Two Novel Determinants of Etoposide Resistance in Small Cell Lung Cancer

Abstract

Patient survival in small cell lung cancer (SCLC) is limited by acquired chemoresistance. Here we report the use of a biologically relevant model to identify novel candidate genes mediating in vivo acquired resistance to etoposide. Candidate genes derived from a cDNA microarray analysis were cloned and transiently overexpressed to evaluate their potential functional roles. We identified two promising genes in the DNA repair enzyme DNA Polymerase β and in the neuroendocrine transcription factor NKX2.2. Specific inhibition of DNA Polymerase β reduced the numbers of cells surviving treatment with etoposide and increased the amount of DNA damage in cells. Conversely, stable overexpression of NKX2.2 increased cell survival in response to etoposide in SCLC cell lines. Consistent with these findings, we found that an absence of nuclear staining for NKX2.2 in SCLC primary tumors was an independent predictor of improved outcomes in chemotherapy-treated patients. Taken together, our findings justify future prospective studies to confirm the roles of these molecules in mediating chemotherapy resistance in SCLC.

Introduction

Small Cell Lung Cancer (SCLC) accounts for 4% of cancer deaths in the USA per annum (1). The standard of care is a platinum agent and Etoposide doublet chemotherapy (with radiotherapy in limited disease) with a typical disease trajectory of initial good response followed by rapid relapse with chemoresistant tumour and death (2, 3). Since surgical resection is not part of standard care SCLC research using primary tumour tissue is limited to small diagnostic samples. A mouse model has been described (4) but the bulk of SCLC research has used cell lines. Understandably, chemoresistance has been a major area of research in SCLC but it has proved challenging to translate the results obtained using cell lines to patient samples. For example the in vitro gene expression of Topoisomerase II has been negatively correlated with the IC₅₀ of SCLC cell lines to Etoposide (5). However, a subsequent study of the immunohistochemical expression of Topoisomerase IIα in diagnostic samples found that higher expression was associated with a worse outcome (6), the opposite of what would be predicted based on the in vitro results. Other work has been based on the study of cell lines rendered resistant to chemotherapy by long-term exposure to increasing concentrations of various drugs (7). This assumes a model of clonal evolution as the basis for the development of chemoresistance in vivo. However, this may be compromised when a cell line such as NCI-H69, derived from a patient already exposed to chemotherapy, is used.

The cell lines GLC-14, GLC-16 and GLC-19 were derived from a single patient at different time points during her treatment for SCLC (Figure 1A) and were originally described as having differing resistance to Etoposide treatment (8). We have used these lines as a biological model for the development of chemoresistance in vivo and from it we have identified the novel Etoposide resistance factors DNA Polymerase β and NKX2.2.

Figure 1. Characterisation of the cell lines confirms that GLC-16 is most resistant to Etoposide induced cytotoxicity.

(A) Timeline describing the derivation of the GLC cell line series based on Berendsen et al. (B) Cell survival after Etoposide treatment was determined for all 3 cell lines. Dose-response curves to Etoposide at the 48 hour time point were constructed for the three cell lines using the CellTiter Glo assay. GLC-16 cells had the highest IC₅₀ (n=4-8). Cell survival at 48 hours after treatment with 42μM Etoposide was determined using CellTiter Glo. Significantly more GLC-16 cells survived than the other cell lines (n=12, Kruskal Wallis test with Dunn’s multiple comparison test). (C) Apoptosis was measured after 48 hours of treatment with 42μM Etoposide using the Caspase Glo assay (n=8, Kruskal Wallis and Dunn’s tests) or by detecting TUNEL staining by flow cytometry (n=5, Mann-Whitney U-test). GLC-16 cells exhibited the least apoptosis. (D) aCGH was performed to determine genetic copy number (CN) and heterozygosity status (Loss of Heterozygosity, LOH). Results are shown for Chromosome 2 from 150-242mb for GLC-14 (top), GLC-16 (middle) and GLC-19 (bottom).

Materials and Methods

Cell Lines and Cell Culture

The GLC-14, GLC-16 and GLC-19 cells were the kind gift of Dr Nina Pedersen (Copenhagen, Denmark). Short Tandem Repeat profiling was performed to confirm their origin from a single case and they were characterised in comparison to the original description (8). The COR-L47 cells were obtained from CRUK cell services (London, UK). All cells were grown in suspension culture using RPMI 1640 with L-glutamine media (Gibco, Paisley, UK) supplemented with 10% Fetal Calf Serum, Penicillin 100units/ml and Streptomycin 100μg/ml at 37°C in 5% CO₂.

Western Blotting

Whole cell lysates were prepared with RIPA buffer (1% NP40, 1% Triton X-100, 0.1% SDS, 0.5% Deoxycholate, 150mM Sodium Chloride, 50mM Tris, pH7.4) supplemented with Complete protease inhibitor (Roche, Mannheim, Germany). Protein was quantitated using the Bicinchoninic Acid assay and equal quantities loaded into appropriate concentration acrylamide gels. After electrophoresis of denatured samples the gels were blotted onto nitrocellulose membrane, probed with appropriate antibodies (NKX2.2 Clone P-20, Santa Cruz Biotechnology, Santa Cruz, CA or Pol β, Ab1831, Abcam) and developed using enhanced chemiluminescence.

Luminescence Assays

Experiments were prepared in black flat-bottomed 96-well plates (Nunc, Roskilde, Denmark) with 1500 cells in 100μl of media per well. Etoposide or DMSO was added at an appropriate concentration and the experiments incubated for 48 hours. In parallel 2.25×10⁵ cells were prepared in the same experimental conditions in a flask. Cells were then harvested from the flask and a single cell suspension created. From the single cell suspension 500μl were taken for cell counting on Vi-Cell XR 2.03 (Beckman Coulter, Miami, FL) and 100μl were pipetted in technical quadruplicate into spare wells of the experimental plate in order to convert luminescent signal to cell number. There were blank wells, containing normal growth medium, on each plate. The plate was loaded into a Clarity luminometer (Biotek, Potton, UK) and the CellTiter Glo assay (Promega) prepared by mixing the buffer and substrate. The luminometer was programmed to automatically inject 100μl of assay into each well of the plate and then shake the plate for 10 seconds before incubating it at RT for 10 minutes and reading the luminescent signal. The data were extracted and the raw luminescent signal converted into a cell count for statistical comparison after subtraction of background signal. For the NKX2.2 stable expressing cells starting cell concentration was determined by number of available cells.

For Caspase Glo assays samples were prepared as for the CellTiter Glo experiments in 96-well plates. Cell counting was also performed so that the caspase signal could be corrected for cell number. Analysis was performed on the Clarity luminometer.

Cloning

PCR primers used to amplify the coding sequence of the gene of interest from cDNA were designed using Lasergene 8 software (DNASTAR, Madison, WI) and the NCBI Gene Nucleotide database and supplied by Sigma Aldrich (Supplementary Table 1). Amplification was performed with Phusion High-Fidelity DNA polymerase (Finnzymes, Keilaranta, Finland) on a Tetrad 2 thermal cycler (Bio-Rad Laboratories, Hercules, CA). The reaction mixture contained 10μl buffer, 1μl of 10mM dNTPs, 0.25μl of 100μM solution of each primer, 1μl of template, 1.5μl of DMSO and 0.5μl of Phusion (which was added last). The cycling conditions used were: initial denaturation for 30 seconds at 98°C then 35 cycles of 10 seconds denaturation at 98°C, 30 seconds of primer annealing at an appropriate Tm (Supplementary Table 1) and 15 seconds of extension per kilobase of amplified product at 72°C, followed by a final extension step of 72°C for 10 minutes. PCR products were gel purified by electrophoresis in a1.5% agarose gel and the product purified using a Gel DNA Recovery Kit (Zymo Research, Orange, CA). Restriction digestion was performed and the insert ligated into the vector using a Quick Ligation Kit (NEB). 50ng of vector was combined with a 3-fold molar excess of insert. The product was used to transform 10G DUO Escherichia cloni cells (Lucigen, Middleton, WI). Plasmids were purified using a Midiprep kit (Qiagen) and the sequence was verified using a 3130XL genetic analyser (Applied Biosystems).

Transfections

1×10⁶ cells in media without antibiotics were transfected using Lipofectamine 2000 (Invitrogen) in a ratio of 5μl to 1μg of DNA. After 48 hours the cells were used in experiments.

Stable expressing cell lines were generated using the pLPC vector and the protocol of the laboratory of Dr S Lowe (9).

Genomics

DNA was extracted from cultured cells using a DNeasy Blood and Tissue kit (Qiagen, Dorking, UK) as per the protocol for cultured human cells. RNA was extracted from cells or human tissue samples using the RNeasy Plus Mini kit (Qiagen) following the animal cells protocol provided. Nucleic acids were quantified using a nanodrop spectrophotometer and ND-1000 v3.3 software (Nanodrop Technology Inc., Wilmington, DE). RNA quality was measured using a 2100 Bioanalyzer (Agilent Technologies, Walbronn, Germany) following the RNA 6000 Nano protocol. RNA was hybridised to Illumina Human WGv3.0 Expression BeadChips. ACGH was performed using the Affymetrix SNP 6.0 platform.

Flow Cytometry

Cells treated with Etoposide or DMSO as for the luminescent assays were harvested and a single cell suspension created using 0.05% Trypsin EDTA (Gibco) and gentle mechanical dissociation by passage through a 23G needle. For TUNEL staining cells were permeablized in ice cold 70% ethanol before staining using an APO-BrdU TUNEL kit (Molecular Probes, Eugene, OR) and flow cytometry using an LSR II (BD Biosciences, San Jose, CA). Results were processed using FlowJo v8.8.4 software (Tree Star Inc., Ashland, OR).

For active caspase 3 staining cells were fixed and permeablized before staining with anti-active caspase 3 (Ab13847, 1:1000, Abcam) followed by an APC conjugated secondary (A-10931, 1:100, Invitrogen). Cells were analysed on an LSR II.

For γH2AX foci counting cells were treated for 90 minutes with 50μM Etoposide then washed and allowed to recover for 24 hours in media with 10% serum. Cells were fixed and permeablized in 2% Paraformaldehyde with 0.5% Triton X-100 for 30 minutes on ice then stained with anti-γH2AX antibody (Clone 2F3, Abcam, Cambridge, UK) at a 1:5000 dilution for 60 minutes on ice. An Alexa 488 conjugated secondary antibody was used (Molecular Probes) at a dilution of 1:1000 for 60 minutes on ice. The cells were counterstained with 1μg/ml Propidium Iodide (Sigma Aldrich, Poole, UK) and flow cytometry was performed with an ImageStream imaging flow cytometer (Amnis). The data were analysed using IDEAS v4.0 software (Amnis). The single cell building block was used to define the population of interest. A spot mask was created on the γH2AX channel with 4-fold pixel intensity over background and 2 pixel minimum diameter. Subsequently a 2.5-fold peak mask was created on the spot mask. A feature was created to apply spot counting analysis to the peak mask. The feature was batch applied to all image files.

TMA Immunohistochemistry

The TMA was stained on a BondMax Autostainer (Leica, Milton Keynes, UK). Antigen retrieval was performed at 100°C in Bond ER2 diluent (for NKX2.2, HPA003468, 1:50, Atlas Antibodies, Stockholm, Sweden) or Bond ER1 diluent (for Pol β, Ab26343, 1:200, Abcam), followed by a 15 minute incubation with primary antibody at room temperature (RT), 8 minutes incubation using a polymer secondary system (Leica) followed by developing with Diaminobenzidine (DAB) using copper enhancement. Haematoxylin counterstaining was performed automatically on the Bond system and finally the slides were dehydrated, cleared and mounted using a Leica ST5020 attached coverslipper CV5030 (Leica). The slides were then scanned onto the Ariol system (Genetix, Gateshead, UK) for scoring. Consensus scoring was performed by two observers (one a Senior Pulmonary Histopathologist) and data analysis performed with SPSS v17.0 (SPSS Inc., Woking, UK). Scoring was performed blinded to the clinical data relating to the case. Where more than one core was present the modal score was taken for each tumour and where this was not possible the highest score was taken.

Statistics

Data analysis was performed using GraphPad Prism 5.01 (GraphPad Software Inc.). Where appropriate, parametric statistics were used to analyse differences between experimental groups. If an assumption of Normality was not valid non-parametric analyses were performed. In a small number of experiments the sample size was low but the magnitude of the difference between experimental groups was considered to be of importance so parametric analysis was performed. Where multiple groups existed within a single experiment multiple between group comparisons were made with the Kruskal-Wallis test and Dunn’s post-test, or ANOVA with Bonferroni’s post-test, to reduce occurrence of type 1 errors. A p value <0.05 was considered significant in 2 sided testing. Data are presented as the mean of ‘n’ independent experiments except for the experiments with the NKX2.2 stable overexpressing cells and the γH2AX foci counting experiments where single representative experimental results are shown from three independent experiments.

Survival analysis was performed using SPSS v17.0 (SPSS Inc). Univariable survival analysis used the Log Rank test and multivariable analysis was performed by Cox Regression using a forward stepwise model based on likelihood ratios.

Array CGH quality control was performed using the Affymetrix Genotyping Console (GTC) v3.0.1 software. The data was pre-processed and normalised using the CN5 approach in GTC (10). Both genotyping and loss of heterozygosity data were determined using GTC and the normalised data was then segmented using the circular binary algorithm, which is implemented in DNAcopy, a Bioconductor package. GenomeGraphs was used to create genome plots for both copy number and LOH data and the Broad Institute’s Integrative Genome Browser (IGV: http://www.broadinstitute.org/igv) was also used to view the results.

Gene expression microarray data pre-processing, quality assessment and analysis was performed using the Bioconductor packages beadarray and limma in the R computing environment (11, 12). The quality of bead-level data was assessed using various MA and box plots and any spatial artefacts identified and removed using BASH in beadarray (13). The raw bead-level data was log₂ transformed, bead summarised and then quantile normalised. A linear model was fitted to the normalised data using limma and computation of any differential gene expression involved calculating the log₂ fold change in expression of genes for pair wise contrasts between the groups of interest. Significance of the observed gene expression changes was calculated using the empirical Bayes approach, which calculated the modified t-statistic, the modified F statistic and the B statistic (the Bayes log posterior odds). Gene lists were constructed based on genes that were observed to have a log₂ fold change of > 1 or < - 1 for each paired comparison between all three GLC cell lines in monoculture and co-culture conditions with a false discovery rate (FDR) adjusted p-value cut off of 0.01.

Results

The GLC cell lines

The GLC-14, GLC-16 and GLC-19 cell lines were obtained from the lab of Dr Nina Pederesen (Copenhagen, Denmark). Short Tandem Repeat profiling confirmed that all three lines were derived from a single person. In the original description of the lines the GLC-16 line, derived after chemotherapy, was the most resistant to Etoposide treatment (8). In this study the phenotype was confirmed using a luminescent cell viability assay (CellTiter Glo, Promega, Madison, WI). After 48 hours of treatment with Etoposide the IC₅₀ for the GLC-14 line was 12.4μM (95% confidence interval (CI) 8.9-17.3μM), for GLC-16 was 51.8μM (30.7-87.3μM) and for GLC-19 was 13.1μM (8.8-19.5μM, Figure 1B). Nearly twice as many GLC-16 cells compared to GLC-14 cells were viable after a dose of 42μM of Etoposide for 48 hours. With the same exposure there was more active caspase 3 and 7 per cell in the GLC-14 line compared to the GLC-16 line and more TUNEL positive cells by flow cytometry (Figure 1C).

There are several models that have been proposed to describe the development of chemoresistance in tumours as highlighted by Agarwal and Kaye (14) in the context of ovarian cancer. The original description of the GLC-14, GLC-16 and GLC-19 cell lines implies that they evolved in a linear manner from earliest to latest under the selection pressure of treatment, i.e. that the GLC-19 line was descended via the GLC-16 line from the GLC-14 line. The patterns of DNA copy number and heterozygosity within the three lines were compared using array comparative genomic hybridisation (aCGH, SNP 6.0, Affymetrix, Santa Clara, CA). Using this approach we were able to show that while the lines are very similar, it is clear that all three lines were present as distinct clones prior to the isolation of the GLC-14 line. In Figure 1D copy number and loss of heterozygosity (LOH) are plotted for chromosome 2q from 150Mb to 242Mb, for each of the cell lines. If the GLC-14 cell line had become the GLC-16 line through evolution under the selection pressure of chemotherapy then any areas of LOH in the GLC-14 line would have to be preserved in subsequent lines (Figure 1D top). However, as the GLC-16 line (Figure 1D middle) has an area of heterozygosity around 240Mb that is not present in the GLC-14 line then they must have been distinct from each other prior to the isolation of GLC-14. The GLC-19 line has very little LOH within it (Figure 1D bottom) and neither of the main areas of loss from the GLC-14 or GLC-16 lines are present so it too must have been a distinct clone prior to the derivation of the GLC-14 line.

Determinants of Etoposide response

In order to identify genes that might be involved in Etoposide response in SCLC, RNA was extracted from each of the cell lines and gene expression analysed using an Illumina Human WGv3.0 Expression BeadChip microarray. Gene expression was compared by pairwise subtraction of normalised expression levels for each gene in order to determine differential expression between cell lines (Figure 2). Genes whose expression pattern was consistent with a role in determining the pattern of Etoposide response of the cell lines were identified and a shortlist of eight candidate genes was established based on known or putative gene function (Table 1). These eight genes represented a pragmatic compromise between the practical constraints of the detailed study of genes where little is known about their function and tools such as antibodies may not be available, and the power of the microarray to generate entirely novel candidates and potentially new insights into chemoresistance. Several other interesting hits came out of the microarray data, for instance c-Myc was expressed at a higher level in the GLC-16 and GLC-19 cells whereas n-Myc was expressed at the highest level in the GLC-14 cells. ALDH1A1 was expressed at low level in GLC-14 cells with the expression level rising in the GLC-16 cells and at it’s highest in the GLC-19 cells, this gene has been used as a stem cell marker but is also associated with cyclophosphamide resistance (15).

Figure 2. Identification of genes associated with Etoposide response in the GLC-14, GLC-16 and GLC-19 cell lines.

Gene expression profiling was performed in triplicate on RNA extracted from each of the cell lines using the Illumina BeadChip platform. Results were analysed and a heatmap constructed using log₂ intensity for clustering to allow direct comparison of gene expression levels between all the cell lines.

Table 1.

Candidate genes selected based on known or putative function

Gene Symbol	Fold Change (GLC-14 – GLC-16)	FDR Adjusted p-value
BEX2	−17.271	20×10−15
FAM3B	−4.72	1.31×10−12
NKX2.2	−5.78	7.66×10−09
POLB	−4.76	2.90×10−06
CAMK2N1	30.06	2.91×10−10
QPCT	4.44	9.63×10−10
HOXB5	29.24	5.78×10−08
CDO1	7.41	8.4×10−07

In order to determine if the candidate genes were able to alter the phenotypic response of the cancer cells to Etoposide a transient overexpression experiment was performed. Each candidate gene was cloned from the cell line in which it was expressed at the highest level, inserted into an IRES-GFP expression vector (Clontech, Mountain View, CA) and transiently expressed into the cell line in which it was normally expressed at the lowest level. Cells expressing the gene of interest were treated for 48 hours with 42μM Etoposide and the proportion of GFP expressing cells was determined by flow cytometry. Genes that promoted Etoposide resistance would be expected to lead to enrichment of the GFP positive cell population compared to empty vector transfected cells and genes that increased sensitivity to Etoposide induced cytotoxicity would reduce the proportion of GFP expressing cells. From this screen of all the candidate genes NKX2.2 and DNA Polymerase β (Pol β) were selected as the most promising for further study.

DNA Polymerase β

The gene expression pattern of Pol β in the GLC-14 and GLC-16 cell lines was confirmed at the protein level by Western blotting (Figure 3A). The gene was cloned from the GLC-16 cell line and transiently expressed in the GLC-14 cell line (Figure 3A). When the transiently expressing GLC-14 cells were treated for 48 hours with either Etoposide or vehicle control there was enrichment of the transfected cells as measured by GFP expression. Simultaneous detection of active caspase 3 showed that more of the Etoposide treated, Pol β transfected cells were non-apoptotic compared to the empty vector controls (Figure 3B), but there was no difference between the numbers of apoptotic cells. This suggests that Pol β is protecting the cells from Etoposide-induced apoptosis.

Figure 3. DNA Polymerase β contributes to the repair of DNA damage caused by Etoposide.

(A) Pol β is expressed at the highest level in untreated GLC-16 cells, consistent with the gene expression data. A pIRES-GFP construct was used to express Pol β cloned from the GLC-16 cell line into the GLC-14 cells. (B) GLC-14 cells transiently transfected with either pIRES POLB or the pIRES empty vector (EV) were treated with 42μM Etoposide for 48 hours and then expression of GFP and active caspase 3 were detected by flow cytometry. There were more active caspase 3 negative (non-apoptotic) Pol β transfected cells compared to empty vector transfected controls (n=5, paired t-test) but there was no difference in the number of active caspase 3 positive (apoptotic) cells. This suggests that Pol β expression makes the GLC-14 cells better able to survive treatment with Etoposide. (C) To see if inhibition of Pol β rendered the cells more sensitive to Etoposide GLC-16 cells were either treated with the selective Pol β inhibitor Pamoic Acid (PA, 300μM) or left untreated prior to treatment with Etoposide or DMSO control for 48 hours. PA reduced the number of cells surviving treatment with Etoposide but had no effect on controls (n=6). The experiment was repeated a further 4 times using a concentration of 25μM Etoposide with the same result. (D) To confirm that Pol β promoted cell survival by protecting the cells from Etoposide induced DNA damage GLC-16 cells treated with PA or untreated controls were exposed to 50μM Etoposide or DMSO control for 90 minutes. Cells were then washed and allowed to recover for 24 hours. The number of γH2AX foci per cell was counted using an imaging flow cytometer. PA treatment resulted in more foci per cell after Etoposide treatment (H, n=657-1332 cells, one way ANOVA and Bonferroni’s multiple comparison tests).

Pamoic acid (PA) is a natural inhibitor of Pol β that has previously been characterised (16). GLC-16 cells were treated with Etoposide or vehicle control and the effects of Pol β inhibition were determined by treatment with PA. PA sensitised the cells to Etoposide-induced cell death with lower luminescence generated by the cells exposed to PA (Figure 3C). The cytotoxic effect of Etoposide is caused by the generation of double strand breaks in DNA, which can be visualised by the detection of the phosporylated form of histone H2AX (γH2AX), recruited to the site of the stand breaks (17). In order to confirm that the protective effect of Pol β was through prevention of Etoposide promoted-DNA double strand breaks, GLC-16 cells were exposed to Etoposide for 90 minutes and then allowed to recover for 24 hours before having the number of unrepaired double strand breaks per cell ascertained on an imaging flow cytometer (Imagestream, Amnis, Seattle, WA). Inhibition of Pol β using PA prior to treatment with Etoposide resulted in significantly more γH2AX foci being detected per cell (Figure 3D).

NKX2.2

The protein levels of NKX2.2 in untreated GLC-14 and GLC-16 cells conformed to the pattern of gene expression in the cell lines, with more NKX2.2 in the GLC-16 line (Figure 4A). The gene was cloned from the GLC-16 line and inserted into the pIRES-GFP vector. This was then used to transiently overexpress NKX2.2 in the GLC-14 cell line (Figure 4A). GLC-14 cells transfected with either the pIRES-GFP NKX2.2 construct or pIRES-GFP empty vector were treated with 42μM Etoposide or vehicle control for 48 hours. There was significant enrichment of GFP positive cells in the NKX2.2 expressing cells compared to the empty vector transfected cells, but this enrichment consisted of an enrichment of both apoptotic and non-apoptotic cells (Figure 4B).

Figure 4. Increased NKX2.2 expression increases the number of viable cells after Etoposide treatment.

(A) NKX2.2 is expressed at the highest level in untreated GLC-16 cells, consistent with the gene expression data. A pIRES-GFP construct was used to express NKX2.2 cloned from the GLC-16 cell line into the GLC-14 cells. (B) GLC-14 cells transiently transfected with either pIRES NKX2.2 or the pIRES empty vector (EV) were treated with 42μM Etoposide for 48 hours and then expression of GFP and active caspase 3 were detected by flow cytometry. The number of both active caspase 3 negative (non-apoptotic) and positive (apoptotic) cells was greater in the NKX2.2 transfected Etoposide treated group compared to empty vector transfected Etoposide treated controls (n=5, paired t-test). (C) Stable NKX2.2 overexpressing GLC-14 cells were created and treated with 50μM Etoposide. After 48 hours there were more viable NKX2.2 expressing cells compared to empty vector controls (n=6, Mann-Whitney U-test). (D) Stable NKX2.2 overexpressing COR-L47 cells (a chemotherapy naïve SCLC cell line unrelated to the GLC cells) were created and treated with 50μM Etoposide. After 48 hours there were more viable NKX2.2 expressing cells compared to empty vector controls (n=6, Mann-Whitney U-test).

To further study the effect of NKX2.2 stable overexpression of the gene in the GLC-14 cell line was achieved using retroviral mediated gene transfer in a pLPC vector (9) (Figure 4C). When these cells were treated with Etoposide 50μM for 48 hours significantly more NKX2.2 expressing cells survived compared to empty vector control cells. To demonstrate that this was not a cell line specific effect an unrelated SCLC cell line, COR-L47, was obtained. This cell line was originally derived from a patient who had not previously been treated for SCLC (18). A stable NKX2.2 overexpressing cell line was established as for the GLC-14 line and this too showed increased cell survival after Etoposide treatment (Figure 4D).

Tissue Microarray

We have previously described the construction and use of a tissue microarray (TMA) of SCLC cases to determine the prognostic or predictive influence of tumour proteins in SCLC patients (19). This TMA was used to determine whether Pol β or NKX2.2 expression had relevance to the survival of patients with SCLC. Most cases stained for Pol β were positive (86%) but there was no relationship detectable between Pol β staining extent or intensity and survival or response to chemotherapy. For NKX2.2 there was a significant difference in survival between patients treated with chemotherapy whose diagnostic biopsy had nuclear staining compared to those where there was no nuclear staining (Figure 5). Absence of nuclear staining for NKX2.2 in the diagnostic biopsy sample was an independent predictor of improved survival in this cohort with a hazard ratio for death of 0.52 (0.33-0.82, p=0.005, Figure 5B).

Figure 5.

Absence of nuclear staining for NKX2.2 in diagnostic biopsy samples is an independent predictor of improved outcome after chemotherapy in SCLC. (A) Kaplan-Meier Plot of chemotherapy treated patients stratified on the basis of nuclear or non-nuclear localisation of NKX2.2 staining on the SCLC TMA. Number of patients at risk at the start of each time interval is shown below the plot. (B) Cox regression analysis using Age (as a continuous variable), Sex, Stage, Performance Score and NKX2.2 staining as input variables identified NKX2.2 and Stage as independent predictors of survival.

Discussion

It has proven difficult to study the development of chemoresistance in SCLC and this is, at least partly due to the inadequacies of the models available and the poor availability of sufficient quantities of primary tumour material. In this study, the use of a cell line series where chemotherapy resistance was acquired in vivo has allowed the identification of several potential resistance factors and the characterisation of two of these as novel Etoposide resistance genes.

Cell viability was selected as the most clinically relevant measurement of chemotherapy response as it is the viable cells that repopulate the tumour after chemotherapy. Overall viability can be seen as a compound measurement of both the proportion of cell death and the rate of cell growth. A standard Etoposide dose of 100-140mg/m² would result in a peak plasma concentration of 10-20μg/ml and 800mg/m² gives a peak plasma concentration of 100μg/ml (20). Therefore, in this study Etoposide was used at either 42μM (25μg/ml) or 50μM to represent clinically relevant dosing.

The GLC-14, GLC-16 and GLC-19 cell lines were derived from a single patient during the course of her treatment for SCLC. Using genomic techniques similar to those of Cooke and colleagues (21) we have obtained similar results to their findings in paired ovarian carcinoma cell lines. It is clear that the three cell lines, with their differing phenotypic responses to Etoposide treatment (and also to irradiation (22)) were all present as distinct clones prior to the derivation of the GLC-14 line. This supports a model of genetic heterogeneity with multiple sub-populations of cells within a tumour, possibly with one dominant clone and the others at relatively low numbers. During treatment the dominant population may be eradicated or merely supplanted by a clone more suited to the changed environment. The concept that intra-tumoural heterogeneity in lung cancer may give rise to subpopulations of tumour cells with divergent phenotypes is not new and the term endophenotypes has recently been used to describe the multiple possible phenotypes within a tumour (23). Whilst it is impossible to quantify the amount of genetic change that occurred in the tumour cell population that became GLC-16 and GLC-19 lines after they diverged from the GLC-14 clone, the data support the concept that SCLC is genetically heterogeneous with multiple endophenotypes within the primary tumour. This suggests that resistance to chemotherapy might not be acquired by the tumour through evolution of a homogenous population but instead represents emergence of a new dominant cellular phenotype from a heterogenous pool of cells during treatment. This may have fundamental implications for the future study and clinical management of chemoresistance.

The DNA single strand break repair enzyme DNA Polymerase β is implicated in the repair of DNA lesions induced by Etoposide and consequent cell survival for the first time. It has already been established that Pol β is an important enzyme for DNA maintenance in cell replication, in resistance to ionizing radiation (24) and in resistance to monofunctional alkylating agents such as methyl methanesulfonate (25). Through overexpression and knockdown studies Pol β has been implicated in Cisplatin resistance (26-28) where it is able to promote error prone synthesis across the platinum adduct. It has been demonstrated recently that inhibition of Pol β potentiates the cytotoxity of the alkylating agent Temozolomide by promotion of DNA double strand breaks through conversion of excess unrepaired single strand DNA breaks (29). Ionizing radiation produces several types of DNA lesion, many of which might require Pol β for their repair. In fact Pol β deficient cells are not especially sensitive to ionizing radiation during growth as alternative pathways are able to compensate for the deficiency (24, 30). Etoposide is a DNA Topoisomerase II poison that prevents the religation of Topoisomerase II induced single strand breaks. It has been shown that at clinically relevant doses (as used in the current study) Etoposide causes approximately 2.6 single strand breaks for every double strand break induced (31). This ratio is higher at lower doses of the drug (31, 32) consistent with the ‘two-drug model’ proposed by Bromberg. In this model one molecule of drug will stabilize the cleavage complex at one active site within the Topoisomerase II homodimer to prevent religation. For a double strand break to occur there must be Etoposide at both active sites and this depends on the concentration of both the drug and the Topoisomerase II (31). In the current study it has been shown that the GLC-16 cell line has higher levels of Pol β than the GLC-14 or GLC-19 lines (Figures 2 and 3A), that more GLC-16 cells survive treatment with Etoposide than the other two lines (Figure 1B) and that Etoposide induces less DNA damage (as measured by TUNEL staining) in the GLC-16 line than the GLC-14 line (Figure 1C). Overexpression of Pol β promoted cell survival during treatment with Etoposide (Figure 3B). Furthermore inhibition of Pol β with Pamoic acid reduced cell survival after Etoposide treatment (Figure 3C) and promoted the generation of DNA double strand breaks (Figure 3D). Therefore we propose that Pol β is able to prevent Etoposide induced cytotoxicity through its single strand break repair function. This hypothesis is strengthened by recent data demonstrating the conversion of excess unrepaired single-strand breaks to cytotoxic double strand breaks after inhibition of Pol β and treatment with Temozolomide (29). Other Topoisomerase II targeting drugs such as Teniposide produce an excess of single-strand DNA breaks (33) as do other agents such as Bleomycin (34) and Pol β may be a determinant of cytotoxicity for these drugs as well. The data do not allow conclusions to be drawn about the clinical importance of this, as the TMA did not demonstrate a difference in patient outcome based on Pol β expression. However the high levels of Pol β expression in the cases on the TMA and the fact that Pol β is also implicated in resistance to Cisplatin, the other major chemotherapeutic used to treat SCLC, may make it a worthwhile drug target particularly in relapsed drug resistant disease.

Little is known about the transcription factor NKX2.2 in cancer. It is a member of the Nk-2 family of homeobox transcription factors and is important in the development of the endocrine pancreas and central nervous system (35, 36). It is chromosomally linked to NKX2.4 and the linked pair has a parologous relationship with NKX2.9 and NKX2.1 (37). NKX2.1 (also known as Thyroid Transcription Factor-1, TTF-1) is used in the diagnosis of pulmonary adenocarcinomas and has been shown to be predictive of poor outcome in Non-Small Cell Lung Cancer (38).

In the current study NKX2.2 has been shown to be expressed at a higher level in cells that are more resistant to Etoposide (Figures 1B, 2 and 4A, Table 1). Overexpression of NKX2.2 in the GLC-14 cell line (Figures 4C) and in an unrelated chemotherapy naïve SCLC cell line (Figure 4D) resulted in an increase in cell survival after Etoposide treatment compared to empty vector transfected cells. In patient samples, those with nuclear expression of NKX2.2 (where the transcription factor would be expected to exert its effect) had a worse outcome after chemotherapy than those without nuclear expression (Figure 5). NKX2.2 has been shown to be necessary for oncogenesis in Ewing’s Sarcoma where it acts as a transcriptional repressor (39). In a mouse model of prostate cancer Nkx2.2 expression was associated with advanced disease and neuroendocrine differentiation (40) and in Nkx2.2 null mice there are fewer neuroendocrine cells in the gut (41). This leads to the hypothesis that NKX2.2 may be an important regulator of the neuroendocrine phenotype in SCLC, potentially making it a key effector in the biology of SCLC.

In this study we have provided evolutionary evidence of the existence of endophenotypes in primary SCLC, some of which have the capacity to cause chemotherapy resistant relapse. Using a cell line model of Etoposide resistance that was developed in vivo we have identified and characterised two genes, Pol β and NKX2.2, which confer resistance to Etoposide in vitro. We have shown, for the first time, that DNA single strand break repair by Pol β can prevent Etoposide-induced cell death. This finding has potential implications for future chemotherapy for SCLC, which may be enhanced by inhibition of this pathway. The transcription factor NKX2.2 may be important in the generation of the neuroendocrine phenotype of SCLC but the exact role in SCLC and Etoposide resistance is currently unknown. Our studies show that it is able to promote cell survival in vitro and that nuclear expression of the transcription factor in diagnostic biopsy samples defines a population of patients with poorer outcome from chemotherapy.

This study has highlighted several potential Etoposide resistance factors (Table 1) and we have been able to further characterise two of them. This implies that the overall resistance phenotype may be a consequence of the interactions of multiple resistance modifiying factors rather than being dependent upon one easily targeted factor and this has consequences for treatment.