Ex vivo culture of cells derived from circulating tumour cell xenograft to support small cell lung cancer research and experimental therapeutics

Abstract

Background and Purpose

Small cell lung cancer (SCLC) is an aggressive disease with median survival of <2 years. Tumour biopsies for research are scarce, especially from extensive‐stage patients, with repeat sampling at disease progression rarely performed. We overcame this limitation for relevant preclinical models by developing SCLC circulating tumour cell derived explants (CDX), which mimic the donor tumour pathology and chemotherapy response. To facilitate compound screening and identification of clinically relevant biomarkers, we developed short‐term ex vivo cultures of CDX tumour cells.

Experimental Approach

CDX tumours were disaggregated, and the human tumour cells derived were cultured for a maximum of 5 weeks. Phenotypic, transcriptomic and pharmacological characterization of these cells was performed.

Key Results

CDX cultures maintained a neuroendocrine phenotype, and most changes in the expression of protein‐coding genes observed in cultures, for up to 4 weeks, were reversible when the cells were re‐implanted in vivo. Moreover, the CDX cultures exhibited a similar sensitivity to chemotherapy compared to the corresponding CDX tumour in vivo and were able to predict in vivo responses to therapeutic candidates.

Conclusions and Implications

Short‐term cultures of CDX provide a tractable platform to screen new treatments, identify predictive and pharmacodynamic biomarkers and investigate mechanisms of resistance to better understand the progression of this recalcitrant tumour.

Abbreviations

CDX

circulating tumour cells derived explant

CTC

circulating tumour cells

EdU

5‐Ethynyl‐2′‐deoxyuridine

ES

extensive stage

GEMM

genetic engineered mouse model

GFP

green fluorescent protein

HR

hazard ratio

LOH

loss of heterozygosity

OS

overall survival

PDX

patient‐derived xenograft

SCLC

small cell lung cancer

RPKM

Reads Per Kilobase of transcript, per Million mapped reads

SNV

single nucleotide variant

SOC

standard of care

VSVG

vesicular stomatitis virus glycoprotein

What is Already Known

• Circulating tumour cells derived explants (CDX) are new preclinical models that replicate patient disease and response to chemotherapy.

• Cells derived from CDX tumours can be grown ex vivo.

What this Study Adds

• Characterization of CDX‐derived tumour cells growing in tissue culture versus mice for cancer pharmacology studies.

What is the Clinical Significance

• CDX cultures can be used to screen compound libraries and identify effective treatments.

• CDX cultures can be manipulated to validate/discover mechanisms of therapy resistance.

Introduction

Lung cancer is responsible for the highest cancer mortality worldwide, causing 19% of total tumour‐associated deaths (Ferlay et al., 2015). Of all the lung cancers, small cell lung cancer (SCLC) accounts for 15–20% of cases globally (Gustafsson et al., 2008; van Meerbeeck et al., 2011). SCLC is a highly aggressive disease characterized by rapid cell proliferation and early dissemination. For this reason, most patients have already developed extensive stage (ES) disease with distant metastases by the time of diagnosis. Standard of care (SOC) treatment for ES SCLC is a combination of a platinum‐based agent and a topoisomerase inhibitor, and while most patients initially respond to treatment, 95% relapse with resistant disease within 3–18 months (Lally et al., 2007; van Meerbeeck et al., 2011; Alvarado‐Luna and Morales‐Espinosa, 2016). Despite multiple SCLC clinical trials, there have been no improvements in overall survival (OS) in the past 30 years (Koinis et al., 2016), and in 2013, the National Cancer Institute (2014) placed SCLC into the recalcitrant tumour category and made it a research priority. The lack of progress has been attributed to an inadequate understanding of the genomic landscape and biology of SCLC, the limited availability of tumour tissue for research and a paucity of patient‐derived preclinical models for target validation and therapy testing. More recently, whole‐genome sequencing of 110 SCLC patients (Peifer et al., 2012; Rudin et al., 2012; George et al., 2015) revealed an almost universal loss of the tumour suppressor genes TP53 and RB1. In contrast to similar studies investigating non‐SCLC, no frequently recurrently mutated oncogenes were identified making genome‐driven personalized medicine particularly challenging (Peifer et al., 2012; Rudin et al., 2012; George et al., 2015; Koinis et al., 2016; Santarpia et al., 2016). Only 5% of all SCLC patients undergo surgical resection (Koinis et al., 2016), with the remainder undergoing biopsies, which are highly invasive, poorly tolerated procedures resulting in small and often necrotic specimens with as few as a hundred tumour cells. The establishment of over 50 SCLC cell lines has been instrumental to research progress and the initial characterization of the SCLC genomic landscape (Carney et al., 1985; Phelps et al., 1996; Pleasance et al., 2010); however, SCLC cell lines were typically generated from small biopsies that cannot replicate the tumour heterogeneity, thus potentially explaining the discordance between laboratory research and clinical efficacy (Santarpia et al., 2016). Drawbacks of cell lines include limited clinical details of donor patients, lack of paired germline samples and a strong clone selection after decades of propagation on plastic (Garnett and McDermott, 2014). In vivo models, such as genetically engineered mouse models (GEMM) and patient‐derived xenograft (PDX), reproduce SCLC disease better than established cell lines as they develop in the host tissue micro‐environment. GEMMs have resulted in the clarification of important steps in SCLC progression (Meuwissen et al., 2003), but simultaneous deletion of Trp53 and Rb1 cannot entirely reproduce the complex genomic instability caused by tobacco smoke, the leading cause of SCLC. PDXs are derived directly from human SCLC and maintain the donor patient's tumour characteristics and can be used to test therapies (Bertotti et al., 2011). However, as the establishment of PDX requires tumour tissue that is limited in ES SCLC, a bias towards limited rather than ES models is present, resulting in PDXs representing the minority rather than the majority of SCLC patients at diagnosis.

Recently, we demonstrated that SCLC circulating tumour cells (CTC) are prevalent (Hou et al., 2009) and prognostic (Hou et al., 2012), with at least a subpopulation being tumorigenic in immune‐compromised mice resulting in CTC patient‐derived explant (CDX) models (Hodgkinson et al., 2014). The CDX models closely resemble the donor patient's tumour with similar cellular phenotypes and responses to SOC chemotherapy (Hodgkinson et al., 2014). Furthermore, CDX models have allowed the expansion of SCLC patient tissue for molecular characterization, drug target validation and drug testing (Potter et al., 2016). Additionally, the fact that CDX cultures are derived from minimally invasive blood samples allows the generation of paired longitudinal models from a patient throughout the course of his or her disease. This is an important advantage of the CDX approach as we can assess and compare both biology and therapy responses before and after acquired chemoresistance.

Despite the clear advantages offered by this approach, the establishment of CDX models and in vivo pharmacology studies are expensive and typically take several months. Moreover, mechanisms of drug resistance cannot be readily validated functionally in vivo. This prompted us to investigate whether CDX cells can be grown as short‐term ex vivo cultures, which maintain their clinical relevance, to expedite studies of SCLC biology and pharmacology. With this short report, we hope to add a new patient‐relevant, tractable platform to the current portfolio of methodologies for SCLC research that complements the models available and facilitates investigations to understand the biology of this aggressive disease.

Methods

Cell culture

U2OS, H69 and HCT116 osteosarcoma, SCLC and colorectal cancer cell lines, respectively, from the American Type Culture Collection (ATCC, Manassas, VA) were cultured in RPMI 1640 (GIBCO, Life Technologies, Paisley, UK) supplemented with 10% FBS (Biosera, Labtech International Ltd, East Sussex, UK); SW480, colorectal cancer cell line (ATCC, Manassas, VA), were cultured in DMEM (GIBCO, Life Technologies) supplemented with 10% FBS and glutamine (GIBCO, Life Technologies).

CDX‐derived cells were grown in HITES media (RPMI 1640 supplemented with 50 μg·mL⁻¹ insulin, 100 μg·mL⁻¹ transferrin, 100 nM ß‐oestradiol, 300 nM sodium selenite and 100 nM hydrocortisone, all from Sigma‐Aldrich, Poole, UK) supplemented with 5 μM of Y‐27632 (Selleckchem, Bio‐Techne, Abingdon, UK) with or without the addition of 2.5% FBS. When CDX cultures were passaged, cells were dissociated for 5–10 min with StemPro Accutase (Thermo Fisher Scientific, Paisley, UK) at 37°C. After being dissociated, cells were washed once with PBS and then resuspended in fresh HITES media supplemented with a ROCK inhibitor and FBS.

Animal studies

All procedures were carried out in accordance with the Home Office Regulations (UK) and the UK Coordinating Committee on Cancer Research guidelines using approved protocols (Home Office Project licence no. 70/8252 and Cancer Research UK Manchester Institute Animal Welfare and Ethical Review Advisory Board). Animal studies are reported in compliance with the ARRIVE guidelines (Kilkenny et al., 2010). Disaggregated tumour cells were resuspended in an ice‐cold 1:1 mixture of RPMI and Matrigel (VWR, International Ltd, Lutterworth, UK), and 100–200 μL were injected s.c. into one or both flanks of 6‐ to 10‐week‐old female SCID‐bg mice (CB17.Cg‐Prkdc ^scid Lyst ^bg‐J/Crl, Harlan, Envigo Laboratories Ltd, Bicester, UK and Envigo RMS Sarl, Gannat, France). Mice were kept under 12 h light/12 h dark environment cycle and maintained at uniform temperature and humidity. They were housed in individually‐vented caging systems in groups of five mice per cage. Mice were monitored for signs of tumour growth, and once a palpable tumour was present, this was measured twice a week by calipers. Tumour volumes were calculated as 0.5 × tumour length × (tumour width)². When the total tumour burden was greater than 600 mm³, tumours were harvested for disaggregation. The animals were sacrified following the Schedule 1 regulation under the Animals (Scientific Procedure) Act 1986: neck dislocation followed by confirmation of death, either by rigor mortis or removal of heart. Necropsy was performed after the animals had been killed to check internal organs for evidence of metastasis or other pathologies.

For assessment of tumourigenicity and SOC sensitivity following culture, NSG mice were implanted with CDX3‐derived cells after 4 weeks of culture. Tumour monitoring and efficacy studies were carried out with cisplatin/etoposide, as previously described (Hodgkinson et al., 2014). Five animals were randomized in each cohort. Randomization was performed when tumours reached 150–250 mm³ and mice were assigned to each group based on the following sequence: “1‐2‐2‐1‐1‐2‐2‐1…” where 1 and 2 are the drug and vehicle treated cohorts. Similarly, CDX3 and CDX3‐GFP⁺ tumours were monitored twice a week by calipers.

For the efficacy study with TH1579, SCID‐bg mice were injected with viable CDX3 tumour cells. When the total tumour burden was between 150 and 250 mm³, mice were randomized to either vehicle or TH1579 treatment, as described above. Nine mice were assigned to the drug treated cohort while eight mice received the vehicle only treatment. Mice received 90 mg·kg⁻¹ TH1579 by oral gavage, twice daily (6 h apart) for three cycles consisting of 3 days on and 4 days off. The formulation was prepared daily as a 9 mg·mL⁻¹ solution (administering 10 mL·kg⁻¹). The vehicle was the same, but with the addition of 10% ethanol. Tumour volume was monitored twice a week until tumours reached four times their initial tumour volume (4 × ITV) or until animal health deteriorated (censored in survival analysis).

For the efficacy study with GDC‐0941, CDX4‐bearing mice were randomized, when the tumour reached 150–250 mm³, into a vehicle or GDC‐0941‐treated group, as described above. Ten mice were assigned in each cohort. Treatments were administered as previously described (Potter et al., 2016) with 21 consecutive days of 75 mg·kg⁻¹ GDC‐0941 dosing. Tumour volume was monitored twice a week until tumours reached 4 × ITV or until animal health deteriorated (censored in survival analysis).

Pharmacodynamic studies were performed to assess the target inhibited by TH1579 and GDC‐0941. For TH1579, mice were treated for four consecutive days with either vehicle or 90 mg·kg⁻¹ TH1579. On the fourth day, mice were killed 2 h after dosing, and relevant tissue was collected. For GDC‐0941, the endpoint was at 4 or 24 h after a single dose of vehicle or GDC‐0941.

CDX‐derived cell generation

Harvested tumours were disaggregated with the gentleMACS dissociator (Miltenyi Biotec, Bisley, UK) as previously described (Jahchan et al., 2016). After the procedure, tumour cells were pelleted at 250 g for 5 min and resuspended in HITES media alone. The cells were counted and checked for viability with trypan blue dye exclusion (Sigma‐Aldrich).

Proliferation assay

CDX‐derived cells were seeded in 96‐well plates to test different media conditions. HITES media were used as standard alone or with the addition of 5 μM of Y‐27632 (ROCK inhibitor) with or without 2.5% of FBS. To evaluate cell proliferation, every 2 days, the cells were assessed for ATP content with the Cell‐Titer Glo 3D viability assay (Promega, Southampton, UK), as per the manufacturer's instructions. Every experiment was performed in triplicate and repeated at least twice.

Drug screening

CDX‐derived cells and established cell lines were seeded in the appropriate media formulation; 24 or 48 h after seeding, cells were treated with increasing concentrations of the compound of interest. After 3 or 7 days, the viability of the cells was assessed using the Cell‐Titer Glo 3D viability assay. DMSO or saline was used as a vehicle to the dissolve drugs. For combinatorial screening, single dose–response curves and viability matrix were computed with the open source CRUK software Combenefit (Di Veroli et al., 2016).

Flow cytometry

Cells were grown for 1 to 2 weeks in appropriate media conditions. After treatment with 2.5 μM 5‐Ethynyl‐2′‐deoxyuridine (EdU) for 14 h, CDX‐derived cells were analysed with Click‐iT Plus EdU Alexa Fluor 488 Flow Cytometry Assay kit (C10632, Life Technologies, Thermo Fisher) and propidium iodide (PI)/RNAse staining solution (4087S, Cell Signaling Technologies New England Biolabs, Hitchin, UK) according to the manufacturer's recommendations. After being stained, cells were run on the flow cytometer (LSRFortessa II), and data were analysed with FlowJo software (v10). Cells were gated to remove doublets. In the single cell gate, 10 000 events were recorded and evaluated to calculate the percentage of EdU‐positive cells in the PI population.

Western blotting

Forty‐eight hours after being seeded, cells were treated with 0.5 μM of TH1579, 0.5 μM of ABT‐737, 1 μM of GDC‐0941, GDC‐0941/ABT‐737 in combinations or untreated (0.01% DMSO). Total cellular protein was extracted after 8 h of ABT‐737 and GDC0941 treatment or 72 h of TH1579 and DMSO treatment, by solubilizing the cells in cold lysis buffer (10 mM HEPES pH 7.1, 50 mM NaCl, 0.3 M sucrose, 0.1 mM EDTA, 0.5% Triton X‐100 and 1 mM DTT) in the presence of Protease Inhibitor Cocktail and Phosphatase Inhibitor Cocktail I and III (Sigma‐Aldrich). Protein concentration was determined by the BCA protein assay reagent kit (Thermo Scientific). Protein extracts were resuspended in NuPAGE LDS sample buffer 4× (Invitrogen, Thermo Fisher) and 10× NuPAGE sample reducing agent (Invitrogen). Total protein was resolved on a NuPAGE 4–12% Bis‐Tris 1.0 mm gel (Invitrogen). After transfer onto PVDF membranes (Perkin‐Elmer, Boston, MA), blots were incubated with the appropriate primary antibody and the corresponding HRP‐coupled secondary IgG (Dako, Agilent Technologies, Cheadle, UK). Detection was obtained with the Western Lightning Chemiluminescence Reagent Plus‐ECL (Perkin‐Elmer Beaconsfield, UK). Details of antibodies can be found in Supporting Information Table S1.

Immunohistochemistry and image analysis

Chromogenic staining was performed on formalin‐fixed, paraffin‐embedded 4 μm tumour sections and on CDX‐derived cell pellets using antibodies listed in Supporting Information Table S2. Staining for CD56, synaptophysin, cytokeratins and cleaved PARP was performed using the BondMax autostainer and reagents (Leica Microsystems). Vimentin staining was performed using the Ventana autostainer and reagents (UltraView Universal DAB Detection Kit, Ventana). For these antigens, digital images of whole tissue sections were acquired using a Leica SCN400 histology scanner (Leica Microsystems Milton Keynes, UK). Images were analysed using Definiens Developer XD and the Tissue Studio Portal version 3.51 (Definiens AG, Munich, Germany) excluding the stromal regions.

Manual 8‐oxoguanine (8‐oxo‐G) staining employed the Ultra Vision LP Detection System (Thermo Scientific) according to the manufacturer's instructions. Evaluation of 8‐oxo‐G was performed by counting the percentage of cancer cells exhibiting nuclear positivity. Three independent observers carried out slide examination for 8‐oxo‐G staining, with minimal inter‐observer variability. Five different fields per slide were analysed, with 100 cells counted per field (total number = 500 cells).

Cas9‐GFP CRISPR system

To generate green fluorescent protein (GFP)‐expressing CDX3 cells, cells were infected with the lentiCas9‐EGFP plasmid, which was a gift from Phil Sharp and Feng Zhang (Addgene, Watertown, US, plasmid #63592). For virus production, 750 000 H293‐LentiX cells were seeded into a single well of a six‐well plate in 2 mL of DMEM (10% FBS, gentamicin and glutamicin). On the following day, transfection was performed with 880 ng lentiCas9‐EGFP, 572 ng pMDL, 308 ng vesicular stomatitis virus glycoprotein (VSVG) and 220 ng REV using FuGENE (Promega). The day after, medium was replaced with DMEM (10% FBS, 5% glutamine and gentamicin). Forty‐eight hours post‐transfection, the virus was harvested, centrifuged for 5 min at 250× g and supernatant filtered through a 0.45 μm acrodisc syringe filter (VWR). Afterwards, 5 × 10⁶ CDX3 cells were spin infected with 1 mL of virus by spinning cells at 700× g for 45 min at 37°C in the presence of 6 mg·mL⁻¹ polybrene (Sigma). The day after infection, virus‐containing medium was removed. Forty‐eight hour post‐infection, the top 40% of GFP⁺ cells were sorted using a BD Aria II (BD Biosciences). CDX3 cells were recovered for 1 week in plastic and then implanted into 5 NOD scid γ (NSG) mice using 100 000 cells per flank.

To perform the touch preparations, a piece of tumour generated from the CDX3‐GFP⁺ cells was gently touched to the surface of a cell adhesion slide (Marienfeld Superior, Lauda‐Königshofen, Germany). Slides were air dried and imaged for GFP expression.

Sanger sequencing

Genomic DNA was extracted from cell pellets with the QIAmp DNA mini Kit (Qiagen, Manchester, UK). The extracted genomic DNA was amplified by PCR using primers flanking relevant TP53 mutations and subsequently Sanger sequenced with an ABI3130xl 16 capillary system with appropriate primers: CTCACAACCTCCGTCATGTG (CDX2), CCCTTAGCCTCTGTAAGCTT (CDX3) and TCCCAAGACTTAGTACCTGAAG (CDX4).

RNAseq

Total RNA was extracted with the RNeasy mini kit (Qiagen). RNA with RNA integrity number >7.6 (Agilent 2100 Bioanalyzer) was used to generate libraries with SureSelect poly A samples (Agilent) and sequenced on the NextSeq 500 using 75 bp paired end. Tumour RNAseq data were filtered via a novel algorithm to distinguish human and mouse transcripts (Khandelwal et al., 2017). Subsequently, both the tumour and the CDX‐derived cell culture samples were aligned to human GRCh38 assembly using Mapsplice (Version 2.1.6) (Wang et al., 2010). The aligned reads were then used to generate count data using an Rsubread package (Version 1.16.1) (Liao et al., 2013) in R using Ensembl 77 GTF file. The counts were converted into Reads Per Kilobase of transcript, per Million mapped reads (RPKM) values using the edgeR package in R (Robinson et al., 2010). Differential expression analysis was performed on the data obtained using the limma package in R (Ritchie et al., 2015). The P‐values were corrected using the false discovery rate (FDR) correction using the p.adjust() function in R. Using an FDR cut‐off of P ≤ 0.05, the final list of differentially regulated genes was obtained. Gene ontology and pathway overrepresentation analysis were performed using the gProfiler package in R (Reimand et al., 2016). All RNAseq data were derived from three independent batches of cells derived from the same CDX tumour or from three independent CDX tumours.

Cellular thermal shift assay

Tumours were pulverized in liquid nitrogen and resuspended in 300 μL TBS including complete protease inhibitor cocktail (Roche). Aliquots were incubated for 3 min at 63, 62.1, 60.5, 57.5, 53.9, 51, 49, 48 or 25°C. Samples were then frozen at −80°C and thawed three times to lyse the cells and centrifuged at 17 000× g for 20 min; 4 × Laemmli buffer was added to the supernatants before heating to 95°C for 10 min. Samples were run on SDS‐PAGE and transferred to nitro‐cellulose membranes. Membranes were blocked in TBS/0.1% Tween20/5% milk for 1 h, incubated with primary antibodies for 1 h, washed three times in TBS/0.1% Tween20 and incubated with secondary antibodies for 1 h. Membranes were analysed using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific), and measurements were performed using Odyssey Imaging System (Li‐Cor). The bands were scanned and normalized to the band of the 25°C sample.

Statistical analysis

All statistical analyses were performed using Prism v6.05 (GraphPad) or in R (v. 3.2.1) (R Development core Team, 2015). All data from the proliferation assay and the drug screening are presented as means ± SD of at least three independent experiments, each with three experimental replicates. The only exception is the drug screening of CDX4 culture treated with cisplatin and etoposide or the combination, where only two biological replicates were performed. Doubling times were calculated by non‐linear curve fitting of an exponential growth equation. In vivo experimental data points represent individual relative tumour volumes (percentage of change from baseline). Event‐free survival was defined as the time for the tumour to reach four times the initial tumour volume. Survival analysis was performed with comparison of survival curves by log–rank (Mantle Cox) test. Hazard ratio (HR) has been calculated with the log–rank approach on Prism. Briefly, the software compute the number of deaths in each group and the number of expected events assuming a null hypothesis of no difference in survival. The values obtained define if the death rate was increased (>1) or decreased (<1) in the treated group versus the vehicle control. For pharmacodynamics analysis, an unpaired Wilcoxon test was applied to compare the vehicle versus the treated arm. Box and whisker plots are used to show the distribution of the sample. Boxes indicate median, first and third quartiles, while whiskers indicate maximum and minimum values.

Materials

Cisplatin and etoposide were purchased from Sigma‐Aldrich; GDC0941, ABT737 and Y‐27632 were purchased from Selleckchem. TH1579 (Warpman Berglund et al., 2016a) was supplied by Thomas Helleday, Karolinska Institute, Stockholm.

All compounds were dissolved in DMSO (Sigma‐Aldrich), except cisplatin, which was dissolved in saline. Drugs were aliquoted and stored according to the manufacturer's guidelines.

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 (Alexander et al., 2017a, 2017b).

Results

SCLC cells derived from CDX models can be isolated and cultured

To determine if CDX‐derived cells can be grown ex vivo, we chose three well‐characterized CDXs from our library of CDX models. CDX2, CDX3 and CDX4 encompass the wide range of responses to SOC chemotherapy (cisplatin/etoposide) in vivo (Hodgkinson et al., 2014). CDX3 was generated from CTCs collected from a chemosensitive patient (OS 9.7 months) and demonstrated sensitivity to cisplatin/etoposide in vivo (Hodgkinson et al., 2014). CDX2 and CDX4 were derived from chemorefractory patients, a clinical definition based on disease progression occurring within 90 days of finishing treatment (van Meerbeeck et al., 2011). Of note, patient 2 showed a better OS than patient 4 (3.5 vs. 0.9 months, respectively), and this was reflected in CDX2, which exhibited 50% tumour regression after cisplatin/etoposide, compared with no treatment effect in CDX4 (Hodgkinson et al., 2014).

CDX tumours were dissociated, and human tumour cells were cultured in HITES medium (Gazdar and Oie, 1986) supplemented with ROCK inhibitor Y‐27632 to improve the viability of the primary cells as already demonstrated by others (Watanabe et al., 2007). FBS was added to support cell growth (Supporting Information Figure S1A–C). CDX‐derived cells grew as non‐adherent cell clusters of variable size (Figure 1A), reminiscent of the majority of established SCLC cell lines (Gazdar et al., 1980). They showed a doubling time of ~4 days, similar to what has been observed in SCLC cell lines (Gazdar et al., 1985) and an average proliferative index of 8–18%, as defined by EdU incorporation (Figure 1B,C).

Figure 1.

CDX‐derived cells grow ex vivo. (A) Representative images of ex vivo cultures derived from CDX2 (blue), CDX3 (pink) and CDX4 (green) as well as the established SCLC cell line H69 (black). Magnification: 20×. (B) Doubling time of CDX2, 3 and 4 ex vivo. Each point represents independently derived cultures and includes data from all plating densities tested. Error bars represent SD. (C) Proliferation of CDX2, CDX3 and CDX4 cultures was assessed via EdU incorporation after 15 h of labelling. Graphs depict the percentage of viable cells that were EdU positive in three independent experiments. Error bars represent SD.

Short‐term culture of CDX‐derived cells maintains in vivo characteristics

We previously reported that CDX tumours mirror the pathology, genetics and responses to standard chemotherapy of the donor patient tumours (Hodgkinson et al., 2014). Our aim was to maintain these similarities in ex vivo CDX cultures. However, cells derived from SCLC PDXs are reported to substantially change their gene expression profile after 6 months of culturing on plastic when compared with the corresponding PDX (Daniel et al., 2009). To assess whether culturing markedly affected the differentiation status of the cells, CDX2, CDX3 and CDX4 cultures were analysed by chromogenic for the expression of the typical SCLC diagnostic markers. All three models expressed CD56 and synaptophysin (neuroendocrine markers), and cytokeratins (epithelial markers), both in the original CDX tumours and after 2 weeks of ex vivo culture (Figure 2A). CDX4 tumours showed a mixed population of vimentin positive and negative cells, which was maintained in the CDX4‐derived cells grown ex vivo, whereas CDX2‐ and CDX3‐derived cultures were negative for vimentin, consistent with the staining of their respective CDX tumours.

Figure 2.

CDX‐derived cells maintain phenotypic and molecular characteristics of the donor tumour. (A) Analysis of typical SCLC markers. After 2 weeks in culture, CDX cultures were paraffin‐embedded and then stained for the indicated proteins alongside representative CDX tumours. Pan‐CK, pan‐cytokeratin; SYN, synaptophysin; VIM, vimentin. Scale bar: 25 μm. (B) Targeted TP53 sequencing. Relevant regions of TP53 were sequenced to examine concordance between single nucleotide variants (SNVs) previously identified in the CDX and the corresponding model after 2 weeks in culture; SNVs are highlighted in yellow. The wild‐type TP53 allele does not appear to be due to loss of heterozygosity (LOH). (C) Multidimensional scaling correlates CDX2, CDX3 and CDX4 ex vivo cultures with the corresponding tumour tissue. The scale on the figure indicates root mean square of the log2 fold changes between the samples. Dark red, blue and green dots represent CDX2, CDX3 and CDX4 tumours, respectively, while light‐coloured dots represent the corresponding CDX ex vivo cultures after 2 weeks. (D) Schematic for the samples collected for the RNAseq in (E) and Supporting Information Figure S2D. (E) Comparison of the expression profiles of original CDX3 tumours, CDX3‐derived culture and CDX3 tumours after culture (see schematic D). Hierarchical clustering of the top 947 differentially expressed protein‐coding genes (FDR < 0.05) is depicted in the heat map. The colour key represents the z‐score scale of the log2 RPKM values.

Afterwards, we examined the p53 status of each culture. The tumour suppressor gene TP53 is almost universally inactivated in SCLC (George et al., 2015), and we previously identified missense mutations in CDX2 and CDX3 (c.440A>C and c.263A>G, respectively) and a nonsense mutation in CDX4 (c.892G>T) (Hodgkinson et al., 2014). All of these mutations were confirmed in the corresponding CDX cultures, indicating that this salient genomic feature of SCLC was maintained ex vivo (Figure 2B).

To investigate the changes imposed by our culture protocol more comprehensively, RNAseq analysis was performed to compare the gene expression profile of CDX tumours with matched CDX cultures. For all models tested, multidimensional scaling showed that the ex vivo cultures closely resembled the corresponding CDX tumour (Figure 2C). To evaluate further the effects of the culture, we analysed the transcriptomic profile of CDX3 harvested at different time points during ex vivo growth. Of the 41 720 expressed transcripts investigated, 13 258 (32%, FDR ≤ 0.05) were differentially expressed between the tumour and the cultures, and the majority were non‐coding (~60%, Supporting Information Figure S2A). Most of these changes appeared in the first week of culture with minimal trascriptional adaptation thereafter (Supporting Information Figure S2A). This suggests that short‐term culturing of these cells (4–5 weeks) could limit the strong clonal selection imposed by long‐term cultures, as observed in a previous report (Daniel et al., 2009).

To determine whether ex vivo CDX cultures maintain their tumourigenicity, 4 week CDX3 cultures were reinjected into immuno‐compromised mice. All mice developed tumours after an average of 6 weeks, similar to what was obtained when passaging CDX tumours in the absence of culturing (Supporting Information Figure S2B). Chromogenic analysis of CDX3 tumours derived after short‐term culture and matched original CDX3 tumours demonstrated similar cell morphology, tissue organization and expression of neuroendocrine and epithelial markers (Supporting Information Figure S2C). We then compared the CDX derived directly from the patients blood sample (here called as original CDX tumour) with the CDX‐derived culture and CDX tumours derived after culture (here called as CDX tumour after culture) (Figure 2D and Supporting Information Figure S2D). We observed that the number of genes differentially expressed between the original CDX3 tumours that have never been in culture and the CDX3 tumours derived after culture are less than the genes differentially expressed between original CDX3 and the CDX3 derived cultures (18.9% vs. 28.4%, respectively, Supporting Information Figure S2D). Almost all of the top 947 differentially expressed protein‐coding genes between CDX3 culture and the original CDX3 tumour were re‐acquired when the cultured cells were reinjected in immuno‐compromised mice, suggesting that most of these changes are reversible in the short term (Figure 2E).

Overall, these data demonstrate that CDX cultures maintain the majority of molecular features of the patient's CDX tumour from which they were derived, with some disparity as expected from the fundamental change in the environment in which they grow. Been aware of similarities and differences between the cultures and the donor CDX tumour, we decided to test whether this ex vivo model can support drug development for SCLC.

CDX‐derived cells can be used as a platform to test novel compounds

To test whether our CDX ex vivo cultures are predictive of the CDX tumour and donor patient treatment responses, we first challenged the cultures with the SCLC SOC chemotherapies, cisplatin and etoposide. The efficacy of this treatment has been established in both the corresponding patients and their CDX models (Hodgkinson et al., 2014). When tested ex vivo, CDX3 cultures exhibited the highest sensitivity to cisplatin and etoposide monotherapies as well as the combination (GI₅₀ = 80, 50 and 0.8 nM, respectively) (Figure 3B, Supporting Information Figure S3D–F). Conversely, CDX4 cultures were relatively resistant to both drugs alone or in combination (GI₅₀ = 5.0, 1.4 and 2.5 μM, respectively) (Figure 3C, Supporting Information Figure S3G–I). CDX2 displayed intermediate resistance to cisplatin (GI₅₀ = 1.5 μM) but was relatively sensitive to etoposide (GI₅₀ = 20 nM) (Figure 3A, Supporting Information Figure S3A–C). These responses correlated with the response of CDX tumours to cisplatin/etoposide in vivo (Figure 3D). Interestingly, CDX2 was much more sensitive to etoposide than cisplatin in vitro. This phenomenon may explain the partial response to the combination observed in vivo (Hodgkinson et al., 2014). To further validate our ex vivo cultures, we reinjected CDX3 cells after 5 weeks of culture into immune‐compromised mice and treated them with cisplatin/etoposide or vehicle. CDX3 tumours derived from the cultures demonstrated similar sensitivity to this treatment as was formerly observed in the original CDX3 tumours (Hodgkinson et al., 2014) (Supporting Information Figure S3J).

Figure 3.

CDX‐derived cells mimic the response to therapies of the corresponding CDX in vivo. (A–C) In vitro response of CDX2 (A), CDX3 (B) and CDX4 (C) cultures to cisplatin (red), etoposide (black) or the combination (blue) after 1 week of treatment. For combination experiments, increasing amounts of etoposide were added to 0.2 μM cisplatin. One of at least three representative experiments for each model is shown. (D) Maximum in vivo tumour regression correlates with in vitro drug concentration necessary to reach 50% of growth inhibition (GI₅₀) values for the combination. (E) In vitro response of CDX2 cultures to GDC‐0941 (dark red), ABT‐737 (blue) or the combination (light blue) after 1 week of treatment. For combination experiments, increasing amounts of GDC‐0941 were added to 0.15 μM ABT‐737. One of three representative experiments is shown. (F) CDX2 cultures were treated with DMSO or 0.5 μM of the indicated compound for 8 h. Protein lysates were probed for cleaved PARP and pAkt S473 as biomarkers for ABT‐737 and GDC‐0941 activity respectively.

In addition to SOC studies, we have previously reported the use of targeted agents in CDX models, including the combination of the PI3K inhibitor GDC‐0941 and the BH3 mimetic ABT‐263 in CDX2 tumours (Potter et al., 2016). Although GDC‐0941 had very limited activity against CDX2, ABT‐263 treatment elicited a robust, albeit transient tumour regression. The combination of the two drugs enhanced and prolonged the effect of ABT‐263 monotherapy and significantly improved the OS of mice bearing CDX2 tumours. Following on from these findings, we tested the effect of this combination on CDX cultures. CDX2 ex vivo cultures were treated with GDC‐0941, ABT‐737 (the tool compound equivalent to the bioavailable clinical candidate ABT‐263) or a combination of the two. The GI₅₀ values for GDC‐0941, ABT‐737 and the combination were 0.33 μM, 30 nM and <5 nM, respectively, mirroring the trend seen in the in vivo findings (Figure 3E, Potter et al., 2016). Immunoblot analysis confirmed that PI3K inhibition enhanced ABT‐737‐induced apoptosis as determined by increased cleaved PARP as was previously observed in vivo (Potter et al., 2016) and also increased Akt degradation (Figure 3F). Given the emerging body of evidence that this treatment combination is effective in SCLC (Gardner et al., 2014; Faber et al., 2015; Potter et al., 2016), we questioned whether CDX3 and CDX4 ex vivo cultures were also sensitive to this combination. Both CDX3 and CDX4 ex vivo cultures exhibited sensitivity, re‐enforcing the hypothesis that this combination may warrant clinical investigation in SCLC (Supporting Information Figure S4A,B). Notably, the most chemorefractory of all CDX models generated to date, CDX4, was approximately sixfold more sensitive to PI3K inhibition compared with the chemosensitive model CDX3 (GI₅₀ = 0.19 vs. 1.17 μM), further confirmed by increased cleaved PARP levels upon GDC‐0941 treatment of CDX4 cultures (Supporting Information Figure S4C). We observed elevated baseline pAkt levels in CDX2 and CDX4 cultures compared with CDX3 (Figure 3F, Supporting Information Figure S4C), implying that CDX2 and CDX4 may be more dependent on PI3K/Akt signalling than CDX3 and suggesting a new therapeutic option for chemorefractory SCLC. Overall, these data showed that CDX cultures recapitulate the in vivo CDX responses to both cytotoxic chemotherapies and targeted agents.

Identification of novel therapeutic targets in SCLC

Our results clearly demonstrate that a retrospective comparison of treatment responses in CDX in vivo and CDX culture ex vivo correlates; however, it remained to be determined whether the CDX cultures could predict in vivo efficacy. Based on the data obtained ex vivo from CDX4 treated with the PI3K inhibitor, GDC‐0941, we decided to test if a similar phenotype was observed in vivo. CDX4 tumours were treated with GDC‐0941 or vehicle for 21 consecutive days. The tumours did not regress upon treatment with the PI3K inhibitor, but disease progression was slowed down until the drug was discontinued (Figure 4A). The event‐free survival of CDX4 tumour‐bearing mice treated with GDC‐0941 increased compared with vehicle‐treated mice (32 vs. 26 days, P = 0.0095, Supporting Information Figure S5A), and the rate of death (calculated as HR between the treated and vehicle group) was lower for mice that received GDC‐0941 treatment than mice that received cisplatin/etoposide combination (HR 0.37 vs. 0.68, respectively, Hodgkinson et al., 2014). Pharmacodynamics analysis showed inhibition of the PI3K pathway with reduction in the phosphorylation of both Akt and S6, but there was no increase in the percentage of apoptotic cells, defined by cleaved PARP level (Figure 4B–D). These data suggested a cytostatic rather than cytotoxic effect of GDC‐0941 in CDX4.

Figure 4.

CDX4 progression is stabilized upon GDC‐0941 treatment. (A) Tumours were treated with GDC‐0941 (red) or vehicle (blue) for 21 days. Individual relative tumour volumes are depicted. Yellow‐shaded box demarcates the treatment period. (B) Chromogenic analysis reveals no increase in c‐PARP levels 24 h after GDC‐0941 (red) compared with vehicle (blue). Box and whisker plot shows median, first and third quartiles and maximum and minimum values. Unpaired Wilcoxon test has been applied for statistical analysis. One representative chromogenic image is shown for each cohort. Scale bar: 50 μm. (C) Chromogenic analysis reveals reduced intratumoural pS6 levels 4 h after of GDC‐0941 (red) compared with vehicle (blue). Box and whisker plot shows median, first and third quartiles and maximum and minimum values. Unpaired Wilcoxon test has been applied for statistical analysis. Due to the small sample size, we are statistically underpowered; however, there is a trend of reduced S6 phosphorylation in the GDC‐0941 treated samples. One representative chromogenic image is shown for each cohort. Scale bar: 50 μm. (D) Western blot analysis reveals reduced pAkt S473 and pS6 levels 4 h after GDC‐0941 (red) compared with vehicle (blue). Two independent tumours were used from each cohort.

To further validate the predictive value of our ex vivo culture system for in vivo sensitivity to novel therapies, we tested a novel compound, TH1579. TH1579 is a compound that inhibits the human mutT homologue MTH1 (NUTD1) (Yoshimura et al., 2003) and showed promising results in several cancer cell lines (Warpman Berglund et al., 2016b). To test the predictive potential of our cultures, we first examined the in vitro activity of TH1579 against CDX cultures compared with other cell lines reported to be sensitive to this novel agent. CDX2 and CDX3 cultures had GI₅₀ values lower than the control drug‐sensitive cell lines, whereas CDX4 cultures were similarly sensitive (Figure 5A). Immunoblot analysis of cleaved PARP confirmed that TH1579 treatment promoted apoptosis in CDX3 cultures (Figure 5B). We subsequently examined the in vivo efficacy of TH1579 in mice bearing CDX3 tumours. Mice were treated either with vehicle or TH1579 for 21 days. Tumour regression was observed during treatment; however, all tumours relapsed once the treatment was stopped (Figure 5C). Despite this, TH1579 treatment significantly extended the survival time of mice compared with those treated with vehicle alone (53 vs. 33 days, respectively) (Supporting Information Figure S5B). Consistent with its role in inhibiting MTH1, TH1579 treatment led to an accumulation of oxidative DNA damage and resulting increased apoptosis (Figure 5D,E). Furthermore, we observed MTH1 target engagement, suggesting a proper on‐target effect of the compound (Supporting Information Figure S5C).

Figure 5.

CDX3 cultures predict in vivo response to TH1579. (A) In vitro response of CDX2 (green), CDX3 (blue), CDX4 (purple), U2OS (black), SW480 (grey) and HCT116 (light grey) cells to TH1579 after 1 week or 72 h of treatment (CDXs and cancer cell lines respectively). One of at least three representative experiments for each model is shown. (B) CDX3 cultures were treated with DMSO or 0.5 μM TH1579 for 72 h. Protein lysates were probed for cleaved PARP as a measure of apoptosis. Akt was used as loading control. (C) Tumours were treated with TH1579 (red) or vehicle (blue) for 21 days. Each line represents an individual relative tumour volume from all mice. Yellow‐shaded box demarcates the treatment period. (D, E) Chromogenic analysis reveals elevated intratumoural 8‐oxo‐G (E) and cleaved PARP (F) levels after 4 days of TH1579 (red) compared with vehicle (blue). Box and whisker plot shows median, first and third quartiles and maximum and minimum values. Paired t‐test has been applied for statistical analysis, and P‐value is indicated. One representative chromogenic image is shown for each cohort. Scale bar: 50 μm.

These results confirm the potential of CDX cultures to predict response to novel compounds and identify candidate pathway dependencies. Additionally, we showed that TH1579 may have promising therapeutic implications for chemosensitive SCLC patients, while inhibition of the PI3K pathway in a chemorefractory setting should be considered, especially when proper biomarkers are identified.

Manipulation of CDX ex vivo cultures for functional analysis

We demonstrated that cells derived from the CDX models can grow in culture while maintaining the main characteristics of the donor tumour as well as drug sensitivity. One of the advantage of common established cell lines is that they can be manipulated to functionally characterize tumour evolution and response to therapies. To extend the utility of CDX cultures, we examined the feasibility of labelling CDX cells ex vivo. CDX‐derived cultures proved to be resistant to electroporation and chemical transfection methods but can be transduced with lentiviral vectors. CDX3‐derived cells were permanently transduced with a lentiviral vector containing a GFP cassette (Figure 6A). The GFP⁺ cells were sorted and implanted in immune‐compromised mice where they formed green fluorescent tumours, suggesting that there was no negative selection for GFP⁺ cells during culture and subsequent growth in mice (Figure 6B, Supporting Information Figure S6). Moreover, the growth rate of the CDX3‐GFP⁺ tumours was comparable with the growth rate of CDX3 tumour directly passaged into mice (Figure 6C). This preliminary experiment shows that CDX cells ex vivo can be labelled and reinjected in mice where they can be exploited as tool to dissect mechanisms of tumour dissemination and metastasis.

Figure 6.

CDX3 cultures can be labelled with fluorescent protein. (A) CDX3‐derived cells infected with a lentivirus containing an eGFP cassette and growing ex vivo. Scale bar: 200 μm. (B) CDX3 GFP⁺ cells grown in vivo form tumour and keep the viral eGFP cassette. Scale bar: 200 μm. (C) CDX3 cells directly passaged from CDX3 tumours (red) or maintained in culture and infected with the lentiCas9‐EGFP virus (black) were implanted s.c. into SCID‐bg mice. Mice were monitored for tumour growth, and tumour dimensions were measured biweekly. Each line represents an individual tumour.

Discussion

SCLC is the most aggressive type of lung cancer, killing globally more than 220 000 patients p.a., characterized by late diagnosis, early dissemination, a 5 year survival of <5% and no improvement beyond SOC chemotherapy for the past 30 years (Alvarado‐Luna and Morales‐Espinosa, 2016). Nevertheless, as the genomic landscape of SCLC continues to be uncovered (George et al., 2015), a number of new targeted therapies are beginning to show some promise in early clinical trials (Bunn et al., 2016). The acquisition of tumour biopsies for clinical and preclinical studies is severely limited, delaying the progress of research and drug development. This obstacle in tumour acquisition also negatively impacts the ease with which patient‐derived preclinical models (such as SCLC PDX) can be developed, especially at disease progression when chemotherapy has failed. Panels of established SCLC cell lines (Gazdar et al., 1980) are easy to use and have been cultured directly from the patient; however, prolonged time in culture selects for fitter clones, and, unfortunately, for most of them, linked patient information is minimal (Daniel et al., 2009; Pleasance et al., 2010; Garnett and McDermott, 2014; Santarpia et al., 2016). In this manuscript, we reported the development of a new preclinical SCLC research platform able to combine the benefits of in vitro cell lines and the patient specificity of in vivo models. To reach this goal, we used CTC xenograft tumours (CDX) derived from readily obtained liquid biopsies (Hodgkinson et al., 2014) as a source of SCLC cells. To date, of the 20 SCLC CDX‐derived cells tested, 16 of them were able to grow ex vivo in the described conditions. To better characterize the system, we directly compared the gene expression profile of the cells growing ex vivo with the donor CDX tumour from three different models. The culture system imposes transcriptional changes that can be observed after the first week of culture. However, those variations do not seem to affect the predictive potential of our models, and in contrast to a previous publication (Daniel et al., 2009), we decided to limit our culture to a maximum of 5 weeks to reduce the risk of strong clonal selection. Indeed, the variations observed during the first week of culture were maintained over 4 weeks, and some of those changes were reversible when the cells were re‐implanted into immune‐compromised mice. This is pertinent for function testing studies in which CDX cells can be manipulated in vitro and then reinjected into mice. Of note, the majority of the differentially expressed genes between the original CDX tumour and the derived culture are in non‐coding regions of the genome. Their functional significance is not known, and they should be taken into account when the role of specific non‐coding RNAs is investigated with this system.

Akin to our approach, Drapkin et al. (2018) used a short‐term CDX culture to test the combination of olaparib with temozolimed although they did not examine the impact of the culture on the cells. Here, we described and characterized CDX‐derived cultures compared with the donor CDX tumours, highlighting their potential and limitation. The data we presented as well as the data shown by Drapkin et al. (2018) support the use of this ex vivo system as a platform to assess drug targets for SCLC and support drug development as evidenced by our data confirming the impact of PI3K and MTH1 inhibition in chemorefractory and chemosensitive CDXs respectively. In vitro and in vivo efficacy of GDC‐0941 against CDX4 is of particular interest due to its complete resistance to SOC chemotherapy and requires further investigation. The Akt pathway has already been linked to chemoresistance in SCLC (Cardnell et al., 2016); however, clinical trials with PI3K inhibitors showed disappointing results (Santarpia et al., 2016), likely due to the absence of patient stratification. Combining the patient information of the CDX models with in vitro functional analysis of the derived CDX cells can facilitate the identification of novel biomarkers to be tested in the clinic.

The data with TH1579 are promising; however, more studies are warranted to screen TH1579 against our panel of CDX cultures to determine its broader utility in SCLC and, if encouraging, examine longer treatment schedules in vivo with additional interrogation of mechanisms of action. The results with TH1579 and GDC‐0941 are only representative examples of candidate target discovery. Mid‐throughput screening with compound libraries using our panel of CDX cultures would be an interesting platform to identify novel candidate targets for SCLC and is currently under investigation.

CDX cultures can also be exploited to characterize mechanisms of drug response or resistance. Moreover, with the increased development of CRISPR tools (Wang et al., 2016), we aim to exploit the CDX ex vivo cultures to functionally validate drug targets and identify putative mechanisms of drug sensitivity/resistance. As mentioned before, the observation that tumours derived from the CDX cultures reacquire some of the features lost during in vitro growth suggests that these cells can be genetically manipulated or labelled in vitro and subsequently re‐implanted into mice to study tumour evolution, metastasis and response to treatment using a model that is close to the patient (Hodgkinson et al., 2014; Drapkin et al., 2018). Finally, the possibility to generate longitudinal CDX models from the same patient will enable the development of paired models ex vivo to study patient‐derived mechanisms of resistance and tailor the treatments based on this specific background.

At the present time, the generation of CDXs and the resultant ex vivo cultures generally extend beyond the lifespan of most SCLC patients. Culture of CTCs directly derived from the blood of SCLC patient is under evaluation, and it is showing promise. However, the limitation of the direct culture of CTCs is the low number of cells obtained from a single blood draw (mean ± SD = 1589 ± 5565 in 7.5 mL of blood, Hou et al., 2012). Disaggregated CDX tumours produce between 5–50 million cells, significantly improving their practicality as preclinical models for functional testing. Our aim was to generate and characterize the reliability of an in vitro system that can be used in mid‐throughput therapy testing and for function testing hypotheses about SCLC biology. This approach is particularly attractive in a disease where the availability of tumour tissue is limited yet CTCs are relatively prevalent allowing CDX generation, and there is a clear unmet need to improve outcomes for those patients.

Author contributions

A.L., K.K.F. and C.D. conceived or designed the study. A.L., M.W.S., U.W.B., B.B., I.S.P., C.P.E.C., T.P. and C.K. collected the data. A.L., S.G., M.W.S., G.K., U.W.B, K.K.F., F.B., V.G.G., C.M., T.H. and C.D. analysed and interpreted the data. A.L., K.K.F. and C.D. drafted the article. A.L., S.G., U.W.B., I.S.P., K.K.F., F.B., V.G.G., C.M., T.H. and C.D. critically revised the article. All authors approved the final version to be published.

Conflict of interest

A patent has been filed with data generated in this manuscript where T.H. is listed as inventor.

Declaration of transparency and scientific rigour

This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research recommended by funding agencies, publishers and other organisations engaged with supporting research.

Supporting information

Table S1 List of antibodies for western blot.

Table S2 List of antibodies for chromogenic staining and protocols used.

Figure S1 Related to Figure 1. Effect of the different media conditions on proliferation of CDX‐derived cells cultures.

Figure S2 Related to Figure 2. CDX culture‐derived tumours resemble original CDX tumours.

Figure S3 Related to Figure 3. Dose response and combinatorial matrix response to standard of care.

Figure S4 Related to Figure 3. CDX‐derived cells as a platform to test novel compounds.

Figure S5 Related to Figure 4 and Figure 5. General pharmacodynamic analysis of GDC‐0941 and TH1579.

Figure S6 Related to Figure 6. CDX3 can be infected with lentivirus containing eGFP cassette.

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Lallo, A. , Gulati, S. , Schenk, M. W. , Khandelwal, G. , Berglund, U. W. , Pateras, I. S. , Chester, C. P. E. , Pham, T. M. , Kalderen, C. , Frese, K. K. , Gorgoulis, V. G. , Miller, C. , Blackhall, F. , Helleday, T. , and Dive, C. (2019) Ex vivo culture of cells derived from circulating tumour cell xenograft to support small cell lung cancer research and experimental therapeutics. British Journal of Pharmacology, 176: 436–450. 10.1111/bph.14542.