Functional synapses between neurons and small cell lung cancer

Abstract

Small cell lung cancer (SCLC) is a highly aggressive type of lung cancer, characterized by rapid proliferation, early metastatic spread, frequent early relapse and a high mortality rate^1-3. Recent evidence has suggested that innervation has an important role in the development and progression of several types of cancer^4,5. Cancer-to-neuron synapses have been reported in gliomas^6,7, but whether peripheral tumours can form such structures is unknown. Here we show that SCLC cells can form functional synapses and receive synaptic transmission. Using in vivo insertional mutagenesis screening in conjunction with cross-species genomic and transcriptomic validation, we identified neuronal, synaptic and glutamatergic signalling gene sets in mouse and human SCLC. Further experiments revealed the ability of SCLC cells to form synaptic structures with neurons in vitro and in vivo. Electrophysiology and optogenetic experiments confirmed that cancer cells can receive NMDA receptor- and GABA_A receptor-mediated synaptic inputs. Fitting with a potential oncogenic role of neuron-SCLC interactions, we showed that SCLC cells derive a proliferation advantage when co-cultured with vagal sensory or cortical neurons. Moreover, inhibition of glutamate signalling had therapeutic efficacy in an autochthonous mouse model of SCLC. Therefore, following malignant transformation, SCLC cells seem to hijack synaptic signalling to promote tumour growth, thereby exposing a new route for therapeutic intervention.

Fig. 1. Genome-wide analysis of SCLC across species.

a, Circos plot displaying the transposon integration pattern of an unbiased piggyBac insertional mutagenesis screen in 303 mouse tumours. The chord plot in the centre shows the transpositions from the donor loci (empty triangles) on chromosomes 5 and 10 to the 100 genes with the most significant enrichment in transposon insertions. The middle layer shows the chromosome labels. The scatterplot in the outer layer includes all genes with a significant enrichment in transposon insertions (q < 0.1, Poisson distribution with false discovery rate (FDR) correction). Selected genes are annotated, and genes previously linked to SCLC have label boxes. b, Top 20 most significantly enriched GO terms in the piggyBac dataset and human genetic data. Significance was determined by two-sided Fisher’s exact test with FDR correction. Mod., modulation; reg., regulation. c, Force-directed graph of GO analysis, showing gene sets enriched for genes upregulated in SCLC compared with other types of cancer from the TCGA dataset and with healthy tissue types from the GTEx dataset. Significance was determined by two-sided Fisher’s exact test with FDR correction. d, Scatterplot of the gene sets in c. On the y axis is the RB–E2F score, calculated using ChIP–seq data from the CISTROME database. A high score indicates strong ChIP–seq signal in experiments with antibodies against RB1, RBL2, E2F1, E2F2, E2F3, E2F4 or E2F5 near the promoter of the upregulated genes included in the gene set. On the x axis is the fold change in expression on the log₂ scale for PNECs versus other lung cell types in published scRNA-seq data. A high fold change indicates that the upregulated genes in the gene set are also upregulated in healthy PNECs. Significance was determined by two-sided Fisher’s exact test with FDR correction. Source data

Fig. 2. Detection of nerve fibres in mouse SCLC tumours.

a, Confocal image of an intrapulmonary airway from an RP mouse. Two small tumours (ST) and a normal NEB (open arrowhead) are visualized with CGRP (green) and can be observed to bulge in the airway lumen (L). VGluT1-immunoreactie nerve terminals (red) are detected contacting the NEB and arborizing (arrowhead) in one of the tumours. CGRP-positive nerve fibres (open arrows) can be observed at the base of the tumours and NEB. E, epithelium. b, P2X3-positive nerve terminals (red, arrowheads) can be seen to arborize between the CGRP-positive (green) neuroendocrine cells of a small tumour. CGRP-positive nerve fibres (open arrows) can be observed at the base of the tumour. c, Immunolabelling of a CGRP-positive (green) large tumour (LT). The connective tissue between the tumour and the epithelium harbours many GAP43-positive nerve fibres (red, arrows), which do not appear to penetrate the tumour mass. d, Confocal image of an NEB in an RPC mouse. Two cells are positive for eGFP (blue), indicating successful recombination and incipient transformation. VGluT1-positive fibres (green) arborize between the neuroendocrine cells (red). e, Initial proliferation of eGFP-positive neuroendocrine cells (blue) in an NEB. VGluT1-positive fibres (green) arborize between the transforming cells. f, Small SCLC tumour positive for eGFP (blue) and CGRP (red). VGluT1-positive fibres (green) arborize between the tumour cells. g, Immunolabelling of SCLC cells (expressing DsRed) transplanted into the hippocampus of Thy1-eGFP mice. The inset shows that the core of the tumour is devoid of eGFP-positive fibres. h, 3D reconstruction of SCLC cells located in the tumour periphery surrounded by eGFP-positive axonal varicosities. i, Co-localization analysis of eGFP- and VGluT1-positive boutons contacting a DsRed-expressing SCLC cell.

Fig. 3. Structural evidence for bona fide synapses in SCLC cells.

a, 3D STED images of SCLC (expressing mNeonGreen)–neuron co-cultures stained for presynaptic VGluT1 and postsynaptic HOMER1. The magnified views on the right show regions of marker co-localization. b, Analysis of the number of VGluT1 and HOMER1 single-positive and double-positive puncta per SCLC cell. n = 29 cells derived from three independent cultures and two x10ht experiments. Kruskal–Wallis one-way ANOVA test, ****P < 0.0001. NS, not significant. c, Overview of a representative 3D-reconstructed confocal image of an SCLC cell in a neuronal co-culture subjected to x10ht. Bottom panels depict magnified regions of contact between the neuron (VGluT1 positive) and SCLC cell (HOMER1 positive). d, Two-colour 3D ONE image of region 3 in c. e, Three-colour 2D ONE image of a representative putative synapse showing presynaptic (VGluT1-positive) and postsynaptic (HOMER1-positive) markers at points of contact between neurons and SCLC cells. f, Line scan of the neuron–SCLC contact in e showing the distance between VGluT1- and HOMER1-positive puncta. g, VGluT1–HOMER1 apparent distance measured in neuron–SCLC cell contacts. n = 15 contacts. h, VGluT1–HOMER1 apparent distance measured in neuron–neuron contacts. n = 20 contacts. i, CLEM of SCLC cells (expressing tdTomato) grafted into the mouse hippocampus. The left two panels depict the registered overlay between the fluorescence signal and electron microscopy (EM) image. The third panel shows the electron tomogram of an identified synaptic contact. The tomogram (single slice) depicts a presynaptic bouton (yellow pseudocolour) filled with vesicles contacting a tdTomato-positive cancer cell (red pseudocolour). Blue pseudocolour indicates the nucleus. The rightmost panel shows an enlarged view of the synaptic cleft and a pool of vesicles located within 20 nm of the plasma membrane (green pseudocolour). Source data

Fig. 4. Neuron-to-SCLC synapses are functional.

a, Whole-cell voltage-clamp traces in artificial cerebrospinal fluid (aCSF; control) and following treatment with NMDA receptor (d-AP5) and GABA_A receptor (Gbz) blockers. Representative of seven cells across three experiments. Red asterisks or numbers mark individual events. b, Frequency of currents in H524 cells co-cultured with cortical neurons (untreated or exposed to d-AP5 alone or together with Gbz). Current frequency is compared before and after addition of d-AP5 (paired two-sided Wilcoxon test, n = 6 treated cells). Inset, example of a patched H524 cell. c, Whole-cell voltage-clamp traces of SCLC cells (grey) after a blue-light pulse (5 ms) to stimulate ChR2–enhanced yellow fluorescent protein (eYFP)-expressing neurons. The effects of d-AP5 (n = 12/13) or d-AP5 + Gbz (n = 1/13), compared to aCSF, are shown. d, Amplitude of evoked currents in H524 cells co-cultured with ChR2–eYFP-expressing cortical neurons after optogenetic stimulation. The amplitude before and after addition of d-AP5 is compared (two-sided paired Wilcoxon test, n = 13). Inset, example of a patched tdTomato-expressing H524 cell. e, Retrograde tracing of neurons monosynaptically connected to SCLC cells expressing DsRed, G and TVA after addition of EnvA-pseudotyped (ΔG) RABV-GFP. Lower panels, magnified views of double-positive SCLC starter cells (arrowheads). f, Quantification of RABV-GFP-mediated neuronal labelling following SCLC transduction with virus encoding TVA alone or together with G (n = 5 biological replicates). All conditions are compared to the full experimental system (RABV, DsRed, G, TVA). q values were obtained by two-sided Mann–Whitney test with FDR correction. g, Connectivity ratio per COR-L88 and DMS273 starter cell (n = 4–5 biological replicates). h, Retrograde tracing of neurons monosynaptically connected to G-TVA- and DsRed-expressing SCLC cells grafted into the mouse hippocampus. Right panels, magnified views of GFP-positive presynaptic excitatory neurons. GFP-only-positive axonal fibres contacting SCLC cells are indicated (arrowheads). CA1, cornus ammonis; DG, dentate gyrus; Sub, subiculum. i, Connectivity ratio per SCLC starter cell in mice grafted with TVA- or G-TVA-expressing SCLC cells (n = 6–7 mice per condition), P value obtained by two-sided Mann–Whitney test. j, Number of traced GFP-positive neurons classified as excitatory or inhibitory (n = 7 mice). Source data

Fig. 5. Glutamatergic signalling constitutes an actionable target in SCLC.

a,b, DsRed-expressing COR-L88 SCLC cells cultured for 3 days with (a) or without (b) cortical neurons. c,d, Fold change in total (c) and EdU-positive (d) COR-L88 cells cultured for 3 days with or without neurons and/or TTX (n = 3; two-sided paired t test; centre, mean; error bars, s.d.). e, Growth curves of COR-L88 cells cultured with or without cortical neurons from one experiment (n = 10 and n = 30 wells). f, Quantification of live-cell imaging of SCLC cell lines (n = 8; in order: H526, H1836, H146, H69, COR-L88, DMS273, H211, H524). Cancer cells were cultured with cortical neurons or cortical neuron-conditioned medium. Proliferation was normalized to the growth of monocultures in the same plates. P value were obtained by two-sided Wilcoxon signed-rank test. AUC, area under the curve. g, Growth quantification of NSCLC cell lines (n = 4; in order: HOP62, HCC44, H2291, H1975) in co-culture with cortical neurons, normalized to the growth of monocultures. h, Growth curves of COR-L88 cells cultured with or without nodose ganglia from two experiments (n = 4 and n = 16 wells). i, Growth quantification of SCLC cell lines (n = 3; in order: DMS273, H211, COR-L88) co-cultured for 5 days with nodose ganglion explants, relative to monocultures. j, Response of tumours in mice treated with vehicle (n = 102), DCPG (n = 54) or riluzole (n = 57), expressed as percentage of the initial volume. q values were obtained by two-sided Mann–Whitney test with FDR correction. k, Overall survival of RP mice treated with DCPG (n = 12), riluzole (n = 12) or the relative control (n = 33). q values were obtained by two-sided log-rank test with FDR correction. l, Response of tumours in mice treated with etoposide + cisplatin alone (n = 45) or combined with DCPG (n = 38) or riluzole (n = 36), expressed as percentage of the initial volume. q values as in j. m, Overall survival of RP mice treated with etoposide + cisplatin alone (n = 11) or combined with riluzole (n = 13) or DCPG (n = 10). q values as in k. See Extended Data Fig. 10 for individual replicates of f, g and i. Source data

Extended Data Fig. 1. piggyBac insertional mutagenesis screen.

a) Alleles included in the mouse model b) Mouse lines included in the screen carry the Rb1^fl/fl and Tp53^fl/fl alleles with the addition of the conditional allele to express the piggyBac transposase (Rosa26^LSL-PB). The RPLH line (blue) additionally carries the donor allele ATP1-H39, with 80 copies of the ATP1 transposon on chromosome 5. The RPLS line (orange) additionally carries the donor allele ATP1-S2, with 20 copies of the ATP1 transposon on chromosome 10. c, d) Tumors derived from RPLH and RPLS mice display typical SCLC morphology, including scant cytoplasm, salt and pepper chromatin and positivity for NCAM1 and SYP. e) Tumors harvested from 31 RPLH (blue) and 27 RPLS (orange) mice include lung and metastatic samples and derive from untreated mice, from mice treated with etoposide and cisplatin and from mice treated with anti-PD1 antibody RMP1-14. PB, piggyBac inverted terminal repeat (ITR); SB, Sleeping Beauty ITR; SA, splicing acceptor; pA, polyadenylation signal; CAG, CMV enhancer and chicken beta-actin promoter; SD, splicing donor; NEO, neomycin resistance; iPBase, piggyBac transposase. f-m) Transposon insertions (red arrows) identified in selected genes (horizontal blue lines). The orientation of the exons (vertical obtuse blue angles) point to the direction of transcription. f) Insertions in Crebbp g) Insertions in Pten. h) Insertions in Nfib. i) Insertions in Trp73. j) Insertions in Nrxn1. k) Insertions in Nlgn1. l) Insertions in Dcc. m) Insertions in Reln. Source data

Extended Data Fig. 2. Overview of genetic data from human SCLC patients.

a) Origin and characteristics of human samples from different studies. b) Similar genes are identified in tumor and cell line samples. c) Similar genes are identified in primary and metastatic samples. d) Similar genes are identified in treated and untreated samples. e-l) Selected genes are shown with the corresponding proteins annotated with UniProt Knowledgebase annotations. Mutations identified in SCLC samples are shown as a lollipop chart above the protein. Severe mutations (stop, frameshift, start-loss, and canonical splice-site mutations) are shown in red. Nonsynonymous mutations (amino-acid substitutions, non-frameshift indels) are shown in light blue. e) Mutations in TP53 are either severe or clustered in the DNA-binding domain. f) Mutations in RB1 are almost exclusively severe. g) Mutations in CREBBP are severe or clustered in the HAT domain. h) Mutations in PTEN are severe or clustered on the active site. i) Mutations in NRXN1 are mainly nonsynonymous. j) Mutations in NLGN1 are exclusively nonsynonymous. k) Mutations in DCC are mainly nonsynonymous. l) Mutations in RELN are nonsynonymous or severe. Source data

Extended Data Fig. 3. Cross-validation of genetic datasets.

a) The background rate of genetic events in the piggyBac and human mutation datasets have opposite correlations to expression levels. Genes are binned into equal-sized bins based on their expression level. On the x axis, the bins are plotted on the mean expression of their genes. On the y axis, the bins are plotted on the ratio of total observed/total expected events for the bin. p-value: two sided Spearman correlation test. b) Mean conservation of mutated nucleotides for genes not identified in the human mutation datasets, for genes identified only in the human mutations dataset and for genes identified in both the human mutations and piggyBac datasets. q-values: two-sided Mann-Whitney test with FDR correction, both compared to non-significant genes c-e) Hematoxylin-eosin stains of individual RPR2TC mice induced with lentiviral vectors carrying a non-targeting sgRNA or sgRNAs targeting Reln. Representative of 7, 5 and 5 mice, respectively. f) The mean area of tumors identified in mice induced with sgRNAs targeting Reln is significantly larger than the area of tumors induced with the non-targeting sgRNA. N = 7 mice for non-targeting sgRNA and sgReln-1, n = 5 mice for sgReln-2. q-values: Mann-Whitney test with FDR correction, both compared to sgNT controls. g) The size of individual tumors from the mice in f is significantly greater in mice induced with sgRNAs targeting Reln. q-values as in f. h) Force-directed graph of gene ontology analysis, showing gene sets enriched in both the piggyBac dataset and the analysis of human genetic data. Most gene sets are related to synaptic and neuronal functions (light blue). Source data

Extended Data Fig. 4. Expression of synaptic gene sets in SCLC.

a-h) Selected genes highly expressed in SCLC. The expression levels of individual SCLC samples are shown on the left of each panel. The median expression levels in cancer types included in TCGA and Neuroblastoma as positive control are depicted in the middle. The median expression levels of healthy tissues are on the right. a, b, c, d) The expression levels of TOP2A, CCNE2, RRM2 and UBE2C, representative of genes involved in cell-proliferation, are higher in SCLC than in any other cancer or healthy tissue. e, f, g, h) The expression levels of NRXN1, NLGN1, DCC and RELN, representative of synaptic and neuronal genes, are higher in SCLC than in most other cancers and tissues. i) Leiden clustering of snRNA-seq data from six murine tumors derived from Rb1^fl/fl;Trp53^fl/fl mice. j) The leiden clusters from panel i show markers of SCLC cells (Calca, Chga, Syp, Ncam1) or of one of four broad cell types expected in the lung (Ptprc for immune cells, Col1a2 for stromal cells, Sftpb for epithelial cells and Cdh5 for endothelial cells). k) Visualization of cell types based on markers identified in panel j. l) Genes within the Synaptic Membrane GO term are enriched in the cancer cells. m) Genes within the Glutamatergic Synapse GO term are enriched in the cancer cells. n) Comparison of murine SCLC cells to other lung cell types revealed an enrichment in neuronal and cell proliferation GO terms, with striking resemblance to the analysis of bulk human RNA-seq data (Fig. 1c). o) Comparison of SCLC cells to other cell types in published human lung scRNA-seq data revealed an enrichment in neuronal and cell proliferation GO terms, with striking resemblance to the analysis of bulk human RNA-seq data (Fig. 1c). Source data

Extended Data Fig. 5. Nerve fibers in the SCLC microenvironment.

a-h) Confocal images of lung cryostat sections of RP mice. L: lumen of the airways. E: airway epithelium. a) Intraepithelial VGluT1+ nerve terminals (arrowheads) branch between the CGRP+ PNECs. b) Intraepithelial P2X3+ nerve terminals (arrowheads) protruding between the CGRP+ (green) neuroendocrine cells of a NEB. c) GAP43+ nerve fibers (arrows) branch and protrude (arrowheads) between the CGRP+ PNECs. d) A GAP43+ nerve fiber (arrow) branches (arrowheads) between the CGRP + SCLC cells (green). CGRP+ nerve fibers (open arrows) are seen close to the base of the tumor. e) Subepithelial SYP+ and CGRP+ nerve terminals (arrows) innervate a NEB. Remarkable is that the subepithelial area adjacent to a large tumor appears devoid of nerve fibers. f) Small tumor (ST) from an RPC mouse, with no visible innervation from VGluT1+ fibers. Varicose CGRP+ fibers are visible below the tumor (open arrowheads). g) Large tumor (LT) surrounded by varicose CGRP+ and substance P+ (SP) nerve fibers. h) SYP+ (arrows) and CGRP+ (open arrows) nerve fibers can be seen in the epithelium at the base of a small tumor (ST). i) Electron micrographs showing a cancer cell surrounded by long axon-like fibers near the periphery of a tumor in the lung of an RP mouse. j, k) Magnifications showing the presence of enlarged structures along identified fibers (yellow pseudocolor) containing multiple vesicles and mitochondria (M) near the cancer cell (red pseudocolor). l-p) DAB staining of biopsies from three SCLC patients. All sections are counterstained with hemalum. l) NF-H-positive nerve fibers near an intratumoral vessel in the biopsy from the first patient. m, n) NF-H-positive nerve fibers at the borders of a SYP-positive tumor in a biopsy from the second patient. o, p) NF-H-positive nerve fibers at the borders of a SYP-positive tumor in a biopsy from a third patient.

Extended Data Fig. 6. Cancer-to-neuron contacts in vitro.

a) Co-culture of murine cortical neurons (immunolabeled against MAP2) and SCLC cells (COR-L88, expressing DsRed) showing the appearance of dense VGLUT1-positive puncta onto SCLC cells contacted by neuronal terminals b-d) Different views of a 3D-reconstruction of 3D-STED for co-cultures immunolabeled against axonal marker SMI-312, mNeonGreen to mark SCLC cells, dendritic marker MAP2, and postsynaptic marker HOMER1, showing that that the contacts on cancer cells are predominantly axonal. Representative of 3 experiments. e) Co-culture of human iPSC-derived cortical neurons and SCLC cells (COR-L88, expressing tdTomato), immunostained for the pre- and post-synaptic markers BSN and HOMER1. Right panels show a single confocal stack (top) and 3D reconstruction (bottom) of an SCLC cell contacted by a GFP-positive axonal fiber (white arrowheads) exhibiting BSN and HOMER1 co-localizing puncta (yellow arrowheads) located outside and inside the SCLC cell surface. f) Confocal overview of SCLC cells (mNeonGreen + , shown in white) co-cultured with murine nodose ganglia. g) A detailed view of the boxed region from panel f, followed by individual magnified regions, which indicate the arrangements of VGluT1 (presynaptic, neuronal) and HOMER1 (postsynaptic, within SCLC cell) molecules. h) Line scans were drawn automatically across HOMER1 spots, starting in their intensity maxima, and moving towards the periphery. The signal drops, as expected; a similar drop is seen in the VGluT1 signal, confirming their close apposition (N = 4 independent experiments, n = 782 line scans. Colocalization tested using a two-sided Pearson correlation test. Error bars: standard error of the mean). i) Correlative intensity scatter plot of SCLC mNeonGreen signal vs HOMER1 signal (N = 4 independent experiments, N = 782) indicates that a substantial proportion of the HOMER1-marked spots are formed on SCLC cells, and therefore show a measurable mNeonGreen signal (62.02%). Source data

Extended Data Fig. 7. Cancer-to-neuron contacts in vivo.

a) Confocal imaging of grafted DsRed-expressing SCLC cells in the hippocampus of a Thy1-GFP mouse (top-left), depicting GFP-positive fibers contacting SCLC cells in the tumor periphery (lower-right). On the right and below are orthogonal views on a point of contact between a putative axonal bouton and a DsRed/HOMER1 double-positive punctum in the SCLC cell (arrowheads). b) 3D-STED image of a lung section immunolabeled against the presynaptic marker VGluT1, the postsynaptic marker HOMER1, and an axonal marker (SMI312/SMI311 epitopes). The right panels show magnifications of putative synapses on cancer cells in the marked regions. c) Automatic line scans from the intensity maxima of HOMER1 spots towards the periphery. The signal drops and a similar drop is seen in the VGluT1 signal (N = 3 independent experiments, n = 609 line scans, of which 213 represented putative synapses. Two-sided Pearson correlation test. Error bars: standard error of the mean). d) Correlative intensity scatter plot of SCLC mNeonGreen signal vs HOMER1 signal (N = 3 independent experiments, n = 609 measurements) indicates that most HOMER1 spots in these regions are within SCLC cells. e) Quantification of synapses contacting tdTomato-positive cancer cells in brain allografts. For each mouse (n = 3), 90-96 perimeters in 12-14 consecutive ultrathin sections were examined, for a total of 280 cell perimeters. f) CLEM of COR-L88 SCLC cells (expressing DsRed) co-cultured with cortical neurons. The left panels depict the registered overlay between fluorescent and EM images. The third panel shows the electron tomogram of a synapse, with a presynaptic bouton (yellow pseudocolor) filled with vesicles, contacting the cancer cell (red pseudocolor). The right panels show a 3D reconstruction of the tomogram (250 nm thick), depicting cancer cell (red), axonal bouton (yellow) with vesicles (white), and vesicles located within 20 nm from the plasma membrane (PM) (green). Source data

Extended Data Fig. 8. Electrophysiology of SCLC cells.

a) Example of a patched DsRed-expressing SCLC cell (COR-L88) under whole-cell configuration in cortical neuron-SCLC co-cultures. b) Whole-cell, voltage-clamp traces of sPSCs in SCLC cells (COR-L88) in the presence or absence of neurons. c) Quantification of sPSC frequency in co-culture in the presence or absence of the indicated blockers (TTX, CNQX, D-AP5, Riluzole and Bicuculline) (n = 7-30 cells per condition). All conditions are compared to untreated co-cultures. q values: two-sided Mann-Whitney with FDR correction d-g) Whole-cell voltage-clamp traces of H524 cells d) Traces recorded at three different voltages (−70 mV, 0 mV, and +40 mV) in mono-culture. e) Traces recorded at +40 mV in co-culture with cortical neurons. The synaptic events (red stars and numbers) can be completely abolished by the application of the NMDA receptor blocker D-AP5 and display a long decay time lasting several hundred milliseconds. f) Traces recorded at −70 mV and 0 mV in co-culture with cortical neurons. Note the occurrence of synaptic events at 0 mV, indicating a GABA-A-mediated chloride inward current. g) Traces recorded at +40 mV in co-culture with Channelrhodopsin 2-eYFP expressing (ChR2-eYFP, green) cortical neurons after a short blue light pulse (5 ms). Note the partial decrease in event amplitude during NMDA receptor blockade with D-AP5 (orange), followed by complete abolishment after additional GABA-A receptor blockade with Gbz (lower trace). h) Whole-cell, voltage-clamp recording in an acute hippocampal slice of grafted DsRed-expressing murine SCLC cells. i) Quantification of sPSCs in grafted cancer cells in acute slices in the absence or presence of the indicated blockers (TTX, CNQX, D-AP5 and Bicuculline). All conditions are compared to untreated slices. q values: two-sided Mann-Whitney with FDR correction, n = 8-17 cells per condition. Source data

Extended Data Fig. 9. RABV-tracing of SCLC cells to presynaptic neurons.

a) RABV-GFP-based tracing of neurons monosynaptically connected to DMS273 SCLC cells expressing DsRed. Right panels show enlarged views of the boxed area containing double-positive starter SCLC cells (arrowheads). b) 3D reconstruction of double-positive starter cells in a cluster of DsRed-expressing SCLC cells (COR-L88) following RABV-GFP-based tracing. c) Magnification of the panel boxed in b, showing the profuse expression of VGluT1-positive puncta in GFP-positive neuronal fibers (yellow arrowheads) contacting starter SCLC cells. d) Time-lapse of RABV-traced neurons in neuron-SCLC co-cultures over 48 h. Selected frames at the indicated time points show the initial presence of starter cancer cells (double-positive for the retrovirally-encoded DsRed and the RABV-encoded GFP, yellow arrowheads), which proliferate over time, and the emergence of GFP+ neurons at 48 h (white arrowheads). e) Example of RABV-GFP-based tracing of morphologically identified inhibitory GABAergic neurons located in the stratum oriens (SO) and pyramidale (SP) of CA1, following transplantation of G-TVA-expressing murine SCLC cells (dashed area). Right panels show zooms of the boxed areas depicting identified GFP+ neurons (1) and starter SCLC cell (2), contacted by varicosities of a passing axon (arrowheads). SR, stratum radiatum. f) Example of RABV-GFP-based tracing following transplantation of TVA-only-expressing murine SCLC cells (dashed area), showing the virtual absence of GFP-positive presynaptic neurons. Right panels show zooms of the boxed area depicting an identified DsRed/GFP double-positive SCLC cell (arrowhead).

Extended Data Fig. 10. neuron-promoted SCLC proliferation and Grm8.

a) Growth of SCLC cell lines monitored via live cell imaging under different conditions. Each dot represents an individual well. All conditions are compared to the growth in co-culture with cortical neurons. q-value: two-sided Mann-Whitney test with FDR correction. n ≥ 20 wells / condition, n ≥ 4 neuron batches. Red q-values indicate faster growth than neuronal co-cultures. b) Growth of NSCLC cell lines monitored via live cell imaging in mono-culture or co-culture. Each dot represents an individual well. p-values: two-sided Mann-Whitney test. n ≥ 20 wells / condition, n ≥ 4 neuron batches.c, d) Individual wells containing COR-L88 SCLC cells in mono-culture (c) or in co-culture with nodose ganglia (d). e) Quantification of the growth of SCLC cell lines via live cell imaging with and without nodose ganglia. Each dot represents an individual well. p-value: two-sided Mann-Whitney test as in b. n = 4-29 wells/condition, n ≥ 4 individual ganglia. f) SCLC samples are separated into classic and variant subtypes based on the expression of neuroendocrine features. g) SCLC samples of the SCLC-A and SCLC-N subtypes express higher level of genes included in the GO term Glutamatergic Synapse. h) The expression of GRM8 is higher in SCLC than in most other cancers and tissues and is especially high in classic SCLC with strong neuroendocrine features. i) GRM8 protein with annotations from the UniProt Knowledgebase. Mutations in SCLC samples are shown as a lollipop chart. j) Transposon insertions identified in Grm8. k) UMAP plot of published human SCLC and normal lung scRNA-seq. The cells are grouped into differentiation groups. l) GRM8 is specifically expressed in SCLC cells from panel k. m) UMAP plot of snRNA-seq samples from murine RP tumors, characterized in Extended Data Fig. 4. Grm8 is specifically expressed in SCLC cells. Source data

Extended Data Fig. 11. Response of SCLC tumor-bearing mice under anti-glutamatergic treatment.

a-f) Representative MRI scans of tumor-bearing Rb1^fl/fl;Trp53fl/fl mice under different treatments. Tumors are pseudocolored in red. a, b) Mouse treated with PBS, showing a large increase in tumor size after one month. c, d) Mouse treated with DCPG, showing a minor increase in tumor size after one month. e, f) Mouse treated with riluzole, showing a minor increase in tumor size after one month. g) Median tumor burden for RP mice treated with riluzole (n = 12), DCPG (n = 12) or vehicle controls (n = 33). q-values: two-sided Mann-Whitney test with FDR correction. h) Waterfall chart, showing the best response of individual tumors, grouped by mouse. The mice are sorted based on the total best response. i) Time required for tumors to reach a size five fold greater than the size at inclusion for mice treated with riluzole (n = 12), DCPG (n = 11) or the relative controls (n = 32). q-values: two-sided Mann-Whitney test with FDR correction. j) Best response achieved throughout treatment, calculated based on the total tumor burden for each mouse for mice treated with DCPG (n = 12), riluzole (n = 12) or the relative controls (n = 28). q-values: two-sided Mann-Whitney test with FDR correction. k) Best response of individual tumors from RP mice induced with CGRP-Cre. The mice were treated with DCPG (31 tumors from 18 mice), riluzole (19 tumors from 13 mice), or the relative controls (23 tumors from 17 mice for PBS plus 20 tumors from 12 mice for riluzole vehicle). l) Survival of RP mice induced with CGRP-Cre. Riluzole treatment (n = 14) results in significantly longer survival compared to control mice (n = 18 for PBS plus n = 14 for riluzole vehicle). The benefit provided by DCPG (n = 20) is not statistically significant. q-values: two-sided Mann-Whitney test with FDR correction. Source data

Extended Data Fig. 12. Combination treatment with chemotherapy and anti-glutamatergc drugs.

a-f) Representative MRI scans of tumor-bearing Rb1^fl/fl;Trp53fl/fl mice under different treatments. Tumors are pseudocolored in red. a, b) Mouse treated with etoposide and cisplatin (EC), showing the increase in tumor size after two months. c, d) Mouse treated with EC and DCPG (ECD), showing the increase in tumor size after two months. e, f) Mouse treated with EC and riluzole (ECR), showing a stable disease after two months. g) Median tumor burden for RP mice treated with EC (n = 11), ECR (n = 13) or ECD (n = 10). q-values: two-sided Mann-Whitney test with FDR correction. h) Waterfall chart, showing the best response of individual tumors, grouped by mouse. The mice are sorted based on the total best response. i) Time required for tumors to reach a size five fold greater than the size at inclusion for mice treated with EC (n = 9), ECD (n = 9) or ECR (n = 10). q-values: two-sided Mann-Whitney test with FDR correction. j) Best response achieved throughout treatment, calculated based on the total tumor burden for each mouse for mice treated with EC (n = 10), ECR (n = 9) and ECD (n = 10). q-values: two-sided Mann-Whitney test with FDR correction. Source data