Intrinsic electrical activity drives small-cell lung cancer progression

Abstract

Elevated or ectopic expression of neuronal receptors promotes tumour progression in many cancer types^1,2; neuroendocrine (NE) transformation of adenocarcinomas has also been associated with increased aggressiveness³. Whether the defining neuronal feature, namely electrical excitability, exists in cancer cells and impacts cancer progression remains mostly unexplored. Small-cell lung cancer (SCLC) is an archetypal example of a highly aggressive NE cancer and comprises two major distinct subpopulations: NE cells and non-NE cells^4,5. Here we show that NE cells, but not non-NE cells, are excitable, and their action potential firing directly promotes SCLC malignancy. However, the resultant high ATP demand leads to an unusual dependency on oxidative phosphorylation in NE cells. This finding contrasts with the properties of most cancer cells reported in the literature, which are non-excitable and rely heavily on aerobic glycolysis. Additionally, we found that non-NE cells metabolically support NE cells, a process akin to the astrocyte-neuron metabolite shuttle⁶. Finally, we observed drastic changes in the innervation landscape during SCLC progression, which coincided with increased intratumoural heterogeneity and elevated neuronal features in SCLC cells, suggesting an induction of a tumour-autonomous vicious cycle, driven by cancer cell-intrinsic electrical activity, which confers long-term tumorigenic capability and metastatic potential.

Fig. 1. Electrically active NE cells generate spontaneous and evoked calcium waves.

a, SCLC classification and models used in this study. Matched-colour stars depict paired cell lines/PDX models originally derived from the same tumours. b,c, Patch-clamp recordings in paired NE (b) and non-NE (c) mSCLC cell lines (top) or PDX models (bottom). Voltage responses to current steps of several amplitudes, as indicated. Graphs shown representative of n = 9 (MGH1505-1A_fl), 6 (MGH1505-1A_adh), 30 (AD984LN_fl) and 6 (AD984LN_adh) cells. d, Excitability (Methods) of paired NE and non-NE cells originally derived from the same hSCLC parental line (NCI-H69) and PDX model (MGH1505-1A). NCI-H69: n = 15 NE and 14 non-NE cells; MGH1505-1A: n = 4 NE and 6 non-NE cells. e, Resting membrane potentials (V_rest) of SCLC cells immediately following breakthrough into the whole-cell mode in current clamp. The average V_rest values for each cell line are indicated as individual points; n = 7 NE cell lines and 4 non-NE cell lines (Extended Data Fig. 2j). f, Evoked, propagating calcium waves in GCaMP6m-expressing NE cells (AD984LN_fl). A patch pipette filled with KCl solution was pressed against a cell membrane within a cluster (left), and a train of extracellular voltage-stimulating pulses was applied, initiating a slowly propagating wave of intracellular calcium (right). Representative of n = 7 separate cultures. g,h, Simultaneous, paired patch-clamp recording in adjacent NE cells (g) did not detect synaptic-dependent currents or gap junction coupling (h). Representative of n = 6 pairs of AF3062C cells and 4 pairs of AF1165 cells. i, Propagating waves can be initiated by extracellular voltage-stimulating pulses given through patch pipette to the first cell (1) and traverse cell-free regions in the culture. Representative of n = 5 separate cultures. Mean ± s.e.m. shown in all graphs. Two-tailed unpaired t-test applied in d and e. Scale bars, 200 ms (b,c), 50 μm (f,i), 40 μm (g), 500 ms (h). Schematic in a was created using BioRender (https://biorender.com).

Fig. 2. Cholinergic signalling triggers depolarization and initiates calcium transients in NE cells.

a, nAChR subunit expressions in NE (n = 5) and non-NE (n = 3) cell lines. b, 100 µM CCh application to NE cells (AF1165). Top, depolarizing current-clamp responses, representative of n = 10 cells. Bottom, inward currents in two cells. Holding potential = −70 mV. Representative of n = 5 cells. c–f, In vivo labelling of calcitonin gene-related peptide (CGRP)-expressing cells (c). Representative images of PNEC/NEB (n = 10 from 3 mice) (d), early/small (n = 7 tumours per 3 mice) and advanced/large SCLC (n = 10 tumours per 4 mice) (e) with quantifications (f). g–i, Co-staining of β3-tubulin and VAChT in PNEC/NEB (n = 16 from 3 mice), normal lung (n = 3 mice) and early-stage (n = 10 tumours per 3 mice) and late-stage (n = 10 tumours per 4 mice) SCLC. Representative images (g) and orthogonal projection in early SCLC (h). Nuclei: DAPI (blue). Quantification of the percentage of VAChT⁺ fibres (i). j,k, Axons extending into the core of early/small SCLC (j). Arrowheads indicate terminal buttons of nerve ending (k). Representative of n = 10 tumours per 3 mice. l, β3-Tubulin staining in an advanced SCLC tumour (left) and quantification of β3-tubulin⁺ in tdTomato⁺ PNEC/NEB (n = 11 from 3 mice), early (n = 9 tumours per 5 mice) and advanced (n = 14 tumours per 5 mice) SCLC cells (right). m, Ex vivo imaging of GCaMP6f-expressing mSCLC. n, Quantification of m. n = 11 PNECs/NEBs and 3 SCLC tumours (3 mice per group). Mean ± s.e.m. shown in all graphs. Two-tailed unpaired t-test applied in a (top) and n. Mann–Whitney test applied in a (bottom). Ordinary one-way analysis of variance (ANOVA) applied in f, i and l. Scale bars, 8 μm (k (right)), 10 μm (l (right),k (left)), 20 μm (d,m (bottom),g (right)), 30 µm (h), 50 μm (d,e (top),l (left),j), 100 μm (e (bottom),g (left and middle),m (top)). NS, not significant. Schematics in c and m were created using BioRender (https://biorender.com). Source data

Fig. 3. High ATP demands of NE cells increase OXPHOS dependency.

a, Glucose catabolism and OXPHOS. b,c, Glycolysis index (lactate secretion/glucose consumption) in paired SCLC NE and non-NE cell lines (n = 2 pairs; 3 replicates per cell line) (b) and in SCLC NE (n = 4) and LUAD/PDAC (n = 4) cell lines (c) (Extended Data Fig. 7a). Cells cultured in DMEM with 2% fetal bovine serum (FBS). d, Three genes encoding mitochondrial-related proteins (see Methods for the selection criteria) show SCLC-preferential vulnerabilities (n = 4 cell lines per group). e,f, Rotenone sensitivity (CellTiter-Glo (CTG) assay) of SCLC NE (n = 4) and LUAD/PDAC (n = 3) cell lines (e), as well as of paired SCLC NE and non-NE cell lines (n = 2 pairs; pair 1: AD984LN and pair 2: AF3291LN) (f) cultured in DMEM without pyruvate; n = 3–4 technical replicates per line. g, OCR in NE (n = 4) and non-NE (n = 2) SCLC cell lines; n = 6 technical replicates per line. h, Radioactive 2-deoxyglucose uptake assay (n = 2 pairs of SCLC NE and non-NE cell lines; 3 technical replicates per condition). i, Sulforhodamine B (SRB) assay of three independent NE cell lines in different conditioned media: NE conditioned media and two independent non-NE conditioned media with and without dialysis. Dotted red lines: cell number in NE conditioned media; n = 3 cell lines; 4 technical replicates per line. j, Representative gas chromatography–mass spectrometry results of conditioned media from paired NE and non-NE cell lines; n = 3 technical replicates per line. k, Western blot for MCT4 in NE (n = 5) and non-NE (n = 3) cell lines. l,m, Changes in lactate (l) and glucose (m) concentrations in media after incubation of SCLC NE cells with different concentrations of lactate; n = 3 technical replicates. Mean ± s.e.m. shown in all graphs. Two-way ANOVA; Sidak’s multiple comparison test applied in b, h and i. Two-tailed unpaired t-test applied in c. Schematic in a was created using BioRender (https://biorender.com).

Fig. 4. Metabolite support from non-NE cells sustains the ATP demand of NE cells.

a, Proposed metabolite shuttle model. b–d, Calcium imaging of GCaMP6m⁺ NE cells, cultured alone or with non-NE cells. Calcium signal traces from the ten most active NE cells in each condition (arrowheads: calcium spikes) (b) and quantification of the percentage of NE cells with active calcium signalling (c) and when treated with diclofenac (0.5 mM) (d); n = 6–9 fields of view examined over 3 independent experiments, and each dot represents a field of view. e–h, Patch-clamp recording of NE cells (AF1165). Resting membrane potential (V_rest) (e) and predicted ATP consumption/demand (f) in different culture conditions: starvation (n = 15, 31 and 7 cells per time point); with 5 mM glucose (n = 14, 30 and 6 cells per time point); with 10 mM lactate (n = 14, 29 and 7 cells per time point). V_rest (g) and excitability (h) when incubated overnight in lactate alone (g, n = 5 cells; h, n = 6 cells) or with SR-13800 (5 μM) (n = 10 cells). i, Western blot of HES1 and SOX1 in SCLC cell lines. j, Representative images and quantifications of SOX1 (n = 25 PNEC or SCLC, respectively; 6 mice per group) and MCT4 (n = 19 PNEC/SCLC; 6 mice per group) staining in mPNEC and small mSCLC tumours. k, Representative staining of SOX1 and MCT4 in mSCLC (PRP130) and hSCLC tumours; n = 5 mice (1 tumour per mouse) and 44 patients (1 tumour per patient). Pearson’s coefficient from Extended Data Fig. 8j. Red area: negligible correlation. l, Calcium transient quantification in lung slices from PRP130-Salsa6f mice treated with either control (culture medium) or diclofenac (0.5 mM); n cells per slice reported in graph from two independent imaging sessions. Mean ± s.e.m. shown in all graphs. Kruskal–Wallis test with Dunn’s multiple comparison test applied in c (middle, right), e and f. Two-tailed Mann–Whitney test applied in c (left), d, g, h and j. Two-tailed unpaired t-test applied in l. Scale bars, 50 µm (j (SOX1 and MCT4 (left)),k), 10 µm (j (MCT4, right)). Schematic in a was created using BioRender (https://biorender.com). Source data

Fig. 5. Electrical activity promotes SCLC progression.

a–e, Colony formation assays. a, Representative images after blue light stimulation in parental or ChR2⁺ NE cells (AF3062C) (top) and their quantification (bottom). b–e, TTX treatment (1 µM) in mSCLC NE (n = 3) and non-NE (n = 2) cell lines (b,c) and other excitable (n = 5) and non-excitable (n = 2) cancer cell lines (d,e). Representative images (b,d) and quantification (c,e). f–h, Liver metastasis assays. f, Experimental design (left) and normalized liver weight. NE cells were either untreated or pretreated (pre-Tx) with 1 µM TTX. AF3062C: n = 8 (untreated) and 7 (pre-Tx) mice. AD984LN_fl: n = 4 mice per group. g,h, Quantification of tumour area (g), representative Ki67 (h, left) and cleaved caspase 3 staining (h, right) of the same liver lobes from the control (n = 35 tumours per 4 mice) and pre-Tx (n = 29 tumours per 3 mice). i, Colony formation in AF3062C NE cells either untreated (control) or pretreated for 24 h with TTX before seeding. Representative images (left) with quantification (right). j–m, Chemogenetic suppression of NE electrical activity in liver metastasis assay. Mice were transplanted with AD984LN_fl cells ± iPSAM⁴, treated with vehicle or uPSEM817 (n = 5 mice per group) (j). Bioluminescent imaging (k,l) and survival analysis (m). i.p., intraperitoneal. n, (α3)₂(β4)₃ nAChR subunits (left) are encoded by CHRNA3 and CHRNB4; their impact on prognosis is shown in the Kaplan–Meier curves (right) (n = 104 patients with SCLC from a published cohort); 95% confidence intervals shown. For all colony formation assays, n = 3 technical replicates, repeated at least twice per cell line in n = 2–3 cell lines. Mean ± s.e.m. shown in all graphs. Two-way ANOVA and Sidak’s multiple comparison tests applied in a and c. Two-tailed unpaired t-test applied in e and g–i. Two-tailed Mann–Whitney test applied in f and l. log-rank test applied in m and n. Schematics in f, j and n were created using BioRender (https://biorender.com). Source data

Extended Data Fig. 1. Generation of matched PDX-derived SCLC cell lines.

(a) Schematic illustration of the workflow used to generate PDX-derived SCLC cell lines with two distinct morphologies from the same tumor. As previously described, patient MGH1505 was treated with standard first-line chemotherapy with an excellent response, and at first relapse received an experimental therapy to which he also responded. 461 days into his treatment course, after his second relapse and just before the beginning of 3rd-line paclitaxel, live circulating tumor cells (CTCs) were isolated from a routine blood draw and injected into the right flank of an NSG mouse. After a prolonged latency, a CTC-derived xenograft, MGH1505-1A, emerged and was resected 162 days after injection. Fragments of MGH1505-1A were passaged serially to two NSG mice, with significantly shorter latencies (50–70 days). The third passage was into an NSG-GFP mouse, which enabled identification of murine stromal cells by GFP⁺ signal. After reaching a size of 1500 mm³, this xenograft was manually dissociated for cell line derivation. Numbers next to the arrows represent the latency times between two events (in gray) or the time of palbociclib treatment (in blue). (b - f) Live images of cell cultures at different time points during the cell line derivation process, as annotated in the schematic in (a). GFP⁺ cells are murine, while GFP⁻ cells are human SCLC cells. (b) Image of the culture 56 days after tumor dissociation showing a mixed culture consisting of tightly and loosely adherent cells. Some tightly adherent cells are GFP⁻, indicating that they are tumor cells. After this image was taken, the culture was treated with palbociclib for 7 days to deplete the GFP⁺ murine cells. (c, d) Separated suspension (c) and adherent (d) cultures imaged 28 days after initial palbociclib treatment. Only a few GFP⁺ cells remain, and these were fully depleted by palbociclib treatment for an additional 28 days. (e, f) Cell lines with two distinct morphologies, floating clusters (e) and adherent (f), were generated from a single tumor and were free of any GFP⁺ murine cells. Scale bars: 200 µm. Schematic in a was created using BioRender (https://biorender.com).

Extended Data Fig. 2. Transcriptional and electrophysiological characterization of NE and non-NE cells.

(a) Heatmap of differentially expressed genes in mSCLC cell lines (n = 5 NE cell lines and 3 non-NE cell lines). (b) Clustering analysis of NE and non-NE cell lines showing separate clustering, even when two pairs of NE and non-NE cell lines originally derived from the same tumors were compared. (c, d) GSEA of rank-ordered differential gene scores of non-NE versus NE mSCLC cell lines showing an enrichment of NE with Hes1^negative mSCLC tumor cells (c, top) and GO_Synaptic_Signaling (d, top); and non-NE enrichment in Hes1^high (c, bottom) and GO_Extracellular_Structure_Organization pathways (d, bottom). Benjamini-Hochberg method applied for correction of multiple testing. False Discovery Rate (FDR) shown. (e) Families of representative current-clamp recordings from a mouse cortical interneuron. (f) Families of representative current-clamp recordings from paired hSCLC NE (top) and non-NE (bottom) cell lines originally derived from the same parental cell line NCI-H69. Three different current step amplitudes as indicated. Illustrated voltage responses are representative of 20 (NCI-H69_fl) and 15 (NCI-H69_adh) cells. (g) Families of current-clamp recordings from mSCLC NE (top) and hSCLC NE (bottom) cell lines. Membrane potential responses to indicated current steps are shown. For each cell line, the illustrated voltage responses are representative of 20 cells patched. Note a rebound spike on the termination of the hyperpolarizing step for AF1165, commonly observed in some NE SCLC cells. (h) Voltage-clamp currents in response to a series of voltage steps (middle panel) show the presence of inward fast Na⁺ and maintained outward K⁺ currents in NE, but not in non-NE cells. Results were consistent in all mSCLC NE (n = 5) and non-NE (n = 2) cell lines tested. (i) Peak early and late currents during depolarizing voltage-clamp steps as a function of membrane potential in NE cells (AF1165). TTX: inward current during application of 1 μM tetrodotoxin. The results are representative of 3 independent NE lines: n = 3 (AF1165), 4 (AD984LN_fl), 5 (AF1281m1) cells. (j) Resting membrane potentials (V_rest) of SCLC cells immediately following breakthrough into the whole-cell mode in current-clamp. Individual lines are shown in the left panel: n = 71, 12, 26, 13, 6, 6, 8, 6 cells/line, from left to right, respectively. (k) Left: Calcium signal monitored by Cal-520 indicator (expressed as fluorescence increment relative to baseline, ΔF/F) in the cell body (top) while simultaneously recording membrane potential (middle) and injecting a series of increasing current steps (bottom). Right: Expanded-time view of the action potential responses indicated in the upper panel. Representative of recordings in 4 AD984LN_fl cells. (l) Linear relationship between the number of spikes in each response and the amplitude of the calcium increment relative to k. (m) Non-invasive cell-attached recording of spontaneous action potentials in an AD984LN_fl mSCLC NE cell. Middle and bottom panels: expanded time scale views as indicated by dashed rectangles above. In these extracellular recordings, capacitive currents reflecting action potential upstroke are plotted downwards and repolarization upwards. Spontaneous firing has been observed consistently in more than 10 cells per cell line and across at least three different NE lines examined. (n, o) Lightsheet microscope imaging of 3D-cultured GCaMP6m⁺ AD984LN_fl cells. 12 planes, (side view shown at left), were imaged every 2.8 s (n). Scale bar: 50 µm. Representative of 12 tumoroids in 6 different 3D cultures. (o) Maximum intensity projections across all planes show the propagation of a spontaneous wave starting from the invasive outgrowth at the left, which shows ongoing rapid activity (location 1) (middle and right panels). (p) Membrane capacitance increases following a train of depolarizing voltage-clamp steps in NE cells as monitored by current phase shift in response to a 1 kHz sinusoidal voltage command signal, as illustrated in (q). Representative of recordings in n = 8 AF1165 cells. Schematic in q was created using BioRender (https://biorender.com).

Extended Data Fig. 3. Cholinergic signaling in SCLC.

(a, b) Boxplots of normalised mRNA expression values of CHRNA3 (a) and CHRNAB4 (b) in tumour and normal adjacent tissue (NAT) samples from a previously published cohort of SCLC patients (n = 112). Wilcoxon matched-pairs signed rank two-sided test was applied. Boxplots show the median (line) and IQR (interquartile range), with whiskers no more than 1.5× IQR. (c) Boxplot of differentially expressed nAChR subunits between a group of n = 43 hSCLC NE (SCLC-A, SCLC-N) and n = 11 non-NE (SCLC-Y, SCLC-P) cell lines. Significance was evaluated with the two-tailed Mann-Whitney test. Boxplots show the median (line) and IQR (interquartile range), with whiskers no more than 1.5× IQR. (d) Left: Fluorescence responses in multiple cells in a field of view to carbachol superfusion (CCh, 40 μM). Results representative of 5 cultures of AD984LN_fl-GCaMP6m cell line. Right: Membrane channel currents in low-noise whole-cell patch-clamp recordings activated by CCh (40 μM), showing channels with amplitudes and lifetimes characteristic of nAChR. Representative of 5 recordings from AF3062C NE cells. (e) Currents during a slow depolarizing ramp of membrane before, during and after application of CCh (100 μM). Representative of recordings from 5 cells of the AF1165 NE cell line. (f) Comparison of current level measured at −70 mV during ramps as in (e), for n = 5 cells shows significant activation of inward current by CCh. One-way ANOVA was performed to determine statistical significance, (left*) p = 0.0391, (right*) p = 0.0308, ns=non-significant. Mean ± SD shown.

Extended Data Fig. 4. Single channel characterization of SCLC innervation.

(a–e) Single channel images of immunostaining of PNEC (a, from Fig. 2d, upper panel; b, from Fig. 2d, lower panel) and SCLC (c, from Fig. 2e) innervation. (a, b) Normal pulmonary neuroendocrine cells marked by CGRP (magenta) are innervated by β3 tubulin positive neurons (yellow). (b) A fraction of these nerve axons was CGRP⁺, suggesting their sensory origin and relative axon density quantification (d). Nuclei are stained by DAPI (blue). Scale bars: (a) 50 µm; (b) 20 µm; (c) 50 µm (top); 100 µm (bottom). (e) Correlation plot between cumulative axon length around the tdTomato⁺ PNEC/SCLC and the perimeter of tdTomato⁺ cells. Normal airways: n = 10 PNEC/NEB (3 mice); early SCLC: n = 7 tumors (3 mice); late SCLC n = 10 tumors (4 mice). (f) Single channel images of immunostaining from Fig. 2g, left and middle panels. Scale bars from top to bottom: 50, 50, 30, 100 µm. PNEC/NEB: n = 16 PNEC/NEB (3 mice); normal lung: n = 3 mice; early SCLC: n = 10 tumors (3 mice); late SCLC: n = 10 tumors (4 mice). (g, h) Single channel images of the immunostaining zoom-ins (right panels) from Fig. 2g (g, from the early/small SCLC; h, from the advanced/large SCLC). Scale bars: 20 µm (g), 50 µm (h). Early SCLC: n = 10 tumors (3 mice); late SCLC: n = 10 tumors (4 mice). (i) Single channel images of the immunostaining zoom-ins (right panels) from Fig. 2h. Scale bars: 30 µm. n = 10 tumors (3 mice). (j) Single channel images of immunostaining of early SCLC innervation from Fig. 2j. Scale bars: 50 µm (left panel), 10 µm (right panel). n = 10 tumors (3 mice). (k) 3D reconstruction of axons (β3-tubulin: yellow) inside a small SCLC (magenta). Scale bars: 8 µm. n = 10 tumors (3 mice). Mean ± SEM shown in all graphs. Ordinary one-way ANOVA applied in d. ns: not significant. Source data

Extended Data Fig. 5. Expression levels of cholinergic markers across human cancers.

(a, b) Pan-cancer analysis of different cell lines from various human tumor types obtained from DepMap. Human SCLC cell lines (shown as “Lung Neuroendocrine Tumor” pointed by a red arrow on the boxplots) are among the top 4 or top 3 of the highest CHAT- and SLC18A3-expressing cancer types, respectively. CHAT encodes choline acetyltransferase (ChAT), which is a key enzyme in acetylcholine synthesis; SLC18A3 encodes vesicular acetylcholine transporter (VAChT), which mediates acetylcholine storage in synaptic vesicles. Boxplots show the median (line) and IQR (interquartile range), with whiskers no more than 1.5× IQR. Number of cell lines per tumor type specified next to their corresponding labels.

Extended Data Fig. 6. Analysis of the energy cost of maintaining the resting potential.

(a) Mathematical modeling of the energy cost of maintaining the resting potential. Average values of input conductance and resting potential (V_rest) for four different NE lines predict the relationship between membrane potential and ATP consumption (curves), and the specific amounts associated with the observed resting potentials (points). (b) Illustration of how the rising phase (gray bar) of the action potential in a NE SCLC cell is driven by sodium current: the resulting charge influx charges the membrane capacitance by an amount ΔV, requiring a sodium charge influx of C_mΔV, ultimately requiring restoration by sodium pumping. (c) Indicative energy budget for electrical signaling, using mean parameters for AF1165 cells as an example. (d) Scatter plot of V_rest with input conductance in a panel of mouse (n = 7) and human (n = 6) cell lines measured by patch clamp recording. Lower panel: Zoom in on the area within the dotted square in the left lower part of the upper panel.

Extended Data Fig. 7. Metabolic characterization of NE and non-NE cells.

(a) Glycolysis index in SCLC NE (n = 3), LUAD (n = 2) and PDAC (n = 2) cell lines, measured by the amount of lactate secretion relative to glucose consumed in DMEM with 2% FBS (n = 3 technical replicates/cell line). (b) Gene ontology (GO) enrichment with the top eight candidates from a previous CRISPR screening to identify SCLC preferential vulnerabilities. Fisher’s exact test and Benjamini-Hochberg method applied for correction of multiple testing (adjusted p-values shown) (c) CRISPR-screen results of genes involved in glycolysis^,. Gene scores (log₂ fold change, L2FC) for the indicated genes for SCLC, LUAD and PDAC (n = 4 cell lines/cancer type). (d) Normalization of luminescence signals to the control condition (DMEM without rotenone and pyruvate: dotted red line). SCLC-NE (n = 4 cell lines, 4 technical replicates/cell line) and non-NE (n = 2 cell lines, 4 technical replicates/cell line) were treated with 5 or 50 nM rotenone for NE or non-NE, respectively (in the presence or absence of 2 mM pyruvate). Two-way ANOVA with Sidak’s multiple comparisons test. (e) CTG assay in paired NE and non-NE SCLC cell lines (top), and in a panel of SCLC NE, LUAD and PDAC cell lines (bottom) cultured in DMEM with 2 mM pyruvate, both treated with different concentrations of rotenone (dose-response curves). Pair #1: AD984LN; pair #2: AF3291LN. NE #1-4: AD984LN_fl, AF1165, AF1281m1, AF3062C; LUAD: KP1233; PDAC #1, 2: MDM1402, MDM1403. n = 4 technical replicates/line. (f) WB of FLAG in AD984LN_fl and AF3062C cells engineered to express a FLAG-tagged version of LbNOX. (g) CTG assay in SCLC NE cell lines (NE #1: AD984LN_fl; NE #2: AF3062C) with or without the LbNOX construct and treated with different concentrations of rotenone (dose-response curves). n = 4 technical replicates/line. (h-j) Quantification of the basal (h), ATP-linked oxygen consumption rates (OCRs) (i), and coupling efficiency (j) of SCLC cell lines, being the last one the percentage of mitochondrial OCR dedicated to ATP synthesis. n = 6 technical replicates/cell line, 4 NE-SCLC cell lines and 2 non-NE cell lines. (k) Radioactive 2-deoxyglucose uptake in SCLC NE (n = 3), LUAD (n = 2) and PDAC (n = 2) cell lines (n = 3 technical replicates/condition). (l) SRB assay of mSCLC NE cell lines cultured in conditioned media from two pairs of NE and non-NE cells, originally derived from the same tumors (AD984LN and AF3291LN). n = 3 technical replicates/condition. Two-way ANOVA with Sidak’s multiple comparisons test. (m) CTG assay of the 3 SCLC NE cell lines cultured in conditioned media from 2 pairs of NE and non-NE cells, originally derived from the same tumors (AD984LN and AF3291LN). The luminescence value in each condition normalized to those cultured in DMEM with 2% FBS (dotted red lines). n = 4 technical replicates/condition. Two-way ANOVA with Sidak’s multiple comparisons test; ns: not significant. (n) SRB assay in the mSCLC NE cell line AF3062C cultured with conditioned media from 2 pairs of NE and non-NE cells, originally derived from the same tumors (AD984LN and AF3291LN) using different concentrations of FBS. n = 3 technical replicates/condition. Two-way ANOVA with Sidak’s multiple comparisons test applied. (o) Lactate and pyruvate concentrations in 2 pairs of NE and non-NE cell lines were normalized to cell number across time (n = 3 technical replicates/cell line). Two-way ANOVA with Sidak’s multiple comparisons test applied. (p) WB of MCT1 expression in paired NE and non-NE SCLC cell lines. (q) Changes in lactate concentration in media with 20 mM lactate of paired NE and non-NE SCLC cell lines (n = 3 technical replicates/line). Two-way ANOVA with Sidak’s multiple comparisons test applied. Mean ± SEM shown in all graphs.

Extended Data Fig. 8. Metabolic influences on electrophysiological properties of SCLC.

(a) Frequency of calcium transients in GCaMP6m⁺ NE cells cultured either alone or with non-NE cell lines. n = number of cells reported in each column over 3 independent experiments. (b) Calcium peak frequency in NE cells (left) and fraction of NE cells with active calcium signaling (“peaks”) (right) when GCaMP6m⁺ NE cells (AF3062C) were cultured in conditioned media (CM) from NE or non-NE cells (non-NE cells: #1: AD984LN_adh, #2: AF3291LN_adh). Left panel: n =number of cells reported in each column over 3 independent experiments; right panel: n = 6 fields of view were analyzed for each condition, each dot represents percentage of active cells per field of view. Frequency of transients is normalized to the number of time frames acquired at 0.5 Hz. (c) Frequency of calcium transients in GCaMP6m⁺ NE cells co-culture with AD984LN_fl after Diclofenac (0.5 mM) administration. n=number of cells reported in each column over 3 independent experiments. Frequency of transients is normalized to the number of time frames acquired at 0.5 Hz. (d) Principle of estimating the energetic cost of the resting potential, by balancing the Na pump current and the Na and K components of the resting leak current. See Methods for details. (e) Membrane conductance (g_m) in AF1165 NE cells in different conditions. Starvation: n = 15, 31, 7 cells/time point; 5 mM glucose: n = 14, 30, 6 cells/time point; 10 mM lactate: n = 14, 29, 7 cells/time point. (f) g_m, V_rest and estimated ATP consumption in AF1165 cells incubated with pyruvate for different duration. n = 20, 5, 9 cells/time point. (g) Left: SOX1 normalised mRNA counts of human SCLC cell lines (n = 54), human LUAD cell lines (n = 79), and non-cancerous human cell lines (n = 6) from the CCLE (Cancer Cell Line Encyclopedia) provided by the DepMap 23Q2 database. Right: SOX1 mRNA expression levels across the different hSCLC subtypes (n = 54 cell lines). (h) Correlation analysis of SOX1 mRNA levels and the NE score of previously published PDX models (n = 50 PDX models, Pearson correlation. Two-tailed P-value shown). (i) SOX1 and SLC16A3 (encoding MCT4) normalised mRNA reads in human PDX cell lines previously published. n = 46 of NE-like cell lines (with NE-Score >0.8), n = 4 of non-NE-like cell lines (with NE-Score <0.2). (j) Co-localization analysis between MCT4 and SOX1 in tumors from a GEMM of SCLC. The analysis was performed by ImageJ plugin JaCoP; the intensity of each pixel is plotted on x-axis (SOX1) or on y-axis (MCT4) (left panel). The Pearson correlation coefficient between these two events was calculated (red line, right panel). n = 5 mice (1 tumor/mouse). Scale bar= 20 µm. Mean ± SEM shown in all graphs. Two-tailed Mann-Whitney test applied in a (“AD984LN” panels), c, i (SOX1). Kruskal-Wallis test with Dunn’s multiple comparisons test in a (other panels), b, e, f. Ordinary one-way ANOVA applied in g. Two-tailed unpaired t-test applied in i (SLC16A3). ns: not significant.

Extended Data Fig. 9. Effects of optogenetic stimulation on SCLC.

(a) Normalized liver weight after liver metastasis assay. NOD-SCID mice were intrasplenically injected with either NE or non-NE cells. Liver weight was normalized to animal weight at injection and compared to non-injected controls. Ordinary one-way ANOVA, Dunnett’s multiple comparisons test. *: p = 0.0104; ***: p = 0.0008; ****: p < 0.0001; ns: not significant. n = 6, 4, 6, 7, 6, 5 mice from left to right. (b) Induction of FOS expression and CREB phosphorylation are key downstream events following neuron activation and its subsequent calcium influx. (c, d) Illustration of optogenetics (c) and membrane potential responses recorded from ChR2-expressing SCLC NE cells (AD984LN_fl) in current-clamp mode upon blue light stimulation. Representative of recordings in n = 8 cells. A similar result was observed in ChR2⁺ AF1165 as well (d). (e, f) Current-clamp membrane potential recording of ChR2⁻, parental AD984LN_fl NE cells (representative of n = 3 cells) (e) and ChR2⁺ AD984LN_adh non-NE cells (representative of n = 4 cells) (f) upon blue light stimulation. While depolarization was observed in the AD984LN_adh cells due to ChR2 expression, no action potential firing could be elicited. (g) Western blot of FOS and p-CREB at different time points after blue light LED stimulation (1 Hz for 10 min) of AD984LN_fl and AF3062C NE cells. (h, i) Representative bright field images of NE cells (AD984LN_fl) three days after blue light exposure at different frequencies for 5 min (h) and at different frequencies and durations (i) Scale bars: 1 mm. n = 3 technical replicates/condition. (j) Normalised proliferation rates measured by SRB assay in the mSCLC NE cell lines AD984LN_fl and AF1165 after different exposure times (left) or different pulse frequencies (right) of blue light. The dotted line represents the unstimulated condition. n = 3 technical replicates/cell line. (k, l) WB for markers of ferroptosis (the lipid peroxidation adduct: 4-HNE (k); reduction of GPX4 (l)) and autophagy (LC3B-I/II forms, (k)) in parental and ChR2-expressing AD984LN_fl NE lines, with and without blue light exposure. (m) SRB assay in AD984LN_fl cells (parental and ChR2) after 10Hz-blue light stimulation. n = 3 technical replicates/condition, also other two mSCLC-NE lines showed similar trends. Two-way ANOVA, Sidak’s multiple comparison tests applied. (n, o) Blue light exposure had no effect on the ChR2⁺ non-NE cells, both in the short-term proliferation assay with SRB (n) and in the long-term colony formation assay (o) n = 3 technical replicates/cell line/condition. Groups were compared with two-way ANOVA, Sidak’s multiple comparison tests, ns: non-significant. (p) SRB assay in SCLC cells incubated with 1 μM TTX for three days (n = 3 technical replicates/condition). Two-way ANOVA, Sidak’s multiple comparison tests applied. Mean ± SEM shown in all graphs. Schematics in b and c were created using BioRender (https://biorender.com). Source data

Extended Data Fig. 10. Evaluation of electrical activity in SCLC progression.

(a) Representative images of livers dissected from animals intrasplenically injected with control and TTX-pretreated AF3062C cells relative to Fig. 5f. Scale bar: 1 cm. (b, c) Representative H&E images relative to Fig. 5g of livers from (a), the same lobes were used for comparison. Scale bars: 2 mm (b), 50 μm (c). Control (n = 35 tumors/4 mice) and Pre-Tx (n = 29 tumors/3 mice). (d) Western blot for p-CREB in control (untreated) and TTX washout (Pre-Tx) AF3062C NE cells. (e) Bioluminescence quantification of mice injected with AD984LN_fl cells±iPSAM⁴ expression, treated with uPSEM817 or vehicle control (n = 5 mice/group). Mean ± SEM shown. (f) Immunohistochemistry (IHC) analysis of p-CREB in an mSCLC tumor from a classical PRP130 GEMM (top), and a large airway from its normal adjacent tissue (NAT) (bottom). Representative images of n = 5 mice ( > 20 tumors). Scale bars: 20μm. (g, h) p-CREB IHC in PDX models (g) and clinical SCLC tumor samples (h); PDX models with the highest (240) and lowest (90) H scores (g) and hSCLC tumor with the highest H score (300) (h) were shown. Scale bars: 20 μm. n = 20 patients, 10 PDX models. (g, right) Quantification of p-CREB IHC in PDX models (n = 10). (i) Percentage of SOX1⁺ cells was increased in advanced-stage hSCLC tumors; percentage of MCT4⁺ cells remained stable. Data are presented as the mean ± SEM, stage I: n = 14; stage II: n = 20; stage III: n = 10. Groups were compared by ordinary one-way ANOVA. P-values shown, ns= not significant. (j) Boxplots of normalised protein expression values of SOX1 (left) and CREB (right) in tumour and normal adjacent tissue (NAT) samples from a previously published cohort of SCLC patients (n = 112). Wilcoxon matched-pairs signed rank two-sided test was applied. Boxplots show the median (line) and IQR (interquartile range), with whiskers no more than 1.5× the IQR. (k) Correlation analysis of SOX1 and CREB protein levels in tumour (left) and NAT (right) samples from SCLC patients (n = 112). Spearman correlation (two-sided) applied, p-values shown. (l, m) Kaplan–Meier curves of a published cohort of SCLC patients (n = 104) according to the mRNA expression of different nAChR (l) and mAChR (m) subunits. Log-rank test applied, 95% confidence intervals and p-values shown. Source data