Introduction
Advances in breast cancer therapy are facilitated by molecular characterizations of tumors and tumor subtypes. For example, breast tumors that express estrogen receptor (ER-positive, ER+) and progesterone receptor (PR-positive, PR+) and are dependent on the female sex hormone estrogen for growth and proliferation are treated by drugs that target ER either directly (tamoxifen, raloxifene, fulvestrant) or indirectly (letrozole, anastrozole, exemestane) by disrupting estrogen production [
1,
2]. Tumors that overexpress human epidermal growth factor receptor 2 (HER2/ErbB-2/neu+) on cell surfaces are targeted by monoclonal antibodies (trastuzumab) or tyrosine kinase inhibitors (lapatinib), which block receptor activation and tumor cell proliferation [
3]. Some tumors, however, are refractory to these targeted therapies or develop resistance over time. Blocking estrogen production and functions also imparts menopausal symptoms and increases corresponding health risks in premenopausal women. Moreover, a significant number of patients have triple negative (ER-, PR-, and HER2-) breast cancers and require alternative targeted chemopreventive and therapeutic strategies [
4]. Improvements of current breast cancer therapeutics and development of new ones necessitate the discovery and characterization of novel target mechanisms and targeting agents.
Both ER and PR belong to the nuclear receptor (NR) superfamily of ligand-dependent transcription factors, which function in normal development and physiology and in a number of human diseases [
5]. NRs are highly druggable targets of synthetic and natural ligands, and accumulating insights on NR structures and functions and advances in medicinal chemistry have provided receptor-selective, full, partial and inverse agonists and antagonists, as well as compounds that activate only a subset of the NR functions or in a tissue-specific manner. Discovery and characterization of other NRs that may also play important roles in breast cancer biology are, therefore, likely to yield other promising target mechanisms and agents.
Liver × receptors (LXRs) are NRs that are activated by oxysterols, synthetic ligands, and dietary phytosterols and have been well characterized as regulators of cholesterol, glucose, and fatty acid metabolism and inflammatory responses [
6‐
12]. At the molecular level, two receptor subtypes, LXRα and LXRβ, function in heterodimers with 9-cis retinoic acid receptors (RXRs), and their activities as transcriptional regulators are modulated by ligand binding and post-translational modifications mediated by cell signaling pathways [
13]. A number of LXR ligands have been developed for the treatment of atherosclerosis, diabetes, Alzheimer's disease, and inflammation. Published reports of anti-proliferative effects of LXR ligands on breast, prostate, ovarian, lung, skin, and colorectal cancer cells suggest that LXRs are potential targets in cancer prevention and treatment [
14‐
17]. Observations of increased proliferation markers in colon tissues and pre-neoplastic lesions in the gallbladder of LXRβ knock-out animals further suggest a role for LXRs and their ligands in cancer initiation and progression [
18,
19]. We have previously shown that synthetic LXR ligands can block the proliferation of both ER+ and ER- breast cancer cells through downregulation of some cell cycle and growth-associated genes [
20]. To additionally determine the effects of LXR ligands on breast cancer cells and to identify their potential mechanisms of action, we carried out microarray analysis of gene expression following treatment with the synthetic LXR ligand GW3965 in multiple breast cancer cell lines. Here, we report our findings regarding the effects of LXR activation on breast cancer transcriptomes, the potential role of E2F2 in mediating the anti-proliferative effects in ER+ breast cancer cells, and association of ligand-responsive gene networks with disease outcomes in breast cancer patients.
Materials and methods
Cell culture and treatment
MCF-7 ER+ breast cancer cells were cultured in DMEM (Invitrogen, Carlsbad, CA, USA) supplemented with 10% FBS (Saveen Werner, Limhamn, Sweden or Hyclone, Logan, UT, USA). ER+ T-47D cells were cultured in DMEM:F12 (Invitrogen) supplemented with 5% FBS. ER- SK-BR-3 and MDA-MB-231 cells were grown in Roswell Park Memorial Institute (RPMI) 1640 medium (Invitrogen) supplemented with 10% FBS. For microarray analysis, cells were treated with 10 μM of the synthetic LXR ligand GW3965 for 48 hours before harvest and RNA isolation with EZNA Total RNA Kit I (Omega Bio-Tek, Norcross, GA, USA) according to the manufacturer's protocol. Current institutional and governmental safety guidelines were followed in the performance of these experiments, and no additional ethical approval was required from institutional review boards.
Microarray and data analysis
We amplified 250 ng of RNA, which was converted to cRNA using the Illumina TotalPrep-96 RNA Amplification kit (Ambion, Carlsbad, CA, USA): 750 ng of cRNA was used for hybridization onto the Illumina Whole-Genome Gene Expression Direct Hybridization microarray (Illumina, San Diego, CA, USA), and 25,559 probes from the microarray were included for analysis. Probes for multiple genes were eliminated. The R software running the lumi and limma packages were used to determine differentially expressed genes in ligand-treated cells. Normalized intensity values were log-2 transformed. To correct for false discovery, we implemented the Benjamini-Hochberg correction [
21]. An additional filter for differentially expressed genes was set at 1.5-fold change in expression in either direction. Array data have been deposited in the NCBI Gene Expression Omnibus database [GSE34987].
Data mining
Bioinformatic analyses of enriched gene sets were made in Pathway Studio (Ariadne Genomics, Rockville, MD, USA). Fisher's exact test was applied to determine significantly enriched pathways. The gene sets for analysis of transcription factor (TF) target enrichment, TFT version 3.0, were downloaded from the Broad Institute [
22], whereas the gene ontology (GO) categories used were provided within the software. TFT 3.0 contains sets of genes that share a common predicted TF binding site as defined in TRANSFAC version 7.4. The enrichment of E2F motifs was compared to a random sampling of promoters for all transcription factors in the MSigDb database. The promoter region was defined to be 2kb up- or downstream of the transcription start site.
Quantitative PCR
For time-course experiments, cells were plated in 6-well plates and treated with 5 μM GW3965 for 6 to 72 hours before harvest and RNA isolation. cDNA synthesis was performed using SuperScript II reverse transcriptase (Invitrogen). Quantitative PCR (qPCR) was carried out using Fast SYBR Green Master Mix (Applied Biosystems, Carlsbad, CA, USA) in the 7500 fast real-time PCR system (Applied Biosystems). Primers (Additional file
1) for specific genes were designed using Primer3 software. Relative transcript levels were calculated using the ΔΔct (cycle threshold) method with 36B4 as the reference gene. Statistical significance was determined by Student's
t-test.
RNA interference
MCF-7, T-47D, and MDA-MB-231 cells were seeded in 6-well plates (9.6 cm2, area per well) in appropriate media mentioned above. After 24 hours, cells were washed with PBS and transfected with either 10 nM E2F2 siRNA (Dharmacon, Lafayette, CO, USA) or 10 nM non-targeting control (Dharmacon), using DharmaFECT I transfection reagent (Dharmacon). Cells were counted using the trypan blue staining method and the Countess automated cell counter (Invitrogen). To validate gene knockdown, RNA isolation, cDNA synthesis and qPCR were performed in the same way as described above.
Western blot analysis
After 48 hours of E2F2 siRNA or non-targeting control treatment, cells were lysed with RIPA lysis buffer. Protein concentrations were then determined using Qubit® Protein Assay Kit and fluorometer (Invitrogen, New York, NY, USA). Total protein (100 μg from each sample) was loaded onto a 10% polyacrylamide gel. After electrophoresis, proteins separated by SDS-PAGE were transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore, Billerica, MA, USA). The membrane was blocked with 5% nonfat milk in Tris-buffered saline and Tween 20 (TBST) and then incubated with antibodies against E2F2 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) or β-actin (catalog number: A2228, Sigma, St. Louis, MO, USA) in TBST overnight. The membrane was then reprobed with appropriate secondary antibodies conjugated with horseradish peroxidase for 1 hour. Blots were processed using an ECL kit (Thermo Fisher Scientific, Rockford, IL, USA) and exposed to film.
Chromatin immunoprecipitation
Chromatin immunoprecipitation (ChIP) assays were performed as described previously [
23,
24]. Immunoprecipitations were carried out with E2F2 antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA) or corresponding pre-immune IgG (Santa Cruz Biotechnology). Binding of E2F2 to response elements was measured by qPCR using specific primers (Additional file
1).
Clustering of clinical samples and survival analysis
Clinical and microarray gene expression data from a previously published breast cancer study cohort from Uppsala, Sweden, were used to determine the correlation of the expression profiles of the 83 commonly responsive genes with clinical parameters and disease outcomes [
25]. Dendrograms and heat maps were generated using the Eisen Cluster and Treeview software. Survival data were analyzed using Kaplan-Meier plot functions (log-rank test) of the SigmaPlot software. These analyses utilized previously published and publicly available clinical and experimental data from patient samples and therefore no consent or institutional review was required.
Discussion
In this study, we performed microarray experiments to determine changes in gene expression associated with the anti-proliferative effects of LXR ligands in breast cancer cells. Our goals were to carry out more comprehensive characterization of the functions of LXRs and their ligands in cancer cells, and to identify their mechanisms of action in order to further evaluate their utility as a potential target mechanism and therapeutic agent, respectively, in the treatment of breast cancer. The four cell lines selected for this study are well-characterized and represent molecularly and clinically diverse cellular models of breast cancer, and the proliferation of all four was blocked by LXR ligand treatment [
20]. In addition to their genotypic differences, including both heritable variations and somatic mutations, these cell lines differ in their initial tumor histology, ER status, p53 status, HER2/neu status, responses to targeted therapies and chemotherapeutic compounds, and their tumorigenic and metastatic potential. It is not surprising then that they exhibited vastly different molecular responses to LXR ligand treatment in terms of the number and identity of genes that were responsive to the ligand (Figure
1), as well as the magnitude of change for those genes that were commonly responsive between cell lines (Additional file
3). These differences likely reflect genetic and epigenetic variations between the cell lines, which affect the chromatin structure, the repertoire and functions of co-regulatory proteins and TFs, including LXR subtypes and expression levels, and post-transcriptional regulatory mechanisms, such as those involving microRNA.
In spite of their significant differences in transcriptional responses, however, the four breast cancer cell lines in this study do share a core set of 83 LXR ligand-responsive genes (Figure
1), including those that might be responsible for the anti-proliferative effects of the ligand observed in the cell lines [
20]. These genes can be divided into two groups (Table
1) based on their up- or downregulation by ligand treatment, and each group appears to carry out distinct LXR functions. The list of 23 upregulated genes includes known LXR target genes that function in lipid and cholesterol transport and metabolism, whereas the 60 downregulated genes include many that function in cell cycle regulation and DNA replication. This clear delineation of responses and functions suggests the involvement of discrete mechanisms that can potentially be exploited in evaluating existing ligands and in developing novel ligands specifically for cancer treatment. For example, an ideal LXR ligand for targeting cancer cells would have minimal effects on the 23 upregulated responsive genes and their metabolic functions, thus bypassing the undesired increases in plasma and liver triglyceride levels seen in mouse atherosclerosis models following treatments with some LXR agonists [
34,
35], and elicit robust responses in the 60 downregulated responsive genes associated with cancer cell growth and proliferation.
To further determine the mechanisms of action of LXR ligands on cancer cell proliferation, we analyzed the
cis-regulatory sequences of the 83 universal LXR ligand-responsive genes identified in our microarray study for the presence of TF binding-site motifs, which may provide clues to factors that may be involved in the observed changes in gene expression. Only the E2F binding-site sequence motif is significantly enriched and only in the promoter regions of the set of 60 downregulated responsive genes. Correspondingly, one of the most downregulated responsive genes was E2F2 (approximately 2.5- to 8.0-fold decrease following ligand treatment). Knockdown of E2F2 expression by RNA interference disrupted ER+ breast cancer cell proliferation (Figure
3D), indicating that downregulation of E2F2 expression by ligand activation of LXRs is a potential mechanism of action for LXR ligands and their anti-proliferative effects in ER+ breast cancer cells. E2F2 RNA interference showed significant decrease of some of the predicted E2F target genes in both ER+ (Figure
3E) and ER- cell lines (Figure
3F). Further ChIP anaylsis showed that indeed in ER+ MCF-7 cells, E2F2 binds to the response element of CDC25A (Figure
3G), and this binding is disrupted following treatment with the LXR ligand GW3965. CDC25A is a cell-division cycle gene whose phosphatase action is required for cell progression from the G1 phase of the cell cycle to the S phase. It has previously been shown that decreases in proliferation of both MCF-7 and Vcr-R, another human breast cancer cell line, after treatment of natural tetrasulfides, are due to inhibition of CDC25 [
36]. Furthermore, overexpression of CDC25A in small breast carcinomas is associated with poor survival in patients [
37].
It is not clear, however, whether E2F2 is directly targeted by LXRs or through an indirect mechanism. When we examined the genomic regions proximal to the E2F2 gene for LXR response-element sequence motifs, we did not detect any candidate LXR binding-site. This can be due to the sensitivity of the position weight matrix model we used, or the response element for E2F2 regulation might be located in a distal enhancer region not included in our analysis. Activated LXRs may also indirectly regulate E2F2 expression by tethering to another TF or by regulating the expression of another factor, which in turn affects the expression of E2F2. Alternatively, the expression of E2F2 and the other downregulated responsive genes may merely reflect the disruption of cell cycle progression and the observed accumulation of G1/G
0 cells compared to the cycling vehicle-treated control cells as seen in a previous study. This, however, is unlikely because the changes in gene expression assayed in this study were carried out 48 hours after ligand treatment, whereas the differences in cell cycle progression were observed at 72 hours [
20]. The effects of LXR ligands on cancer cells may also involve non-genomic mechanisms, which can regulate cell proliferation and gene expression via post-translational modifications and signal transduction pathways. These hypotheses regarding the mechanisms of action of LXR ligands in cancer cells and the role of E2F2 and other E2F family members in mediating their anti-proliferative effects await further testing and investigation.
The ultimate goals of this study are to understand and exploit the potential impact of LXR ligands on breast cancer progression and patient survival. This study has defined a set of genes whose expression levels were altered in response to LXR ligand treatment in four cell line models of breast cancer. Expression profiles of these genes in breast tumors from a clinically diverse cohort of 258 breast cancer patients were examined to assess their
in vivo relevance [
25]. Hierarchical clustering of patients with the expression profiles of the 60 downregulated responsive genes separated them into two groups with statistically significant differences in disease outcomes (Figure
5). Patients whose tumors expressed lower levels of these 60 genes experienced longer survival times than patients in the higher expression cluster. In addition to disease outcomes, patients in these groups also differed in important clinical parameters, such as ER status, PR status, lymph node status, and tumor grade (Table
2), consistent with the known association of these parameters with disease outcomes. This strong association of the 60-gene signature with patient survival and clinical parameters indicates that LXR ligand treatment elicited transcriptional responses in breast cancer cells similar to expression profiles observed in tumors from patients with significantly better outcomes. Whether LXR ligand treatment of patients will alter the transcriptional programming in tumors and, ultimately, tumor growth and disease outcome, remain to be determined clinically.
Table 2
Association of patient clustering by downregulated responsive genes with clinical parameters
ER+, n | 88 | 131 | 0.00001128 |
PR+, n | 71 | 121 | 0.00000411 |
LN+, n | 49 | 35 | 0.00183468 |
Tumor grade 3, n | 50 | 5 | 0.0 |
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
TNV performed the microarray data acquisition, data analysis and mining, experimental validation and hypothesis testing, and contributed to the drafting of the manuscript. LLV obtained the samples and performed western analysis. KL designed and carried out functional assays. PJ was responsible for data mining using pathway analysis tools. JZL participated in sample preparation and microarray data acquisition and analysis. NRC performed the ChIP experiments. LPC performed qPCR validation of microarray data. SA contributed to the validation of microarray data. CW participated in the conception and design of the study and drafting of the manuscript. JÅG was involved in the conception of the study and the drafting of the manuscript. KRS conceived the study, coordinated sample acquisition, and helped draft the manuscript. CYL conceived the study, provided overall coordination of the study, and drafted the manuscript. All authors read and approved the final manuscript.