Introduction
Progesterone and estrogen, naturally occurring hormones, are known to modulate the progression and disease outcome of breast cancer [
1‐
3]. Approximately 70% of breast cancer patients—positive for estrogen receptor (ER) and progesterone receptor (PR)—receive hormone therapy, such as blocking ER to inhibit estrogen signaling, as the first-line treatment for patients with luminal breast cancer [
4,
5]. Previous studies have highlighted the beneficial effects of the progesterone-high luteal phase on surgical outcomes in patients with breast cancer [
6‐
8]. However, how progesterone modulates the downstream signaling remains sparsely understood.
The role of ER has been extensively studied in breast cancer due to its prognostic significance [
9,
10], along with its role in increasing the invasion and migration of breast cancer cells [
11]. The PR, on the other hand, is a known ER target. The presence of PR is described as an indication of ER activity [
12]. In vitro studies suggest that progesterone inhibits the invasion and migration of breast cancer cells [
13,
14]. Progesterone also induces cell cycle arrest and mild apoptosis in the cells mediated by PR that can function as a transcription factor to induce gene expression [
15‐
17]. Additionally, PR alters ER binding sites in the genome in response to progesterone, and thus, could modify the expression pattern of ER-responsive genes in breast cancer cells [
18].
Long non-coding RNAs (lncRNAs) and microRNAs (miRNAs), non-coding RNAs (ncRNAs), perform diverse regulation of cellular functions by regulating gene expression at transcriptional and post-transcriptional levels [
19‐
25]. For instance, ER regulates the expression of numerous lncRNAs that control cell invasion, migration, proliferation, and apoptosis in response to estrogen [
26‐
28]. Similarly, progesterone regulates the expression of microRNAs in breast cancer cells [
29]. The lncRNAs function as competitive endogenous RNAs (ceRNA) or miRNA sponges to regulate miRNA functions in cancer cells [
30,
31]. Several studies have identified ceRNA activity of lncRNAs, such as
HULC [
32],
HOTAIR [
33,
34],
TRPM2-AS [
35], and
SNHG7 [
36]. However, whether progesterone modulates the expression of lncRNAs in breast cancer cells remains unknown.
Here, we identify DSCAM-AS1 as a progesterone-responsive lncRNAs in breast cancer using an integrated functional genomics approach. DSCAM-AS1 acts as a sponge for miR-130a to regulate the expression of ESR1 in hormonal receptor-positive breast cancer cells. The study also suggests that targeting these ncRNAs may help improve survival outcomes in patients with breast cancer.
Materials and methods
Transcriptome analysis of breast cancer patient samples
Whole transcriptome sequencing data from 30 breast tumors samples were re-analyzed. Ten tumors were derived from patients who were administered a single dose of 500 mg of hydroxyprogesterone within 15 days prior to surgery, with varying duration for individual patients, while 20 tumors were obtained from patients who were not exposed to hydroxyprogesterone [
37]. Gene expression was quantified using Salmon [
38]. Genes with expression > 5 reads in at least 20% of the cancer samples were retained. Design matrices were created based on progesterone treatment, and differential gene expression analyses were performed with progesterone-treated (
n = 10) and control (
n = 20) tumor samples, using DESeq2 [
39]. Data were assessed in the R environment. ENSEMBL IDs were converted using bioconductor packages (org.Hs.eg.db), and gene names not matching the ENSEMBL IDs were obtained from LNCipedia.
Tissue culture and cancer cell line maintenance
T47-D, BT-474, MCF7 and MDA-MB-231 breast cancer cells were procured, confirmed, cultured, and maintained as explained previously [
13,
40]. The human embryonic kidney 293FT cell line was purchased from Invitrogen (Cat No. R70007), cultured in DMEM with 10% FBS, and maintained at 37℃ with 5% CO
2.
Progesterone and mifepristone treatment, RNA isolation, cDNA synthesis, and qPCR
Cells were serum-starved and treated with 10 nM progesterone (6 h), 100 nM mifepristone (2 h), or an equal volume of ethanol (vehicle control) as explained previously [
13]. RNA isolation, DNaseI treatment, and cDNA synthesis for genes/lncRNAs and microRNAs were performed as explained previously [
29,
40]. Further, the cDNAs were used to study gene/miRNA expression patterns by quantitative real-time PCR (qPCR) method using the KAPA SYBR real-time PCR master mix (Sigma, Cat No. KK4601) and QuantStudio 5 real-time PCR system (Applied Biosystems, Cat no. A34322).
GAPDH or
ACTB and
U6 were used as internal controls to normalize the expression of genes and miRNAs, respectively. Differential gene expression changes were calculated as fold change values using the 2
−∆∆CT method. The sequences for qPCR primers were manually designed using SnapGene sequence viewer. Designed primers were tested and optimized using OligoCalc (Sigma), UCSC In Silico PCR, and NCBI blast. The primer sequences for the genes and miRNAs are listed in Additional file
2: Table S1.
RNA-sequencing of progesterone-treated breast cancer cells
Total RNA was isolated from progesterone treated and untreated T47-D and MDA-MB-231 cells. Good quality RNA samples (RNA integration number > 9) were used to prepare the sequencing library using TruSeq library prep kit v2 (Illumina) with ribosomal RNA depletion. Libraries were sequenced on HiSeq4000 with 100 bp pair-end chemistry. A minimum of 60 million paired-end reads were obtained for each RNA sample with good Phred scores (score > 30). Differential gene expression analysis was performed using the salmon-DeSeq2 pipeline. Briefly, all the raw reads were corrected using the trimmomatic version V0.32 [
41], followed by alignment to the human reference pseudo-genome (GRCh38) using Salmon (version: 0.8.2) [
38] and differential expression analysis using DeSeq2 [
39]. Genes/lncRNAs with fold change > 2 and < 0.5 with
p-value < 0.05 were considered to be significantly deregulated in response to progesterone.
ChIP-sequencing data analysis
ChIP-sequencing data with PR, ER, and p300 pulldown for progesterone-treated T47-D cell line were downloaded from the SRA database [
18]. The raw data for these experiments were analyzed as described earlier [
40]. Briefly, reads were aligned to the gencode (v30) human reference genome (GRCh38) using a BWA aligner (version 0.7.17). Peak calling was performed using the MACS tool (version 2.0) [
42]. Aligned reads were used for differential protein binding in the genome using DiffBind (version 3.0) [
43]. The 5 kb upstream and downstream regions for annotated genes/lncRNAs were analyzed for PR, ER, and p300 binding, and annotation of the peaks was performed using Uropa [
44].
DIANA-LncBase v2 [
45] database was used to predict the binding of miRNAs to
DSCAM-AS1. Predicted miRNAs binding to
DSCAM-AS1 were further determined using the "microRNA–ncRNA targets” module of MirWalk v2.0 [
46], which includes prediction algorithms of miRanda [
47], RNAHybrid [
48], and Targetscan [
49]. Further, miRTarBase [
50] and MirWalk v2.0 were mined to extract miRNAs targeting 3’-UTR of
ESR1.
Sense and antisense DNA oligonucleotides with T7 RNA promoter sequences were designed and synthesized by Sigma-Aldrich to prepare siRNAs targeting
DSCAM-AS1,
PGR, and
ESR1. The complete method for synthesis of small RNA transcripts using T7 RNA polymerase has been described previously [
51]. Briefly, sense and antisense strands of DNA oligonucleotides with T7 RNA promoter complementary sequences were annealed in a Thermocycler for synthesizing dsDNA. The dsRNAs were subjected to in vitro transcription reaction (37 ℃ for 2 h) using T7 RNA polymerase (Promega, Cat no. P2075) in 1 × T7 Transcription Buffer (Promega, Cat no. P118B). The single-stranded sense and antisense siRNAs were further annealed to prepare double-stranded siRNAs. The complete list of DNA oligonucleotides for siRNA synthesis is provided in Additional file
2: Table S1. Synthesized siRNAs were transfected in breast cancer cells using Lipofectamine 3000 kit (Invitrogen, Cat No. L3000015) in serum-free media. Progesterone treatment was given to transfected cells 48 h post-transfection and collected for downstream analysis.
Overexpression of genomic elements
For transient overexpression of miR-130a in breast cancer cells, the precursor miRNA sequence with 200 bp flanking gene sequence was amplified from T47-D genomic DNA and cloned in pJET1.2/blunt vector (Thermo Scientific, Cat no. K1232) followed by sub-cloning in pcDNA3.1(-) mammalian expression vector under CMV promoter. XbaI (NEB, Cat no. R0145) and HindIII (NEB, Cat no. R0104) recognition sequences in multiple cloning sites of pcDNA3.1(-) were used for cloning. For transient overexpression of DSCAM-AS1, the complete cDNA sequence was cloned in the pcDNA3.1( +) expression vector using XbaI and BamHI (NEB, Cat no. R0136). Cloned constructs were confirmed by restriction digestion and Sanger sequencing. Further, the overexpression plasmids were transfected into breast cancer cells using Lipofectamine 3000. Empty pcDNA3.1(-) vector was transfected as vector control. Cells were collected 48 h post-transfection and RNA was extracted. Overexpression was confirmed by qPCR analysis.
For stable overexpression of
DSCAM-AS1 in breast cancer cells, a complete cDNA sequence was cloned in the pBABE-puro expression vector using
BamHI and
SalI (NEB, Cat no. R0138). Cloning was confirmed using restriction digestion and Sanger sequencing. The primer sequences used for cloning and Sanger sequencing are provided in Additional file
2: Table S1. The 293FT cells were used for transfection and retrovirus production. Transductions were performed for 16 h in T47-D cells, followed by a selection of positive clones using 1 µg/mL puromycin (HiMedia, Cat no.TC198-10MG). Puromycin-resistant clones were further confirmed for
DSCAM-AS1 overexpression by qPCR analysis.
Transwell cell invasion and migration assay
Transwell cell migration and invasion assays were performed as described previously [
13]. Briefly, cell invasion assay was performed with Matrigel loaded onto the inserts in Boyden chambers; while, cell migration assay was performed without Matrigel. The number of cells that migrated or invaded through the membrane was counted and the total fraction of cells was plotted as percent cell migration or invasion, respectively.
Luciferase reporter assay
Full-length
DSCAM-AS1 cDNA sequence was amplified from T47-D cells and cloned in pJET1.2/blunt vector, followed by sub-cloning in pGL3-promoter vector (Promega, Luciferase expressing vector) downstream to firefly
luciferase using
XbaI.
DSCAM-AS1 mutant construct was generated with mutated overlapping primers by site-directed mutagenesis using
DpnI (NEB, Cat no. R0176) and PrimeSTAR GXL DNA polymerase (TaKaRa, Cat no. R050B).
DSCAM-AS1 mutant construct contained mutations in
miR-130a MRE in
DSCAM-AS1 cDNA to prevent
miR-130a binding. Wild-type and mutated
DSCAM-AS1 constructs were confirmed by Sanger sequencing. A 620 bp fragment of 3’-UTR of
ESR1 with
miR-130a binding site was amplified from T47-D cDNA and cloned in pJET1.2/blunt vector. The cDNA was sub-cloned in the pGL3-promoter vector using
XbaI. The primer sequences for cloning and generating mutant construct are provided in Additional file
2: Table S1.
For luciferase assay, 293FT cells (50,000 cells/well) were co-transfected with pGL3-DSCAM-AS1 (wild-type/mutant) along with pcDNA3.1(-)-miR-130a or pcDNA3.1 empty vector using Lipofectamine 3000 kit. Additionally, co-transfections were performed with pGL3-3’-UTR-ESR1, pcDNA3.1(-)-miR-130a, or pcDNA3.1 empty vector. pEGFP-N2 was transfected to measure transfection efficiency in all wells. Cells were lysed 48 h post-transfection, and luciferase activity was measured using a luminometer (Berthold Luminometer, Germany). Luminescence and fluorescence units were measured from each transfected well. The luciferase activity was calculated by normalization of luminescence units with fluorescence units from the same well and plotted as luciferase activity. Each experiment was performed in triplicates.
Gene–miRNA correlation analysis
The total RNA and miRNA sequencing data for patients with breast cancer were downloaded from The Cancer Genome Atlas (TCGA). Data from 751 breast cancer samples sequenced for total RNA and miRNAs were considered for further analysis. The samples with normally distributed DSCAM-AS1 or ESR1 expression values were segregated into quartiles. The upper and lower quartile samples were compared. The miRNA levels were compared between patients with ESR1-high and -low expression (the upper and lower quartiles, respectively). A similar analysis was performed for miRNAs in patients with DSCAM-AS1-high and -low expression. The significance of differences between both the groups was calculated using the Wilcoxon–Mann–Whitney test.
Survival analysis
The TCGA breast cancer samples with high and low miRNA expression were compared for survival outcomes. The KM plotter [
52] and GEPIA [
53] were used for Kaplan–Meier survival analysis within specified breast cancer groups. Overall and relapse-free survival of patients was calculated based on the levels of lncRNAs and miRNAs in the samples.
Statistical analysis
GraphPad Prism version 8 (GraphPad Software, La Jolla, CA) was used to calculate statistical significance between different experimental groups in qPCR, cell-based assays, and luciferase reporter assays. The student's unpaired t-test was used to investigate statistical significance. A p-value < 0.05 was considered to be statistically significant.
Discussion
Progesterone confers better survival outcomes in patients with breast cancer, especially in those with lymph node involvement [
58]. These early clinical observations have increased interest in researchers globally to investigate the mechanisms by which progesterone affects breast cancer pathophysiology. We have previously shown that progesterone reduces breast cancer cell invasion and migration [
13] by regulating a tight network of protein-coding genes that reduce the activity of kinases that are known to induce cellular stress [
40]. The present study highlights the multiplicity of genomic mediators, especially ncRNAs, recruited by progesterone and PR in breast cancer to abrogate cell invasion and migration.
To begin with, this is the first study to describe progesterone-responsive lncRNAs in breast tumor samples and cell lines. Interestingly, the analyses identified
DSCAM-AS1 as a novel target of progesterone in breast cancer. Progesterone downregulates the expression of
DSCAM-AS1 specifically in PR-positive breast cancer cells, wherein PR modulates the genomic binding pattern of ER, the classical activator of
DSCAM-AS1 [
27], in response to progesterone. This also highlights the importance of PR in clinical outcome of breast cancer prognosis and confirms the previous findings that PR modulates ER binding in breast cancer cells treated with progesterone [
18,
59]. However, recent report suggests that progesterone treatment may have varied response on tumor growth in patient derived xenograft mouse models [
60]. Consistent with this, we also observed variability in
DSCAM-AS1 expression in response to progesterone.
Second, the findings suggest that
DSCAM-AS1 functions as a miRNA sponge to help maintain the high expression of ER in breast cancer cells.
DSCAM-AS1 has previously been shown to function as a miRNA sponge for
miR-101 [
61] and
miR-186 [
62] in osteosarcoma, and
miR-136 in endometrial cancer [
63]. Interestingly, we show that progesterone opposes the
DSCAM-AS-1‐
ESR1 feedback loop, and thus essentially the ER signaling pathway, by employing two synergistic mechanisms—it decreases the expression of
DSCAM-AS1 and increases the expression of
miR-130a that binds to both
DSCAM-AS1 and 3’UTR of
ESR1 in breast cancer cells. This strengthens the role of progesterone in regulating the expression of non-coding genomic elements in breast cancer [
29,
64], in addition to regulating the expression of protein-coding elements. The results of the present study also emphasize the necessity of PR expression in breast cancer cells for progesterone to alter the expression of
DSCAM-AS1 and
miR-130a, as these effects were not observed in PR-negative MDA-MB-231. Additionally, the expression pattern of
miR-130a was found to be inversely correlated with that of
ESR1 and
DSCAM-AS1 in cell lines and patients with breast cancer.
Third, the cellular experiments indicated that silencing of
DSCAM-AS1 or overexpression of
miR-130a led to a significant reduction in breast cancer cell migration and invasion than that in vehicle control cells, comparable to the effect induced by progesterone-alone. Furthermore, progesterone treatment of cells with high
miR-130a levels led to a greater reduction in cell invasion and migration than progesterone treatment of vehicle-treated control cells; this result demonstrates that variation in expression of these ncRNAs modifies other genomic components that augment the effects of progesterone on breast cancer cells, as described previously [
13,
29,
40]. Further,
miR-130a has been reported to be involved in mitigating progression in breast cancer stem cells [
65], and its expression has been reported to be downregulated in breast cancer [
66,
67]. Finally, using the TCGA datasets, we show that patients with breast cancer with high
miR-130a levels correlate with a tendency toward better overall survival (that could not attain statistical significance). Therefore, the findings may help clinicians to better categorize patients with luminal A/B subtype based on the expression of
DSCAM-AS1 or
miR-130a to receive appropriate care and aid in prolonging their survival outcomes.
In conclusion, this study elucidates an underlying mechanism for a clinical consequence in response to progesterone treatment among patients with breast cancer. Progesterone downregulates the expression of DSCAM-AS1, a known ncRNA member of the ER signaling pathway, and increases the expression of miR-130a that inhibits ESR1, to suppress breast cancer cell invasion and migration. Additionally, high miR-130a levels are associated with improved overall survival outcomes in patients with breast cancer, similar to that observed in the randomized controlled trial with preoperative progesterone. Thus, progesterone treatment under hormonal therapy in the adjuvant and neoadjuvant settings may help in impeding cell migration and invasion of breast cancer cells, and in improving the overall and relapse-free survival outcomes in patients with breast cancer.
Acknowledgements
We acknowledge the overall help from Dr. Rajendra Badwe and Dr. Sudeep Gupta and specifically for designing and generating the whole transcriptome sequencing data of surgically resected breast cancer samples treated with progesterone.We thank all members of the Dutt laboratory for critically reviewing and suggesting corrections in the manuscript. N.Y., S.D., and P.C. are supported by senior research fellowship from ACTREC-TMC. B.D. is supported by senior research fellowship from CSIR. M.G. is supported by emoluments from MIT World Peace University. The funders had no role in study design, data collection, and analysis, decision to publish, or preparation of the manuscript.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.