A network of communicating tumour cells that is connected by tumour microtubes mediates the progression of incurable gliomas. Moreover, neuronal activity can foster malignant behaviour of glioma cells by non-synaptic paracrine and autocrine mechanisms. Here we report a direct communication channel between neurons and glioma cells in different disease models and human tumours: functional bona fide chemical synapses between presynaptic neurons and postsynaptic glioma cells. These neurogliomal synapses show a typical synaptic ultrastructure, are located on tumour microtubes, and produce postsynaptic currents that are mediated by glutamate receptors of the AMPA subtype. Neuronal activity including epileptic conditions generates synchronised calcium transients in tumour-microtube-connected glioma networks. Glioma-cell-specific genetic perturbation of AMPA receptors reduces calcium-related invasiveness of tumour-microtube-positive tumour cells and glioma growth. Invasion and growth are also reduced by anaesthesia and the AMPA receptor antagonist perampanel, respectively. These findings reveal a biologically relevant direct synaptic communication between neurons and glioma cells with potential clinical implications.
This is a preview of subscription content, access via your institution
Open Access articles citing this article.
Advanced immunotherapies for glioblastoma: tumor neoantigen vaccines in combination with immunomodulators
Acta Neuropathologica Communications Open Access 10 May 2023
Glioblastoma remodelling of human neural circuits decreases survival
Nature Open Access 03 May 2023
Considerations for modelling diffuse high-grade gliomas and developing clinically relevant therapies
Cancer and Metastasis Reviews Open Access 01 April 2023
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$29.99 / 30 days
cancel any time
Subscribe to this journal
Receive 51 print issues and online access
$199.00 per year
only $3.90 per issue
Rent or buy this article
Get just this article for as long as you need it
Prices may be subject to local taxes which are calculated during checkout
Bulk RNA-seq data have been deposited in the Sequence Read Archive (SRA) database under the accession number PRJNA554870. Clinical data of patient samples can be found in Supplementary Table 2. All other data and code that support the findings of this study are available from the corresponding authors on reasonable request.
Custom-written MATLAB and Igor code is available upon reasonable request.
Scherer, H. J. A critical review: the pathology of cerebral gliomas. J. Neurol. Psychiatry 3, 147–177 (1940).
Venkatesh, H. S. et al. Neuronal activity promotes glioma growth through neuroligin-3 secretion. Cell 161, 803–816 (2015).
Venkatesh, H. S. et al. Targeting neuronal activity-regulated neuroligin-3 dependency in high-grade glioma. Nature 549, 533–537 (2017).
Ishiuchi, S. et al. Blockage of Ca2+-permeable AMPA receptors suppresses migration and induces apoptosis in human glioblastoma cells. Nat. Med. 8, 971–978 (2002).
Takano, T. et al. Glutamate release promotes growth of malignant gliomas. Nat. Med. 7, 1010–1015 (2001).
Savaskan, N. E. et al. Small interfering RNA-mediated xCT silencing in gliomas inhibits neurodegeneration and alleviates brain edema. Nat. Med. 14, 629–632 (2008).
Rzeski, W., Turski, L. & Ikonomidou, C. Glutamate antagonists limit tumor growth. Proc. Natl Acad. Sci. USA 98, 6372–6377 (2001).
Li, L. & Hanahan, D. Hijacking the neuronal NMDAR signaling circuit to promote tumor growth and invasion. Cell 153, 86–100 (2013).
Li, L. et al. GKAP acts as a genetic modulator of NMDAR signaling to govern invasive tumor growth. Cancer Cell 33, 736–751 (2018).
Osswald, M. et al. Brain tumour cells interconnect to a functional and resistant network. Nature 528, 93–98 (2015).
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. 19, 1316–1326 (2017).
Zhu, Z. et al. Targeting self-renewal in high-grade brain tumors leads to loss of brain tumor stem cells and prolonged survival. Cell Stem Cell 15, 185–198 (2014).
Harris, K. M. & Weinberg, R. J. Ultrastructure of synapses in the mammalian brain. Cold Spring Harb. Perspect. Biol. 4, a005587 (2012).
Gray, E. G. Axo-somatic and axo-dendritic synapses of the cerebral cortex: an electron microscope study. J. Anat. 93, 420–433 (1959).
Venteicher, A. S. et al. Decoupling genetics, lineages, and microenvironment in IDH-mutant gliomas by single-cell RNA-seq. Science 355, eaai8478 (2017).
Darmanis, S. et al. Single-cell RNA-seq analysis of infiltrating neoplastic cells at the migrating front of human glioblastoma. Cell Reports 21, 1399–1410 (2017).
Maas, S., Patt, S., Schrey, M. & Rich, A. Underediting of glutamate receptor GluR-B mRNA in malignant gliomas. Proc. Natl Acad. Sci. USA 98, 14687–14692 (2001).
Sommer, B., Köhler, M., Sprengel, R. & Seeburg, P. H. RNA editing in brain controls a determinant of ion flow in glutamate-gated channels. Cell 67, 11–19 (1991).
Burnashev, N., Monyer, H., Seeburg, P. H. & Sakmann, B. Divalent ion permeability of AMPA receptor channels is dominated by the edited form of a single subunit. Neuron 8, 189–198 (1992).
Dalva, M. B., McClelland, A. C. & Kayser, M. S. Cell adhesion molecules: signalling functions at the synapse. Nat. Rev. Neurosci. 8, 206–220 (2007).
John Lin, C. C. et al. Identification of diverse astrocyte populations and their malignant analogs. Nat. Neurosci. 20, 396–405 (2017).
Mosbacher, J. et al. A molecular determinant for submillisecond desensitization in glutamate receptors. Science 266, 1059–1062 (1994).
Traynelis, S. F. et al. Glutamate receptor ion channels: structure, regulation, and function. Pharmacol. Rev. 62, 405–496 (2010).
Bergles, D. E., Diamond, J. S. & Jahr, C. E. Clearance of glutamate inside the synapse and beyond. Curr. Opin. Neurobiol. 9, 293–298 (1999).
Korber, V. et al. Evolutionary trajectories of IDHWT glioblastomas reveal a common path of early tumorigenesis instigated years ahead of initial diagnosis. Cancer Cell 35, 692–704.e612 (2019).
Patel, A. P. et al. Single-cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. Science 344, 1396–1401 (2014).
Chaichana, K. L., Parker, S. L., Olivi, A. & Quiñones-Hinojosa, A. Long-term seizure outcomes in adult patients undergoing primary resection of malignant brain astrocytomas. J. Neurosurg. 111, 282–292 (2009).
Weller, M., Stupp, R. & Wick, W. Epilepsy meets cancer: when, why, and what to do about it? Lancet Oncol. 13, e375–e382 (2012).
Ohtaka-Maruyama, C. et al. Synaptic transmission from subplate neurons controls radial migration of neocortical neurons. Science 360, 313–317 (2018).
de Groot, J. & Sontheimer, H. Glutamate and the biology of gliomas. Glia 59, 1181–1189 (2011).
Izumoto, S. et al. Seizures and tumor progression in glioma patients with uncontrollable epilepsy treated with perampanel. Anticancer Res. 38, 4361–4366 (2018).
Venkatesh, H. et al. Electrical and synaptic integration of glioma into neural circuits. Nature https://doi.org/10.1038/s41586-019-1563-y (2019).
Gibson, E. M. et al. Neuronal activity promotes oligodendrogenesis and adaptive myelination in the mammalian brain. Science 344, 1252304 (2014).
Buckingham, S. C. et al. Glutamate release by primary brain tumors induces epileptic activity. Nat. Med. 17, 1269–1274 (2011).
Huberfeld, G. & Vecht, C. J. Seizures and gliomas-towards a single therapeutic approach. Nat. Rev. Neurol. 12, 204–216 (2016).
Verhaak, R. G. et al. Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell 17, 98–110 (2010).
We thank M. Kaiser, S. Hoppe, J. Grosch, S. Weil, I. Sonntag, M. Osswald, K. Gunkel, C. Kocksch, I. Frommer, A. Schlicksupp, R. Rosauer, H.-Y. Nguyen, L. Doerner, M. Schmitt, U. Lindenberger, H. Zheng and S. Wendler for scientific discussion, support and assistance. We thank A. Hotz-Wagenblatt for help with bioinformatic analysis pipelines. We thank M. Suva and I. Tirosh for support with the analysis of single-cell RNA-seq databases from human gliomas, C. Steinhäuser and R. Jabs for advice on how to perform electrophysiological recordings from non-neuronal postsynaptic cells and the EM Core Facility of University Heidelberg for general support. We thank M. Monje for stably transducing our S24 glioma cell line with ChR2(H134R). We thank C. Watts for generating and providing the E2 primary glioblastoma cell line. A.A. was supported by the Chica and Heinz Schaller research foundation and the grant from the Deutsche Forschungsgemeinschaft (AG 287/1-1). V.V. was supported by the MD/PhD program of the Medical Faculty Heidelberg and the Stiftung für Krebs- und Scharlachforschung. D.I.T. was supported by the Deutsche Krebshilfe. W.W. and F.W. were supported by a grant from the Deutsche Forschungsgemeinschaft (SFB 1389). F.W. was supported by a grant from the Deutsche Forschungsgemeinschaft (WI 1930/6). T. Kuner was supported by the CellNetworks Excellence Cluster (EXC 81). F.W. and T. Kuner acknowledge their children Jakob and Manili, respectively, for seeding this collaboration.
F.W. and W.W. are inventors of the patent WO2017020982A1 ‘Agents for use in the treatment of glioma’. This patent covers new treatment strategies that all target the formation and function of TMs in glioma. F.W. reports research collaboration with DC Europa Limited, Glaxo Smith Kline, Genentech and Boehringer.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Peer review information Nature thanks Andrés Barría, Michael Taylor and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
Extended Data Fig. 1 Sensitivity and specificity of staining methods to detect glioma cells, TMs and NGS.
a, S24 PDX cells constitutively expressing tdTomato, co-labelled with a human epitope-specific antibody against nestin, with the latter sparing the nucleus (arrows). b, Nestin labelling is not detectable in Thy1-mGFP mice that express monomeric green fluorescent protein (mGFP) in a subset of neurons (arrowheads). c, On average, more than 90% of glioma cells in the S24 xenograft model are nestin-positive (n = 5 samples of 3 mice). d, e, Nestin specifically labels somata and TMs excluding the nucleus as revealed by DAB labelling using wide-field light microscopy (d) and transmission electron microscopy (e) (see Fig. 1f and Extended Data Fig. 2a–d for sample sizes). Arrowheads denote DAB-positive cytosolic regions of cells; arrows arrows denote DAB-spared nuclei. f, Immunogold labelling of a nestin-positive glioma TM (blue) in a S24 PDX mouse (n = 13 samples in 2 mice). Sepia staining denotes presynaptic bouton. Circles denote 5-nm immunogold particles, with an associated line profile through a 5-nm immunogold particle (bottom left) illustrating its high electron density. Arrowheads denote postsynaptic density of an asymmetric synaptic contact; arrow denotes docked vesicle at the active zone. g, Electron micrograph of an exemplary NGS as depicted in Fig. 1a, showing the electron-dense precipitate without colouring (see Fig. 1f for sample size). h, Nuclear DAB precipitate in a glioma cell expressing histone–GFP stained with an anti-GFP antibody (n = 3 samples in 2 mice). After DAB precipitation, the nucleus is prominently more electron-dense (arrowheads) than the nuclei from neighbouring non-malignant cells (arrows). The cell is outlined by a blue line. i, Nestin specifically labels glioma cells in IDH1(R132H)-mutated human gliomas. Glioma cells are unambiguously identified with the IDH1(R132H)-specific antibody. Nestin was stained on an adjacent section. The anti-nestin antibody specifically labels large subsets of IDH-positive areas. Nestin-positive microregions that are negative for IDH1(R132H) are not detectable (n = 3 patient samples). j, Immunostaining as in i, with an anti-EGFRvIII tumour-cell-specific antibody in wild-type IDH human GB, confirming the high degree of specificity of nestin staining that is only detectable in brain microregions that contain tumour cells (n = 3 patient samples). EGFRvIII denotes deletion variant III of EGFR. Data are mean ± s.e.m.
Extended Data Fig. 2 Ultrastructural identification of NGS across models.
a–d, Immuno-electron microscopy images of NGS in different mouse glioma models. Ultrastructural NGS identification with an anti-human nestin antibody and consecutive DAB precipitate in T1 PDX (a, n = 3 tissue blocks in 3 mice), E2 PDX (b, n = 5 tissue blocks in 1 mouse), T269 PDX (c, n = 3 tissue blocks in 1 mouse) and with an anti-GFP antibody and consecutive DAB precipitation in the syngeneic mouse glioma model where tumour cells express GFP (d, n = 5 samples in 3 mice). Arrowheads denote the synaptic cleft; arrows denote docked vesicles. Scanning electron microscopy (SEM) (a, b, d) and transmission electron microscopy (TEM) (c) images are shown. e, Electron tomography showing an overview of S24 PDX NGS with glioma cells marked by the anti-nestin antibody and consecutive 5-nm immunogold labelling. Arrows denote immunogold particles (8 tomograms from 2 S24 mice). f–i, Serial virtual 20-nm sections of e through the active zone of a NGS. In i, arrows denote postsynaptic density. Arrowheads denote the synaptic cleft. j, Overview electron microscopy image of S24 PDX marked by nestin and DAB precipitate, illustrating the infiltration zone of the tumour (n = 12 FOVs with a total area of approximately 13,400 µm2 in 2 S24 mice were evaluated). All samples analysed in this study showing NGS were obtained from such infiltration zones. Magnifications show normal synapses (left) and NGS (right). k, Overview and magnification of wild-type cortex labelled with nestin (n = 17 FOVs with a total area of approximately 12,400 µm2 in 2 wild-type mice were evaluated). As expected, in the absence of tumour cells, no DAB precipitate can be seen, indicating the tumour cell-specificity of the nestin staining in the analysed samples. l, Four examples of fully reconstructed NGS. m, Four fully reconstructed normal synapses located next to NGS. n, Immunogold labelling of nestin in T1 PDX (7 tissue blocks from T1 PDX from 3 mice). Small arrows depict immunogold particles; arrowheads point towards the synaptic cleft; long arrows denote docked vesicles. o, Stereologically determined synaptic density in wild-type neocortex and in glioma cell infiltration zone. No significant difference of the density of normal synapses could be found (two-sided unpaired t-test; n = 12 and 17 FOVs in 3 mice each). p, Stereologically quantified NGS density in the S24 PDX model (n = 12 FOVs in 2 S24 mice). q, Electron microscopy sections of a fully reconstructed NGS (representative example from n = 66 glioma and 52 healthy synaptic boutons in 3 samples from S24 mice). Consecutive serial sections are denoted ss01 to ss32. r, Different better view angles of the 3D reconstructed NGS from q. s, Electron microscopy image of the main tumour mass in the S24 PDX model showing no apparent NGS (n = 2 samples in 2 S24 mice). t, Immunofluorescence staining using an IDH1(R132H)-specific antibody (red) and a presynaptic bouton marker (synapsin, green) in the main tumour mass (left, n = 3 samples from 3 patients with IDH1(R132H)-mutated glioblastoma) and in the infiltration zone (right, n = 3 samples from 3 patients with IDH1(R132H)-mutated glioblastoma). u–x, Representative immuno-electron microscopy images of oligodendroglioma (u, v) and meningioma (w, x) PDX models. No neuron-tumour synapses were found (n = 6 samples TS603 with 361 oligodendroglioma cell sections from 3 mice and n = 6 samples IOMM with 309 meningioma cell sections from 3 mice were analysed). Data are mean ± s.e.m. Colour code in all panels: presynaptic boutons (sepia), glioma cell somata and TMs (blue), postsynaptic dendrites of neurons (green)
Extended Data Fig. 3 NGS classification in PDX glioma models and human tumour samples.
a–c, Three different morphological categories of glioma TM-associated synaptic contacts: single synaptic contact on a glioma TM (a); multisynaptic contact to both a glioma TM and a neuron (b); and a TM approaching an existing neuronal synapse with contact to the synaptic cleft, but without showing ultrastructural details of a bona-fide NGS (c). The corresponding original data with serial electron microscopy sections are shown underneath the schematic drawings. d1–d12, Morphometric synaptic parameter quantification of normal brain synapses versus NGS. d1, Synaptic vesicle diameter. d2, Synaptic cleft diameter. d3, Number of docked vesicles per bouton. d4, Area of postsynaptic density (PSD). d5, Synaptic profile length. d6, Number of endosomes per bouton. d7, Number of mitochondria per bouton. d8, Presynaptic bouton volume. d9, Number of vesicles per PSD area. d10, Percentage of perforated versus unperforated synapses. Perforation refers to a hole formed within the PSD area. d11, Vesicles per volume. d12, Vesicles per bouton. In conclusion, NGS are morphometrically similar to neuronal synapses. (d2, d5 two-sided unpaired t-test, d1, d3, d4, d6, d7, d8, d9, d10, d11, d12 two-sided Mann–Whitney test; n = 52 normal synapses and n = 66 NGS). e, T1 gadolinium contrast-enhanced MRI image of a patient diagnosed with an IDH1(R132H)-mutated astrocytoma WHO grade II (arrow). f, Representative example of a NGS imaged with SEM in resected tumour material of the patient in e. g, Fully reconstructed 3D model of the NGS in f. h, T1 gadolinium contrast-enhanced MRI image of a patient diagnosed with wild-type IDH GB (arrow). i, Representative example of a NGS imaged with SEM in resected tumour material of the patient in h. j, Fully reconstructed 3D model of the NGS in i. k–q, Quantification of synaptic morphometry in three examples of fully reconstructed NGS in resected material from a patient with IDH1(R132H)-mutated astrocytoma WHO grade II. k, Presynaptic bouton volume. l, Number of mitochondria per bouton. m, Number of endosomes per bouton. n, Synaptic cleft diameter. o, Number of vesicles per bouton. p, Number of docked vesicles. q, Number of vesicles per volume. The synaptic morphometry of human NGS was comparable to normal synapses. r, SEM images of S24 GB cell spheroids showing no synaptic ultrastructural differentiations at contact sites (see Supplementary Data Table 3 n = 8 S24 spheroid blocks, n = 3 T1 spheroid blocks). s, Example of a SEM image of a co-culture of S24 GB cells with cortical neurons: NGS that can be found in glioma cells stained with a human-specific anti-nestin antibody with consecutive DAB precipitation (n = 2 samples). t, Serial sections of the NGS depicted in s. u, Confocal MIP images of 10 µm of glioma cell co-cultures with cortical neurons. Putative synaptic contacts in GBSC line S24 (left, n = 12 putative synapses) and GBSC line BG5 (right, n = 8 putative synapses) in two independent experiments. Arrowheads denote postsynaptic HOMER1–3 cluster; arrows denote presynaptic VGLUT1. Data are mean ± s.e.m. Colour code in a–t: presynaptic boutons (sepia), glioma cell somata and TMs (blue), postsynaptic dendrites of neurons (green)
Extended Data Fig. 4 Extended molecular characterization of NGS.
a, Confocal maximum intensity projection (900 nm) showing co-localization of AMPAR (GluR1, arrowheads) with TMs (nestin) and presynaptic glutamatergic boutons (VGLUT1, arrows) (n = 3 mice for S24 and n = 3 human GB samples, see b). b, Quantitative co-localization analysis for the S24 and BG5 PDX models and human GB (n = 19 FOVs with 248 clusters in 3 mice for S24 PDX, n = 20 FOVs with 384 clusters in 1 mouse for BG5 PDX, n = 32 FOVs with 215 clusters in 3 human GB samples). P values determined by two-sided Kruskal–Wallis test with post hoc Dunn’s multiple comparisons test. c, Expression levels of neurotransmitter receptors detected by single-cell RNA-sequencing obtained from resected material of human IDH-mutated astrocytomas (n = 4 patients). d, Heat map with single-cell expression of neurotransmitter receptors in the single-cell glioblastoma dataset. Each column depicts a single cell. Glioblastoma cells are hierarchically clustered by receptor expression as depicted by the dendrogram. e, Chromosomal 1p/19q co-deletion and non-co-deletion determines classification into either IDH-mutated non-co-deleted astrocytomas or co-deleted oligodendrogliomas. Astrocytomas show significantly lower relative amount of GRIA2 mRNA editing than 1p/19q co-deleted oligodendrogliomas (n = 521 patient samples). P < 0.0001, two-sided Mann–Whitney test. f, Relative level of GRIA2 mRNA editing in human GBs according to their gene expression subtype37 (n = 60 GB patient samples and 24 LGG patient samples). See Supplementary Table 7 for P values; ordinary one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test. g, h, Single-cell expression analysis of synaptogenic genes in human IDH-mutated astrocytomas (g) and GBs (h). Most tumour cells express at least one of the five genes that have been most robustly associated with synaptogenesis (NLGN1 (neuroligin 1), NLGN2 (neuroligin 2), LRFN1 (leucine-rich repeat and fibronectin type III domain-containing protein 1), CADM1 (cell adhesion molecule 1) and EFNB2 (ephrin B2)). The green colour on top of the graph indicates that at least one of the synaptogenic genes is expressed in the particular cell. Each vertical column represents one cell with the normalized expression of the synaptogenic genes. The order of the x axis is determined by hierarchical clustering without showing the dendrogram. i, j, Confocal MIP of 10 µm of the BG5 PDX model (i, n = 3 samples in 1 mouse) and human GB (j, n = 3 samples in resected material from 3 patients) demonstrating connections between synaptic and intratumoural TM-connectivity. Arrows denote presynaptic VGLUT1 clusters (blue); arrowheads denote GluR1 clusters (red) located on crossing, connected TMs (green). k, Electron microscopy images of a PDX S24 glioma cell showing 2 NGS (top right and bottom right) in the vicinity of a TM crossing (top left) (7 observations in n = 3 experiments). l, Quantification of relative SR101 signal intensity in single glioma cells (n = 116 TM connected versus n = 36 TM unconnected cells in 3 S24 PDX mice). P value determined by two-sided Mann–Whitney test m, Single-cell gene expression analysis of a human IDH-mutated astrocytoma showing a distinct gene expression pattern in cell clusters positive for ApoE. n, Connectivity score of single tumour cells (see Supplementary Methods) is significantly correlated with ApoE cluster assignment (n = 1,911 single cells from 4 patients with IDH mutated astrocytomas). P value determined by two-sided Mann–Whitney test. o, GRIA1 expression of single tumour cells in the ApoE-negative versus ApoE-positive cluster (n = 1,911 single cells from 4 patients with IDH mutated astrocytomas). P value determined by two-sided Mann–Whitney test. p, q, Gene Ontology (GO) term analysis of representative ApoE-positive clusters from patients with IDH-mutated astrocytoma (p) and GB (q) associated with neuronal and synapse-associated processes (top 15 GO terms shown). All associated GO terms of all ApoE-positive clusters as well as ApoE-negative clusters can be found in Supplementary Tables 5 and 6. The false discovery rate was kept below 0.05. P values determined by Fisher’s exact test. For sample size, see ‘Statistics and reproducibility’ in Supplementary Methods. Data are mean ± s.e.m
Extended Data Fig. 5 Electrophysiological characterization of EPSCs and SICs in glioma cells.
a, Representative glioma cells expressing GFP (green) filled with Alexa 594 via the patch pipette (red, one example of n = 720 filled cells). Note that only the bottom of the two cells marked with arrows was recorded in whole-cell configuration. The top glioma cell was presumably filled via gap junctions. b–d, Quantitative properties of EPSCs in different glioma models resemble normal AMPAR kinetics (n = 171 EPSCs in 7 different models; Supplementary Methods for details). n.s., not significant; two-sided Kruskal–Wallis test with post hoc Dunn’s multiple comparisons test. e, Representative traces showing inhibition of spontaneous EPSCs by CNQX in BG5 co-culture. f, Representative traces of spontaneous EPSC inhibition with NASPM in S24 co-culture (n = 2 cells). g, Quantification of spontaneous SIC amplitude and duration across different cell lines (S24 n = 203 SICs, BG5 n = 39 SICs, T1 n = 55 SICs). **P < 0.01, ****P < 0.0001, two-sided Kruskal–Wallis test with post hoc Dunn’s multiple comparisons test. h, i, Representative traces of spontaneous SIC inhibition with TTX in S24 co-culture (h) and corresponding quantification (i) (n = 8 cells). P value determined by two-sided Wilcoxon matched pairs test. j, k, Representative traces of spontaneous SIC inhibition with CNQX in S24 co-culture (j) and corresponding quantification (k) (n = 8 cells). P value determined by two-sided Wilcoxon matched pairs test. l, m, Representative traces of spontaneous SIC inhibition with NASPM in S24 co-culture (l) and corresponding quantification (m) (n = 10 cells). P value determined by two-sided paired t-test. n, Quantification of spontaneous SIC amplitude in baseline compared to NASPM (n = 100 spontaneous SICs baseline versus n = 98 spontaneous SICs under NASPM). P value determined by two-sided Mann–Whitney test. o–q, Pharmacological characterization of evoked SICs in S24 co-culture. Note the differing effects of TBOA and barium on individual cells (black denotes baseline; red denotes drug; green denotes washout). r, Quantification of access resistance in different cell lines, grouped by spontaneous slow inward current positivity. In all cell lines, sSIC-positive (sSIC+) cells show significantly lower access resistance (two-sided unpaired t-test with Welch’s correction, S24 n = 12 sSIC+ cells versus n = 28 sSIC− cells; two-sided Mann–Whitney test, T1 n = 4 sSIC+ cells versus n = 13 sSIC− cells, BG5 n = 2 sSIC+ cells versus n = 38 sSIC− cells, pooled n = 18 sSIC+ cells versus n = 79 sSIC− cells). s, t, Representative traces (s) and corresponding quantification (t) showing inhibition of sSIC frequency with GAP26 and GAP27 in S24 co-culture (n = 5 cells). u, v, Representative traces (u) and corresponding quantification (v) showing inhibition of spontaneous spontaneous SICs by meclofenamate (n = 8 cells). P value determined by two-sided paired t-test. w, Quantification, showing meclofenamate significantly decreases the leak current of tumour cells (two-sided Wilcoxon matched-pairs signed rank test, n = 8 cells). x, Example traces from two different cells showing that spontaneous SICs and sEPSCs can occur simultaneously or separately. EPSCs are marked by asterisks in the top trace. Magnification of EPSCs occurring on top of slow inward currents is shown in the bottom trace and marked by arrows. y, Quantification of different electrophysiological subgroups of glioma cells in different cell lines (S24 n = 30 cells, T1 n = 18 cells, BG5 n = 39 cells) and summary of electrophysiological subgroups across different glioma cell lines in co-culture. Data are mean ± s.e.m
Extended Data Fig. 6 Neuronal input triggers glioma network activation.
a, Spontaneous SIC with temporally correlated calcium transient in high magnification. b, c, Representative traces of voltage response to current injection simulating slow inward currents (top) and simultaneous calcium trace (bottom) in two different time scales. d, Representative current-clamp recording from a PDX S24 glioma cell selected from a total of n = 318 recordings from different models. Voltage responses to current injections (−80 to 360 pA; step size, 40 pA). e, Individual glioma cells of one tumour microregion respond to ChR2 stimulation with different latencies. Calcium traces of individual glioma cells were normalized to values between 0 and 1. f, Standard deviation projection of calcium time-series imaging of GCaMP6 fluorescence in S24 cells co-cultured with neurons. Arrow denotes bipolar stimulation electrode; numbers 1 to 4 correspond to individual tumour cells in g. g, Different latencies of calcium responses to neuronal stimulation. h, Representative coactivity maps of functionally neuron-connected glioma cells (blue) and non-neuron connected glioma cells (red) after neuronal ChR2 stimulation. i, Quantification of connections between functionally neuron-connected glioma cells and non-neuron connected glioma cells before and after neuronal ChR2 stimulation, normalized to the number of functionally neuron-connected glioma cells (n = 10 experiments in 5 S24 PDX mice, see Supplementary Methods). P value determined by two-sided Mann–Whitney test. j, Representative traces showing gabazine-induced stacking of slow currents into prolonged events (n = 8 cells in S24 co-culture). Data are mean ± s.e.m
Extended Data Fig. 7 General anaesthesia silences glioma networks and effects of neuron-induced calcium transients on invasion speed.
a, In vivo calcium transient frequency under control conditions and deep anaesthesia. Calcium transient frequency of S24 PDX glioma cells imaged during deep anaesthesia with isoflurane compared with control (minimal dose anaesthesia) conditions (n = 274 cells per group in 3 S24 PDX mice). P value determined by two-sided Mann–Whitney test. b, Representative coactivity maps of glioma cells under control conditions versus deep anaesthesia. c, Calcium transient coactivity indices under both conditions (n = 6 experiments in 3 mice). P value determined by two-sided Mann–Whitney test. d, Three-step anaesthesia experiment of calcium transient frequency during different stages of isoflurane anaesthesia (n = 160 cells per group; in 3 S24 PDX mice. Data are s.d.). ****P < 0.0001, two-sided Kruskal–Wallis test with post hoc Dunn’s multiple comparisons test. e, Calcium transient frequency before, during and after isoflurane anaesthesia (n = 106 cells per group; in 2 S24 PDX mice). *P < 0.05; ****P < 0.0001, two-sided Kruskal–Wallis test with post hoc Dunn’s multiple comparisons test. f, In vivo calcium transient frequency under control conditions and deep anaesthesia. Calcium transient frequency from S24 PDX glioma cells imaged during deep anaesthesia with a mixture of midazolam, medetomidine and fentanyl compared with control (minimal dose anaesthesia) conditions (n = 154 cells per group in 3 S24 PDX mice). P value determined by two-sided Mann–Whitney test. g, Representative coactivity maps of glioma cells under control conditions versus deep anaesthesia. h, Calcium transient coactivity indices under both conditions (n = 6 experiments in 3 mice). P value determined by two-sided Mann–Whitney test. i, Quantification of invasion speeds of glioma cells in vivo under control conditions versus isoflurane anaesthesia with respect to glioma cells possessing TMs (control: n = 21 cells without TMs in 4 mice and n = 234 cells with TMs in 3 mice, isoflurane anaesthesia: n = 16 cells without TMs in 3 mice and n = 127 cells with TMs in 3 mice). ***P < 0.001, ****P < 0.0001, two-sided Kruskal–Wallis test with post hoc Dunn’s multiple comparisons test. j, Temporal correlation between individual calcium fluctuations (bottom trace) of Rhod2-AM-loaded glioma cells and individual cell movement during migration. Note that the cessation of calcium transients (marked by the dashed red line) correlated with a decrease in cell mobility. k, Relationship between the frequency of calcium transients and the migration speed of glioma cells (n = 51 cells in S24 co-culture). P < 0.0001, two-sided linear regression; Pearson correlation coefficient is 0.57, r2 value is 0.327. l, Differential interference contrast (DIC) (left) and fluorescent (right) images of HEK-293T cells transfected with a tdTomato-GluA2 (tdT-GluA2) plasmid. Cells expressing tdT-GluA2 are indicated by red arrowheads. m, DIC (left) and fluorescent (middle and right) images of HEK-293T cells transfected with tdTomato-GluA2 and DN-GluA2-GFP plasmids. Cells expressing eGFP-dnGluA2 or both eGFP-dnGluA2 and tdTomato are indicated by green or yellow arrowheads, respectively. n, o, Example trace showing GluA2 homotetramer-mediated currents, in whole-cell voltage clamp mode, in response to uncaged glutamate (blue box) on cells expressing GluA2 (n) and both GluA2 and dnGluA2 (o). p, Peak current in GluA2 cells and in cells expressing both GluA2 and DN-GluA2 subunits (n = 5 GluA2 cells and n = 6 GluA2/dnGluA2 cells). P value determined by two-sided Mann–Whitney test. Data are mean ± s.e.m. q, Quantification of invasion speed of DN-GluA2-GFP and tdTomato cells compared with control cells with respect to glioma cells possessing TMs (control without TMs n = 7 cells, control with TMs n = 94 cells; DN without TMs n = 15 cells, DN with TMs n = 114 cells in 5 mice). **P < 0.01, ****P < 0.0001, two-sided Kruskal–Wallis test with post hoc Dunn’s multiple comparisons test. Data are mean ± s.e.m
Extended Data Fig. 8 Validation of functional NGS effects in vitro.
a, Demonstration of depolarizing inward current in S24 ChR2 cells in response to 10-Hz pulses of blue light as measured in voltage clamp. b, Comparison of proliferation with the CellTiter-Glo assay after three days of stimulation (BG5 calcium-translocating channelrhodopsin (CatCh) and S24 ChR2 glioblastoma cells) in monocultures (BG5 CatCh n = 24 samples per group; S24 ChR2 n = 24 samples per group). P values determined by two-sided Mann–Whitney test. c, In vitro growth analysis of GluA2-DN-GFP and tdTomato S24 glioma cells (DN) and control cells (WT) in monoculture reveal no significant difference in proliferation rate between cell lines (n = 16 technical replicates for each group and day, one representative experiment out of three, two-sided Mann-Whitney test). d, Representative epifluorescence images of S24 glioma cells co-cultured with neurons (left) or cultured alone (right). The area marked in the top panel is shown magnified in the central panels. Thresholded images are used for automated quantification (bottom; see Supplementary Methods). e, g, Co-cultures of S24 glioma cells and neurons treated with the AMPAR antagonist 20 µM CNQX for 8 days grow more slowly and therefore show lower glioma cell densities (e) and smaller areas covered by glioma cells (g) than controls (e, n = 13 coverslips control and n = 14 coverslips CNQX; g, n = 30 coverslips control and n = 28 coverslips CNQX). f, h, Monocultures of S24 glioma cells generally grow more slowly than co-cultures, but show no significant effect of treatment with 20 µM CNQX on cell densities (f) or area covered by glioma cells (h) (f, n = 6 coverslips control and n = 7 coverslips CNQX; h, n = 6 coverslips control and n = 8 coverslips CNQX). i, j, Quantification of relative T1 glioma cell area in co-culture with neurons (i) and in monoculture (j), after 8 days of treatment with CNQX, TTX, SAS or SAS plus CNQX, leading to a significant decrease in the proliferation of T1 tumour cells cultured with neurons while not significantly affecting cells in monoculture (i, control: n = 27 coverslips, CNQX: n = 23 coverslips, TTX, SAS, SAS + CNQX: n = 12 coverslips, two-sided Kruskal-Wallis test, post hoc Dunn’s multiple comparisons test; j, control: n = 17 coverslips, CNQX, SAS + CNQX: n = 13 coverslips, TTX, SAS: n = 12 coverslips, two-sided ordinary one-way ANOVA, post hoc Holm–Sidak’s multiple comparisons test). k, l, Quantification of relative T1 glioma cell area in co-culture with neurons (k) and in monoculture (l) after 8 days of treatment with neurexin 1β, CNQX or both neurexin 1β and CNQX (n = 4 coverslips for each group). Only CNQX treatment leads to a significant decrease in proliferation of T1 co-cultured tumour cells. m, The effects of different concentrations of glutamate on cell proliferation were tested in monocultures of S24 glioma cells. The pro-proliferative effect of glutamate can be partially abrogated by CNQX (n = 10 for 500 µM glutamate and 500 µM glutamate/CNQX; n = 5 for all other groups). See Supplementary Methods for further details. P values determined by two-sided unpaired t-test (e, g, h), two-sided Mann–Whitney test (f), two-sided Kruskal–Wallis test, post hoc Dunn’s multiple comparisons test (i, k–m), or two-sided ordinary one-way ANOVA, post hoc Holm-Sidak’s multiple comparisons test (j). Data are mean ± s.e.m
Extended Data Fig. 9 Inhibition of glioma growth by perturbation of AMPAR signalling.
a, BG5 PDX glioma regions imaged in vivo on days 0 and 14 under control conditions and after treatment with AMPAR antagonist perampanel (n = 25 control and n = 29 perampanel FOVs in 5 (control) and 6 (perampanel) mice; see Fig. 5k). b, Cell viability of S24 and BG5 glioma cells exposed to different concentrations of perampanel (1–100 µM) in monocultures in vitro (n ≥ 3 wells per condition, 2 biological independent experiments)
Extended Data Fig. 10 Schematic illustration of neurogliomal synapses and their functional role in brain tumour progression.
The left of the scheme illustrates the neuronal network, the gliomal network, and their interconnectivity. A magnified scheme of a neurogliomal synapse is shown in the centre. The consequence of NGS function for brain tumour biology is illustrated on the right.
This file contains a Supplementary Discussion, Table Legends, References and the Supplementary Video Legends.
This file contains the Methods section of this manuscript.
This zip file contains Supplementary Tables 1-7. Table legends are provided in Supplementary Information file.
Video 1: 3D reconstruction of a NGS in the S24 PDX model with serial section SEM. Electron microscopic overview of a NGS in the infiltration zone of the S24 xenograft model in the mouse brain. Note that the NGS (sepia-blue) lies next to normal synapses (sepia-green) and holds all hallmarks typical for chemical synapses, including a presynaptic vesicle cluster, docked vesicles, a synaptic cleft and, a postsynaptic density. Serial sections and a full 3D reconstruction of a NGS reveal the configuration of the presynaptic bouton, including area of the active zone (turquoise), total number of vesicles (white) and, docked vesicles (red). Note that the postsynaptic part is a glioma TM.
Video 2: 3D reconstruction of a NGS in a syngeneic glioma model with serial section SEM. Electron microscopic overview of a NGS in the infiltration zone of the syngeneic glioma model in the mouse brain. Note that the NGS (sepia-blue) lies next to normal synapses (sepia-green) and holds hallmarks typical for chemical synapses, including a presynaptic vesicle cluster, docked vesicles and a synaptic cleft. Serial sections and 3D reconstruction reveal the configuration of the presynaptic bouton, including area of the active zone (turquoise), total number of vesicles (white) and, docked vesicles (red). Note that the postsynaptic part is a glioma TM.
Video 3: 3D reconstruction of a NGS in a human LGG with serial section SEM. Electron microscopic overview of a NGS in human LGG (IDH-mutated astrocytoma WHO grade II) resection material. Note that the NGS (sepia-blue) holds hallmarks typical for chemical synapses, including a presynaptic vesicle cluster, docked vesicles and a synaptic cleft. Serial sections and 3D reconstruction reveal the configuration of the presynaptic bouton, including area of the active zone (turquoise), total number of vesicles (white) and, docked vesicles (red). Note that the postsynaptic part is a glioma TM.
Video 4: 3D reconstruction of a NGS in a human GB with serial section SEM. Electron microscopic overview of a NGS in human GB resection material. Note that the NGS (sepia-blue) holds hallmarks typical for chemical synapses, including a presynaptic vesicle cluster, docked vesicles and a synaptic cleft. Serial sections and 3D reconstruction reveal the configuration of the presynaptic bouton, including area of the active zone (turquoise), total number of vesicles (white) and, docked vesicles (red). Note that the postsynaptic part is a glioma TM.
Video 5: Glioma cell calcium imaging before and after neuronal channel-rhodopsin stimulation. Calcium activity of S24 PDX cells transduced with the genetic calcium indicator TWITCH3A (green) measured with in vivo MPLSM. Neurons expressing channel-rhodopsin (not visible) were stimulated with blue light. After stimulation with 20 Hz for 1 s (disruptions/blue screens) calcium activity in glioma cells increases significantly (n=8 experiments in 4 S24 PDX mice).
Video 6: Effects of high-dose isoflurane on glioma cell calcium activity. Calcium activity of S24 PDX cells transduced with the genetic calcium indicator TWITCH3A (green) measured with in vivo MPLSM. First half shows calcium transients under low-dose isoflurane as control condition. Activity significantly decreases when isoflurane anesthesia is increased (second half; n=6 experiments in 3 mice).
Video 7: Effects of a mixture of medetomidine, midazolam and, fentanyl on glioma cell calcium activity. Calcium activity of S24 PDX cells transduced with the genetic calcium indicator TWITCH3A (green) measured with in vivo MPLSM. First half shows calcium transients under low-dose isoflurane as control condition. Activity significantly decreases under deep anesthesia with a mixture of medetomidine, midazolam, and fentanyl (second half; n=6 experiments in 3 mice).
Rights and permissions
About this article
Cite this article
Venkataramani, V., Tanev, D.I., Strahle, C. et al. Glutamatergic synaptic input to glioma cells drives brain tumour progression. Nature 573, 532–538 (2019). https://doi.org/10.1038/s41586-019-1564-x
This article is cited by
Advanced immunotherapies for glioblastoma: tumor neoantigen vaccines in combination with immunomodulators
Acta Neuropathologica Communications (2023)
Spatial transcriptomics reveals niche-specific enrichment and vulnerabilities of radial glial stem-like cells in malignant gliomas
Nature Communications (2023)
The neural addiction of cancer
Nature Reviews Cancer (2023)
Considerations for modelling diffuse high-grade gliomas and developing clinically relevant therapies
Cancer and Metastasis Reviews (2023)
Neuronal Mechanisms Govern Glioblastoma Cell Invasion
Neuroscience Bulletin (2023)
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.