Background
Breast cancer is the leading cause of cancer in women, and its incidence continues to rise, particularly in developed countries [
1]. Strong evidence exists to support the role of aberrant epigenetic mechanisms in breast tumorigenesis, of which the most intensively investigated are changes in DNA methylation [
2-
4]. DNA methylation is evolutionarily the oldest and perhaps best studied mechanism of epigenetic transcriptional regulation, whereby a methyl group is covalently added to the 5-carbon of cytosine bases in a cytosine-guanine dinucleotide (CpG site). CpG sites tend to cluster into non-random CpG islands (CGIs) around the transcription start sites (TSS) of approximately 60% of genes. Dogma states that methylation of the promoter region-associated CGIs leads to conformational changes in the DNA strand and regional chromatin [
5,
6], which inhibit the initiation of the transcriptional machinery and prevent the recruitment of RNA polymerase II. If the CGI is unmethylated, the gene should be actively transcribed.
A recent review indicated that the differential methylation of intragenic variable regions may have important implications for transcription and cell-specific differentiation [
7]. Changes in intragenic methylation (IGM) levels may represent the consequences of the transcriptional machinery [
8], or a functionally relevant mechanism that affects transcriptional efficiency or gene stability [
9-
11]. It is likely that there are gene-to-gene subtleties in such mechanisms, and functionally important genes in breast cancer therefore warrant closer investigation as the transcriptional regulation of genes during breast tumorigenesis and throughout the disease course remains poorly understood. One such gene, oestrogen receptor alpha (
ESR1), is of crucial importance in terms of both diagnostic and prognostic implications in breast cancer [
12-
14]. A previous study from our group indicated that regions of DNA methylation variability (MVRs) exist across the
ESR1 gene in peripheral blood cells from breast cancer patients compared to healthy matched controls [
15], but the functional implications of this variability remains unknown.
Based on the hypothesis that IGM may play an important role in transcription [
16-
19], we aimed to ascertain whether IGM patterns differed in human breast cancer cells lines that were positive (n = 3) or negative (n = 3) for ESR1 expression. We also explored the effects on the cells in terms of the methylation and transcription of
ESR1 after treatment with a demethylating agent, decitabine (DAC), Furthermore, methylation levels across the
ESR1 gene were assessed in 155 samples of human breast cancer, and in 89 samples of exfoliated breast epithelial cells from donated expressed breast milk (EBM) from healthy women.
Methods
Cell lines
Six cell lines were obtained from stocks at the Hammersmith Hospital or purchased (ATCC, VA, USA). Of these, three were confirmed as ESR1-positive (T47D, MCF7, and BT474) and three were ESR1-negative (MDA-MB-231, BT549, and SKBR3), verified by STR profiling. Cells were cultured in sterile conditions at 37°C in a humidified atmosphere with 5% carbon dioxide, and maintained in either DMEM (Sigma-Aldrich, Poole, UK) or RPMI (Sigma) supplemented with 10% fetal calf serum (FCS; Sigma) and 5 ml L-glutamine. Cells were passaged when their confluence exceeded 70%.
Decitabine treatment
The effect of increasing concentrations of DAC on the six cell lines was assessed using the MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) dye reduction assay. Decitabine (DAC; Sigma-Aldrich) was re-suspended in 2.2 ml 100% dimethyl sulphoxide (DMSO; Sigma-Aldrich), and made up to 0.5, 1, 5, 10, or 20 μM compared to growth medium (0 μM) alone as the negative control. Assays were performed in triplicate, and the MTT assay was performed using 20 μl CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, WI, USA) according to the manufacturer’s protocol. The results indicated that cell viability was preserved for each cell line at ≤5 μm DAC. Therefore, 1 μm DAC was chosen for the subsequent cell culture experiments to prevent DAC cytotoxicity.
Fresh aliquots of DAC and DMSO were used for each experiment. Each cell line was cultured in 75 cm3 flasks in 10 ml DMEM + 10% FCS with 1 μM DAC or DMSO for 7 d in triplicate, and at three separate time points. After the appropriate duration of incubation, cells were trypsinised and counted. Cell pellets were collected after three PBS washes and centrifugation at 1,500 rpm for 5 min, and divided in half for DNA and RNA extraction. DNA was extracted using the QIAamp® DNA Mini Kit (Qiagen, Crawley, UK), and concentration and quality was assessed using a Nanodrop1000 spectrophotometer (ThermoScientific, UK). DNA was stored at −20°C until bisulphite conversion.
Methylation analyses
Bisulphite conversion changes all unmethylated cytosine bases into uracil, therefore allowing the identification of unconverted cytosines as those that are methylated by pyrosequencing [
20]. DNA samples were bisulphite-converted using the EpiTect kit according to the manufacturer’s protocol (Qiagen). Bisulphite-treated DNA was then desulphonated, washed and eluted prior to its use in PCR.
PCR assays were designed using a semi-nested approach to avoid the amplification of repetitive elements, such as long-interspersed nuclear elements (LINE) segments, which are often present in the MVRs across
ESR1 [
15]. A biotinylated tag was placed on one of the primers, and a common biotinylated primer was used for all reactions as described in previous reports [
15,
21]. The list of PCR and sequencing primer sequences is given in Additional file
1: Table S1. The prepromoter region assayed was found between −4839 and −3904 bp upstream of the transcription start site (TSS), while the promoter region assayed comprised CpG sites from the TSS to 178 bp into the gene. Reactions took place in a thermal cycler under the following conditions: incubation at 95°C for 10 min; an initial 20 sec incubation at 95°C followed by 10 cycles of a 20 s incubation at 60°C (temperature decreased by 1.0°C every cycle) and incubation at 72°C for 20 s; second round PCR steps were performed using nested primers as follows: 30 cycles at 95°C for 20 s, 50°C for 20 s and 72°C for 20 s followed by a final incubation of 72°C for 5 min, with the exception of MVR 7b which only required a single-step PCR amplification. Products were assessed for quality by agarose gel electrophoresis and stored at 4°C until pyrosequencing.
Bisulphite-converted DNA samples were pyrosequenced using specific sequencing primers designed with the use of the PyroQ assay design software (Pyromark MD, Qiagen), and assay were performed on a Pyromark MD pyrosequencer using standard protocols and controls. Assays were repeated if any the inbuilt quality control measures were flagged.
RNA isolation, cDNA synthesis and qRT-PCR assays of cell line RNA
RNA was isolated from cell pellets using the Qiagen RNeasy® Mini kit (Qiagen), according to the manufacturer’s instructions. The concentration of each RNA sample was assessed with the Nanodrop and all OD260/280 ratios were >1.8. cDNA was synthesised from 2 μg of each RNA sample using the SuperScript™ III First Strand Synthesis System for RT-PCR (Invitrogen, Carlsbad, CA, USA). Negative controls were prepared without Superscript™ III RT for each group of samples. All samples were stored at −20°C prior to RT-PCR.
Each qRT-PCR analysis was performed in triplicate for each of the duplicate experimental sets of cDNA from the six cell lines. Each qRT-PCR run was performed in duplicate using primers that were specific for
ESR1 mRNA and for the housekeeping gene,
GADPH (forward, 5′-TCCCATCACCATCTTCCA-3′ and reverse, 5′-CATCACGCCACAGTTTCC-3′) [
22]. The details of primers used are given in Additional file
1: Table S2, and assays were checked using gel electrophoresis to confirm the expected amplicon sizes were valid. All primers were 100% specific for the region of interest. The plate was centrifuged briefly and placed in a C1000™ Thermal Cycler (BioRad, UK). The PCR conditions were established using the Bio-Rad CFX Manager software as follows: 95°C for 3 min denaturing step; 42 cycles of 10 s at 95°C, 10 s at 56°C and 30 s at 72°C; 10 s at 95°C and a melt curve cycle of 5 min that ranged from 72°C to 95°C. Cycle threshold (Ct) values were recorded at a logarithmic threshold of 10
3, and the relative quantitative expression of
ESR1 mRNA in each sample was calculated by the
-∆∆Ct conversion.
Breast tumour samples
Power calculations based on the observed differences in cell lines suggested that group sample sizes of n = 45 would be sufficient to reach >90% power at alpha = 0.01 to detect the maximum difference observed (p6), and >80% power at alpha = 0.05 to detect a significant difference of >40% methylation (observed at other sites across ESR1). We received samples from the Breast Cancer Campaign Tissue Bank, comprising 10 formalin-fixed paraformaldehyde slides per tumour for 135 tumours (45 ER-negative tumours samples, 45 ER-positive grade 2 tumours and 45 ER-positive grade 3 tumours, as defined from histopathological review by MEB). Furthermore, we received 20 samples of fresh frozen (FF) tumours matched to FFPE samples for quality control purposes. This study was approved by the Ethics Committee of the Breast Cancer Campaign Tissue Bank (Approval no. BCC-TB00001). All H&E stained slides were reviewed by a pathologist to define the percentage of tumour with a minimum cut-off of >70%.
Slides were dewaxed for 10 min in Histoclear, followed by 10 min in 100% ethanol and another 10 min in fresh 100% ethanol. Slides were prepared with Levi buffer using standard techniques, and DNA was extracted using the phenol:chloroform technique. DNA concentrations and quality were assessed by the Nanodrop. Bisulphite conversions and pyrosequencing analyses were performed as described above. For the DNA extracted from FFPE slides, different primers had to be described with amplicons of <120 bp owing to the relative fragmentation of the DNA after formalin treatment, as DNA quality was poorer in these 135 samples (Additional file
1: Table S4).
Slides were stained immunohistochemically with antibodies against Ki67 according to standard protocols to assess the rate of cell proliferation within tumour sections. Briefly, 2-μm-thick sections from formalin-fixed, paraffin embedded tissue blocks were prepared, deparaffinised and rehydrated. Immunohistochemical staining and detection was performed using an automated Leica Bond 3 machine according to the manufacturer’s protocol. Antibodies raised against Ki67 (Leica, Cat No: NCL-L-Ki67-MM1, 1:100) and ER (Leica, Cat No: NCL-ER-6 F11, 1:500) were used. Stain detection was performed using a bond polymer refine detection kit. Tonsil sections were used as a positive control for Ki67 staining and breast tissue was used for ER staining. Negative controls were processed in the same manner but with the substitution of PBS for the primary antibody. All sections were examined by light microscopy to assess the presence and scoring of expression. The percentage of tumour cells with nuclear expression of Ki67 was estimated. The Allred scoring system was used to assess ER staining.
Methylation data in BCC Tissue Bank tumour biopsies was validated using the TCGA breast tumour for which 450 K Illumina Infinium Beadchip Array data was publically available (n = 365 ER-positive tumours, n = 109 ER-negative tumours). Data was extracted using R software and logistic regression analysis was performed to assess the relationship between ER-status and methylation beta-values at each ESR1 CpG locus, with histology as an independent variable. The Wilcoxon signed rank sum test was performed with false discovery rate correction as the data was non-parametric.
Extraction and processing of breast epithelial cell samples from expressed breast milk
Ethics approval for this part of the study was obtained from the Hammersmith Hospital Human Imperial NHS Tissue Bank access committee (reference R13020). Cells were pelleted from frozen 20-ml samples of expressed breast milk in a series of centrifugation and wash steps, and analysed using flow cytometry with a FITC labelled antibody against epithelial membrane antigen (EMA; CD227, Sigma Aldrich; n = 60) and a Cyp5.5-labelled antibody against intracellular ESR1 (Sigma Aldrich; n = 6), according to standard techniques using a FACScalibur flow cytometer (BD Biosciences). DNA was extracted using a phenol:chloroform technique, with duplicate phenol and chloroform steps to optimise yields and reduce phenol contamination, respectively. DNA was bisulphite converted (500 ng) using the EZ-96 Methylation-Gold™ kit (Zymo) and samples from 36 donors were used for hybridisation onto the Infinium HumanMethylation 450 BeadChip array, using the Illumina Infinium HD methylation protocol (conducted by UCL Genomics). The methylation scores from samples on all three chips were processed using standard quality control measures, and normalised (colour correction) and batch adjusted using COMBAT, resulting in beta methylation values according to the fluorescent intensity ratio that ranged from 0 (unmethylated) to 1 (completely methylated). R was also used to analyse the probes related to ESR1. Pyrosequencing was subsequently performed for the ESR1 regions described above on 250 ng bisulphite converted DNA samples (n = 53) to validate the ESR1 regional methylation.
RNA was found to be highly fragmented from frozen milk samples according to assessment by the Bioanalyser 2100 (Agilent Technologies, Santa Clara, CA, USA) using standard techniques, and was unsuitable for further use. A small number of fresh breast milk samples were collected and high quality RNA, according to the results of the Bioanalyser 2100 (RIN score >7), was obtained from 11 samples using a standard Trizol technique. Furthermore, the OD260/280 was >1.8 for all 11 samples. cDNA was prepared and qPCR assays for ESR1 were performed according to the techniques and primers described above for the cell line analysis. EBM samples were normalised against MCF7, and MDA-MB-231 RNA was used as the negative control, along with a negative reverse transcriptase sample.
Statistical analysis
All experiments were performed in triplicate unless otherwise stated. The mean ± standard deviation (SD) was calculated from each triplicate repeat of the pyrosequencing and qRT-PCR experiments. The mean ± SD were calculated after each replicate, and the standard error of the mean (SE
x) was then calculated. Parametric data, such as the methylation levels in cells incubated with DMSO and DAC or DMSO alone, were compared using paired t-tests. Non-parametric data, including average methylation levels across the gene body, were compared using unpaired Wilcoxon signed rank sum tests. All statistical tests were two-sided and performed using Microsoft Excel (Microsoft, USA). To validate the expression changes of ESR1 after DAC treatment, two publically available expression datasets were mined for data regarding DAC treatment of two breast cancer cell lines used in this current study (gse10613 and gse13733) [
23,
24]. The software programmes, R v2.15 and Microsoft Excel, were used to analyse all data.
Discussion
This study identified marked and reproducible differences in the pattern of IGM across ESR1 in vitro in ER-positive compared to ER-negative cell lines, which was supported by similar methylation patterns in the strongly ER-positive breast epithelial cells from breast milk samples. Promoter regions were uniformly methylated in ER-negative cell lines, and unmethylated in ER-positive cells. As predicted, demethylation with DAC treatment increased the transcription of ESR1 in all three ER-negative cell lines, but the most surprising finding from this study was that DAC resulted in decreased levels of expression in ER-positive cell lines, via a mechanism independent from promoter methylation. Of note, the patterns of promoter and IGM established in tumour cell lines and a homogenous population of breast epithelial cells from EBM samples were highly similar, but differed markedly from those generated from two sources of ER-positive or ER-negative tumour biopsy samples (BCC Tissue Bank and the TCGA database). These observations are likely to reflect various caveats, including cell type heterogeneity and the tissue microenvironment, which results in a mixed epigenetic signal in tissue samples. This finding was in contrast to the artificial nature of cell lines grown on plastic in the presence of high concentrations of growth factors and has important implications on the choice of tissue for epigenetic analyses.
The principal focus to date of the transcriptional effects of DNA methylation has been on promoter-associated CpG islands. The results of this current study were in accordance with recent findings by Yang et al., which indicated that DAC treatment reduced the expression levels of overexpressed genes [
26]. The functional mechanism by which IGM exerts a transcriptional effect remains unknown, but the transcriptional changes observed for ESR1 in ER-positive cells may represent a therapeutic target. With the advent of array techniques that examine greater proportions of the genome, including high-density microarrays and next generation sequencing based DNA methylation analyses, the functional roles of IGM are becoming more apparent [
27] and are linked to gene expression [
17,
18,
28]. IGM levels have been shown to change markedly during carcinogenesis [
27], but in the absence of precise roles in the normal state, the effects of disrupted IGM levels in aberrant cells cannot be predicted or quantified. Several mechanisms by which IGM may be functionally important have been proposed, including the prevention of transcription from alternative start sites in the gene body, chromatin regulation, the inhibition of transposable elements, and the control of alternative splicing [
29,
30]. Furthermore, high IGM levels may prevent the transcription of non-coding RNA in the antisense direction, although this finding was not supported by the results of this current study. Moreover, the methylation of intragenic transposable elements may affect transcription efficiency by impeding RNA polymerase II along the gene body [
31], although recent evidence suggests that intragenic DNA methylation represents a by-product of the chromatin assemblies related to transcription, and has no direct impact on transcription efficiency [
8].
DAC is currently used as a clinical treatment for myelodysplastic syndrome and acute myeloid leukemia [
32]. DAC is incorporated into double-stranded DNA during cell replication, and therefore more rapidly dividing cells might show greater levels of demethylation. The ER-positive breast cancer cell lines in this study were passaged more frequently than ER-negative ones, and all cell lines were cultured in media that contained oestrogen, fuelling a higher rate of replication in the ER-positive cells. A link between DNA methylation and proliferation has previously been proposed by our group and others [
19,
33]. Aran et al. observed that proliferating cells and tissues tended to have higher levels of IGM [
19]. This observation was corroborated by our Ki67 and methylation data in tumour biopsies where the ER-negative tumours had a higher proportion of proliferating cells than the ER positive tumours (Table
1). This leads us to hypothesise that intragenic methylation levels may be influenced by the cell proliferation rate. In terms of the DNA from tumours biopsies, ER-negative tumours had a much faster rate of cell proliferation, as indicated by the significantly higher levels of staining with Ki67 compared to ER-positive cells (Table
1). It is likely that the higher levels of IGM observed in ER-negative tumours were influenced by the higher levels of cell proliferation in these tumours. Velicescu et al. noted that serum starvation stalls cells at the G0/G1 phase of the cell cycle, preventing cell division and that DNA methyltransferases were predominantly expressed during the S phase [
34]. Active demethylation of promoter regions is known to be initiated by the same enzymes that induce methylation (DNMT3A and 3B) [
35,
36], but is a long and energy expensive process [
37]. To establish if cell proliferation does actively affect IGM levels, and the subsequent effects on transcription, larger studies that investigate different cell lines and clinical samples with a genome-wide microarray approach may be required.
The genomes of cancer cells undergo massive epigenetic changes with the loss and redistribution of methylation. During neoplastic change, the CGI-associated promoter regions of multiple genes across the genome become focally hypermethylated [
3,
38,
39], which may occur concurrently with genome-wide hypomethylation in tumour cells from a variety of cancers, including breast cancer [
40,
41]. These IGM changes may follow a distinct order during carcinogenesis and provide biomarkers of breast cancer risk in healthy women [
42,
43], as has already been proposed in ovarian cancer [
44]. However, the pathological mechanisms and implications of genome-wide demethylation are not understood, but may result in the reactivation of repetitive elements that are usually hypermethylated, with consequent genomic instability [
45].
The observed intragenic
ESR1 methylation in epithelial cells extracted from human breast milk confirmed the
in vitro cell line findings of high levels of IGM in ER-positive tumour cell lines, and low levels of promoter methylation. This suggested that if a homogenous cell type is investigated, the epigenetic profile is also more homogenous compared to the tumour biopsy material, where the cell-specific signatures create a mixed signal [
46]. From the limited number of samples of RNA available from EBM (n = 11), no correlation was found between the highly variable methylated region at p2 in
ESR1 and ESR1 expression, however, further studies in a larger number of samples will be required to investigate such associations.
Of note, this is the first study to show the feasibility of using DNA extracted from cells in EBM for 450 K DNA methylation analyses. All 36 samples passed the quality control procedures, with a relatively low level of excluded probes. Although the DNA from such frozen milk samples is relatively fragmented (data not shown), sufficient quality is retained to enable both array and PCR-based assays to be performed. Given that these cells are in a highly proliferative state during lactation [
47], and represent the most oestrogen responsive breast epithelial cell type, it will be important to collect and characterise further samples of these cells in the future to gain a greater understanding of the normal biology of ductal epithelial cells, and how their epigenetic status differs from cancer cells [
48]. They also represent an important resource in which the impact of environmental and intrinsic cancer risk factors can be assessed.
Limitations
This study had several limitations. Firstly, we have not investigated the effect of passive demethylation via decitabine on distant enhancers or the many alternate promoters of
ESR1. Secondly, the results from breast tumour biopsies could have been confounded by the presence of 5’hydroxymethylation, which is present in primary tissue but not in
in vitro cultured cell lines (and high passage number cell lines in particular). Future studies that use novel techniques such as oxidative bisulphite sequencing for the detection of this epigenetic mark will be needed to assess this. Thirdly, the quality of DNA extracted from FFPE tumour sections was relatively poor compared to that of fresh frozen samples, leading to higher technical variation which may have further confounded the analysis. The technical variation in gene expression may have been reduced by using alternative control genes to GAPDH [
49]. Finally, the number of samples of epithelial cells from EBM was relatively small, particularly for freshly expressed samples from which it was possible to extract RNA for further assays and larger studies are warranted. In accordance with most previous reports, this study demonstrated a correlation between methylation and expression levels. While it is possible to remove methylation with decitabine and show the reciprocal changes in expression, we have not shown the reverse of adding intragenic methylation to a gene and showing a reciprocal increase in expression. Only now, with advances in the use of CRISPR technology, is there promise that such an experiment might be possible, and future investigations will examine whether this technique can work reliably for DNA methylation studies [
50].
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
Experiments were performed by NSS, EB and EG. KJF participated in sample preparation. M-EB performed pathological review of all tumours. CWB and WD assisted with bioinformatics support and statistical analysis. JMF, RB and NSS conceived the study design and coordination. GW coordinated the provision of breast milk samples from the Queen Charlotte and Chelsea Hospital Milk Bank. NSS and JMF drafted the manuscript. All authors read and approved the final manuscript.