Intrinsic electrical activity drives small-cell lung cancer progression – Nature

Animal studies

F₁ mice used for intrasplenic injections of SCLC cells were obtained by crossing 129S6/SvEvTac males to C57BL/6J females imported, respectively, from Taconic (Germantown, NY, USA) and The Jackson Laboratory (JAX; Bar Harbor, ME, USA). Both male and female mice between 9 and 19 weeks of age were used, and age-, litter- and sex-matched mice were randomly assigned to experimental groups. NOD-SCID mice (Prkdc^scid; JAX strain 001303) used for intrasplenic injection SCLC cells were obtained from The Jackson Laboratory. Cells expressing iPSAM⁴ were injected in males between 7 and 11 weeks of age. Age- and litter-matched mice were randomly assigned to experimental groups. Paired NE and non-NE SCLC cells were injected in 8- to 12-week-old animals. Age-, litter- and sex-matched mice were randomly assigned to experimental groups. The animals were anaesthetized with a gas mixture of oxygen-enriched air and 3–3.5% isoflurane (Zoetis UK Ltd.), and an ultrasound-guided (Vevo 3100; FUJIFILM VisualSonics Inc.) transdermal intrasplenic injection of SCLC cells (50 μl of a 10⁶ cell per millilitre suspension in Dulbecco’s PBS (DPBS)) was performed, as described previously⁴². For all intrasplenic injections of cancer cells with different treatments, the experiment was conducted by a researcher who was blind to the experimental groups.

For the transplantation of SCLC NE cells after TTX pretreatment, cells were treated with 1 µM TTX for 24 h before the injection. Liver was sampled 25 days (AF3062C) or 34 days (AD984LN_fl) after injection. Liver weight was normalized on the baseline body weight before injection. After weighing, the livers were fixed in 10% neutral buffered formalin (NBF) for 24–48 h.

Autochthonous SCLC tumours were obtained from mice harbouring Trp53^fl/fl, Rb1^fl/fl, Rbl2^fl/fl and Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze} alleles (PRP130 model, as described previously¹²) or Trp53^fl/fl, Rb1^fl/fl, Rbl2^fl/fl and Gt(ROSA)26Sor^{tm1.1(CAG-tdTomato/GCaMP6f)Mdcah/J} alleles (PRP130-Salsa6f model). The PRP130 animals were on a mixed 129S5/C57BL/6J background. SCLC tumour formation was induced by intratracheal administration of Ad5-CGRP-Cre adenoviruses (2 × 10⁸ pfu per animal; VVC-Berns-1160; University of Iowa Vector Core; plasmid origin: A. Berns and K. Sutherland) according to the method described previously⁶¹. Tumour growth was monitored by regular computed tomography (CT) scans (Quantum GX2; PerkinElmer), starting from 5 months after virus administration. Animals that reached the humane end point (development of moderate breathing signs or weight loss below 15% of the maximum weight), typically in 7–10 months, were killed by an overdose of anaesthetic and processed for precision-cut lung slices or perfused transcardially with 20 ml of PBS containing 20 U ml⁻¹ of heparin (H3400; Sigma-Aldrich) and 20 ml of 10% NBF. Dissected lungs were incubated in 10% NBF for 24–48 h and then stored in a 0.02% solution of sodium azide in PBS at +4 °C until use.

SCLC tumour samples from the PRM model were obtained from mice harbouring Trp53^fl/fl, Rb1^fl/fl and Igs2^{tm1(CAG-Myc*T58A/luc)Wrey} alleles (JAX strain 029971) on a C57BL/6J background. SCLC tumour formation was induced by intratracheal administration of Ad5-CGRP-Cre adenoviruses (1 × 10⁸ pfu per animal; VVC-Berns-1160) at 11 weeks of age. Tumour growth was monitored by regular computed tomography scans (Quantum GX2), starting from 46 days after virus administration. Upon reaching humane end point at day 89 following virus administration, the animals were culled by cervical dislocation for tumour collection.

Pulmonary NE cells were labelled by intratracheal administration of Ad5-CGRP-Cre adenoviruses (2 × 10⁸ pfu per animal; VVC-Berns-1160) in 8- to 10-week-old Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze} homozygous mice (Ai14 allele; JAX strain 007908). Lungs were collected 1 week after adenovirus instillation, either for precision-cut lung slices or for immunofluorescence after transcardial perfusion.

The mice were group-housed (up to five mice per cage) at a specific pathogen-free facility at The Francis Crick Institute in individually ventilated cages (GM500; Tecniplast) at an ambient temperature of 22 ± 2 °C, relative humidity of 55 ± 10% and standard 12-h light/12-h dark cycles. The animals received standard rodent food (2018 Teklad Global; ENVIGO) and water ad libitum. Humane end points (moderate clinical signs, clinical signs of suffering/distress or loss of more than 15% body weight) for animals in tumour studies were not exceeded in any of the experiments. All procedures were conducted in accordance with the UK Animals (Scientific Procedures) Act 1986, approved by the Institutional Animal Welfare and Ethical Review Body (The Francis Crick Institute PPL Review Committee), conducted under the authority of the UK Home Office approved project licence PP4103600 and approved by the Massachusetts Institute of Technology (MIT) Institutional Animal Care and Use Committee.

Cell culture

Mouse SCLC lines

The following cell lines were derived from GEMMs, as described previously⁴². In particular, AF1165 was derived from a primary tumour of a PR; Rosa26^LSL-Tom/+ mouse. AF3062C was derived from liver metastases of a male PR; Rosa26^LSL-Tom/+ mouse. AF1281m1 was derived from a relapse tumour after chemotherapy in a PR; Rosa26^LSL-Tom/+ mouse. TP2031T2 was derived from a primary PR tumour. AD984LN_fl/AD984LN_adh were derived from the same lymph node metastasis in a PR mouse⁶². AF3291LN_fl/AF3291LN_adh were also derived from the same lymph node metastasis in a PRPTEN mouse.

The PRM2.1a cell line was derived in-house from a primary tumour in a Trp53^fl/fl; Rb1^fl/fl; Myc^LSL-T58A/+ (PRM) mouse, as described previously⁶³. Briefly, the primary tumour was dissected, cut into small pieces and incubated for 30 min at 37 °C in 6 ml digestion solution (10% TrypLE (12605010; Gibco), 1 mg ml⁻¹ of Collagenase IV (17104019; Gibco) and 1 mg ml⁻¹ of Dispase II (D4693; Sigma-Aldrich) in Hanks’ balanced salt solution (HBSS)). The digestion reaction was quenched by adding 4 ml of ice-cold quenching medium (10% FBS (11320033; Gibco) and 18.75 µg ml⁻¹ of DNase I (DN25; Sigma-Aldrich) in DMEM). The cell suspension was passed several times through an 18G needle before being filtered with a 100-µm strainer. Cells were centrifuged at 800g for 5 min, resuspended in 1× RBC Lysis Buffer (420301; BioLegend) and incubated for 3 min at 37 °C. After washing with ice-cold PBS, the cells were plated in a complete culture medium on tissue-treated culture vessels, selecting for cells growing as floating aggregates.

NE cells were maintained in DMEM-F12 with GlutaMAX (10565018; Gibco) with 1× NEAA (11140050; Gibco), 10% FBS (11320033; Gibco) and 100 U ml⁻¹ of penicillin–streptomycin (15140122; Gibco). NE cells were grown as floating aggregates or were allowed to adhere to culture vessels coated with 50 µg ml⁻¹ of growth-factor-reduced Matrigel (356231; Corning) or Cultrex BME (3432-010-01; Bio-Techne) in HBSS. NE SCLC cells were also grown 3D-embedded in a Matrigel drop for calcium imaging experiments. Non-NE cells were cultured in DMEM (10-013-CV; Corning) with 10% FBS, 2 mM GlutaMAX (35050038; Thermo Fisher Scientific) and 100 U ml⁻¹ of penicillin–streptomycin.

Human SCLC lines

NCI-H889, NCI-H82, COR-L47, COR-L279 (purchased from ATCC), NCI-H69_fl and NCI-H69_adh (a gift from J. Minna) cell lines were cultured in DMEM-F12 with GlutaMAX, 1× NEAA, 10% FBS and 100 U ml⁻¹ of penicillin–streptomycin.

Mouse LUAD lines

KP1233 and KP1234 were derived from the primary tumours of KP mice (Kras^LSL-G12D/+ and Trp53^fl/fl), as described previously⁴² and cultured in DMEM (10-013-CV; Corning) with 10% FBS and 100 U ml⁻¹ of penicillin–streptomycin.

Mouse pancreatic lines

MDM1402 was previously derived from a primary tumour of a female KPC mouse (Kras^LSL-G12D/+, Trp53^LSL-R172H/+, Rosa26^LSL-Tom/+ and Pdx1-cre⁺); MDM1403 was previously derived from a primary tumour of a male KPC mouse (Kras^LSL-G12D/+, Trp53^fl/+, Rosa26^LSL-Tom/+ and Pdx1-cre⁺)⁴² and cultured in DMEM (10-013-CV; Corning) with 10% FBS and 100 U ml⁻¹ of penicillin–streptomycin.

Mouse βTC-B6 PaNET cell line was previously derived from a RIP1-Tag2 mouse model and cultured in DMEM (10-013-CV; Corning) with 10% FBS and 100 U ml⁻¹ of penicillin–streptomycin^1,64.

Other cell lines

HEK293T cells were supplied by the Cell Services facility at The Francis Crick Institute and cultured in DMEM (11995073; Gibco) with 10% FBS and 100 U ml⁻¹ of penicillin–streptomycin.

All the cell lines were regularly cultured under standard conditions (humidified 5% CO₂ atmosphere at 37 °C) and were routinely tested for mycoplasma contamination and authenticated by short tandem repeat (STR) profiling. For all conditioned medium experiments and metabolic assays comparing across different types of cancer cells, the same culture medium was used, mostly pyruvate-free DMEM (10-017-CV; Corning) with 2% or 10% dialysed FBS unless specified otherwise.

MGH1505-1A cell line derivation

The PDX model MGH1505-1A was established from circulating tumour cells isolated from patient MG1505, a 59-year-old man with relapsed SCLC²⁷. Two cell lines with distinct morphologies were derived from this PDX model over 260 days as follows (Extended Data Fig. 1). A fragment of an MGH1505-1A xenograft was implanted in the right flank of an NSG-GFP mouse (JAX strain 021937), in which all mouse cells express GFP. When the xenograft reached 1,500 mm³, the NSG-GFP mouse was euthanized, and the xenograft was resected and manually dissociated in ice-cold PBS. Live-cell clusters were isolated through serial gravity sedimentation at room temperature in 15-ml conical tubes, with periodic evaluation of supernatant to identify fractions containing the highest ratio of live-cell clusters to cell debris. Red blood cells were lysed using ACK lysis buffer (A1049201; Gibco). The enriched cell clusters were grown in a modified HITES medium (DMEM-F12 with GlutaMAX, 5% FBS, 1× NEAA, 1× ITS -G supplement (Gibco, 41400045), 10 nM hydrocortisone (07925; STEMCELL Technologies), 10 nM β-estradiol (E2758; Sigma-Aldrich) and 100 U ml⁻¹ of penicillin–streptomycin), and the culture was monitored two to three times weekly. Adherent GFP⁺ mouse fibroblasts began to proliferate within 3 days of dissociation, and the culture growth pattern was dynamic over the first 8 weeks, with initial suspension clusters transitioning to a mixture of tightly and loosely adherent tumour cells interspersed with GFP⁺ murine cells with fibroblast morphology (Extended Data Fig. 1b). To deplete the murine cells, we leveraged the absence of functional RB1 and consequent lack of CDK4/6 dependence that characterizes most SCLC, including MGH1505-1A. We treated the MGH1505-1A culture with the CDK4/6 inhibitor palbociclib (10 µM) (508548; Thermo Fisher Scientific) for 1 week, resulting in significant depletion of GFP⁺ cells and a mixed culture of floating tumour clusters and adherent tumour cells. The adherent and suspension components were separated and cultured independently without palbociclib; however, residual GFP⁺ cells were observed in both cultures after 4 weeks (Extended Data Fig. 1c,d), prompting an additional 4 weeks of palbociclib treatment. During this treatment, suspension clusters were dissociated by means of gentle pipetting, and adherent cells were dissociated using TrypLE digestion twice weekly to expose GFP⁺ cells to the drug. Following palbociclib treatment, the suspension (MGH1505-1A_fl) and adherent (MGH1505-1A_adh) cultures were grown at low density and screened manually for GFP⁺ cells two to three times weekly to confirm complete depletion of mouse cells (Extended Data Fig. 1e,f). The total culture time of 260 days exceeded 50 population doublings to establish both as cell lines with consistent and stable morphologies. The origins of both cell lines from patient MGH1505 were confirmed by STR profile comparison with patient germline genomic DNA.

Molecular cloning and cell line engineering

ChR2-expressing cell lines

ChR2 coding sequence was PCR amplified from FCK-ChR2-GFP (Addgene plasmid 15814, a gift from E. Boyden)⁶⁵ and cloned into a piggyBac transfer plasmid under the constitutive CAG promoter in frame with a P2A-PuroR cassette⁶⁶ to generate the plasmid pPB CAG ChR2-P2A-PuroR. SCLC, LUAD and PDAC mouse cell lines were reverse-transfected using Lipofectamine 3000 (Invitrogen) or TransIT-LT1 (Mirus Bio) with pPB CAG ChR2-P2A-PuroR and a plasmid encoding for a hyperactive form of the piggyBac transposase (pCMV HAhyPBase, a gift from the Wellcome Sanger Institute) at a ratio 4:1. Stable ChR2-expressing cells were established after puromycin selection.

GCaMP6m-expressing cell lines

GCaMP6m coding sequence was PCR amplified from pGP-CMV-GCaMP6m (Addgene plasmid 40754, a gift from D. Kim and GENIE Project)²⁹ and cloned into a lentiviral transfer plasmid under the constitutive EFS promoter in frame with a P2A-PuroR cassette to generate the plasmid pLenti EFS GCaMP6m-P2A-PuroR. Lentiviral particles were generated and used to transduce mouse SCLC NE cell lines, as described below.

LbNOX-expressing cells

FLAG-tagged LbNOX coding sequence was PCR amplified from pUC57 LbNOX (Addgene plasmid 75285, a gift from V. Mootha)⁴⁴ and cloned into an all-in-one doxycycline inducible lentiviral vector to generate the pTL LbNOX-FLAG plasmid. The LbNOX-FLAG sequence was cloned under a tight TRE promoter in the pTL lentiviral backbone, which also contains a cassette expressing PuroR-P2A-rtTA Advanced under a constitutive EF1a promoter. Lentiviral particles were generated from pTL LbNOX-FLAG and used to transduce AD984LN_fl and AF3062C cells, as described below.

iPSAM⁴-expressing cell lines

The coding sequence of the inhibitory chemogenetic receptor PSAM⁴-GlyR (iPSAM⁴) was PCR amplified from pCAG PSAM⁴ GlyR IRES EGFP (Addgene plasmid 119739, a gift from S. Sternson)⁵⁵ and cloned in frame with a cassette containing the T2A linker, firefly luciferase coding sequence linked through P2A to PuroR. Gibson assembly was used to generate the piggyBac transfer plasmid pPB CAG PSAM⁴-GlyR T2A Luc P2A PuroR. A control plasmid (pPB CAG Luc P2A PuroR) was assembled by cloning firefly luciferase coding sequence linked through P2A to PuroR in the same piggyBac backbone. AD984LN_fl cells were reverse transfected with either the iPSAM⁴-containing or control piggyBac transfer plasmid (together with pCMV HAhyPBase; 4:1 ratio) with TransIT-LT1 (Mirus Bio). Engineered cells were established after puromycin selection.

Generation of lentiviral particles, titration and transduction

HEK293T cells were co-transfected with lentiviral packaging plasmids psPAX2 (Addgene plasmid 12260, a gift from D. Trono) and pMD2.G (Addgene plasmid 12259, a gift from D. Trono) and the appropriate transfer plasmid using the calcium phosphate method. Supernatants were harvested 60 h after transfection, filtered using 0.45-μm filters and stored at −80 °C. The viral titre was estimated using a lentiviral standard of known titre (IU ml⁻¹) and the SYBR Green I-based PCR-enhanced reverse transcriptase assay (SG-PERT) method⁶⁷. SCLC NE cells were transduced by resuspending them in a complete culture medium supplemented with the appropriate lentivirus (multiplicity of infection (MOI) = 10) and 10 µg ml⁻¹ of polybrene (Sigma-Aldrich, TR-1003-G). Cells–virus suspensions were plated in a coated well of a six-well plate and spun at 1,200g for 2 h at 30 °C. The medium was changed 24 h after transduction.

In vivo chemogenetic treatment, bioluminescence imaging and survival analysis

Mice transplanted with iPSAM⁴-expressing or control cells were injected intraperitoneally with 0.3 mg kg⁻¹ of uPSEM817 tartrate (6866; Bio-Techne) in sterile saline or vehicle alone (daily for 5 days a week), starting from 3 days after cell transplantation. On the same day, tumour burden was monitored by in vivo bioluminescence imaging and followed up weekly. The animals were imaged 10 min after intraperitoneal administration of 150 mg kg⁻¹ of d-luciferin (1-360223-200; Regis Technologies). The animals were anaesthetized with 2% isoflurane, and the bioluminescent signal was captured using the IVIS Spectrum system (PerkinElmer). Bioluminescent signals were quantified using the Living Image software (v.4.8; PerkinElmer): a 7-cm² square region of interest (ROI) was drawn on each mouse abdominal area, and the average radiance (p s⁻¹ cm⁻² sr⁻¹) was calculated. Background signal was removed from each measurement by subtracting the average value in a similar ROI placed on an empty area within the same field of view. Survival analysis was performed by calculating the lifespan after cell transplantation in days of every mouse in each experimental group. Mice that died due to causes unrelated to the study were censored in the analysis. Data were displayed using the Kaplan–Meier format, and the statistical significance of the results was tested using the log-rank (Mantel–Cox) test.

Patch-clamp electrophysiology

Patch-clamp recordings were carried out by standard methods⁶⁸. Cells were plated after passaging onto 35-mm tissue-culture-treated plastic Petri dishes and cultured for 1–4 days before recording. For NE cell lines, dishes were pre-coated with a 5% solution of Matrigel for 20–30 min to allow adhesion. Cells were visualized with phase-contrast microscopy (Olympus IX71) using 10× or 20× objectives.

Current-clamp and voltage-clamp recordings were carried out with an Axon MultiClamp 700B amplifier (Molecular Devices), and 16-bit waveforms were generated and sampled at 10–20 kHz using an X Series data acquisition interface (National Instruments), following low-pass analogue filtering (four-pole Bessel) at a 4-kHz cutoff and analysed with custom software in R or MATLAB. Before recording, the culture medium was exchanged for Ringer solution composed of 140 mM NaCl, 4 mM KCl, 1.4 mM CaCl₂, 1 mM MgCl₂ and 10 mM HEPES, balanced to pH 7.4 with NaOH. In most recordings, 5 mM glucose was further added but was omitted or replaced by lactate (10 mM) in some recordings, as described in the text.

Recordings were established using 1.5-mm-outer-diameter borosilicate glass capillary pipettes (Harvard Apparatus), fire-polished at the tip and filled with an intracellular perfusion solution composed of either 105 mM potassium gluconate, 30 mM KCl, 10 mM HEPES, 1 mM EGTA, 4 mM ATP·Mg, 0.3 mM GTP and 10 mM phosphocreatine·Na₂, balanced to pH 7.3 with KOH, or 125 mM potassium gluconate, 10 mM HEPES, 4 mM ATP·Mg, 0.3 mM GTP and 10 mM phosphocreatine·Na₂, balanced to pH 7.3 with KOH. Recordings were carried out at room temperature (23–25 °C).

Membrane potentials were adjusted for prenulling of a calculated liquid junction potential of 10 mV before seal formation, with the open pipette immersed in the bath solution. Pipettes had resistances of 5–10 MΩ. To estimate the unperturbed resting membrane potential, before significant dialysis of the intracellular compartment by pipette solution, current-clamp recordings were made immediately following the rupture of a cell-attached patch to establish whole-cell recording mode. This was usually carried out in current-clamp mode to allow time-resolved recording of the zero-current potential during rupture, and the initial value within 0–3 s was measured. If rupture was carried out in voltage clamp, the potential was measured in the current clamp within approximately 4 s. Cell input conductance and capacitance were evaluated from fits of the current-clamp responses to small (2–4 pA), hyperpolarizing step current stimuli. For voltage-clamp recordings, the built-in circuitry of the MultiClamp 700B was used to compensate for pipette and whole-cell capacitance and series resistance (80%).

To measure excitability, the overshoot or maximum positive difference between the peak voltage and steady-state depolarized voltage in a run was determined. If the cell shows a rebound spike, this is always the largest and the one used for the measurement. Excitability is defined as the maximum positive difference between the peak voltage and steady-state depolarized voltage in a sequence of progressively larger current step responses. Excitability is zero if there is no clear, consistent overshoot in more than one sweep in a run.

For pre-incubation in conditions of different energy supplies, the culture medium was exchanged for Ringer solution containing 0–10 mM of glucose or 5 mM lactate for periods from less than 2 h up to more than 16 h (overnight), and dishes were incubated at room temperature in a humidified chamber.

Biophysical modelling of ATP consumption required for electrical activity

To estimate the energetic cost of maintaining the negative resting potential required to enable action potential generation, a simple model was used, which assumes that the membrane conductance at rest is composed solely of sodium-selective and potassium-selective fractions, which is reasonable because these contribute to the dominant electrochemical gradients at the resting potential. The model solves for the electrogenic sodium pump current, which equals the sum of inward sodium and outward potassium fluxes through the resting conductance of the membrane at steady state⁶⁹.

where V is the resting membrane potential, and and are the equivalent sodium-selective and potassium-selective fractions, respectively, of the total input leak conductance: \({g}_{{\rm{in}}}={g}_{{\rm{Na}}}+{g}_{{\rm{K}}}=1/{R}_{{\rm{in}}}\), which was measured from the response to small hyperpolarizing current steps in current-clamp mode. Nernst equilibrium potentials and were assumed to equal +53.8 and −108.4 mV, respectively, as calculated from the compositions of pipette and Ringer solutions, because in NE cells, there was typically a shift of only a few millivolts in the resting potential from the initial breakthrough potential, implying that the sodium and potassium contents of the unperturbed cytoplasm were close to those of the pipette solution.

Because the sodium pump transports three Na⁺ ions out and two K⁺ ions in for each ATP molecule hydrolysed,

From this, it can be shown that the inward sodium current at rest is given by

and the ATP consumption rate at rest for the cell is

where F is Faraday’s constant (equations (3) and (4) from ref. ⁶⁹). Thus, by measuring and V and assuming the values of and \({E}_{{\rm{K}}}\), one can estimate the energetic cost of maintaining the resting potential.

To estimate the cost of each action potential, the difference between resting potential and the peak of the action potential (assumed to be a typical value of 60 mV for AF1165 cells) is multiplied by the membrane capacitance (assumed 20 pF, a typical value for AF1165 cells), as estimated from the measured passive time constant of the membrane, to give the amount of sodium charge required to depolarize the membrane during each action potential. This gives an estimate of the ATP required of 2.5 × 10⁶ ATPs per action potential, which is a conservative lower bound because it assumes complete efficiency of the action potential (non-overlap in time of sodium and potassium voltage-gated currents⁷⁰).

To estimate the cost of vesicular release, we assumed a typical value of 20 F increase in the plasma membrane capacitance, which was measured using the phase shift of current during voltage clamp to a 1-kHz sinusoidal command voltage, following a brief depolarizing pulse^71,72. Assuming a specific membrane capacitance of 1 μF cm⁻², this translates to the release of about 400 vesicles (40-nm diameter). We have no specific information about the energetics of packaging of the vesicles in these cells, but using the rough estimate⁷³ of 23,400 ATPs per 40-nm vesicle in neurons yields a cost of 5 × 10⁷ ATPs per action potential for the vesicular release.

These costs can be compared to reported measurements of ATP production in a panel of epithelial cancers⁷⁴, which showed an upper limit of around 20 pmol μg⁻¹ of protein per minute. Assuming a dry weight fraction of 25% and that each cell has a volume of 1 pl (equivalent to a cube with sides of 10 μm) yields 4,000 cells per microgram of protein. This equates to an ATP production rate of 5 × 10⁷ ATPs per cell per second.

Blue light stimulation

For blue light stimulation, we used Bluecell, a device with in-house built hardware and software, as detailed in a previous study⁷⁵. This device operates with an array of blue LED strips (460 nm), as well as temperature sensors and cooling fans that prevent overheating the cell culture plates upon prolonged blue light exposure. The LED array is controlled by a Raspberry Pi microcomputer, which receives a light sequence that was previously programmed using the BLUECELL.ijm file provided by J.-P. Vincent’s laboratory. The file was run in ImageJ2 (v.2.9.0), and different illumination sequences were designed. All the illumination sequences had a blue light intensity of 10 mW cm⁻² with different frequency patterns (1, 10 or 100 Hz) maintained during a range of time points (1, 5 or 10 min). Specifically, for the 1-Hz stimulation sequence, an additional step of 5 s without blue light exposure was set up in every cycle of stimulation to prevent exhaustion of the cells.

Drug and metabolite treatments

The following drugs were used: tetrodotoxin citrate (TTX; ab120055; Abcam), rotenone (R8875; Sigma-Aldrich), oligomycin (O4876; Sigma-Aldrich), FCCP (HY-100410; MedChemExpress), antimycin A (A8674; Sigma-Aldrich), SR-13800 (5431; Tocris Bioscience), diclofenac sodium salt (D6899; Sigma-Aldrich) and CCh (C4382; Sigma-Aldrich). The metabolites used for different metabolic assays were sodium pyruvate (P2256-25g; Sigma-Aldrich) and sodium lactate (L7022-5G; Sigma-Aldrich). Concentrations were specified in each of the experiments. For LbNOX induction, cells were treated with 1 µg ml⁻¹ of doxycycline hyclate (D9891; Sigma-Aldrich).

CellTiter-Glo assay and dose–response curves

Cells were plated in 96-well assay plates (3903; Corning) (3,000–6,000 cells per well for NE SCLC cells in wells pre-coated with Matrigel/HBSS solution; 500 cells per well for non-NE SCLC/LUAD/PDAC cells) in a regular culture medium. The next day, the cells were switched to DMEM with 10% dialysed FBS (F0392; Sigma-Aldrich) containing the drugs of interest. After 3 days of culture or the respective drug treatments, the number of viable cells per well was measured using the CellTiter-Glo Luminescent Cell Viability Assay (G7570; Promega). The results were normalized to control and untreated samples. All assays were performed in four technical replicates per cell line tested. For dose–response curves performed with LbNOX-expressing cells, they were cultured with 1 µg ml⁻¹ doxycycline since the seeding day.

Conditioned medium collection and dialysis

To collect conditioned medium, non-NE cells were seeded in 15-cm plates at 5 × 10⁶ cells per plate, whereas NE cells were seeded in Matrigel-coated 10-cm plates at 10⁷ cells per plate. Seeding was done in the respective regular culture medium. After 24 h of seeding, the medium was changed (20 ml per 15-cm plate and 10 ml per 10-cm plate) to pyruvate-free DMEM (10-017-CV; Corning) with 100 U ml⁻¹ of penicillin–streptomycin and 2% of commercially dialysed serum (F0392; Sigma-Aldrich). Another 24 h after medium change (48 h after seeding), the medium was collected and filtered through a 0.22-μm filter. The medium was either stored at 4 °C for use within a week or aliquoted and stored at −80 °C. The conditioned medium was subject to equilibrium dialysis using 3.5K MWCO cassettes (66330; Thermo Fisher Scientific). The conditioned medium was injected into cassettes and allowed to dialyse against a fresh medium at a volumetric ratio of 1:100 overnight at 4 °C. This process was repeated twice per round of dialysis for an expected dilution ratio of 1:10,000 of all molecules smaller than 3.5K MW. For calcium imaging experiments, FluoroBrite DMEM (A1896701; Thermo Fisher Scientific) was used in the protocol instead of DMEM (10-017-CV; Corning) and subsequently collected.

Radioactive 2-deoxyglucose uptake assay

The day before the experiments, cells were seeded in six-well plates in a regular culture medium (DMEM; 10-013-CV; Corning; with 10% FBS and 100 U ml⁻¹ of penicillin–streptomycin), with two plates per cell line. On the day of the experiment, cells were changed to a 1 ml fresh medium per well followed by incubation in a 37 °C incubator. After 2 h, one plate per cell line was taken out, with 100 µl of radiolabelled 2-deoxyglucose spiked into each well (100 µl of 11 µCi ml⁻¹ [³H]-2DG, for a final concentration of 1 µCi ml⁻¹). The plates were gently rocked for mixture and incubated for exactly 15 min at room temperature. The medium was then aspirated, and the plates were placed on ice and washed four times with ice-cold PBS (5 ml per well). The cells were trypsinized with 500 µl trypsin/EDTA and mixed well using a 1-ml pipette. Then, 400 µl of cells per trypsin mixture per sample was transferred into scintillation vials, and scintillation buffer was added. The vials were swirled until the solution turned clear, and the samples were ready for radioactivity measurement. Cell numbers in the other plate were counted in parallel.

Lactate titration experiment and metabolite quantifications

NE cells were seeded at 10⁶ cells per well in 2 ml regular culture medium in six-well plates the day before the experiment. Mock plates were medium-only controls and treated the same way as other wells. The next day, wells were washed with 5 ml PBS twice before replacing with a fresh medium containing different concentrations of lactate (DMEM; 10-017-CV; Corning; with 10% dialysed serum or 2% dialysed serum for the paired NE/non-NE cell experiments). The next day, the medium was collected after gentle rocking of the plates, centrifuged at 1,000 rpm for 5 min, and then 1 ml of supernatant was collected. The glucose and lactate concentrations were measured in a YSI Bioanalyzer (Yellow Springs Instruments). The degree of secretion/uptake in these metabolites was determined by the difference between concentrations from the control and the cell-containing medium.

Sulforhodamine B assay for proliferation

The sulforhodamine B (SRB) assay was performed, as described previously⁷⁶. For the blue light stimulation experiment, between 6 × 10⁴ and 4 × 10⁵ mSCLC-NE, mSCLC-non-NE, LUAD and PDAC cells were seeded in six-well plates with 1 ml of medium per well. During the following 3 days, the cells were either daily stimulated with 1 or 10 Hz blue light (1, 5 or 10 min), or treated with 1 µM TTX or with the conditioned medium. The day after seeding, one of the plates was used as the t₀ time point of the experiment and was fixed by adding 500 µl of 10% trichloroacetic acid to each well. After 3 days, all plates were also fixed with 500 µl of 10% trichloroacetic acid per well and kept in the cold room for at least 1 h. All plates (including the t₀ plate) were then washed three times with distilled water (dH₂O) and stained with a solution of 0.04% SRB (prepared in 1% acetic acid) for at least 30 min at room temperature. Next, the dye was removed, and the wells were washed three times with 1% acetic acid. After the last wash, the remaining liquid was aspirated completely. The plates were placed with the lids off in an incubator at 37 °C for 10–15 min until dried. Finally, 1 ml of 10 mM Tris (pH 10.5) was added to each well to solubilize the dye. The plates were shaken for 5–10 min at room temperature. Then, 100 µl of each well was transferred to a 96-well plate, and absorbance was read at 510 nm in the TECAN Infinite M1000 microplate reader. The doubling time per day was calculated with the following formula, where X_n and X₀ are the absorbance values at the end and beginning (t₀) of the experiment, respectively. For each experiment, three technical replicates were included.

Colony formation assay of mouse cell lines

Clonogenicity was evaluated by seeding 1,000–10,000 cells of mSCLC-NE, mSCLC-non-NE, LUAD and PDAC cells and 30,000 cells of PaNET cells per well in six-well plates. After 10–14 days, colonies were stained for 15 min with a solution of 0.1% crystal violet, 1% formaldehyde, 1% methanol in 1× PBS. After extensive washing and drying, the staining reagent was resolubilized in 1% acetic acid, and absorbance was measured at 570 nm with a TECAN Infinite M1000 microplate reader as an indirect measure of cell number.

For the experiments with blue light stimulation, cells were seeded for colony formation assay, and the next day, they were stimulated with blue light for 10 min (1, 10 or 100 Hz), including an unstimulated condition in each of the experiments. For the colony formation assays with TTX, two experimental set-ups were designed. In one of them, the cells were seeded in six-well plates, as described above. The next day, the medium was replaced by either a fresh medium or medium with 1 µM TTX. The plates were grown under regular conditions for 10 days before the staining of the colonies. In the other approach, before seeding the colony assays, the cells were pretreated with 1 µM TTX for 24 h. The following day, both untreated and treated cells were washed and replated for colony formation in a regular culture medium without TTX and grown for 10 days.

Colony formation assay of hSCLC cell lines

Soft-agar colony formation assays were used to assess the clonogenicity of hSCLC cell lines, as described in a previous study⁷⁷. In brief, a 1.2% low-melting-point agarose (16520-100; Invitrogen) solution was mixed with an equal volume of 2× medium, which contained 26.74 g l⁻¹ DMEM powder (50-013-PB; Corning), 2.4 g l⁻¹ sodium bicarbonate, 1 mM sodium pyruvate, 20% FBS (11320033; Gibco) and 2× NEAA (11140050; Gibco). Then, 1.5 ml of the agarose-medium mixture was layered onto the bottom of each well in six-well plates and left at room temperature to solidify for 30 min. For the top layer, each of the hSCLC cell lines was resuspended in 2× medium to reach a final density of 15,000 cells per well. Next, these cell suspensions were mixed with 0.6% agarose solution at a 1:1 ratio and added to the top of each well (1.5 ml). After a 30-min incubation at room temperature, 1 ml per well of complete culture medium (with or without 1 µM TTX) was added, and the plates were incubated under regular growth conditions for 3–4 weeks. Every 2 days, a fresh medium (with or without 1 µM TTX) was added to the wells to prevent evaporation. At the end of the experiment, the colonies were stained by adding 200 μl of nitroblue tetrazolium chloride solution (1 mg ml⁻¹; N6876; Sigma-Aldrich) per well and incubating the plates overnight at 37 °C.

Western blot

Total protein was extracted from cells using cell lysis RIPA buffer, supplemented with a phosphatase and protease inhibitor cocktail (78440; Thermo Fisher Scientific); 30 µg of protein lysate was resolved by SDS–polyacrylamide gel electrophoresis (4–15%) and transferred to polyvinylidene fluoride (PVDF) membranes, which were blocked with either 5% BSA in TBS-T or 5% non-fat milk in PBS-T and incubated overnight with the indicated antibody. The primary antibodies used were the following: anti-HES1 (11988S; 1:1,000; Cell Signaling Technology), anti-MCT4 (sc-376140; 1:100; Santa Cruz Biotechnology), anti-MCT1 (AB1286-I; 1:1,000; Sigma-Aldrich), anti-SOX1 (4194S; 1:1,000; Cell Signaling Technology), anti-c-FOS (ab190289; 1:500; Abcam), anti-GPX4 (ab125066; 1:1,000; Abcam), anti-4-HNE (ab46545; 1:1,000; Abcam), anti-LCB3 (43566; 1:1,000; Cell Signaling Technology), anti-alpha-tubulin-HRP (ab40742; 1:3,000; Abcam), anti-Hsp90 (610418; 1:3,000; BD Biosciences), anti-phospho-Ser133-CREB (9198; 1:500; Cell Signaling Technology) and anti-FLAG tag (2368; 1:1,000; Cell Signaling Technology). If needed, the membranes were then incubated with HRP-conjugated anti-rabbit (ab205718; 1:3,000; Abcam), anti-mouse (G-21040; 1:3,000; Invitrogen) or anti-chicken (A16054; 1:3,000; Invitrogen) secondary antibodies. The target protein bands were visualized using ECL Prime Western Blotting Detection Reagents (RPN2232; Amersham) and ChemiDoc XRS+ System (Bio-Rad). For gel source data, see Supplementary Fig. 1.

Immunofluorescence

A single fixed lung lobe was cryoprotected with sucrose (30% sucrose in PBS, overnight), embedded in optimal cutting temperature compound and slowly frozen in dry ice. Human SCLC tissue array LC703a was bought from US Biomax. Samples were cryo-sliced into 100-μm-thick sections using a Leica CM3050 S cryostat. Slices were collected on SUPERFROST PLUS adhesion microscope slides (J1800AMNZ; Epredia), washed in PBS for 30 min at room temperature and quenched in 0.24% NH₄Cl PBS for 10 min. Slides were permeabilized and saturated in a blocking solution (15% donkey serum, 0.20% glycine, 2% BSA, 0.25% gelatine and 0.5% Triton X-100 in PBS) for 2 h at room temperature and then incubated for 24 h in the blocking solution at 4 °C with the following primary antibodies: anti-MCT4 (sc-376140; 1:100; Santa Cruz Biotechnology), anti-SOX1 (AF3369; 1:200; Bio-Techne), anti-VAChT (139105; 1:100; Synaptic Systems), anti-tdTomato (AB8181-200; 1:200; SICGEN) and anti-β3-tubulin (ab52623; 1:250; Abcam). The day after, samples were washed three times in PBS for 10 min each and incubated in PBS for 1 h at room temperature with the following fluorescent conjugated secondary antibodies: donkey anti-mouse Alexa Fluor (AF) 488 (A21202; 1:250; Invitrogen), donkey anti-rabbit AF 488 (A-21206; 1:250; Invitrogen), donkey anti-goat AF 568 (A-11057; 1:250; Invitrogen), donkey anti-goat AF 647 (A32849; 1:250; Invitrogen) and goat anti-guinea pig AF 647 (A21450; 1:250; Invitrogen). The samples were then washed three times in PBS for 10 min and incubated with DAPI (D1306; 1 µg ml⁻¹; Thermo Fisher Scientific) in PBS for 5 min. The slides were mounted using the Dako fluorescence mounting medium (S3023; Agilent Technologies) and Epredia 22 × 50-mm coverslips. Images were captured with a Leica SP8 FALCON inverted confocal microscope with white light laser (470–670 nm), equipped with HC PL APO 20× NA 0.75 CB2, HC PL APO 40× NA 1.30 CS2 and HC PL APO 63× NA 1.40 CS2 oil immersion objectives and HyD detectors. The laser excitation line, power intensity and emission range were chosen according to each fluorophore to minimize bleed-through. Data were collected with LAS X software. Data were analysed using ImageJ software, and JACoP plugin was used to perform co-localization analysis and calculate Pearson’s coefficient. To calculate the cumulative axon length in proximity to the tumour, an ROI was designed 300 µm radially to the tumour, and NeuronJ plugin was used. The axon density score was calculated as the ratio between the cumulative axon length and diameter of the tumour. The number of MCT4- and SOX1-positive cells at the TMA was calculated using QuPath 0.5.0, and 3D visualization, rendering and videos have been generated using the Imaris software.

Immunohistochemistry

For immunohistochemistry analysis, tissue samples were fixed in 10% NBF, dehydrated, embedded in paraffin wax and sectioned at 4 μm using a Leica RM2235 microtome. Slides were dewaxed in xylene twice for 5 min and rehydrated with 100% industrial methylated spirit (IMS) twice for 5 min, followed by 70% industrial methylated spirit for 5 min and dH₂O for 5 min. The samples were transferred to a citrate-based antigen retrieval solution (H-3300; Vector Laboratories), microwaved at 900 W for 8 min, cooled to 50 °C, boiled again for 3 min and finally cooled to room temperature. The slides were incubated in 1.6% H₂O₂ in PBS for 10 min, washed for 5 min in dH₂O, incubated in 1% BSA for 1 h at room temperature and incubated ON at 4 °C with the following primary antibodies: anti-phospho-Ser133-CREB (9198; 1:400; Cell Signaling Technology), anti-Ki67 (ab15580; 1:500; Abcam) and anti-cleaved Caspase 3 (9579S; 1:250; Cell Signaling Technology). The samples were washed three times in PBS-T for 5 min. For the secondary antibody (anti-rabbit polymer) and 3,3′-diaminobenzidine, a BOND Polymer Refine Detection kit (DS9800; Leica) was used. Briefly, the samples were incubated with a Polymer detection system reagent for 30 min at room temperature and washed in PBS-T three times for 5 min each. 3,3′-Diaminobenzidine chromogen solution was applied, washed in dH₂O to terminate the reaction and counterstained in Sakura Tissue-Tek Prisma. Images were captured with an Olympus VS200 slide scanner. The function ‘positive cell detection’ in QuPath 0.5.0 was used to quantify Ki-67 and cleaved Caspase 3-positive cells in liver sections.

Gas chromatography–mass spectrometry

NE cells were seeded at 10⁶ cells per well, and non-NE cells were seeded at 250,000 cells per well in six-well plates in 2 ml regular DMEM per well. Six wells per cell line were seeded for two time points and three technical triplicates per time point. The controls were DMEM in mock plates without cells. The next day, the plates were washed with 5 ml PBS twice for each well, including the mock plates, and then replaced with exactly 2 ml of the fresh medium (10-017-CV with 10% dialysed FBS; Corning) using P1000 pipettes. In the meantime, three wells per cell line were counted for cell number at T0. After 24 h, 10 µl of medium was collected for gas chromatography–mass spectrometry, and cell counts were obtained for T24. Growth rates of NE cells and non-NE cells were established by fitting an exponential growth equation to the initial and final cell counts, and integration of this equation was performed to determine the average cell per unit time over the course of this experiment (area under the curve). The collected medium was stored at −80 °C or proceeded directly to the extraction of polar metabolites. High-performance liquid chromatography (HPLC)-grade methanol containing a norvaline (Sigma) internal standard at a concentration of 10 µg ml⁻¹ was added to samples in 1.5-ml Eppendorf tubes to reach a final concentration of 80% methanol, and the mixture was then vortexed for 10 min at 4 °C then centrifuged at the maximum rpm for 10 min at 4 °C. The supernatant was collected and transferred to new Eppendorf tubes and dried under inert nitrogen gas. Dried samples were derivatized to form methoxime-tBDMS derivatives by initial incubation with 16 µl MOX reagent (2% methoxyamine hydrochloride in pyridine, Thermo Scientific) at 37 °C for 90 min, followed by addition of 20 µl N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide (MTBSTFA) with 1% tert-butyldimethylchlorosilane (t-BDMCS) (Regis Technologies) and incubation at 60 °C for 30 min. The samples were then centrifuged at maximum speed for 10 min at 4 °C. Then, 16 µl of supernatant was transferred to a gas chromatography–mass spectrometry vial. Gas chromatography–mass spectrometry analysis was performed using an Agilent 6890 GC equipped with a 30 m DB-35MS capillary column connected to an Agilent 5975B MS operating under electron impact ionization at 70 eV. One microlitre of sample was injected in splitless mode at 270 °C, using helium as the carrier gas at a flow rate of 1 ml min⁻¹. The gas chromatography oven temperature was held at 100 °C for 3 min and increased to 300 °C at 3.5 °C min⁻¹. The mass spectrometer source and quadrupole were held at 230 °C, and the detector was run in scanning mode, recording ion abundance in the range of 100–605 m/z. Annotation of lactate and pyruvate peaks were determined by reference to comparison to lactate and pyruvate standards derivatized using the same procedure described above. The metabolite secretion rates were calculated by dividing the total number of extraction ions per metabolite by the average cell per unit time.

Seahorse measurement of OCR

An Agilent Seahorse Bioscience Extracellular Flux Analyzer was used to measure OCRs. Cells were seeded in Agilent Seahorse XF96 cell culture microplate (TC-treated; 102416-100; Agilent Technologies) for experiments, and the wells were coated with Cultrex BME (3432-010-01; Bio-Techne) in HBSS. SCLC NE cells were seeded at 10⁵ cells per well, whereas non-NE cells were seeded at 10⁴ cells per well in 100 µl appropriate regular culture medium. The next day, the medium was replaced by 180 µl per well Seahorse XF DMEM Medium (103575-100; Agilent Technologies) supplemented with 10 mM d-glucose, 1 mM pyruvate and 2 mM l-glutamine. The plates were incubated at 37 °C for 1 h without CO₂. The XF Cell Mito Stress Test protocol was carried out after injecting oligomycin (1.5 µM, final concentration), FCCP (1.5 µM, final concentration) and rotenone–antimycin A (0.5 µM, final concentration). After running the assay, the protein content of each well was measured with the Pierce BCA Protein Assay Kit (23225; Thermo Fisher Scientific). The total milligrams of protein were used for normalization. Data analysis was performed using the online software Seahorse Analytics (v.1.0.0-699) to calculate the basal respiration rates, ATP-coupled respiration and coupling efficiency.

In vitro calcium imaging

Cal-520, AM (ab171868; Abcam)-loaded SCLC NE cells, GCaMP6m-expressing mSCLC-NE cells alone or GCaMP6m-expressing mSCLC-NE and non-NE cells in a ratio of 4:1 were seeded in 35-mm glass-bottom dishes (81218-200; ibidi) coated with Matrigel (356231; Merck) and cultured overnight in FluoroBrite DMEM (A1896701; Thermo Fisher Scientific) supplemented with GlutaMAX (35050038; Thermo Fisher Scientific) and penicillin–streptomycin (15140122; Thermo Fisher Scientific) with 2% dialysed FBS (26400044; Thermo Fisher Scientific). They were imaged on the next day using an Olympus IX73 inverted epifluorescence microscope equipped with a pe-300 Ultra standard fluorescence illumination system with TTL trigger control and a Prime BSI Express sCMOS Camera 4.2 MP with a 95% quantum efficiency. A 40× Plan-Neofluar objective (0.75NA) was used to acquire images of cells that were stimulated with an LED light source using the green channel (excitation at 470 ± 20 nm) and a GFP filter set, and controlled by the Micro-Manager software (v.2.0.0). Recordings were made at 0.5 frames per second (Hz) for 5 min. The exposure time was 200 ms. For experiments, including the MCT inhibitor treatments, SR-13800 (5 µM) and diclofenac (0.5 mM) were added to the cell culture medium and incubated for 5 min before imaging.

Calcium imaging of 3D-cultured NE SCLC cells was performed using a custom-built upright light sheet fluorescence microscope at the Cambridge Advanced Imaging Centre.

Ex vivo calcium imaging

Precision-cut lung slices were obtained from mice using a protocol adapted from a previous study⁷⁸. A solution of 2% low-melting-point agarose (16520-100; Invitrogen) in HBSS was used for lung inflation. For imaging of tumour cells expressing Salsa6f reporter, the lung lobes harbouring the tumours were isolated and cut transversely at 300 µm using an automated vibratome (Leica VT1200S) in ice-cold HBSS/HEPES buffer.

The slices were placed in a 12-well plate in serum-free DMEM (21063029; Gibco) supplemented with 1% penicillin–streptomycin (15140122; Thermo Fisher Scientific) and incubated at 37 °C with 5% CO₂ for 30 min before mounting for imaging. The slices that were used for imaging of NEBs were incubated with Oregon Green 488 BAPTA-1-AM (O6807; Invitrogen) for 30 min before imaging.

The slices were mounted between two thin layers of a low-melting-point agarose gel, solidified at room temperature in a 24-well glass-bottom imaging plate with 1.5 cover glass (P24-1.5H-N; Cellvis). They were imaged using an Olympus CSU-W1 SoRa spinning disk confocal microscope with an environmental chamber (at 37 °C, 5% CO₂). A 30× silicon immersion objective (1.05 NA) and 488 and 561 nm laser excitation wavelengths were used for image acquisition. Recordings were made at 0.5 frames per second for 5 min for each field of view.

For experiments including treatment with the MCT4 inhibitor diclofenac (0.5 mM), time lapses were recorded from the same field of view before treatment with the drug and after 5-min incubation with the drug.

Calcium imaging analysis

For all calcium imaging experiments, individual fields of view were analysed in ImageJ (v.1.54f). The TrackMate plugin and Cellpose detector pretrained models cyto and cyto2 were used for automated segmentation of cells and tracking during the time lapse recorded for each field of view. Only cells that have been segmented and tracked at all time points of the time lapse were included in the subsequent calcium peak analysis. The mean intensity fluorescence values were obtained for each segmented cell, and the background measured in each field of view was subtracted from individual values recorded at each time point. The individual fluorescence intensity traces for each segmented cell were generated using a custom MATLAB script⁷⁹, and peak metrics were generated using the findpeaks function. The fluorescence intensity at each time point was normalized to the median of all fluorescence intensity values for a particular segmented cell. A cutoff of 0.1 was used for peak prominence. Only cells with at least one peak with a minimum peak prominence of 0.1 were considered active cells.

RNA sequencing

Total RNA from mSCLC-NE/non-NE cell lines was isolated using the miRNeasy Kit (QIAGEN). RNA libraries were prepared for sequencing with the Illumina TruSeq kit following the manufacturer’s instructions. Illumina HiSeq 2000 50-nt single-ended reads were mapped to the UCSC mm9 mouse genome build (http://genome.ucsc.edu/) using RSEM⁸⁰ (v.1.2.12) and bowtie (v.1.0.1) with default options. Raw estimated expression counts were upper-quartile normalized to a count of 1,000 (ref. ⁸¹). Given the complexity of the dataset in terms of a mixture of different biological backgrounds, a high-resolution signature discovery approach was used to characterize global gene expression profiles. Independent component analysis, an unsupervised blind source separation technique, was used on this discrete count-based expression dataset to elucidate statistically independent and biologically relevant signatures, as detailed previously¹. All RNA sequencing analyses were conducted in the R Statistical Programming language (http://www.r-project.org/). GSEA was carried out using the pre-ranked mode with default settings⁸². Heat maps were generated using the Heatplus package in R.

Analysis of published datasets

To further investigate the presence of vulnerabilities of SCLC related to the characteristic metabolic requirements observed in our experiments, we used previously published CRISPR screens performed with SCLC, LUAD and PDAC cells^42,83. The criteria chosen to identify specific vulnerabilities of SCLC were median log₂ fold change (L2FC) 84,85,86.

To extend our observations to other datasets generated with PDXs of SCLC, we analysed an RNA sequencing dataset from a cohort of 51 PDXs²⁷. Specifically, we performed a Pearson correlation analysis using their normalized RNA sequencing data in fragments per kilobase of transcript per million mapped reads and the NE scores they assigned to these PDXs. Moreover, we divided these PDXs into two groups according to their NE score: NE with NE score greater than 0.8 and non-NE with NE score less than 0.2. SOX1 and SLC16A3 (encoding MCT4) mRNA levels were analysed in these two groups. The normality of the data was assessed using the Shapiro–Wilk test, and data transformations and statistical tests were chosen accordingly, as described in figure legends.

Additionally, we analysed the RNA sequencing and proteomic datasets of a cohort of 112 patients with SCLC (with paired tumour and NAT samples) recently published⁵⁶. Further details about data acquisition, normalization and clinical information can be found in the original paper. For differential expression analyses between tumours and paired NATs, the Wilcoxon matched-pair signed-rank test was applied. For correlation analyses, the Spearman test was used. For survival analyses, we split the patients into two groups (‘high’ or ‘low’) based on whether their normalized and log₂-transformed expression was above or below the median for each of the genes of interest. We plotted Kaplan–Meier curves for ‘high’ and ‘low’ patients and compared both groups using a univariate log-rank test.

Statistics and reproducibility

Unless stated otherwise, all statistical analyses were performed in GraphPad Prism using the recommended tests and post hoc tests from the software. No data have been excluded from the analyses. The number of replicates for each experiment is reported in the corresponding figure legend and methods. Western blotting was performed with at least two biological replicates. No statistical methods were used to calculate the sample size. Sample sizes were chosen based on preliminary experiments aimed at capturing biological effects in line with similar research in the field.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.