Background
The pathogenesis of breast cancer is a complex, multistep process involving multiple genetic changes. A major risk factor associated with the development of the disease is the duration of exposure to estrogens, the length of which is increased in women experiencing early menarche and/or late menopause. Estrogens are steroid hormones that play important roles in the growth and development of the mammary gland and it is well established that the growth of breast cancer cell lines in culture or in ovariectomized nude mice is stimulated by estrogens [
1‐
3].
Approximately two-thirds of all breast cancer tumours are ER-positive [
4‐
6] and more than 50% of these are also PR-positive [
7]. Both receptors are useful in predicting response to endocrine therapy [
5,
7‐
9] and in general ER-negative tumours are associated with early recurrence and poor patient survival relative to those that are ER-positive [
5,
8,
9]. Despite clinical advances of ER-targeted therapy,
de novo and acquired resistance to all forms of endocrine therapy remains a great obstacle [
8,
9]. Complicating matters, we and others have shown in mostly retrospective studies, that expression of ER and PR are unstable during tumour progression from a primary lesion to its corresponding metastasis [
10‐
13].
Long-term estrogen deprived (LTED) cell lines can serve as an
in vitro model mimicking the hormonal milieu of breast cancer cells in oophorectomized pre-menopausal women, postmenopausal women and/or patients treated with primary endocrine therapy, in particular aromatase inhibitors (AIs) [
14]. Of note, the use of AIs in place of traditional endocrine treatments results in a statistically significant survival gain (HR 0.90, 95% CI 0.84 to 0.97) [
15].
Whilst previous studies have examined ER, PR and HER-2/neu expression in an LTED setting, no comprehensive gene and protein analysis has been performed on all three markers. As such, our descriptive study addresses this knowledge gap by determining the levels of ER, PR and HER-2/neu gene and protein expression in two ER-positive and one ER-negative cell line at multiple time points, coupled with gene expression array profiling, all in a well-described LTED model [
16‐
20]. Adding further clinical relevance to our analysis, we related our expression array findings to publicly available array data of breast cancer patients treated with an aromatase inhibitor. Our work highlights the unstable nature of ER and PR expression under conditions of estrogen deprivation, and demonstrates the significant overlap of genes altered in LTED cell lines and AI-treated patients.
Methods
Cell culture
A long-term estrogen deprivation (LTED) model was used to study the three commonly used breast cancer cell lines MCF7, BT474 and MDA-MB-231 [
7,
8]. MCF7 and MDA-MB-231 cells were newly purchased from Sigma-Aldrich and BT474 cells from the American Type Culture Collection (ATCC). Control and LTED cells were routinely maintained in phenol red containing MEM or DMEM supplemented with 10% fetal bovine serum (FBS) or phenol red-free MEM or DMEM supplemented with 10% dextran-coated charcoal-stripped FBS (DCC-FBS) to remove substantial amounts of estrogen, respectively. Each culture medium was further supplemented with 100 IE/ml penicillin and 100 μl/ml streptomycin. All cells were grown at 37°C in a humidified atmosphere of 5% CO
2 and 95% air.
Immunocytochemistry
50 000 cells per cell line (MCF7, BT474 and MDA-MB-231 cells) were attached to slides (ChemMateTM Capillary Gap Microscope Slides, DAKO) by centrifuging them in a Cytospin 3 centrifuge (Shandon, Thermo Electron corporation, Waltham, Massachusetts), at 1000 rpm for 4 minutes in room temperature. The slides were then fixed in 4% formalin for 10 minutes at room temperature, followed by PBS for 10 minutes, methanol for 4 minutes in -20°C, and acetone for 1 minute in -20°C, before being placed in TBS. Automatic immunostaining was performed in a DAKO Tech Mate instrument (DAKO, Glostrup, Denmark). Staining of ER and PR was done using the recommended DAKO ChemMate Detection Kit (Peroxidase/DAB Rabbit/Mouse). The MDA-MB-231 cell line served as negative control for ER, PR and HER-2/neu expression. MCF7 cell line was used as positive control for ER expression, while BT474 cell line served as positive control for PR and HER-2/neu expression.
Immunoslides were assessed in a microscope by counting of positive cells and degree of staining. We used a modified H score system, using the formula: H score = (0 ×% tumour cells negative) + (1.5 ×% tumour cells moderately positive) + (3 ×% tumour cells strongly positive), giving a range 0–300. Five hundred cells were counted per slide. Two observers (JM and JK) evaluated the immunoslides, and the final score was calculated by taking the mean score. If the ratio between two scores was higher than 1.5, the slides were re-evaluated to reach consensus.
The following primary antibodies were used for immunocytochemical analyses: Monoclonal mouse anti-human progesterone receptor (PR) antibody (Clone PgR 636, DAKO, Glostrup, Denmark), diluted 1:1000, monoclonal mouse anti-human estrogen receptor antibody NCL-ER-6 F11 (NovoCastra Laboratories Ltd, Newcastle, UK), diluted 1:50, monoclonal mouse anti-human HER-2 (c-erbB-2 Oncoprotein) antibody (NCL-CB11), diluted 1:250 (Novocastra Laboratories Ltd., Newcastle, UK).
RNA isolation
RNA extraction was performed according to the RNeasy mini protocol (Qiagen, Germany). Briefly, one to five millions cells were collected for isolation of RNA from each sample before being applied to the MicroSpin affinity columns in the Qiagen kit. The quality of RNA was assessed using an Agilent 2100 bioanalyzer (Agilent Technologies, Rockville, MD, USA).
Quantitative real-time PCR analysis
The mRNA expression levels of ESR1, PGR, ERBB2 and an endogenous housekeeping gene encoding for 18S ribosomal RNA as a reference were quantified using TaqMan® technology on an ABI PRISM 7500 sequence detection systems (PE Applied Biosystems). Sequence-specific primers and probes were selected from the Assay-on-Demand products (Applied Biosystems), including ESR1 (assay ID: Hs01046817_m1), PGR (Hs00172183_m1), ERBB2 (Hs00170433_m1) and 18S ribosomal RNA (Hs99999901_s1). All qRT-PCR experiments included a no template control and were performed at least in duplicate.
Microarray analysis
A total of eight samples, four from each cell line (control, 2 days, 6 weeks and 10 months), were selected for microarray analysis, performed at the core facility for Bioinformatics and Expression Analysis (BEA) at Karolinska Institutet. Briefly, biotinylated cRNA was hybridized to HG-U133 Plus 2.0 oligonucleotide arrays (Affymetrix Incorporated, Santa Clara, CA, USA), washed and scanned according to the protocol recommended by the supplier. Gene Chip Operating Software (GCOS) was used for calculation of detection calls, signal values and for calculation of the target intensity scaling of each array to an identical value and quantification of the signal log ratio. An average signal log ratio value was calculated for all transcripts in the long-term estrogen deprived cell lines compared to the cell lines cultured in medium containing estrogen. A minimum signal log ratio of 0.7 in each of four pair-wise comparisons was set as a threshold for significant differential expression. The quality of the data was verified by correlation analysis and multidimensional scaling plots in
R statistical environment using Bioconductor packages (
http://www.bioconductor.org). This data has been made publically available at NCBI GEO with series accession number GSE50820. Gene Ontology (GO) terms enriched in the lists of up-regulated and down-regulated genes including the 300 genes with highest SLR, were identified by Fisher’s exact test. For comparison of genes significantly changed in response to estrogen silencing to those significantly altered in our LTED model, we accessed publically available data (Gene expression omnibus number: GSE27473) from the NCBI GEO repository. The data is taken from a publication by Al Saleh
et al. [
21] where gene expression changes are determined in MCF7 cells after estrogen receptor silencing. In order to directly compare with our data, we downloaded and re-analysed the dataset using the statistical parameters outlined above to determine genes significantly changed in response to estrogen silencing.
Statistical analysis
All statistical analysis were performed using SPSS data analysis statistics software system version 17.0 (SPSS Inc., Chicago, IL, USA), the statistics tool in Microsoft Excel or R. ANOVA with post-hoc Tukey was performed on H-score and qPCR data and significance was calculated relative to day 0 control. Experimental results are expressed as mean ± SEM, where applicable. P-values of < 0.05 were considered statistically significant.
Discussion
In spite of the substantial progress that has been achieved in recent years in the treatment of hormone receptor positive breast cancer,
de novo and acquired resistance to endocrine therapy is still a major clinical problem [
8,
9]. In this descriptive study, we employed a LTED model to gain a greater understanding of how estrogen deprivation impacts clinically relevant prognostic markers and gene expression over time. To our knowledge, this is the first report to comprehensively investigate ER, PR and HER-2/neu expression along with qRT-PCR and gene expression array profiles at multiple early and late time points, in breast cancer cell lines after estrogen deprivation. Overall, our data are in line with previous reports showing that breast cancer cells can survive estrogen deprivation and re-grow, creating a phenotype that is likely less responsive to anti-hormonal therapy [
27]. Additionally, due to the multiple consecutive time points examined, we note clear trends in how the expression of ER and PR change over time on both the gene and protein level. Lastly, we underline the similarities between the specific genes changed in our LTED cell lines and patients treated with aromatase inhibitors, demonstrating the strong translational value of this model, as others have also noted [
23,
24,
28].
In order to put our work in the context of other studies and strengthen our findings, we compared our gene expression results to that of Aguilar
et al., who performed a similar study in an MCF7 LTED model [
24]. Through integrated aCGH and gene expression analysis the Aguilar study demonstrated that there may be shift towards a transcriptomic program in LTED cells that is independent of ERα transcriptional function. Whilst we did not perform matching aCGH analysis on our LTED samples, and despite the differences in time points assessed in both studies, we did note similar changes in gene expression probes over time. Specifically, we noted analogous changes in the probes for
ESR1, MKI67,
EGFR and
RAF1 (but not
GATA3), thus lending support to hypotheses proposed by Aguilar
et al.
Recent publications including two prospective studies, indicate lack of stability of ER and PR during tumour progression, in particular they seem to be altered when adjuvant therapies are given [
29‐
31]. This loss of receptors, at least in the examined parts of the biopsies, may be a further factor involved in resistance to endocrine therapies. It is also apparent from these studies that ER and PR seem to be more discordant in patients receiving more abundant adjuvant therapies and a similar finding has been demonstrated with chemotherapy and trastuzumab in the comparison of HER-2/neu status in the primary tumour and the corresponding recurrence [
31]. This clinical instability is reflected in our present cell line model, again underlining the suitability of LTED studies for investigating the time related alteration of receptors during conditions which mimic endocrine therapy with aromatase inhibition.
Previous studies have shown the propensity of breast cancer cells to adapt to conditions of long-term estrogen deprivation by up-regulating expression of ER, but not PR [
19,
32], thus developing hypersensitivity to the mitogenic effect of estradiol. In our experiments, we observed a marked up-regulation of ER in the MCF7 but not BT474 cell line at 10 months after estrogen deprivation. Some reports claim that this estradiol hypersensitivity is not a consequence of ER-mediated gene transcription but rather related to activation of the MAPK/ERK [
19] and EGFR/ERBB/AKT pathways [
24]. Similarly, recent evidence has also implicated a switch from ERα to NOTCH signalling in LTED cells [
28], a finding supported by our analysis where we see an up-regulation of the
NOTCH1 in MCF7 cells relative to control after 6 weeks of LTED culture.
The up-regulation of
NOTCH1 fits well with our findings of increased expression of genes that promote EMT in both LTED MCF7 cells at 6 weeks and AI treated patients. Previous studies have linked induction of EMT under hypoxic conditions to Notch signalling [
33], whilst ectopic expression of Notch1 intracellular domain (N1CD) has been demonstrated to trigger an EMT in epithelial cancer cells [
34]. Of particular note, others have shown that a decrease in estrogen dependency is correlated with an increase of the EMT marker Snail1 in an MCF7 LTED model [
35]. What these results mean in the context of AI treatment of breast cancer patients is difficult to ascertain. One might expect that as induction of EMT leads to an enhancement in the migratory capacity of cells, treating breast cancer patients with AIs would push tumour cells towards a more invasive metastatic phenotype. However, given the high success rates of endocrine treatments and reduced numbers of metastasis seen amongst these patients (relative to those who receive chemotherapy), this hypothesis would seem unlikely.
The down-regulation of PR following estrogen deprivation observed in our experiments could be caused by multiple cellular mechanisms. Cui
et al. have shown that insulin-like growth factor-1 (IGF-1), independent of ER activity, considerably down-regulates PR through the PI3K (Akt/mTOR) pathway [
36]. Along with others, they propose that low PR status may serve as an indicator of substantial activation of the growth factor signalling cascade, leading to hormonal therapy resistance [
37‐
40]. However, our gene array data did not support any significant involvement of the PI3K/Akt pathway and as such the mechanisms governing loss of PR in our model will require further investigations.
Acknowledgements
This work was supported by grants from the Swedish Research Council, Gösta Miltons Donations fond, Karolinska Institutet Foundations, Swedish Cancer Society, Cancer Society in Stockholm, The King Gustaf V Jubilee Fund, Swedish Breast Cancer Association, Märit and Hans Rausing’s Initiative Against Breast Cancer and BRECT.
Competing interests
Professor Bergh receives research funding from Merck, paid to Karolinska Institutet and from Amgen, Roche, Sanofi-Aventis and Bayer, paid to Karolinska University Hospital. The authors have no other potential competing interest to disclose.
Authors’ contributions
JM carried out the cell culturing, immunocytochemistry, H score analysis and qRT–PCR analysis, performed statistical analysis studies, participated in the interpretation of the data and drafted the manuscript. JK carried out H score analysis, participated in the interpretation of the data and helped to draft the manuscript. ALB helped with the immunocytochemistry analysis and isolation of RNA and DNA. TF participated in the qRT–PCR analysis. JB conceived of the study, participated in its design and coordination and revised the manuscript critically. NT participated in the interpretation of the data, design, performed the statistical analysis, and revised the manuscript critically. All authors read and approved the final manuscript.