Development and characterisation of acquired radioresistant breast cancer cell lines

Cell culture reagents were obtained from Gibco Thermo Fisher Scientific (Paisley, UK), unless otherwise stated. Human breast cancer cell lines ZR-751, MCF-7 and MDA-MB-231 were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% foetal calf serum (FCS), 50 U ml^− 1 penicillin and 50 mg ml^− 1 streptomycin and incubated at 37 °C in a humidified atmosphere with 5% CO₂. Cell lines obtained from the American Type Culture Collection (LGC Standards, Teddington, UK) were authenticated by short tandem repeat (STR) profiling performed at Health England (Porton Down, Salisbury, UK). All cell line DNA samples tested matched 9 of 9 tested core alleles in DNA from known cell line samples confirming their identity. All experiments were performed using cells, maintained at low passage number from these frozen stocks.

Cells were irradiated using a Faxitron cabinet X-ray system 43855D (Faxitron X-ray Corporation, IL, USA). Radioresistant cell lines (MCF-7 RR, ZR-751 RR and MDA-MB-231 RR) were developed from their respective parental cell lines (MCF-7, ZR-751 and MDA-MB-231) by weekly exposure to single fractions of radiation. An initial dose of 2 Gy was followed by weekly incremental doses of 0.5 Gy for 12 weeks. During this period cells received a total radiation dose of 57 Gy. Cells were subsequently maintained with further weekly doses of 5 Gy.

Cells were seeded into 96 well plates (500 cells/well) and incubated for 24 h. Cells were drug-treated or exposed to radiation and fixed between 24 and 144 h after treatment by the addition of 50 μl cold 25% trichloracetic acid (Sigma-Aldrich, UK) per well at 4 °C for 1 h. Plates were washed in H₂O and, when dry, 50 μl SRB dye (0.4% SRB dissolved in 1% glacial acetic acid (VWR International)) was added to each well and incubated for 30 min. Plates were washed 4 times in 1% glacial acetic acid and, when dry, 150 μl of 10 mM Tris-NaOH buffer (pH 10.5) was added to each well. The plates were incubated on a shaker for 60 min. Optical density was measured at 540 nm using a Biohit BP800 spectrophotometer (Biohit Ltd., UK) and Wallac 1420 Manager program (PerkinElmer, UK). The half maximal inhibitory concentrations (IC₅₀ values) were calculated using the GraphPad prism 7 package.

Cells were seeded into 75 mm plates (1 × 10³ cells/plate) and incubated for 24 h before radiation treatment. Once visible colonies had formed (approximately 50 cells per colony) in the untreated control group (approximately 10–14 days post-seeding) the plates were washed twice in PBS before fixing the cells with the addition of 5 ml of 1,9-dimethyl-methylene blue zinc chloride double salt (Sigma-Aldrich, UK). After 45 min the plates were washed and allowed to air dry before colonies were counted. Analysis was performed by calculating plating efficiencies and survival fractions for control and treatment plates [8].

Cells were seeded into 6 well plates at a density to achieve 100% confluence after 24 h. Scratch assays were performed as previously described [9]. 0.1% serum-supplemented media was added to each well. Phase contrast images of the cell monolayer were captured (Axiovert DS100, × 5 objective) during migration up to a maximum of 48 h post-scratch. At each time point the area devoid of migrating cells was calculated using FIJI software and expressed as a % of the initial scratched area.

A single cell suspension of each cell line from a T175 flask (approximately 15 × 10⁶ cells) was transferred to a spinner flask (Cellcontrol Spinner Flask, Integra, Switzerland) containing 100 ml of routine DMEM and placed onto a magnetic stirrer platform (Cellspin, Integra, Switzerland). MTS developed over 7 days in normal incubation conditions.

A single MTS was removed with 500 μl of collagen mix (ice cold 0.1% filtered acetic acid, cell matrix type 1-A (Alphalabs), 0.22 M NaOH (Sigma-Aldrich, UK), FCS and 10x DMEM (Sigma-Aldrich, UK) at concentrations of 45, 25, 10, 10 and 10%, respectively) and placed into a 24 well plate. Cultures were incubated for 1 h to allow polymerisation of collagen and 500 μl routine DMEM was then added to each well. Phase contrast images were captured (Axiovert DS100, × 5 objective) at regular intervals up to 120 h post-seeding. MTS invasion was measured using a FIJI macro developed by Matthew Pearson (IGMM Advanced Imaging Resource, University of Edinburgh) at each time point and expressed as a % of the initial MTS area.

Cells were seeded into 75 mm plates (1.0 × 10⁶ cells/plate) and incubated for 24 h. Cells were serum-starved for 2 h and exposed to 2 Gy radiation before undergoing routine lysis collection at 0, 5, 10 and 30 min post-radiation. Lysates were snap frozen on dry ice and stored at − 70 °C for western blot analysis.

Whole cell lysates were prepared as previously described [10] and protein concentration was determined using a bicinchoninic acid (BCA) assay. Equal amounts of protein were separated by sodium dodecyl sulphate (SDS) polyacrylamide gel electrophoresis and transferred to Immobilon-P transfer membrane (Millipore). Ponceau S solution was used to visualise protein bands and confirm equal loading; membranes were blocked using Odyssey Blocking Buffer (LI-COR Biosciences, UK) (1:1 with PBS) for 1 h, before incubating overnight at 4 °C with primary antibodies (Table 1). Signals were detected using IRDye 800CW (Li-Cor, 926–32,210, 1:10,000) and IRDye 680LT (Li-Cor 926–68,021, 1:10,000) with a Li-Cor Odyssey Imager. Membranes required for re-probing were stripped using NewBlot PVDF Stripping Buffer (LI-COR Biosciences, UK).

IHC was carried out on formalin-fixed MTS. Samples were deparaffinised and rehydrated, and antigens retrieved (Table 1). Endogenous peroxidase activity was inhibited with 3% H₂O₂ solution (Dako, UK) for 10 min and non-specific antibody staining was blocked using Total Protein Block (Dako, UK) for 10 min. Primary antibodies were incubated for 1 h (Table 1). 1 drop of Envision labelled polymer (Dako, UK) was added for 30 min, before DAB and substrate buffer (1:50) (Dako, UK) were added to each section for 10 min.

ICC was performed on cells grown in chamber slides (2 well chamber slide, Lab-Tek II, Scientific Laboratory Supplies, UK) seeded to achieve a confluency of approximately 80% at 24 h. Cells were fixed in cold acetone (500 μl/chamber) for 10 min at 4 °C, then washed twice in PBS for 10 min. The same protocol as described for MTS IHC, from the addition of H₂O₂ solution, was then followed.

All slides were counterstained in haematoxylin, dehydrated and mounted with coverslips using DXP mountant (Sigma-Aldrich, UK). Slides were scanned using a NanoZoomer ER slide scanner (Hamamatsu Photonics, UK) and viewed using NanoZoomer Digital Pathology software.

Cells were seeded into 96 or 6 well plates (500 or 2 × 10⁵ cells per well, respectively) in media without antibiotics and incubated overnight to achieve 30–50% confluence. Cells were transfected according to the manufacturer’s protocol using DharmaFECT 1 Transfection Reagent (Dharmacon) with on-target plus ERα siRNA (J-103401-12) (GE Healthcare Dharmacon) at a concentration of 25 nM.

Cells were seeded into 75 mm plates (3.0 × 10⁶ cells/plate) and incubated for 24 h. Cells were serum-starved for 2 h and were then exposed to either 0 or 2 Gy radiation. Pellets of up to 1 × 10⁷ cells were collected by trypsinisation at 0, 2 and 8 h post-radiation, snap-frozen on dry ice and stored at − 70 °C for subsequent RNA extraction. Total RNA was extracted from cells with the RNeasy Mini Kit using QIAshredder technology (UK Qiagen, Ltd). The manufacturer’s protocol for purification of total RNA from animal cells using spin technology was followed. Extracted RNA samples were quantified and assessed for the presence of contaminants using the NanoDropTM Spectrophotometer ND1000 (Thermo Fischer Scientific). Full genome expression read-counts were generated using Lexogen QuantSeq 3′ FWD sequencing technology on an Illumina flow cell which was scanned using an Illumina HiScanSQ system (Edinburgh Clinical Research Facility, University of Edinburgh). NGS reads were generated towards the poly(A) tail and read 1 directly reflects the mRNA sequence. The ZR-751 2 h, 2 Gy sample failed sequencing and was removed from further analysis. RNA integrity number (RIN) was generated for each sample to assess RNA quality (Agilent Bioanalyzer); all samples had RIN values above 9.7 (Additional file 1: Table S1). FASTQ files of raw read-count data were pre-processed using the Lexogen recommended BlueBee high-performance NGS analysis software which implemented poly(A) tail trimming and alignment to the Genome Reference Consortium Human genome build 38 reference genome using the Spliced Transcripts Alignment to a Reference (STAR) algorithm [11]. Prior to analysis, data were log2 transformed and quantile normalised in R (Bioconductor) software and packages [12]. These preliminary transcriptomic data were only used for supervised pathway-focused analyses for the purposes of hypothesis generation and each analysis was subsequently validated by lab-based experimentation. Heatmap and cluster analysis were performed using TM4 MeV (multiple experiment viewer) software [13]. Heatmap clustering was carried out using Pearson correlation with average linkage. For integration of gene expression data with public datasets correction for integration batch effects was performed in R using ComBat as described previously [14, 15]. Hierarchical clustering of parental and RR cell lines was performed using a published list of genes whose expression profile denotes the breast cancer intrinsic subtypes (basal, normal-like, Her2, luminal A and luminal B) [16]. Assignment of individual samples to intrinsic subtypes was performed using the genefu R package [17]. Genefu implements a Single Sample Predictor (SSP) algorithm which is a nearest-centroid classifier. The centroids representing the breast cancer molecular subtypes were identified through hierarchical clustering using the same intrinsic gene list that we used for cluster analysis in this study. All datasets generated and/or analysed during the current study are available in the NCBI’s Gene Expression Omnibus [18] and are accessible through GEO Series accession number GSE120798.

Image analysis software QuPath version 0.1.2 [19] was used to analyse ki67 and ERα target protein expression. Two-way ANOVA with Holm-Sidak’s multiple comparisons test was used to test for differences between 2 groups in CF, SRB, invasion and migration assays and western blot experiments. Unpaired (two tailed) t-test was used to assess differences between 2 groups in IHC analysis. P values < 0.05 were deemed statistically significant. Data is shown as mean ± SEM with all statistical analysis and graphs generated with GraphPad Prism 7. An overview of the samples included in each experiment (including cell line, time points, treatments and number of replicates) is provided in Additional file 2: Table S2.

Radioresistant cell lines (MCF-7 RR, ZR-751 RR and MDA-MB-231 RR) were developed from their parental cell lines (MCF-7, ZR-751 and MDA-MB-231) by weekly exposure to single fractions of radiation, increasing by 0.5 Gy per week over a period of 12 weeks; cells were subsequently maintained by weekly doses of 5 Gy. Although maintenance radiation doses were still accompanied by cell death in all 3 RR cell lines, this was significantly less than that seen during the initial 12 week development period. Radioresistance was confirmed by both CF and SRB assays. The CF ability of all RR cell lines was significantly higher than their respective parental cell lines when exposed to a single dose of radiation up to 6 Gy (Fig. 1a). Significantly less inhibition of proliferation was also seen in the RR cell lines compared to their parental cell lines when exposed to a single dose of radiation up to 10 Gy (Fig. 1b). The IC₅₀ (the dose of radiation required to reduce cell number by 50%) values were higher in the RR cell lines when compared to their parental cells (Table 2). MCF-7 RR and MCF-7 RR cells which had not received radiation for 6 months (MCF-7 rr) were similarly radioresistant (Additional file 3: Figure S1), suggesting the longevity of the changes involved in the acquisition of this phenotype.

SRB assays were used initially to assess proliferation rates in 2D cultures (Fig. 2a). Results showed lower rates of proliferation in the MCF-7 RR and MDA-MB-231 RR and higher rates in the ZR-751 RR cell lines in comparison to their respective parental cell lines. Further investigation of proliferation was performed through 3D assays using IHC staining for the proliferation marker Ki67 in MCF-7 and ZR-751 parental and RR MTS (Fig. 2b) (MDA-MB-231 cell lines failed to develop MTS which could withstand IHC processing). IHC with quantitative analysis showed a lower percentage of positively stained ki67 cells in the RR MTS suggesting lower basal proliferation rates than their parental cell lines. Using gene expression data from 2D cultures, MCF-7 RR and ZR-751 RR cell lines were characterised by lower expression of genes involved in DNA replication and repair and those related to G1/S-phase transition and regulation of cell cycle, including AURKA and TP53, along with the higher expression of cell cycle arrest genes. Fewer changes in these genes were identified in the triple negative MDA-MB-231 parental and RR cell lines, both of which clustered next to each other and the MCF-7 and ZR-751 parental cell lines (Fig. 2c).

Epithelial-to-mesenchymal transition (EMT) is considered a normal feature of embryogenesis involved with cellular movement and morphogenesis during embryonic development [20] and may be an inherent property of normal basal stem cells in the breast [21]. However, the process can also be involved in tumour development with the conversion of early stage non-invasive tumours into invasive malignancies [22]. Acquisition of an EMT phenotype is associated with increased migration and invasion potential. Here, using a 3D invasion assay, we showed that both MCF-7 RR and ZR-751 RR have increased invasive potential compared to parental cells (which demonstrated little or no invasive capabilities); although the MDA-MB-231 RR cell line showed a slight increase in invasiveness at 72 h, this was not statistically significant (Fig. 3a). Similarly, we assayed for migration potential using a 2D scratch assay and showed that all 3 RR cell lines had significantly enhanced migratory ability compared to their parental cells (Fig. 3b).

Parental ER+ cell lines exhibited a typical epithelial-like morphology, consisting of tightly packed cells forming cobblestone-like monolayers, with the cells consisting of a large nucleus and small amount of cytoplasm. By 12 weeks post-radiation, morphological changes were observed in their RR derivatives (Fig. 4a). Individual cells or those growing in small clusters gained a more spindle-shaped morphology with a larger amount of cytoplasm, with the contact between cells now via focal points rather than cell clusters. The MDA-MB-231 cell line already exhibited a mesenchymal-like phenotype, and phenotypic changes in its RR derivate were not obvious.

The observed change in cell morphology with the acquisition of radioresistance is consistent with cells undergoing EMT. We therefore investigated the expression levels of EMT markers using IHC for the MCF-7 and ZR-751 parental and RR cell lines and identified increased expression of vimentin, N-cadherin and SNAIL, along with the partial down regulation of E-cadherin in the RR cell lines. The MDA-MB-231 cell line was characterised by high vimentin along with low N-cadherin and E-cadherin expression; no differences between parental and its RR derivative were identified (Fig. 4b and Additional file 4: Figure S2 and Additional file 5: Figure S3). These results were recapitulated using gene expression analysis from a published EMT signature (Fig. 4c) [23]. This study combined bioinformatic expression data analysis from both The Cancer Genome Atlas and Cancer Cell Line Encyclopaedia databases of seven tumour types which identified a pan-cancer EMT-associated gene expression signature. The lists of genes used have been provided in Additional file 6: Table S3. In our analysis, the MCF-7 and ZR-751 parental cell lines had a pattern of expression consistent with an epithelial genotype, whereas the MDA-MB-231 (both parental and RR) cells had higher expression of the mesenchymal cluster of genes. The MCF-7 RR and ZR-751 RR cell lines had a mixed pattern of expression with relatively high expression of both epithelial and mesenchymal genes, suggesting a transition from an epithelial towards a mesenchymal expression profile (Fig. 4c).

Analysis of transcriptomic data from ER+ parental and RR cell lines in respect of the WNT signalling pathway was investigated due to its reputed role in EMT and radioresistance [24]. The WNT pathway was investigated using panels of WNT signalling pathway members and down-stream targets, taken from the KEGG database [25]. The lists of genes used have been given in Additional file 6: Table S3. In this analysis the MCF7 RR and ZR-751 RR cell lines had a pattern of expression consistent with WNT pathway activation enriched for genes including FRIZZLED family members 1/2/5/7, WNT5a and WNT5b (Fig. 5a). IHC using MCF-7 and ZR-751 parental and RR MTS showed increased expression of WNT5a in the radioresistant cell lines which was predominantly expressed in the cells located in the outer proliferating layer of the MTS (Fig. 5b). Further validation through western blot analysis of serum starved whole cell lysates (in accordance with gene analysis experiments) also showed increased WNT5a expression in the radioresistant cell lines at 0 and 24 h post-serum starvation (Fig. 5c).

Breast cancer subtypes can be characterised by the expression profiles of key signalling receptors (ERα, PgR, HER2 and EGFR). Western blotting and IHC showed that the RR phenotype in the cell lines derived from ER+ cells was characterised by the loss of ERα and PgR expression, along with a gain in total EGFR expression (Fig. 6A & B and Additional file 7: Figure S4 and Additional file 8: Figure S5). ICC with quantitative analysis confirmed reduction in the percentage of cells staining positive for ERα expression in the MCF-7 RR (0.27 ± 0.17%) cell line compared with the parental MCF-7 cell line (81 ± 1.1%) and in the ZR-751 RR (0.52 ± 0.36%) cell line compared with the parental ZR-751 (96.5 ± 1.8%) cell line. No Staining was seen in the MDA-MB-231 cell line or its RR derivative. Three replicates with a minimum of 10 spheroids were included in each analysis. To investigate the consequences of this gain in EGFR expression, we treated parental and RR cells with increasing doses of the EGFR inhibitor, gefitinib, and determined its effect on the proliferation and migration of ER+ parental and RR cell lines. There was a statistically significant decrease in MCF-7 RR proliferation after 72 h with gefitinib concentrations ranging from 0.1–15 μM compared to the parental cell lines (IC₅₀: MCF-7 13.36 μM, MCF-7 RR 6.43 μM) and a significant reduction in migration, with a dose of 5 μM decreasing the migratory potential of the RR cell line to parental levels (Fig. 6Ci & Cii). Similar results were also seen in the ZR-751 parental and RR cell lines when treated with gefitinib (Additional file 9: Figure S6). As we had identified a loss in ERα expression, we also examined the effects of tamoxifen on the cell lines. After 72 h of tamoxifen treatment there was a statistically significant increase in proliferation of the MCF-7 RR cell line compared to the parental cell line at concentrations ranging from 0.01–3 μM (Fig. 6Ciii).

The loss of ERα in the MCF-7 cell line was further investigated using ERα siRNA knockdown. ERα knockdown was associated with increased EGFR expression 96 and 120 h after transfection, as shown through western blot analysis (Fig. 6D). There was also a significant reduction in proliferation of the MCF-7 cells after 72 h treatment with either ERα siRNA (25 nM) or tamoxifen (0.05 μM) (Fig. 6Ei), whereas no reduction in proliferation was seen in the MCF-7 RR cells (6Eii). An additional reduction in MCF-7 proliferation was observed when ERα knockdown or tamoxifen treatment was combined with radiation 24 h after transfection/tamoxifen treatment (Fig. 6Fi&ii), suggesting a radiosensitising rather than a radioprotective effect from ERα loss.

ER signalling was further investigated through the application of a published gene expression signature for ER pathway signalling activity to our transcriptomic data [26]. As expected, the ER+ parental cell lines (MCF-7 and ZR-751) were characterised by high expression of these genes whereas the ER- parental and RR cell lines (MDA-MB-231) had lower expression. Interestingly, the MCF-7 RR and ZR-751 RR cells clustered with the ER- cell lines in hierarchical clustering analysis, although they maintained higher expression levels, similar to their parental cells, for a subset of these genes. Treatment with 2 Gy of radiation for 2 and 8 h was not found to influence the expression of these genes (Fig. 7a).

Gene expression data was integrated with a public gene expression dataset (GSE50811) of 67 breast cancer cell lines. The MDA-MB-231 parental and RR cell lines, both treated and untreated, clustered tightly with each other in the dendrogram branch enriched for the basal breast cancer subtype. Both treated and untreated MCF-7 and ZR-75-1 parental cell lines were classified as luminal A whereas their RR derivates clustered independently and were enriched for normal-like (ZR-751 RR) and HER2-overexpressing (MCF-7 RR) intrinsic subtypes (Fig. 7b and Additional file 10: Figure S7).

Downstream signal transduction pathways of the HER/ERBB tyrosine-kinase receptor family were further investigated following identification that total EGFR expression was increased in the ER+ derived RR cell lines. Western blot analysis of time course experiments up to 30 min after 2 Gy radiation showed increased levels of p-ERK in MCF-7 RR and ZR-751 RR cell lines 5 min post-radiation, whereas a smaller increase was seen in the parental cell lines at a later time point (Fig. 8a). Although p-AKT did not show increased levels in response to radiation, the RR cell lines had overall increased expression levels compared with the parental cell lines (Fig. 8b). No differences were seen in the MDA-MB-231 and RR cell lines. The activity of the MAPK pathway in our cell lines was investigated using a published gene expression signature [27]. PI3K activity was assessed using genes taken from the KEGG pathway database [25] in combination with FOXO-regulated genes (these have an inverse expression pattern to PI3K activity) [28]. Gene expression analysis suggested that active MAPK and inactive PI3K signalling were constitutive in the MDA-MB-231 parental and RR cell lines and expression levels of these signalling pathways were not affected by radiation treatment. The untreated ER+ cell lines were characterised by both inactive MAPK and PI3K activity, whereas their RR derivatives had lower overall expression of the MAPK-associated genes, potentially representing a switch to active MAPK and PI3K signalling (Fig. 8c and d).