subtypes is crucial for assessing whether GBM cells are able to recapitulate the full trajectory of normal brain development or whether this process is perturbed in the tumor context.
Second, the study by Mossi Albiach, Janusauskas and colleagues examined the classification of a mesenchymal-like (MES) state, which is associated with unfavorable outcomes and therapy resistance in individuals with GBM. The authors challenged the previous classification of mesenchymal cells in GBM based on the marker genes CHI3L1 and VIM, which are expressed beyond mesenchymal cells. Aligning MES-classified cells from GBM to healthy mesenchymal stem cells and fibroblasts, the authors found true fibroblasts in gliosarcoma – a rare subtype of GBM. The majority of mesenchymal cells in GBM, however, represent SOX2-positive glial tumor cells with wound-healing signatures similar to wound-healing and/or repair processes reported in the brain and skin.
This new finding also links central nervous system (CNS) tumors to neuroblastic tumors of the peripheral nervous system (PNS). Peripheral glia, i.e. Schwann cells, adopt a special repair phenotype upon nerve injury and promote nerve regeneration. Schwann cells associated with PNS tumors called ganglioneuroma express repair Schwann cell-like transcriptional profiles (Weiss et al., 2016, 2021), thus representing a peripheral glial wound-healing type. It is tempting to speculate that glial wound-healing factors may have context-dependent functional repertoires, fueling tumor progression in GBM, while favoring repair functions in peripheral neuroblastic tumors, which results in terminal ganglionic differentiation. Such a distinction recapitulates the injury response in nervous systems, which results in glial scar formation in the CNS and Schwann cell-fueled nerve regeneration in the PNS.
Third, after examining the spatial organization of tumors with various ratios of glial wound-healing states, the authors suggest dynamical changes upon tumor progression. They place developmental-like processes on the leading edge of tumors, which helps tumors expand, while wound-healing states are associated with the most hypoxic tumor core. The wound-healing response of this subset of tumor cells might be linked to the relapse phenotype observed in some individuals upon radiotherapy, as also noted by Ravi et al. This tumor morphology was found in GBM tumors of various genotypes and is therefore not due to a specific genotype but rather to the control of a combination of developmental process and the wound healing response.
Taking into account that the leading edge responsible for the expansion of GBM mimics the normal developmental trajectory (but in the adult brain), the therapeutic interrogation of this process may be an effective treatment option for individuals with GBM. Direct targeting of GBM cells of the more therapy-resistant, wound-healing phenotype (observed after radiotherapy) or inhibiting the switch towards this phenotype may be instrumental in preventing both uncontrolled expansion and later relapses.
Acknowledgements
We are grateful to Igor Adameyko for insightful discussion and constructive feedback.
Footnotes
Funding
S.T.-M. is supported by St Anna Kinderkrebsforschung.
References
Competing interests
The authors declare no competing or financial interests.