The white matter is a pro-differentiative niche for glioblastoma

Glioblastomas are hierarchically organised tumours driven by glioma stem cells that retain partial differentiation potential. Glioma stem cells are maintained in specialised microenvironments, but whether, or how, they undergo lineage progression outside of these niches remains unclear. Here we identify the white matter as a differentiative niche for glioblastomas with oligodendrocyte lineage competency. Tumour cells in contact with white matter acquire pre-oligodendrocyte fate, resulting in decreased proliferation and invasion. Differentiation is a response to white matter injury, which is caused by tumour infiltration itself in a tumoursuppressive feedback loop. Mechanistically, tumour cell differentiation is driven by selective white matter upregulation of SOX10, a master regulator of normal oligodendrogenesis. SOX10 overexpression or treatment with myelination-promoting agents that upregulate endogenous SOX10, mimic this response, leading to niche-independent pre-oligodendrocyte differentiation and tumour suppression in vivo. Thus, glioblastoma recapitulates an injury response and exploiting this latent programme may offer treatment opportunities for a subset of patients.

Glioblastoma (GBM) is the most common and malignant primary brain tumour1. Stark resistance to current treatments, which include maximal surgical resection, chemo- and radiotherapy, leads to tumour relapse in virtually all patients with a median survival of less than 15 months2. GBM initiation, growth and recurrence are thought to be rooted within a subpopulation of therapy-resistant tumour cells with properties of normal neural stem cells, termed glioma stem cells (GSCs)3,4,5,6,7.

Mounting evidence indicates that GSCs fuel tumourigenesis by recapitulating normal neural lineage hierarchies of quiescence, self-renewal and generation of non-dividing progeny8,9,10. Fate decisions of normal neural stem cells are tightly controlled by the microenvironment, which maintains stemness in specialised niches and directs differentiation along the appropriate lineages11. Similarly, it is well established that GSCs are maintained in perivascular and hypoxic regions of the tumour bulk12,13,14. In contrast, little is known about the differentiation potential of GSCs within tumours and whether pro-differentiative niches exist. However, improved understanding of lineage progression is central to our understanding of the disease and may reveal novel strategies to suppress GBM growth and recurrence by directing GSC fate towards a mature, non-dividing state. Consistent with this idea, manipulation of candidate developmental pathways to promote tumour cell differentiation into astrocyte and neuronal lineages has shown efficacy in preclinical models10,15.

Diffuse infiltration into the normal brain parenchyma is a hallmark of GBM and underlies recurrence by precluding complete surgical resection16,17. As GBM cells infiltrate away from the tumour bulk, they are confronted with new and heterogeneous microenvironments, which would be predicted to affect fate decisions18. Indeed, it has been proposed that invading GSCs may lose stemness, but this remains controversial, with studies both in support and against this idea19,20,21,22. A predominant route of infiltration is the white matter, which consists of bundles of myelinated axons, known as tracts, and represents approximately 60% of the brain16,23. GBMs frequently develop within white matter and spread throughout the brain using myelinated fibres as scaffolds for cell migration24. This includes spread to the contralateral hemisphere, which occurs exclusively along the white matter tracts of the corpus callosum16. Remarkably, the cellular and molecular mechanisms that underpin white matter invasion remain almost entirely unknown.

Here, we investigated the impact of the white matter microenvironment on tumour cell fate by profiling region-specific transcriptomes of GBM cells invaded into white and grey matter, alongside matched bulk cells. Unexpectedly, this analysis revealed that the white matter dominantly suppresses malignancy by directing GSC differentiation towards pre-oligodendrocyte fate. We show that this process recapitulates an injury response and can be harnessed to suppress tumourigenicity in a subset of GBMs.

Region-specific GBM transcriptomes

Despite the fundamental role of white matter in GBM biology, how this specialised microenvironment might modulate tumour cell behaviour remains unknown. To probe white matter phenotypes, we made use of a well characterised patient-derived GSC line with propensity to invade along white matter tracts (G144)25. GFP-labelled G144 cells were stereotactically injected into the striatum of immunocompromised mice and, upon development of clinically apparent disease, the tumour bulk (B) and margin regions extending into the corpus callosum (CC) as a region of white matter, and the striatum (ST) as a region of grey matter, were micro-dissected under fluorescence guidance (Fig. 1a). Tumour tissue was dissociated to single cells and GFP+ tumour cells FACS-purified, pooled and processed for RNA-sequencing (RNA-seq). Bioinformatics analysis of human tumour reads revealed that margin cells are transcriptionally distinct from bulk cells, with invasive cells acquiring markers of lineage progression towards astrocytes, as well as downregulating proliferation signatures, which resulted in decreased EdU incorporation in vivo (Fig. 1b, c, Supplementary Fig. 1a–e, Supplementary Table 1 and Supplementary data 1–3)26,27,28. Strikingly, it also indicated that the transcriptomes of CC and ST tumour cells differed from one another (Fig. 1b). Tumour cells invading into white matter selectively upregulated signatures of oligodendroglia, suggesting that myelinated regions may promote progression along the oligodendrocyte lineage (Fig. 1c, Supplementary Fig. 1f and Supplementary Table 1, and Supplementary data 1–3). In agreement with this, CC cells upregulated a panel of oligodendrocyte lineage marker genes, including the transcription factor SOX10, a master regulator of oligodendrogenesis (Fig. 1d)26,29,30. Notably, upregulation of myelin genes was mild and incomplete in the CC, indicative of partial differentiation, as expected from cancer cells31. To determine whether these gene expression changes also correlated with phenotypic changes, we characterised the differentiation response at the single cell level in two sets of complementary experiments. First, we dissociated G144 cells from the CC, ST and B of xenografts, seeded them acutely in mitogen-free media and assessed expression of SOX10, the pre-oligodendrocyte marker O4 and the myelinating oligodendrocyte marker myelin basic protein (MBP) by immunocytochemistry (Supplementary Fig. 1g, h). Preparations isolated from the CC had a strong increase in the proportion of cells positive for SOX10, the majority of which also expressed O4, but not MBP, confirming that transcriptomic signatures reflect changes in tumour cell fate. Second, we examined lineage marker expression within xenografts in situ by immunohistochemistry. GFP or the human-specific nuclear antigen (HuNu) were used to label tumour cells and distinguish them from endogenous mouse glia (Supplementary Fig. 1i, j). The majority of tumour cells began to express high levels of SOX10 in white matter, including in myelinated fibres inside the tumour bulk (Fig. 1e, f). SOX10+ tumour cells were significantly less proliferative than tumour cells that remained SOX10- within the same region (Fig. 1g,h) and occasionally expressed the immature oligodendrocyte markers CNP and CC1 (Supplementary Fig. 1k, l), but were again negative for the mature marker MBP (Supplementary Fig. 1m). To further assess the white matter specificity of this response, we implanted G144 cells directly into cortical grey matter or in the corpus callosum and examined their differentiation using SOX10 induction as a read-out. Differentiated SOX10+/EdU− tumour cells were only found in the CC, confirming that the white matter selectivey promotes partial GBM differentiation (Fig. 1i,j).

The observed differentiation block of GBM cells is likely due to amplification of PDGFRA in this line, a key negative regulator of developmental oligodendrocyte differentiation which is commonly mutated in proneural/OPC-like GBM cells9. Indeed, overexpression of constitutively active PDGFRA in cultured mouse neural stem cells arrested their differentiation at the O4+ pre-oligodendrocyte stage, confirming that failure to switch off PDGFRA signalling plays an important role in preventing terminal differentiation of tumour cells (Supplementary Fig. 1n–p). Thus, GBM cells differentiate to a pre-oligodendrocyte/immature pre-oligodendrocyte state in white matter, which is marked by induction of SOX10 expression.

To determine the generality of these findings, we examined SOX10 expression in the TCGA dataset and in a panel of primary GSC lines by qPCR. Consistent with previous studies, we found that a subset of patient tumours (~18%) expressed high levels of SOX10. SOX10 high tumours were enriched for the proneural, but also included mesenchymal and classical transcriptional subtypes, as well as IDH1 mutant tumours, suggesting that pre-oligodendrocyte differentiation may occur across all GBMs (Supplementary Fig. 2a, b)9. In addition, a subset of primary GSC lines retained above average SOX10 expression in vitro, confirming cell-intrinsic expression in tumour cells (Supplementary Fig. 2c and Supplementary Table 2). We next xenotransplanted five SOX10+ GSC lines (three of which were freshly dissociated from primary tumours) and examined their behaviour in white matter. One line had lost white matter regulation of SOX10 induction, displaying homogeneous high SOX10 expression throughout the xenograft and was not analysed further. In contrast, the other four lines were similar to G144 cells, with a somewhat variable, but consistent increase in the proportion of SOX10+ cells in white matter as compared to grey matter (Fig. 2a, b and Supplementary Fig. 2d). In addition, as in G144 xenografts, SOX10+ cells proliferated significantly less than SOX10− cells in white matter and partially progressed to CNP+ or CC1+ immature pre-oligodendrocyte cells (Fig. 2c and Supplementary Fig. 1k, l and Fig. 2e). To determine whether the white matter-induced response also occurs in primary patient tumours, we took two complementary approaches. First, we selected three SOX10+ cases based on retained SOX10 expression in derivative GSC lines (GCGR-L12, GL67 and GL23). We found that in all cases the white matter contained SOX2+ tumour cells, which upregulated SOX10 and proliferated less than their SOX10-counterparts, confirming that GSC behaviour in xenograft models reflects disease phenotypes (Fig. 2d, e and Supplementary Table 3)3. SOX2 staining was specific to GBM cells, as the vast majority of endogenous SOX10+ OPCs were negative for SOX2 in areas of tumour infiltration (Supplementary Fig. 3a). Second, we selected 23 white matter-containing cases without prior knowledge of SOX10 status. Remarkably, we found that 15 cases contained at least a subset of SOX10+ cells in white matter, regardless of molecular characteristics (Fig. 2f, Supplementary 3b and Supplementary Table 3). In addition, SOX10+cells were lowly proliferative, progressed to a CNP+/CC1+/MBP− immature pre-oligodendrocyte state and were restricted to MBP+ tumour regions (Fig. 2e and Supplementary Fig. 3c–e). Finally, in a somatic Nf1/Pten/p53 CRISPR-based mouse model, while the vast majority of tumour cells were SOX10+, a much greater proportion of tdTomato+ tumour cells differentiated to CC1+/CNP+/Ki67− immature oligodendrocyte cells in the white matter relative to the grey matter (Supplementary Fig. 3f–j). These experiments confirm that a subset of GBMs progress to pre-oligodendrocyte-like cells in white-matter.

Disrupted white matter drives pre-oligodendrocyte differentiation

Next, we sought to understand what properties of white matter are differentiation-promoting. We noticed that SOX10 induction was most robust in areas where OSP staining appeared disrupted, suggesting that it might be linked to demyelination (Supplementary Fig. 1i). We therefore assessed myelin integrity in xenografts by fluoromyelin staining and by electron microscopy (EM). The proportion of SOX10+/HuNu+ tumour cells was highest in heavily infiltrated, fluoromyelin negative, white matter areas, as compared to dye-positive, low tumour-density, myelin regions in four independent xenografts (Fig. 3a, b and Supplementary Fig. 4a). A similar differentiation pattern was also found in patient tissue, where only few SOX2+ tumour cells upregulated SOX10 in intact myelin (Supplementary Fig. 4b). EM analysis of G144 xenografts further revealed severe axonal pathology and demyelination in highly infiltrated white matter regions, including axonal swelling and vacuolisation, presence of dark axons, myelin decompaction and a significant increase in g-ratios relative to contralateral normal brain (Fig. 3c–i and Supplementary Fig. 4c). In contrast, g-ratios remained normal in the tumour-infiltrated striatal grey matter (Fig. 3h), indicating that demyelination is specific to white matter. Consistent with this, activated microglia increased selectively in tumour-infiltrated white matter and endogenous oligodendrocytes were significantly decreased in the CC (Supplementary Fig. 4d–g). Furthermore, activated microglia with engulfed myelin debris was frequently observed in both xenografts (Supplementary Movie File 1) and human samples, and colocalised with SOX10+ tumour cells (Supplementary Fig. 4h–j). Surprisingly, despite severe myelin dysfunction, the mice did not exhibit overt motor or cognitive deficits.

Super resolution and correlative light and electron microscopy (CLEM) showed that within areas of myelin disruption tumour cells associated closely with both intact myelinated and demyelinating axons (Fig. 3j, k). This did not affect tumour cell survival (Supplementary Fig. 4k, l), but occasionally resulted in tumour cells acquiring oligodendrocyte morphology, in line with a differentiation response (Supplementary Fig. 4m).

Together, this data supports a model whereby tumour infiltration into the white matter induces an injury-like microenvironment, which in turn triggers GBM differentiation in a tumoursuppressive feedback loop. To test this more directly, we carried out a time-course analysis of the response of GBM cells and the CC microenvironment to tumour infiltration. G144 xenografts were collected at 2 weeks intervals from full engraftment at 4 weeks to symptomatic disease at 12 weeks post-implantation and subjected to immunohistochemistry for differentiation and activated glia markers. Tumour infiltration resulted in progressive loss of myelin integrity, which correlated with a gradual increase in astrocyte reactivity, microglia activation, OPC activation and oligodendrocyte death, all hallmarks of the glial response to brain injury (Fig. 3l–r, Supplementary Fig. 5a–d). Importantly, tumour cell differentiation paralleled these changes, with the number of SOX10+/EdU− tumour cells gradually increasing over time and remaining lowly proliferative, indicative of stable pre-oligodendrocyte differentiation (Fig. 3s, Supplementary Fig. 5e).

To determine whether injury is causal to tumour differentiation, we next performed gain-of-function experiments. G144 cells were injected directly into the densely myelinated inner layers of the cortex, a brain region in which tumour cells frequently infiltrate from the striatal tumour bulk, but do not differentiate (Fig. 3t). Remarkably, the stab-wound injury caused by the needle (evidenced by induction of astrocyte reactivity and microglia activation, Supplementary Fig. 5f–h) was sufficient to induce GBM cell differentiation to SOX10+/EdU− cells (Fig. 3t, u). Thus, GBM differentiation is a white matter injury-like response.

Differentiation is niche-dependent

To determine the stability of white-matter induced differentiation, we carried out secondary xenografts. G144 cells were isolated from CC, ST and B xenograft regions as above, and immediately re-injected in the striatum of secondary hosts (Fig. 4a). Survival analysis indicated that all cells, regardless of region of origin, formed tumours with similar latency, indicative of comparable tumourigenic potential (Fig. 4b). Furthermore, independent of initial SOX10 levels, all secondary lesions recapitulated SOX10 expression patterns of the primary tumours, which was high in white matter and low in grey matter regions and bulk (Fig. 4c). This suggests that once removed from white matter, pre-oligodendrocyte tumour cells may de-differentiate back to a GSC state. To directly test this hypothesis, we used the surface marker O4 together with GFP to FACS-purify pre-oligodendrocyte tumour cells from myelin-rich CC and B regions of xenografts, seeded them acutely in vitro in the presence of mitogens and monitored their morphology and proliferation in real-time by time-lapse microscopy (Supplementary Fig. 6a, b). Consistent with the in vivo results, we found that approximately 20% of branched, pre-oligodendrocyte O4+ tumour cells retracted their processes, re-acquired GSC morphology, and re-entered the cell-cycle (Supplementary Fig. 6b and Supplementary Movie Files 2, 3). Thus, maintenance of pre-oligodendrocyte fate in vivo requires continuous exposure to white matter.

SOX10 drives differentiation

In normal oligodendrogenesis, SOX10 is essential for lineage progression, differentiation and myelination32,33,34. We therefore asked whether the increase in SOX10 observed in white matter may also play a causal role in the induction of tumour cell differentiation. To address this, we transduced G144 cells with an inducible Tet-ON SOX10 overexpression construct and profiled vehicle- and doxycycline (Dox)-treated cultures by RNA-seq at 48 h post-induction. This time point was chosen to maximise the specificity of detected transcriptional changes and enrich for direct SOX10 targets. Immunofluorescence analysis confirmed detectable SOX10 protein expression in approximately 60% of the cells (Supplementary Fig. 6c, d). We identified 186 differentially expressed genes, the majority of which were upregulated (157, Supplementary data 4). Approximately 40% of SOX10-induced genes overlapped with genes upregulated in the corpus callosum in vivo, alongside significant overlap with signatures of oligodendrocyte lineage cells (Fig. 5a). These genes included both immature and mature oligodendrocyte markers, such as GPR17, ERBB3 and the myelin genes PLP1, CLDN11, MYRF and UGT8 (Fig. 5b). Furthermore, GO analysis identified enrichment for terms associated with oligodendrocyte differentiation, including axon ensheathment, myelination, cell adhesion and positive regulation of gliogenesis (Fig. 5c and Supplementary data 5). Thus, SOX10 controls a transcriptional programme of oligodendrogenesis in tumour cells.

To functionally assess effects of this programme on tumour cell phenotypes, we overexpressed SOX10 in a panel of patient-derived GSC lines. This included two lines with ability to spontaneously differentiate to pre-oligodendrocytes (G7, G144). Cells were transduced with lentiviral vectors expressing constitutive SOX10 (SOX10 OE), or control empty vectors, and subjected to differentiation by growth factor withdrawal for 7 days35. SOX10 overexpression induced 7 of the 12 lines to generate O4+ pre-oligodendrocytes, regardless of basal differentiation competency, and, remarkably, decreased proliferation in all lines, as measured by EdU incorporation (Fig. 5d–g). This indicates that SOX10 is sufficient for GBM cell differentiation. To determine if it is also necessary, we knocked-out SOX10 by gene editing in G144 cells and carried out differentiation assays as above. We found that differentiation to O4+ cells was fully abolished in the absence of SOX10 (Fig. 5h–j). Together, these experiments demonstrate that SOX10 functions as a master regulator of pre-oligodendrocyte fate in GBM and confirm that corpus callosum phenotypes in vivo are largely mediated by SOX10 upregulation.

Increased SOX10 suppresses tumourigenesis

Our results so far suggest that manipulations capable of increasing SOX10 levels niche-independently should suppress tumourigenesis by promoting stable differentiation. To test this more directly, we performed intracranial transplantations of luciferase and GFP-tagged G144 cells transduced with SOX10 OE or Control lentiviruses and monitored tumour growth and disease-free survival. SOX10 overexpression significantly delayed tumourigenesis, resulting in an overall increase in median survival from 65 to 91 days (Fig. 6a, b). To understand the mechanisms responsible, we examined Control and SOX10 OE tumours at 4 weeks post implantation by immunofluorescence analysis. EdU labelling revealed a marked decrease in the total number of proliferating tumour cells in SOX10 OE compared to Control lesions (Fig. 6c, d). Importantly, time-course analysis of SOX10 levels in the tumours that eventually formed revealed that SOX10 overexpressing cells were progressively outcompeted by cells that had escaped transduction, confirming that high SOX10 levels are tumoursuppressive (Supplementary Fig. 6e).

Although SOX10-mediated differentiation was most pronounced in invasive tumour cells that infiltrate the corpus callosum (Fig. 1e), normal oligodendrocyte maturation is accompanied by a progressive loss of migratory potential36,37. We therefore sought to understand the impact of differentiation on GBM invasion by imaging whole sections of 4 weeks tumours. Surprisingly, SOX10 OE tumours appeared less diffuse and migrated much shorter distances than controls (Fig. 6e and Supplementary Fig. 6f). In addition, within SOX10 OE tumours the majority of invasive tumour cells had low or undetactable SOX10 levels, whereas SOX10 high cells were found predominantly closest to the tumour bulk (Fig. 6f). These results indicate that differentiation also reduces the migration of GBM cells. To determine whether this phenotype was cell-intrinsic, we measured in vitro motility of SOX10 OE and Control G144 cells by live cell imaging and found that increased SOX10 levels profoundly reduced cell motility (Supplementary Fig. 6g–i). This was not an artefact of SOX10 overexpression, as O4+ tumour cells acutely FACS-purified from the corpus callosum of wildtype G144 xenografts were also significantly less motile than O4- cells from the same region (Fig. 6g).

These results suggest that white matter effects are tumour suppressive and might slow down the progression of the primary disease. To test this hypothesis experimentally, we compared the tumourigenicity of parental and SOX10 knock-out G144 cells, in which differentiation along the oligodendrocyte lineage is abolished (Fig. 5h–j). SOX10 knock-out tumours had a significantly faster growth rate than controls, as measured by longitudinal bioluminescence imaging and analysis of EdU incorporation (Fig. 6h–i). These experiments demonstrate that white matter-driven pre-oligodendrocyte maturation suppresses tumourigenesis by inhibiting proliferation and invasion of GBM cells and that in the absence of this response GBMs are more aggressive.

Finally, we explored the translational potential of these findings by testing whether known myelination-inducing pharmacological agents could differentiate GBM cells by increasing endogenous SOX10 levels. We first treated cultured G144 cells with two compounds: the clinically-approved anti-asthma medication Pranlukast and the cell permeable cAMP analogue dibutyril cAMP (db-cAMP). In addition to blocking cysteinyl-leukotriene 1 receptor, Pranlukast inhibits GPR17, a negative regulator of oligodendrocyte development38,39,40. GPR17 is normally expressed in immature oligodendrocytes38 and, importantly, was strongly induced in G144 cells purified from the CC and upon SOX10 overexpression (Figs. 1d and 5b). db-cAMP treatment of normal progenitors results in elevated intracellular cAMP levels, which promote oligodendrocyte differentiation, likely by phenocopying GPCR activity41,42. Treatment of G144 with either compound for 2 weeks was sufficient to increase endogenous SOX10 levels and differentiation to O4+ pre-oligodendrocyte cells, leading to a reduction in proliferation (Fig. 7a–h). These effects were fully dependent on SOX10, as no differentiation was observed in drug-treated SOX10 knock-out cells (Supplementary Fig. 7a, b). Next, we administered Pranlukast to tumour-bearing mice in vivo. As Pranlukast has poor BBB penetration43, we delivered it intrathecally using osmotic mini-pumps. Strikingly, we found a significant increase in the number of SOX10+cells in Pranlukast-treated tumours, which was accompanied by a dramatic reduction in proliferation relative to saline-treated controls (Fig. 7i–l). Pranlukast did not affect endogenous inflammatory glia, indicative of a direct effect on the tumour cells (Supplementary Fig. 7c–e). We conclude that a subset of GBMs can be induced to undergo pre-oligodendrocyte differentiation with small molecules that raise SOX10 levels and propose that such differentiation therapy may be an effective strategy for curtailing tumourigenesis and recurrence.

The failure of both conventional and targeted therapies is commonly attributed to the pervasive molecular intra- and intertumoural heterogeneity of GBM, which occurs at the genetic, epigenetic, transcriptional and functional levels44,45,46,47,48,49,50,51. Despite this staggering diversity however, recent studies are beginning to reveal that all GBMs converge on a finite number of cellular states, which recapitulate normal developmental programmes8,9. Such convergence offers hope of identifying shared biological vulnerabilities that could be exploited for the treatment of many patients, independent of genetic diversity. Here we identified one such vulnerability: the competence of GBM cells to undergo pre-oligodendrocyte differentiation.

Although GSCs retain ability to partially differentiate into mature glial cells in vitro, the extent to which this occurs in tumours has remained unclear8,9,10,52. Our results demonstrate the existence of robust differentiation in vivo, including in patient material, which is controlled by the microenvironment. They further reveal that differentiation occurs in specialised niches within the brain and identify the white matter as one such pro-differentiative niche for tumour cells with oligodendrocyte competency30. Thus, despite the presence of extensive genetic abnormalities, exposure to appropriate environmental cues is sufficient to revert GBM cells to a more normal, differentiated phenotype, underscoring the dominance of the microenvironment in suppressing malignancy53,54.

We found that differentiation is an injury-like response, which results from infiltrating tumour cells disrupting the white matter. Thus, although invasion along white matter is a common mode of infiltration, the ensuing myelin damage creates a tumour-suppressive feedback loop, which paradoxically slows GBM spread. Interestingly, the microenvironment of tumour-infiltrated white matter appeared remarkably similar to that of neuroinflammatory disease states, entailing severe demyelination and axonal pathology55. Our EM analysis suggests that in the tumour context demyelination may be caused by both oligodendrocyte and neuron death. The underlying mechanisms remain to be determined, but a combination of compression injury to the axons, microglia and astrocytes activation, as well as excitotoxins are all likely to contribute55,56.

The finding of OPC activation in tumour-infiltrated white matter suggests that the ensuing demyelinating microenvironment initiates a repair programme in normal glia. It is therefore tempting to speculate that GBM cells may recapitulate a similar oligodendrogenic programme, which results in their partial differentiation to pre-oligodendrocyte cells. While it is clear that an injury-like microenvironment drives differentiation, the signals and cell types responsible remain undefined. These are likely to be complex and combinatorial, but our results suggest that microglia, a key regulator of normal re-myelination, may play an important role57. Indeed, we found that microglia activation was selective to tumour white matter and immediately preceded tumour cell differentiation. It would be of great interest to examine the contribution of microglia to GBM differentiation in future studies.

Our findings are of clinical relevance as we show that exploiting the aberrant GBM injury-like response through niche-independent upregulation of SOX10, locked tumour cells in the differentiated, non-proliferative state and suppressed tumourigenesis in preclinical models of the disease. Furthermore, our work predicts that myelination-promoting compounds, should be particularly effective in tumours with oligodendrocyte lineage competency, providing a potential strategy for patient stratification. Future studies should continue to explore how the intersection of repair and developmental programmes may afford novel translational opportunities for GBM.