Both the perivascular niche (PVN) and the integration into multicellular networks by tumor microtubes (TMs) have been associated with progression and resistance to therapies in glioblastoma, but their specific contribution remained unknown. By long-term tracking of tumor cell fate and dynamics in the live mouse brain, differential therapeutic responses in both niches are determined. Both the PVN, a preferential location of long-term quiescent glioma cells, and network integration facilitate resistance against cytotoxic effects of radiotherapy and chemotherapy—independently of each other, but with additive effects. Perivascular glioblastoma cells are particularly able to actively repair damage to tumor regions. Population of the PVN and resistance in it depend on proficient NOTCH1 expression. In turn, NOTCH1 downregulation induces resistant multicellular networks by TM extension. Our findings identify NOTCH1 as a central switch between the PVN and network niche in glioma, and demonstrate robust cross-compensation when only one niche is targeted.
Malignant gliomas, including glioblastoma, are the most frequent and yet incurable primary brain tumors characterized by early infiltrative growth and high therapy resistance1. Apart from frank angiogenesis, the vasculature of the brain and of the tumor seem to play a special role in glioma biology: a subpopulation of stem-like cells was identified in brain tumors that is potentially responsible for their frequent treatment failure2,3,4,5, and these stem-like cells can inhabit distinct perivascular niches (PVN)6,7,8,9,10. Importantly, those stem-like glioma cells share marker expression and molecular traits with neural stem cells (NSCs)2,3, and NSCs also populate distinct neurogenic niches, which include PVN that provide a unique microenvironment for the preservation of the stem cell pool11. Targeting of the PVN thus emerged as a strategy to eradicate cancer stem-like cells in incurable gliomas, to improve their response to therapy and to prevent tumor recurrence9,12, but at the same time raised the question of the collateral effects associated with it.
In line with the neurodevelopmental origins of glioma13,14 and shared traits between neural progenitor cells and glioma cells, neurite-like cellular protrusions named tumor microtubes (TMs) have recently been characterized in gliomas15,16. TMs are leading structures during glioma cell invasion15,17 and enable glioma cells to form functional, communicating multicellular networks that render network members resistant against cytotoxic therapies, likely due to improved cellular homeostasis15,18.
All in all, while there is increasing evidence that interactions with pre-existing blood vessels in the brain are of importance for tumor progression and resistance in glioma, their exact cellular and molecular mechanisms and overall role in brain tumor biology are not sufficiently understood. By following patient-derived primary glioblastoma cells dynamically for extended periods of time with in vivo two photon microscopy, we here describe mechanisms of tumor progression that depend on interactions with brain microvessels and tumor cell networks, and provide the definitive confirmation that the NOTCH1-dependent PVN is a primary niche of resistance in brain tumors. We also demonstrate that NOTCH1 is an important modulator of TM-network formation, and interference with this pathway leads to a partial cross-compensation between these two prime niches of resistance in glioma.
A perivascular niche for tumor cells in glioma
First, we asked how tumor cells in patient samples from diffusely infiltrating gliomas are distributed in relation to blood vessels in the brain. Mutation-specific immunohistochemical stainings against IDH1-R132H were used, which allow specific detection of tumor cell structures vs. nonmalignant brain cells in IDH1 mutant human oligodendroglioma and astrocytoma samples19,20. As validated before21, nestin staining allowed to detect tumor cells in IDH wild-type glioblastoma. Irrespective of the glioma subtype, only a minority of tumor cells closely associated with blood vessels in a distinct perivascular niche (Fig. 1a; arrowheads in Fig. 1b). The majority of tumor cells populated the parenchymal compartment distant from blood vessels (Fig. 1a, with representative examples given in Fig. 1b where parenchymal, non-perivascular positions are evident on the left side images, but also visible on the right side images). This phenotype and population of different compartments could also be recapitulated in a patient-derived xenograft model growing in the mouse brain (Fig. 1c shows a microregion with primarily parenchymal, Fig. 1d one with primarily perivascular positions). In both human (Fig. 1b, e) and mouse (Fig. 1c, d) tumor samples, many parenchymal but also some perivascular cells were clearly characterized by the extension of TMs, which connected single glioma cells to a network of tumor cells, as described before15,21,22. This TM-connected tumor cell network becomes most evident in 3D reconstructions of thick sections from mouse (Fig. 1c) and man (Fig. 1e).
In principle, the study of individual glioblastoma cells in these two niches—their plasticity, heterogeneity, stroma interactions, contribution to tumor progression, and their response to therapies—would greatly benefit from the ability to dynamically long-term track individual tumor cells23. Thus, to investigate the relative impact and dynamic interplay of the PVN and TM network niches, we used the above mentioned mouse xenograft models that closely reflect the growth pattern of malignant gliomas in humans, combined with intravital two photon-microscopy that allows to follow tumor microregions over long periods (up to months) in the three-dimensional space of the brain in living mice15,17,18,21. With this methodology we first analyzed general characteristics of tumor cells in the different compartments before investigating their responses to therapies, and finally molecular drivers of niche formation.
Long-term quiescence in the perivascular niche
Long-term tracking of glioma cells in invasive tumor regions (Fig. 2a) allowed to identify a subset of tumor cells that did not change their position over days to weeks, although being located at the invasive front. A significant majority of those long-term quiescent cells were found in a strict PVN position (Fig. 2b), closely following the course of blood vessels by stretching out to an elongated shape. A fluorescent reporter for tumor cell nuclei allowed to follow mitotic events in those cells. Here, the adjacent daughter cells remained closely associated with the respective blood vessel after cell division (Fig. 2c). Likewise, in the tumor core which is characterized by a higher cellular density and dense TM-connected tumor cell networks (Fig. 2a), the majority of PVN cells remained resident over weeks, whereas many parenchymal cells were still migrating (Fig. 2d). Histological analyses revealed that perivascular cell clusters were located between astrocytic endfeet and endothelial cells of brain microvessels (Fig. 2e), thus integrating into structures of the blood-brain barrier. Although the tumor cell soma remained at the very same spot over weeks, these perivascular cells actively extended and retracted TMs, scanning the environment and transiently or permanently connecting with neighboring cells (Fig. 2f). Some resident cells were closely associated with remodeled vessels such as capillary loops or glomeruloid‐like bodies, with some cells ultimately migrating away and others remaining resident (Fig. 2g). The majority colonized morphologically unaltered vessels though.
As stem-like, potentially slow-cycling tumor cells are believed to populate the PVN in glioma9, we next wanted to elucidate if cells in the different compartments vary in their proliferative activity. Histological analyses revealed that the proliferation rate of perivascular cells was lower compared to parenchymal cells in our xenograft models (Fig. 3a, b). In vivo EdU incorporation confirmed that most tumor cells in S-phase are located in the parenchyma, whereas perivascular cells are slow-cycling (Fig. 3c, d). In addition, intravital microscopy was used to analyze cell and vessel associations in larger 3D volumes, and demonstrated that the normalized mitotic rate of perivascular cells was indeed lower compared to the parenchymal subpopulation (Fig. 3e). Likewise, in molecularly defined human glioma specimen, the majority of proliferating, ki67-positive cells were located in the parenchymal compartment (Fig. 3f, g). The comparison of the distribution of tumor cells in the different compartments (Fig. 1a) and the proliferating fraction (Fig. 3f) suggests that perivascular tumor cells less actively divide in IDH1 mutant astrocytoma (p = 0.0002), IDH1 mutant high grade astrocytoma (p = 0.0005), and IDH wild-type glioblastoma (p = 0.0064). In contrast, in oligodendroglioma, the distribution of proliferating cells matched the general distribution of tumor cells in the different compartments (p = 0.2339, two-tailed t-tests), arguing against a specific quiescence of cells in the PVN in this very glioma entity. In summary, perivascular cells constitute a major fraction of long-term resident astrocytoma and glioblastoma cells, both at the invasive front and in the established tumor core, and are characterized by a quiescent state.
PVN location, tumor cell networks, and glioma resilience
In neurogenic niches and brain tumors alike, the PVN provides a specific microenvironment for stem-like cells, which has been linked to cellular resilience6,7,12,24. Furthermore, our group recently characterized the integration of glioma cells into TM-connected multicellular networks as a mechanism of resistance to cytotoxic therapies and prerequisite for tumor self-repair15,18. This leads to the question of the interrelation and relative relevance of both key factors of glioma progression and tumor cell persistence. More cells in the parenchymal compartment extended two TMs (Fig. 4a), a phenotype associated with fast invasion17. Moreover, a significant fraction of cells without any TMs was only found in the perivascular compartment (Fig. 4a). The degree of integration into multicellular networks did not appear to be relevantly biologically different between parenchymal and perivascular tumor cells though (Fig. 4b).
After application of radiotherapy, long-term in vivo microscopy revealed that PVN cells are highly radioresistant, whereas the number of parenchymal cells was significantly reduced by therapeutic irradiation as early as 7 days later (Fig. 4c, d). To rule out that the higher resistance of PVN cells was due to their sole integration into the TM-connected tumor network, a subgroup analysis was performed (Fig. 4e). Interestingly, even non-connected cells in the PVN were largely radioresistant, in contrast to the non-connected cells in the parenchyma which significantly regressed after radiotherapy. As expected15, TM-connected cells resisted the cytotoxic effects of radiotherapy, irrespective of their compartment (Fig. 4e).
Likewise, after temozolomide (TMZ) chemotherapy, a higher resistance of PVN glioblastoma cells was confirmed in the S24 GBMSC line, and again this resistance could not be explained by differential tumor cell connectivity, at least not solely (Fig. 4f). In T269 gliomas derived from another MGMT promoter hypermethylated GBMSC line which is more sensitive to chemotherapy than S24 and also more quickly proliferating in vivo18, perivascular glioblastoma cells also resisted TMZ better than parenchymal ones (Fig. 4g).
After surgery and also after targeted ablation of one single tumor cell within the tumor cell network, a repair response with directed extension of newly formed TMs towards the lesion, replacement of network-integrated glioblastoma cells, and repopulation of the damaged tumor region has been described18,25. We hence investigated whether perivascular and parenchymal tumor cells differed in their reactivity to a direct traumatic event. After inducing a microtrauma by ablation of one single glioblastoma cell with a high-power laser beam, an early reaction of a perivascular cell in this tumor microregion very frequently occurred within two hours, while parenchymal cells showed less reactivity (Fig. 5a). Of note, this early reaction was independent of the TM network connectivity of the ablated cell (p = 0.119, Fisher’s exact test), thereby suggesting a paracrine rather than a network-mediated signal. Furthermore, after milder laser damage to a larger tumor area (Fig. 5b) as well as after a surgical lesion (Fig. 5c), a marked increase in tumor cell count in the damaged area was reproducibly inducible. This “malignant healing response” was strikingly accentuated in the PVN (Fig. 5b–d) and preceded the dense repopulation of the lesioned brain area (arrowheads in Fig. 5c–e).
In summary, the perivascular localization of glioblastoma cells was strongly associated with their therapy resistance and resilience, additive to TM connectivity. Despite their static and often dormant behavior under resting conditions, perivascular cells can exhibit a strong reactivity to damage and initiate a dense repopulation of lesioned brain tumor regions.
PVN colonization and TM networks are inversely regulated by NOTCH1
Next we asked the question how both niches of resistance are molecularly interrelated: can glioblastoma cells dynamically change their position between the PVN and TM network niche, cross-compensating a molecular deficiency that deprives them of one or the other? For this we first inhibited network function by knockdown of the gap junction protein Cx43, which is a known connector of TM networks15. Here, morphological tumor cell networks greatly decreased, and tumor cells were driven into a perivascular position with high numbers of perivascular tumor cell clusters (Fig. 6a).
This led to the question whether a defect in PVN colonization in turn drives TM network integration of glioblastoma cells. Several publications suggested that activation of the NOTCH1 pathway is associated with glioma cell properties in the PVN, including cellular stemness7,26,27. Interestingly, the NOTCH1 pathway also plays an important role in neurite outgrowth28, which shares many features with TM formation15,16,17,29. Hence, we sought to elucidate its relevance for both the PVN and TM network niche. Immunohistological analyses demonstrated a NOTCH1 pathway activation in perivascular tumor cells (Fig. 6b), confirming previous results7.
We next used shRNA mediated knockdown of NOTCH1 to investigate the effects on brain and niche colonization as well as tumor cell morphology in vivo. Downregulation of NOTCH1 expression led to decreased vascular co-option (Fig. 6c) and a reduction of the perivascular tumor cell population (Fig. 6d). On the other hand, TM formation was markedly induced, with longer TMs (Fig. 6e), an early shift to highly interconnected cellular subpopulation with more than 4 TMs (Fig. 6f), and the formation of dense multicellular networks (Fig. 6g, h).
Brain tumor cells integrated into multicellular networks and those remaining non-connected in vivo can be separated based on their differential uptake of the fluorescent dye SR10115,21. We separated both cell populations by FACS and performed RNA sequencing of bulk populations. Analyses of the differential expression in two primary glioblastoma cell lines (S24 and T269) confirmed downregulation of the NOTCH1 receptor and of several genes associated with NOTCH1 pathway activation in the connected tumor cell population, and also upregulation of genes associated with NOTCH1 pathway inhibition (Table 1).
In contrast to glioblastomas, 1p/19q-codeleted oligodendrogliomas are characterized by TM network deficiency15 (Fig. 7a). Compared to the TM- and network proficient S24 glioblastoma cells, BT088 oligodendroglioma cells show a higher NOTCH1 expression (Fig. 7a), in line with the suggested role of NOTCH1 for TM network biology. Likewise, by analyzing TCGA gene expression data of molecular glioblastomas vs. oligodendrogliomas15, we found a relative downregulation of the NOTCH1 gene and several downstream targets of the NOTCH1 pathway in glioblastoma, as well as an upregulation of genes associated with NOTCH1 pathway inhibition (Table 2). When compared to astrocytoma and glioblastoma cells, oligodendroglioma cells showed a comparable perivascular niche colonization in the xenograft model (Fig. 7b) and in human oligodendroglioma specimen (Fig. 1a). Finally, after shRNA-mediated knockdown of NOTCH1, BT088 oligodendroglioma cells were not able to give rise to tumors in immunodeficient mice as cells slowly regressed over time (Fig. 7c). All in all, this data further support the importance of NOTCH1 as a regulator of TM network connectivity, but also suggests NOTCH1 as a principal regulator of oligodendroglioma growth, without specifically affecting PVN biology in this glioma subgroup.
NOTCH1-dependent niche positions regulate growth and resistance in gliomas
To understand how these differential changes of tumor cell niche integrations after NOTCH1 downregulation influence the overall sensitivity of 1p/19q intact gliomas to cytotoxic therapy, we analyzed tumor cell responses to radiotherapy, and found it decreased for the total shNOTCH1 glioblastoma cell population (Fig. 8a). Deeper analysis revealed that this can exclusively be attributed to the increased radioresistance of the more abundantly TM-interconnected shNOTCH1 cells, whereas the fewer non-TM connected cells were even more vulnerable to radiotherapy in shNOTCH1 tumors (Fig. 8b). Further subgroup analysis revealed that this effect was due to an increased therapy response of the remaining non-connected perivascular cells (Fig. 8c), suggesting that NOTCH1 proficiency is relevant for the resistance of perivascular cells to radiotherapy. In vitro, NOTCH1 downregulation inhibited proliferation of glioblastoma cells (Fig. 8d); in vivo, tumor bearing mice showed an improved survival when NOTCH1 was downregulated (Fig. 8e). High-field MRI confirmed the brain tumor growth attenuation after NOTCH1 knockdown (Fig. 8f, g). In contrast, after irradiation, the small and growth-deficient brain tumors were highly therapy resistant, with no regression being detectable compared to sham treatment (Fig. 8f, g). In summary, NOTCH1 proficiency plays an important role for the colonization of the PVN and tumor growth, whereas its deficiency results in impaired growth but also induces TM- and network formation that makes the tumor more radioresistant (Fig. 8h).
The dynamics, fate, and relevance of individual glioma cells in the perivascular niche was revealed in this study over weeks and months. In line with previous work investigating tumor cell dynamics in different tumor areas30, tumor cell invasion was not restricted to the invasive front but also prevalent in the dense tumor core. Here, we show that these dynamics are characteristic for parenchymal cells, whereas a majority of perivascular cells shows residency, even at the invasive front. The go-or-grow hypothesis suggests that glioma cells either migrate or proliferate31. Our data suggest that the invading parenchymal cells constitute the proliferating subpopulation, which is in line with recent single cell analyses32, arguing for a go-and-grow (and not go-or-grow) mechanism that should be further investigated in future studies. A minority of the resident perivascular cells was closely associated with remodeled blood vessels forming capillary loops or glomeroloid bodies33, further suggesting reciprocity in the perivascular niche8,10. Previous studies demonstrated that glomeroloid bodies are populated by stem-like tumor cells in human tumor specimen34 and this microvascular pattern is associated with a worse outcome in patients35,36.
During neurogenesis, the PVN provides a specific microenvironment for neural stem cells37. A similar PVN for cancer stem-like cells has been proposed for primary brain tumors6,7,10,12 and these stem-like cells have been associated with therapy resistance and tumor recurrence2,6,7,38. Recently, a metabolic zonation and consequent cellular heterogeneity dependent on the distance from blood vessels have been demonstrated in gliomas and perivascular cells exhibited a more aggressive phenotype and pronounced therapy resistance39. Our data confirm that tumor cells in the perivascular region are highly resistant against cytotoxic therapies, largely independent of the integration into multicellular networks. If this resistance is due to the slower proliferation rate, metabolic differences and/or specific microenvironmental cues, such as NOTCH1 signaling discussed below, remains to be elucidated. Here, despite its residency, perivascular glioma cells demonstrated reactiveness to nearby damage including surgical lesions by inducing an early gliotic reaction and a subsequent (re-)population of the lesioned brain area, thereby sharing traits with neural stem cells40 and reactive astrocytes, which also possess stem cell features41.
Endothelial cells are known to activate NOTCH1 signaling in adjacent glioma cells, including glioma stem-like cells4,7,8,26. NOTCH1 inhibitors have been tried in various preclinical studies, partially demonstrating effects on stem-like cells, although conflicting data exists and evidence for efficacy in clinical trials is still lacking42,43. Making use of the shRNA mediated knockdown of NOTCH1 we provide one possible explanation for those conflicting results: Although downregulation of NOTCH1 led to a depletion of the perivascular cell population, reduced tumor growth and reduced therapy resistance of cells in the PVN in line with previous reports7,26,44,45, it also induced TM- and network formation, thereby rendering the growth-deficient tumors virtually unsusceptible for the cytotoxic effects of radiotherapy. NOTCH1 proficiency is hence a key mechanism for the resilience of cells in the PVN, firstly by allowing population of that niche and secondly by being prerequisite for the protective properties of this specific microenvironment. In line with our results outlined here, many publications demonstrate a growth-inhibitory effect of NOTCH1 downregulation42,46,47,48,49,50. The differences regarding response to cytotoxic therapies46,47,48,49,50 might be explained by the different models used and their TM-proficient phenotype. Our data suggest that NOTCH1 is an important mediator of the resistance of perivascular tumor cells and the resistance promoting effects of NOTCH1 downregulation can be attributed to the induction of TMs. The latter might not be reflected in all models: especially cultivation under serum-containing conditions, which were used for many studies on NOTCH1 in glioblastoma49, leads to TM-deficient tumors17.
The formation of TMs shares many features with axon and neurite outgrowth, which is underlined by the phenotype observed after NOTCH1 knockdown. During neurodevelopment, NOTCH1 activation inhibits neurite sprouting28,51, whereas NOTCH1 inhibition has been shown to promote neurite extension52. NOTCH1 deficiency is only the third known molecular driver of TMs. Despite the caveats in targeting this pathway, this can provide insights into the molecular biology of multicellular networks and how to target them16,53,54.
Our data suggests that NOTCH1 inhibition might be used to slow down tumor growth but should be avoided in combination with cytotoxic therapies due to the induction of tumor cell networks. In TM-deficient oligodendrogliomas, we found an upregulation of the NOTCH1 pathway, both in the BT088 cell line and human samples, which is in line with the principal regulation of TM networks by NOTCH1. However, our analyzes also show fundamental biological differences between 1p/19q-codel oligodendrogliomas and 1p/19q-intact other glioma types: in contrast to astrocytomas and glioblastomas, the NOTCH1 pathway activation in oligodendroglioma seems not to be strongly related to PVN position and cells in the PVN were not slow cycling.
Thus, our data imply that at least two, partially additive and complementary, niches of cellular resilience exist in glioblastoma: a PVN and a TM connectivity niche, that inversely depend on NOTCH1 expression. Phenotypic adaptation55 or switches between both niches must be considered for targeted therapies.
In conclusion, the perivascular compartment accommodates the majority of long-term resident glioma cells and is associated with resistance against cytotoxic therapies. Downregulation of the NOTCH1 pathway depletes this cell pool and diminishes the protective properties of the perivascular niche, but concomitantly induces TM- and network formation with subsequent therapy resistance. All in all, this describes two niches of progression and resistance in gliomas—the TM connectivity niche and the PVN—and speaks for a complex and partially complementary role of these two. To effectively treat incurable astrocytomas and glioblastomas, novel concepts to target both niches seem of high importance, ideally avoiding cross-compensatory tumor cell reactions.
The patient-derived glioblastoma cell lines S24, T269, T325 and P3xx and BT088 oligodendroglioma cells were kept under stem-like neurosphere conditions (37 °C, 5.0% CO2): DMEM F-12 medium (31330-038, Invitrogen), B27 supplement (12587-010, Invitrogen), 5 μg ml−1 insulin (I9278, Sigma-Aldrich), 5 μg ml−1 heparin (H4784, Sigma-Aldrich), 20 ng ml−1 epidermal growth factor (rhEGF; 236-EG, R&D Systems), and 20 ng ml−1 basic fibroblast growth factor (bFGF; PHG0021, Thermo Fisher Scientific). S24 (human, female), T269 (human, male), T325 (human, male), and P3xx (human) were authentificated (Multiplexion GmbH, Germany). S24, T269, and T325 were further authenticated as glioblastoma by Illumina 850k methylation array56. BT088 cells were obtained from ATCC (ATCC CRL-3417, RRID:CVCL_N708) (human, male). The molecular characterization of all used cancer cell lines can be found in Supplementary Tables S1 and S2.
For transduction, cells were incubated with lentiviral particles and 10 μg ml−1 polybrene (TR-1003-G, Merck Millipore) for 24 h. Cell lines were stably transduced with the LeGO-T2 vector (cytoplasmatic tdTomato, Addgene 27342, RRID:Addgene_27342), pLKO.1-LV-GFP (H2B-GFP, Addgene #25999, RRID:Addgene_25999), pLenti6.2 hygro/V5-Lifeact-YFP, plKO.1-puro-CMV-TurboGFP_shnon-target (cytoplasmatic GFP, SHC016, Sigma-Aldrich), plKO.1-puro-CMV-TurboGFP_shCX43 (CX43 knockdown, sequence: GCCCAAACTGATGGTGTCAA, Sigma-Aldrich), and plKO.1-puro-CMV-TurboGFP_shNOTCH1 (NOTCH1 knockdown, sequence: CCGGGACATCACGGATCATAT, Sigma-Aldrich). NOTCH1 and CX43 knockdown were confirmed by Western blot analysis (45% (S24 GBMSCs); 92% (BT088) (NOTCH1) / 80% (S24 GBMSCs) (CX43) reduction in protein expression). Anti-Notch1 (1:1000; #4380, Cell Signaling Technology, RRID:AB_10691684) and anti-GAPDH antibody (1:5000; LAH1064, Linaris) were used for Western blot analyses.
8-to-10-week-old male NMRI nude mice (Charles River) were used for studying the intracranial tumor growth and dynamics. Mice were kept in individual housing after cranial window implantation (constant housing conditions: temperature 22± 2 °C, humidity 55 ± 10%, 12 h light/dark cycles). All animal procedures were performed in accordance with the institutional laboratory animal research guidelines after approval of the responsible animal welfare officer (German Cancer Research Center, Heidelberg, Germany) and the regional council (Referat 35, Regierungspräsidium Karlsruhe, Germany). A chronic cranial window and a titan ring that allows pain-free fixation of the animals during two photon microscopy were implanted as described before15,17. Two weeks after window implantation 30,000 patient-derived and fluorescently labeled glioblastoma stem-like cells suspended in PBS were intracranially implanted in a depth of 500 µm. For survival and MRI studies, 50,000 cells (S24 shControl and shNotch1) were implanted without preparation of a chronic cranial window. T2-weighted rapid acquisition with refocused echoes sequence MRI images were acquired on a 9.4 T horizontal bore MR scanner (BioSpec 94/20 USR, Bruker BioSpin) with a four-channel phased-array surface coil (parameters: TE = 33 ms, TR = 2500 ms, flip angle = 90°, acquisition matrix: 200 × 150, number of averages = 2, slice thickness = 700 μm duration = 2 min 53 s). For radiotherapy experiments, whole brain irradiation with fractions of 7 Gy at a dose rate of 3 Gy min−1 were performed on 3 consecutive days at D60 ( ± 10 days) using a 6 MV linear accelerator (Artiste, Siemens) with a 6 mm collimator adjusted to the window size. As control no radiation was applied (sham radiation). For chemotherapy experiments, mice were treated with 100 mg/kg body weight temozolomide for 3 consecutive days on D85 ± 3 after tumor implantation by oral gavage. For the surgical lesion experiment, an established tumor region was chosen by two photon microscopy, the cranial window was removed and a 26-gauge Hamilton syringe was used to resect a cylindrical volume within the established tumor region as described before18. The perilesional tumor regions were followed repetitively by two photon microscopy afterwards. For two photon microscopy and MRI imaging mice were anesthetized with isoflurane. Mice were scored clinically and rapidly killed if they showed neurological symptoms or a weight loss of >20%.
In vivo multiphoton laser scanning microscopy
Intravital two photon microscopy (2-PM) was performed with a Zeiss 7MP microscope (Zeiss) equipped with a Coherent Chameleon UltraII laser (Coherent) on anesthetized mice (4% isoflurane for initiation, 0.5–2% for maintenance of anesthesia). A custom-made aperture allowed painless fixation of the head for imaging. Body temperature was permanently controlled by a rectal thermometer and was kept constant by a heating pad. Angiograms allowed identification of the same tumor regions over time. For angiograms, fluorescent dextranes (10 mg ml−1, TRITC-dextrane (500 kDa; 52194, Sigma Aldrich) or FITC-dextrane (2000 kDa, FD2000S, Sigma Aldrich)) were applied by tail vein injection. The following wavelengths were used for excitation of different fluorophores: 750 nm (FITC-dextrane, tdTomato), 850 nm (GFP, TRITC-dextrane), and 950 nm (tdTomato, YFP). Appropriate filter sets (band pass 500–550 nm/band pass 575–610 nm) were used. For ablation of single glioma cells, the laser beam was focused on the GFP-labeled nucleus (H2B-GFP) and continuous scanning was performed until disintegration of the cell. For damage of a larger tumor area, repetitive scanning (950 nm wavelength) for ~8 min with high laser power was performed. For the evaluation of therapeutic responses, the same tumor regions were imaged repetitively over 1–3 weeks. For invasion speed measurements and identification of invading tumor cells, single glioma cells were followed with repetitive imaging over 24 h. Short-term residency of individual glioma cells was determined by repetitive imaging over 4 (S24) and 5 days (P3xx), long-term residency by repetitive imaging between D41 and D62.
SR101 staining and separation of connected and non-connected tumor cells
For separation of connected and non-connected tumor cells, mice bearing S24 and T269 GFP tumors were injected with sterilized saline solution dissolving SR101 (S359, Invitrogen) i.p. at a dose of 0.12 mg per gram b57ody weight, which is taken up by tumor cells within 5–8 h. Mice were then perfused with PBS and the brains were harvested. Whole brain single cell suspensions were prepared with brain tumor dissociation kit (130-095-942, Miltenyi Biotec) and gentleMACSTM Dissociator (Miltenyi Biotec). After dissociation the suspension was resuspended in FACS buffer (PBS + 1% FCS, 10500064, ThermoFisher). FACS was performed on a FACSAria cell sorter (BD Biosystems). Cells were stained with Calcein Violet 450 AM (65-0854-39, Invitrogen) and TO-PRO®-3 Iodide (T3605, Invitrogen) for 10 min on ice before sorting for the viable cell population (Calcein Violet 450high and TO-PRO-3neg). Within the viable cell population SR101high, GFPhigh (connected tumor cells), and SR101low, GFPlow (non-connected tumor cells) were separated. The YG586/15 channel was used to visualize SR101 signal. FACS was performed as part of another study57. The exemplary visualization of the gating strategy provided in Supplementary Fig. S1b was adopted from this study57.
Bulk RNA sequencing of connected and non-connected tumor cells
After cell sorting, cell populations were lysed (lysis buffer, RNeasy Micro Kit, 74004, Qiagen). mRNA was isolated and purified in accordance with the manufacturer’s instruction. RNA was pooled from 3 mice for each sample. The conversion of RNA to dsDNA was done with the SMARTer Ultra Low Input RNA for Illumina Sequencing (Clontech), the libraries were then prepared using NEBNext® ChIP-Seq Library Prep Master Mix Set for Illumina (E6240, New England Biolabs) and sequenced on HiSeq2000 v4 (Illumina) in 50 bp single-end mode by our core facility. The quality of bases was evaluated using the FASTX Toolkit. Homertools 4.7 were applied for PolyA-tail trimming58; reads with a length of <17 were removed. The filtered reads were mapped with STAR 2.359 against the human reference genome (GRCh38) and PicardTools 1.78 with CollectRNASeqMetrics were used for quality checking. Count data were generated by htseq-count using the gencode.v26.annotation.gtf file for annotation60. DESeq2 1.4.1 was run with default parameters for the group-wise comparison61. The expression levels were transformed to logarithmic space using log2.
Analysis of differential mRNA expression of human gliomas
RNA sequencing data of molecular (1p/19q-codeleted, IDH mutant) oligodendroglioma (n = 56) and molecular (1p/19q non-codeleted, IDH wild-type) glioblastoma (n = 70) was downloaded from TCGA and analyzed as described previously15. The full list of differentially expressed genes can be found in Osswald et al. (2015)15.
In vivo EdU incorporation
Animals were treated with 50 mg/kg EdU i.p. (BCK488-IV-IM-S, baseclick GmbH) 4 h prior to euthanasia. Animals were cardially perfused with 4.5% phosphate-buffered formaldehyde solution (Roti-Histofix 4.5%, 2213, Carl Roth). Tissue was frozen after cryoprotection with 30% sucrose solution and 10 µm cryosections were cut. The sections were permeabilized with 0.5% Triton X-100 in PBS for 20 min. The EdU Click Kit (BCK488-IV-IM-S, baseclick GmbH) was used for EdU detection. After washing, sections were incubated with the reaction cocktail for 30 min as per the manufacturer’s instructions. Afterwards sections were incubated with anti-nestin (1:400, ab6320, Abcam, RRID:AB_308832) and anti-CD31 (1:50; PA5-16301, ThermoFisher Scientific, RRID:AB_10981955) antibodies.Goat anti-rabbit Alexa Fluor 594 (1:400; A-11037, ThermoFisher Scientific, RRID:AB_2534095) and goat anti-mouse Alexa Fluor 633 (1:400; A-21052, ThermoFisher Scientific, RRID:AB_2535719) secondary antibodies were used. Images were acquired using a Leica TCS SP5 confocal microscope (LAS software version 126.96.36.19923).
Immunofluorescence and confocal microscopy
For immunofluorescence of mouse tissue, tumor bearing mice were cardially perfused with PBS, followed by 4.5% phosphate-buffered formaldehyde solution (Roti-Histofix 4.5%, 2213, Carl Roth). Brain tissue was incubated in 30% sucrose solution overnight for cryoprotection and snap frozen afterwards. Heat-induced epitope retrieval with 0.01 M citrate buffer, pH 6.0, was performed. The following primary antibodies were used: anti-nestin (1:400, ab6320, Abcam, RRID:AB_308832), anti-CD31 (1:100, AF3628, R&D Systems, RRID:AB_2161028), anti-aquaporin 4 (1:200, ab9512, Abcam, RRID:AB_307299), anti-activated Notch1 (1:50, ab8925, Abcam, RRID:AB_306863) and anti-ki67 (1:500, ab15580, Abcam, RRID:AB_443209). Donkey anti-mouse IgG Alexa Fluor 488 (1:400; A-21202, Thermo Fisher Scientific, RRID:AB_141607), goat anti-mouse IgG Alexa Fluor 488 conjugate (1:400; A-11029, Thermo Fisher Scientific, RRID:AB_138404), donkey anti-goat IgG Alexa Fluor 633 (1:400; A-21082, Thermo Fisher Scientific, RRID:AB_141493) and donkey anti-rabbit IgG Alexa Fluor 546 (1:400; A-10040, Thermo Fisher Scientific, RRID:AB_2534016) secondary antibodies were used. Images were acquired using a Leica TCS SP5 confocal microscope.
Immunohistochemical and immunofluorescence staining of human tumor specimen
Tissue sections of resected primary gliomas were obtained from the Department of Neuropathology in Heidelberg in accordance with local ethical approval (Ethikkommission der Universität Heidelberg, Heidelberg University, Heidelberg, Germany). Samples were anonymized prior to analyzes and all patients gave their informed consent for the scientific use of resected tissue samples. For immunohistochemical nestin staining samples from ten patients from the Heidelberg Neuro-Oncology center diagnosed with glioblastoma, IDH wild-type based on histopathological and molecular characteristics were used. Inclusion of patients into this analysis is covered by a local Heidelberg ethics vote (No. S307/2019). 3 µm sections were incubated with the anti-nestin antibody, clone 10C2 (1:200; MAB5326, Merck Millipore, RRID:AB_11211837). Nestin expression was detected using the ultraView DAB protocol on the automated VENTANA BenchMark ULTRA platform (Roche, Switzerland). Ventana standard signal amplification and ultra-wash was followed by counterstaining with Hematoxylin II (790-2208, Roche, Switzerland) and blueing reagent (760-2037, Roche, Switzerland).
Immunohistochemistry of IDH1 R132H and ki67 was performed on 0.5-μm-thick formalin-fixed, paraffin-embedded (FFPE) tissue sections. For IDH1 R132H immunohistochemistry, sections were stained on a BenchMark XT immunostainer, for ki67, staining was performed on a BenchMark Ultra immunostainer (Ventana Medical Systems, Tucson, AZ, USA). Anti-IDH1 R132H (1:20; DIA-H09, Dianova, RRID:AB_2335716) and anti-Ki67 (1:100; M7240, Dako, RRID:AB_2142367) antibodies were used respectively. Slides were scanned on an Aperio AT2 Slide Scanner (Biosystems Switzerland AG, Switzerland).
For immunofluorescence staining of nestin and CD31 in human glioblastoma specimen, 100 µm thick, PFA-fixed sections were used. Heat-induced antigen retrieval was performed with Tris-EDTA buffer. Sections were incubated with anti-nestin (1:400; ab6320, Abcam, RRID:AB_308832) and anti-CD31 (1:50; PA5-16301, ThermoFisher Scientific, RRID:AB_10981955) antibodies for 24 h. Goat anti-rabbit Alexa Fluor 594 (1:400; A-11037, ThermoFisher Scientific, RRID:AB_2534095) and goat anti-mouse Alexa Fluor 633 (1:400; A-21052, ThermoFisher Scientific, RRID:AB_2535719) secondary antibodies were used. Auto-fluorescence was quenched with Sudan Black.
Tissue selection and molecular characterization of human tumor specimen
Tumor tissues were selected based on their methylation data from the database of the Department of Neuropathology Heidelberg62. All methylation data were generated using the Illumina MethylationEPIC (850k) array platform according to the manufacturer’s instructions (Illumina, San Diego, USA). Sample preparation was performed as previously described63. DNA methylation status of 10.000 CpG sites was analyzed on the current version v12b4 of the Classifier (https://www.molecularneuropathology.org/mnp). Of note, all cases presented a calibrated Classifier Score > 0.9.
The probability of MGMT promoter methylation from 850k array data was estimated as previously described64. For cases with an MGMT promoter methylation status not determinable by 850k methylation analysis, additional pyrosequencing was performed using the therascreen® MGMT Pyro® kit (QIAGEN®) and the PyroMark® Q24 system (QIAGEN®) according to the manufacturer’s protocol. Bisulfite conversion was done with the EpiTect fast DNA bisulfite kit (QIAGEN®). In accordance with published studies65,66,67, a mean MGMT promoter methylation percentage < 8% across the investigated CpG sites was considered as non-methylated and a value ≥8% was considered as methylated.
The series used for immunohistochemical analyses consisted of tissues with the following molecular characteristics:
1. Oligodendroglioma: n = 18 (for IDH1 R132H staining), n = 16 (for ki67 staining). Methylation class IDH glioma, subclass 1p/19q codeleted oligodendroglioma, all ATRX retained.
2. Astrocytoma (IDH1 R132H mutant): n = 19 (for IDH1 R132H staining), n = 19 (for ki67 staining). Methylation class IDH glioma, subclass astrocytoma, all ATRX loss, all 1p/19q non-codel.
3. High grade astrocytoma (IDH1 R132H mutant): n = 20 (for IDH1 R132H staining), n = 15 (for ki67 staining). Methylation class IDH glioma, subclass high grade astrocytoma, all 1p/19q non-codel. a) IDH1 R132H staining: 16 ATRX loss, 3 ATRX retained, 1 undetermined. b) ki67 staining: 11 ATRX loss, 3 ATRX retained, 1 n/a.
4. Glioblastoma (IDH1 wild-type): n = 10 (for nestin staining), n = 16 (for ki67 staining). Methylation class glioblastoma, IDH wild-type. a) Nestin staining: subclass: RTK I = 1, RTK II = 7, mesenchymal=1, MYCN = 1. All 1p/19q non-codel, all MGMT promoter unmethylated, all TP53 wild-type, 9 ATRX retained, 1 n/a. b) ki67 staining: subclass: RTK I = 4, RTK II = 9, mesenchymal=3. All 1p/19q non-codel, 6 MGMT promoter methylated, 10 MGMT promoter unmethylated.
Alamar blue proliferation assay
1000 GBMSCs were seeded in 96-well plates and AlamarBlue Reagent (DAL1025, Thermo Fisher Scientific) was added on days 0, 1, 3, and 5. Fluorescence intensity (excitation: 560 nm; emission: 590 nm) was measured using a SpectraMax M5e microplate reader (Molecular Devices) after 3 h.
For image analysis and figures, images were processed using ZEN software (Zeiss, RRID:SCR_018163) and Imaris (Bitlane, RRID:SCR_007370). Image calculation and filtering was performed to reduce background noise. For figures, 3D stacks are transformed into orthogonal 2D maximum intensity projections (MIP).
All quantifications were performed manually in ImageJ (version 1.53, NIH, RRID:SCR_003070) and Imaris (Bitlane, RRID:SCR_007370) in high resolution (1024 * 1024 pixels) 3D z-stacks. Immunohistochemical and immunofluorescence images were analyzed in Qupath68, Aperio ImageScope software (v188.8.131.525, Aperio Technologies, USA) and ImageJ (version 1.53, NIH, RRID:SCR_003070). Cells were regarded perivascular if the whole cell body was directly associated with a blood vessel. For quantification of ki67 positive cells in human tumor specimen, cells were regarded perivascular if the cell was located within 10 µm from a blood vessel and was not separated by another cell body. TM length of individual cells was measured by tracking TMs in z-stacks (3D images). For the mitotic index, mitotic events were analyzed in H2B-GFP expressing cells. The mitotic events were normalized to the distribution of cells in the parenchymal and perivascular compartment in the respective region, thereby assuming equal fractions in both compartments. The early reaction after single ablation was defined as the directed extension of TMs towards the damaged area within 100 min. For the quantification of the tumor/brain area ratio on MRI images, the tumor and the whole-brain area were measured manually in the image with the largest tumor expansion.
Statistics and reproducibility
Statistical analyses were performed using SigmaPlot (version 14.0, Systat Software, RRID:SCR_003210) and Prism 8.4.1 (GraphPad, RRID:SCR_002798). Normal distribution of datasets was assessed with a Shapiro–Wilk test. Statistical significance of normally distributed data was assessed by a two-tailed Student’s t test. Non-normally distributed data were assessed with a two-tailed Mann–Whitney rank-sum test. For datasets with >2 groups ANOVA or ANOVA on ranks with the appropriate post hoc tests (Tukey’s or Student–Newman–Keuls tests) for multiple comparisons were performed. The p < 0.05 was considered statistically significant.
For ANOVA tests, p values are reported in the figure legend, multiplicity adjusted p values are noted in the figures and multiple test tables including confidence intervals can be found in the Source data file. No 95% confidence intervals for each difference or multiplicity adjusted exact p values can be reported in case the Newman–Keuls post hoc test was used for multiple comparisons, as this test works in sequential fashion. Statistical details including group sizes can be found in the respective figure legends. All experiments were repeated at least three times unless otherwise noted.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
The RNA seq. data of connected and non-connected glioblastoma cells have been deposited in the Sequence Read Archive (SRA) database under the accession number PRJNA554870. Uncropped Western blot data are provided in Source data and Supplementary Fig. S1a. Source data are provided with this paper.
Weller, M. et al. Glioma. Nat. Rev. Dis. Prim. 1, 15017 (2015).
Galli, R. et al. Isolation and characterization of tumorigenic, stem-like neural precursors from human glioblastoma. Cancer Res. 64, 7011–7021 (2004).
Ignatova, T. N. et al. Human cortical glial tumors contain neural stem-like cells expressing astroglial and neuronal markers in vitro. Glia 39, 193–206 (2002).
Lathia, J. D., Mack, S. C., Mulkearns-Hubert, E. E., Valentim, C. L. & Rich, J. N. Cancer stem cells in glioblastoma. Genes Dev. 29, 1203–1217 (2015).
Lathia, J. D. et al. Direct in vivo evidence for tumor propagation by glioblastoma cancer stem cells. PLoS One 6, e24807 (2011).
Calabrese, C. et al. A perivascular niche for brain tumor stem cells. Cancer Cell 11, 69–82 (2007).
Charles, N. et al. Perivascular nitric oxide activates notch signaling and promotes stem-like character in PDGF-induced glioma cells. Cell Stem Cell 6, 141–152 (2010).
Lathia, J. D., Heddleston, J. M., Venere, M. & Rich, J. N. Deadly teamwork: neural cancer stem cells and the tumor microenvironment. Cell Stem Cell 8, 482–485 (2011).
Hjelmeland, A. B., Lathia, J. D., Sathornsumetee, S. & Rich, J. N. Twisted tango: brain tumor neurovascular interactions. Nat. Neurosci. 14, 1375–1381 (2011).
Gilbertson, R. J. & Rich, J. N. Making a tumour’s bed: glioblastoma stem cells and the vascular niche. Nat. Rev. Cancer 7, 733–736 (2007).
Alvarez-Buylla, A., Kohwi, M., Nguyen, T. M. & Merkle, F. T. The heterogeneity of adult neural stem cells and the emerging complexity of their niche. Cold Spring Harb. Symp. Quant. Biol. 73, 357–365 (2008).
Charles, N. & Holland, E. C. The perivascular niche microenvironment in brain tumor progression. Cell Cycle 9, 3012–3021 (2010).
Filbin, M. & Monje, M. Developmental origins and emerging therapeutic opportunities for childhood cancer. Nat. Med. 25, 367–376 (2019).
Lee, J. H. et al. Human glioblastoma arises from subventricular zone cells with low-level driver mutations. Nature 560, 243–247 (2018).
Osswald, M. et al. Brain tumour cells interconnect to a functional and resistant network. Nature 528, 93–98 (2015).
Jung, E. et al. Emerging intersections between neuroscience and glioma biology. Nat. Neurosci. 22, 1951–1960 (2019).
Jung, E. et al. Tweety-homolog 1 drives brain colonization of gliomas. J. Neurosci. 37, 6837–6850 (2017).
Weil S. et al. Tumor microtubes convey resistance to surgical lesions and chemotherapy in gliomas. Neuro Oncol. (2017).
Sahm, F. et al. Addressing diffuse glioma as a systemic brain disease with single-cell analysis. Arch. Neurol. 69, 523–526 (2012).
Capper, D. et al. Characterization of R132H mutation-specific IDH1 antibody binding in brain tumors. Brain Pathol. 20, 245–254 (2010).
Venkataramani, V. et al. Glutamatergic synaptic input to glioma cells drives brain tumour progression. Nature 573, 532–538 (2019).
Venkatesh, H. S. et al. Electrical and synaptic integration of glioma into neural circuits. Nature 573, 539–545 (2019).
Koga, T. et al. Longitudinal assessment of tumor development using cancer avatars derived from genetically engineered pluripotent stem cells. Nat. Commun. 11, 550 (2020).
Chen, J. et al. A restricted cell population propagates glioblastoma growth after chemotherapy. Nature 488, 522–526 (2012).
Alieva, M. et al. Preventing inflammation inhibits biopsy-mediated changes in tumor cell behavior. Sci. Rep. 7, 7529 (2017).
Zhu, T. S. et al. Endothelial cells create a stem cell niche in glioblastoma by providing NOTCH ligands that nurture self-renewal of cancer stem-like cells. Cancer Res. 71, 6061–6072 (2011).
Bayin, N. S. et al. Notch signaling regulates metabolic heterogeneity in glioblastoma stem cells. Oncotarget 8, 64932–64953 (2017).
Sestan, N., Artavanis-Tsakonas, S. & Rakic, P. Contact-dependent inhibition of cortical neurite growth mediated by notch signaling. Science 286, 741–746 (1999).
Osswald, M., Solecki, G., Wick, W. & Winkler, F. A malignant cellular network in gliomas: potential clinical implications. Neuro Oncol. 18, 479–485 (2016).
Alieva, M. et al. Intravital imaging of glioma border morphology reveals distinctive cellular dynamics and contribution to tumor cell invasion. Sci. Rep. 9, 2054 (2019).
Giese, A., Bjerkvig, R., Berens, M. E. & Westphal, M. Cost of migration: invasion of malignant gliomas and implications for treatment. J. Clin. Oncol. 21, 1624–1636 (2003).
Wang, L. et al. The phenotypes of proliferating glioblastoma cells reside on a single axis of variation. Cancer Discov. 9, 1708–1719 (2019).
Rojiani, A. M. & Dorovini-Zis, K. Glomeruloid vascular structures in glioblastoma multiforme: an immunohistochemical and ultrastructural study. J. Neurosurg. 85, 1078–1084 (1996).
Chen, J. et al. The pathological structure of the perivascular niche in different microvascular patterns of glioblastoma. PLoS One 12, e0182183 (2017).
Chen, L. et al. Classification of microvascular patterns via cluster analysis reveals their prognostic significance in glioblastoma. Hum. Pathol. 46, 120–128 (2015).
Birner, P. et al. Vascular patterns in glioblastoma influence clinical outcome and associate with variable expression of angiogenic proteins: evidence for distinct angiogenic subtypes. Brain Pathol. 13, 133–143 (2003).
Bjornsson, C. S., Apostolopoulou, M., Tian, Y. & Temple, S. It takes a village: constructing the neurogenic niche. Dev. Cell 32, 435–446 (2015).
Singh, S. K. et al. Identification of a cancer stem cell in human brain tumors. Cancer Res. 63, 5821–5828 (2003).
Kumar, S. et al. Intra-tumoral metabolic zonation and resultant phenotypic diversification are dictated by blood vessel proximity. Cell Metab. 30, e206 (2019).
Arvidsson, A., Collin, T., Kirik, D., Kokaia, Z. & Lindvall, O. Neuronal replacement from endogenous precursors in the adult brain after stroke. Nat. Med. 8, 963–970 (2002).
Robel, S., Berninger, B. & Gotz, M. The stem cell potential of glia: lessons from reactive gliosis. Nat. Rev. Neurosci. 12, 88–104 (2011).
Bazzoni R. & Bentivegna A. Role of notch signaling pathway in glioblastoma pathogenesis. Cancers 11, (2019).
Xu, R. et al. Molecular and clinical effects of notch inhibition in glioma patients: a phase 0/I trial. Clin. Cancer Res. 22, 4786–4796 (2016).
Teodorczyk, M. & Schmidt, M. H. Notching on cancer’s door: notch signaling in brain tumors. Front. Oncol. 4, 341 (2014).
Parmigiani, E., Taylor, V. & Giachino, C. Oncogenic and tumor-suppressive functions of NOTCH signaling in glioma. Cells 9, 2304 (2020).
Han, N. et al. Notch1 ablation radiosensitizes glioblastoma cells. Oncotarget 8, 88059–88068 (2017).
Wang, J. et al. Notch promotes radioresistance of glioma stem cells. Stem Cells 28, 17–28 (2010).
Hai, L. et al. Notch1 is a prognostic factor that is distinctly activated in the classical and proneural subtype of glioblastoma and that promotes glioma cell survival via the NF-kappaB(p65) pathway. Cell Death Dis. 9, 158 (2018).
Gersey, Z. et al. Therapeutic targeting of the notch pathway in glioblastoma multiforme. World Neurosurg. 131, e252 (2019).
Yahyanejad, S. et al. NOTCH blockade combined with radiation therapy and temozolomide prolongs survival of orthotopic glioblastoma. Oncotarget 7, 41251–41264 (2016).
Berezovska, O. et al. Notch1 inhibits neurite outgrowth in postmitotic primary neurons. Neuroscience 93, 433–439 (1999).
Liu, M. et al. ASIC1 promotes differentiation of neuroblastoma by negatively regulating Notch signaling pathway. Oncotarget 8, 8283–8293 (2017).
Dolgin, E. Cancer-neuronal crosstalk and the startups working to silence it. Nat. Biotechnol. 38, 115–117 (2020).
Jung, E., Alfonso, J., Monyer, H., Wick, W. & Winkler, F. Neuronal signatures in cancer. Int. J. Cancer 147, 3281–3291 (2020).
Celiku, O., Gilbert, M. R. & Lavi, O. Computational modeling demonstrates that glioblastoma cells can survive spatial environmental challenges through exploratory adaptation. Nat. Commun. 10, 5704 (2019).
Kessler, T. et al. Methylome analyses of three glioblastoma cohorts reveal chemotherapy sensitivity markers within DDR genes. Cancer Med. 9, 8373–8385 (2020).
Xie, R. et al. Tumor cell network integration in glioma represents a stemness feature. Neuro Oncol. https://doi.org/10.1093/neuonc/noaa275 (2020). (In press).
Heinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589 (2010).
Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013).
Anders, S., Pyl, P. T. & Huber, W. HTSeq—a Python framework to work with high-throughput sequencing data. Bioinformatics 31, 166–169 (2015).
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Capper, D. et al. DNA methylation-based classification of central nervous system tumours. Nature 555, 469–474 (2018).
Capper, D. et al. Practical implementation of DNA methylation and copy-number-based CNS tumor diagnostics: the Heidelberg experience. Acta Neuropathol. 136, 181–210 (2018).
Bady, P. et al. MGMT methylation analysis of glioblastoma on the Infinium methylation BeadChip identifies two distinct CpG regions associated with gene silencing and outcome, yielding a prediction model for comparisons across datasets, tumor grades, and CIMP-status. Acta Neuropathol. 124, 547–560 (2012).
Felsberg, J. et al. Promoter methylation and expression of MGMT and the DNA mismatch repair genes MLH1, MSH2, MSH6 and PMS2 in paired primary and recurrent glioblastomas. Int. J. Cancer 129, 659–670 (2011).
Reifenberger, G. et al. Predictive impact of MGMT promoter methylation in glioblastoma of the elderly. Int. J. Cancer 131, 1342–1350 (2012).
Quillien, V. et al. Comparative assessment of 5 methods (methylation-specific polymerase chain reaction, MethyLight, pyrosequencing, methylation-sensitive high-resolution melting, and immunohistochemistry) to analyze O6-methylguanine-DNA-methyltranferase in a series of 100 glioblastoma patients. Cancer 118, 4201–4211 (2012).
Bankhead, P. et al. QuPath: Open source software for digital pathology image analysis. Sci. Rep. 7, 16878 (2017).
We thank Hrvoje Miletic for providing the P3 cell line. We thank Miriam Gömmel, Cathrin Löb, Chanté Mayer and Susann Wendler for technical assistance. We thank Matthia Karreman for critical reading of the manuscript. This study was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 404521405, SFB 1389 – UNITE Glioblastoma, Work Packages A01 (E.J., F.W.), A03 (W.W., T.K., D.C.H.), A06 (F.S.), and D01 (M.R., A.v.D.), as well as DFG, KU 3555/1-1 (F.K.). F.S. is fellow of the Else Kröner Excellence Program of the Else Kröner-Fresenius Stiftung (EKFS).
Open Access funding enabled and organized by Projekt DEAL.
E.J., M.O., W.W. and F.W. report the patent (WO2017020982A1) “Agents for use in the treatment of glioma”. F.W. is co-founder of DC Europa Ltd (a company trading under the name Divide & Conquer) that is developing new medicines for the treatment of glioma. Divide & Conquer also provides research funding to F.W.’s lab under a research collaboration agreement. All other authors declare no competing interests.
Peer review information Nature Communications thanks Jeffrey Segall and Maria Castro for their contribution to the peer review of this work. Peer reviewer reports are available.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Jung, E., Osswald, M., Ratliff, M. et al. Tumor cell plasticity, heterogeneity, and resistance in crucial microenvironmental niches in glioma. Nat Commun 12, 1014 (2021). https://doi.org/10.1038/s41467-021-21117-3
This article is cited by
Insights and opportunities at the crossroads of cancer and neuroscience
Nature Cell Biology (2022)
Disconnecting multicellular networks in brain tumours
Nature Reviews Cancer (2022)
Mass spectrometry imaging discriminates glioblastoma tumor cell subpopulations and different microvascular formations based on their lipid profiles
Scientific Reports (2022)
Angiocrine extracellular vesicles impose mesenchymal reprogramming upon proneural glioma stem cells
Nature Communications (2022)
Brain Tumor Networks in Diffuse Glioma
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.