Loss of REST in breast cancer promotes tumor progression through estrogen sensitization, MMP24 and CEMIP overexpression

TCGA (The Cancer Genome Atlas) data was analyzed for the expression of REST target genes in many different types of cancer, including breast cancer patient samples, using the Broad Institute FireBrowse portal (http://​firebrowse.​org/​). REST downstream gene expression levels in breast cancer sub-types were investigated in the following GEO series (https://​www.​ncbi.​nlm.​nih.​gov/​geo/​), GSE26304 (normal, DCIS and IDC patients), GSE2034 (ER + and ER- patients), and GSE19615 (TNBC and non-TNBC patients) [34‐36]. Sixteen ChIP-seq datasets (Additional file 1) from ENCODE [37] were used to determine potential REST target genes by examining REST binding to RE1 sites located within or near genes of interest. REST DNA binding sites were identified by the presence of both a strong ChIP-peak and REST consensus sequence within the peak region with at least 80% identity.

MCF-7 cells were obtained from ATCC (Manassas, VA) and authenticated by STR DNA Fingerprinting at MD Anderson. Cells were cultured in DMEM (Corning, Tewksbury, MA) containing 10% FBS (R&D Systems, Inc., Minneapolis, MN) and 1% penicillin/streptomycin (Gibco, Waltham, MA) in a humidified 5% CO₂ atmosphere at 37 °C. MDA-MB-231 cells were obtained from ATCC (Manassas, VA), and authenticated by STR DNA Fingerprinting at Arizona Research labs. They were cultured in Leibovitz’s L-15 medium (Corning, Tewksbury, MA) containing 10% FBS (R&D Systems, Inc., Minneapolis, MN) in a humidified incubator at 37 °C. HCC1937, MDA-MB-468, and T-47D cell lines were obtained from ATCC (Manassas, VA) and HCC1937/WT BRCA1 (hereafter referred to as HCC1937 +) were obtained from Dr. Jensen [38]. Cell culture conditions have been previously described [38]. Cell lines were tested for mycoplasma using the ATCC Universal Mycoplasma Detection kit (Manassas, VA).

Cultured MCF-7 cells were plated in 6-well plates at 300,000 cells per well in antibiotic free media. After 24 h at 37 °C, 5%CO₂, cells were transfected with a 100 pmol mixture of Silencer Select siRNAs (S11934, S11932, S119323) to REST from Ambion (Life Technologies, Carlsbad, CA) using Lipofectamine 2000 transfection reagent (Invitrogen, Carlsbad, CA) according to manufacturer’s protocol. Control experiments included ON-TARGETplus nontargeting scrambled siRNA 2 (ThermoFisher Scientific, Waltham, MA). After 24 h, transfected cells were treated with vehicle (PBS), estradiol (10 nM), progesterone (1uM), or estradiol and progesterone in charcoal stripped serum, phenol red free medium (Lonza, Morristown, NJ). After 24 h of treatment, RNA and protein extracts were collected and analyzed.

Total RNA was isolated from MCF-7 cells using TRIzol Reagent (Invitrogen, Carlsbad, CA) according to manufacturer’s protocol. After quality control, RNA samples were sequenced by the Genome Sequencing Facility at the University of Kansas Medical Center using the Illumina NovaSeq 6000 sequencing machine (Illumina, San Diego, CA). Strand specific 100 cycle paired end reads were generated at around 41.5 to 43.5 million reads per sample of which around 99% mapped to the reference genome giving 39 to 41 million uniquely mapped reads per sample. The read quality was assessed using the FastQC software [39]. On average, the per sequence quality score measured in the Phred quality scale was above 30 for all the samples. The reads were mapped to the human genome (GRCh38/hg38) using STAR software [40]. Transcript abundance estimates were calculated using the RSEM software [41]. Expression normalization and differential gene expression calculations were performed in edgeR to identify statistically significant differentially expressed genes [42]. A false discovery rate (FDR) was calculated for the significantly differentially expressed genes using the Benjamini and Hochberg procedure [43]. A FDR cutoff of 0.05 or less with an absolute fold change difference greater than or equal to 1.5 was used as the cutoff criteria for selecting significantly differentially expressed genes. Biological, functional, and pathway analysis were performed using Ingenuity Systems Pathway Analysis software (IPA, QIAGEN, Germantown, MD) on the significantly differentially expressed genes between MCF-7 cells with siREST and MCF-7 cells with scrambled siRNA as control. Data deposited in the NCBI Gene Expression Omnibus (GEO, GSE173857).

Microarray data downloaded from the GEO database were background corrected, normalized and gene-level summarized using the Robust Multichip Average (RMA) procedure. The normalized log expression data were used for downstream differential expression and clustering analysis. Differential expression analysis was performed using the Partek Genomic suite (v 6.5, Partek Inc., St. Louis, MO). P-values were adjusted (FDR) for multiple hypothesis testing by the Benjamini and Hochberg procedure [43]. Genes with multiple transcripts in the microarray were resolved by taking the transcript with maximum variance across samples. Genes with differential expression statistics between any two conditions in the dataset passing the cutoff criteria of absolute fold difference ≥ 1.5 and FDR ≤ 0.05 were considered significant for further analysis. Prior to clustering, gene expression data were standardized to the geometric average of the three housekeeping genes MRPL19, PSMC4, and PUM1 by subtraction. Only REST target genes and breast cancer genes downstream of them were used for clustering. Independent hierarchically clustering was performed on the Euclidian pair wise distance of the row standardized gene expression values linked using the Ward’s method. All clustering computations were performed in Matlab (R2018b, The MathWorks Inc, Natick, MA).

Cells were lysed in 1 × cell lysis buffer (Cell Signaling Technology, Davers, MA) supplemented with protease and phosphatase inhibitor cocktails (Sigma-Aldrich Co., LLC, St Louis, MO). Samples were sonicated and centrifuged at 18,000xg at 4 °C. Protein was quantified using Bio-Rad Protein Assay (Bio-Rad, Hercules, CA). Western blots were performed as described previously [14]. Proteins were detected using SuperSignal West Pico PLUS Chemiluminescent Substrate (ThermoFisher Scientific, Waltham, MA) and blots were imaged using BioRad ChemiDoc MP (BioRad, Hercules, CA). The following antibodies were used: REST (Millipore, #07–579), Beta-actin (Genescript, #A01865), CEMIP (Novus Biologicals, #45,750,002) and anti-rabbit-HRP (Promega). For REST and CEMIP quantification, ImageJ software was used to measure band intensities and normalized to B-actin. Original western blot images can be found in Additional file 11.

Total RNA was isolated from cells using TRIzol Reagent (Invitrogen, Carlsbad, CA) according to manufacturer’s protocol. After quantitation using Nanodrop spectrophotometer, aliquots of RNA were reverse transcribed using the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems, Foster City, CA). Using the Applied Biosystems QuantStudio 7 Flex real-time PCR system and TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA), TaqMan assays (Additional file 2) were used to quantify gene expression using the delta delta C(T) method in comparison with 18 s rRNA [44].

ChIP assays of cultured MCF-7 were performed using the SimpleChIP Plus Enzymatic Chromatin IP Kit (Cell Signaling Technology, Davers, MA). Anti-REST ChIPAb + antibody (Millipore, #17–641), Anti-Histone H3 (Cell Signaling Technology, #4620), RNA Polymerase II (Millipore, #05–623), H3K27Me3 (Millipore, #07–449), and Rabbit IgG (Cell Signaling Technology, negative control) were used according to manufacturer’s protocol. PCR primers spanning the conserved RE1 elements in MMP24 and CEMIP were designed based on the genomic sequence. The primer sequences are found in Additional file 3. For quantification, ImageJ software was used to measure band intensities and the samples were normalized to their respective inputs. Original gel images are included in Additional file 11.

In an effort to identify cancers in which REST targets were dysregulated, we analyzed gene expression profiling data in TCGA using FireBrowse. Analysis of the various cancers revealed that several putative REST target genes, including Polo Like Kinase 1 (PLK1), ADAM Metallopeptidase Domain 12 (ADAM12), Troponin T1 (TNNT1), and cell migration-inducing and hyaluronan-binding protein (CEMIP, KIAA1199) were highly dysregulated (Additional file 4). Since REST plays a crucial role in the pathogenesis of uterine fibroids [14] and its steroid hormone response (unpublished data), we wanted to investigate if similar mechanisms of steroid hormone sensitivity and tumor pathogenesis exist in breast cancer. When broken down into breast cancer subtypes based on invasiveness including normal like, invasive ductal carcinoma (IDC), and ductal carcinoma in situ (DCIS) we see differential expression of a novel set of REST target genes across all the subtypes (Fig. 1). Additionally, we found the more invasive and aggressive subtype, IDC, has higher REST target gene expression. These observations were similar to those seen by Wagoner et al., where loss of REST function is associated with a unique gene signature [5]. Analysis of REST target gene expression in the presence and absence of hormone receptors showed significant differential regulation of a subset of REST target genes (Additional file 5). This distinct regulation was also seen when we compared triple negative breast cancer (TNBC) to non-TNBC (Additional file 5). The subset of REST target genes that are upregulated in ER- and TNBC include VEGFa, CDKN2a, SFRP1, EGFR, MYC, LIF, MMP9, and KIAA1199 (CEMIP). These genes play important roles in breast cancer progression. CEMIP has recently been shown to contribute to metastasis and invasion in a number of different cancers [23, 24]. Unfortunately, the regulation of these genes are not well understood. In ER + and non-TNBC we see ADAM12, COL1A2, IGF1, AR, MMP2, CAV1, SMAD2, and GATA3 are upregulated. Many of these genes contribute to multiple cancer associated pathways [20, 45, 46]. REST’s regulation of these genes is important to consider as RESTless tumors will behave differently in ER + vs ER- subtypes. This difference could affect RESTless tumors response to hormone-based therapies. Using a RESTless gene signature could allow for hormone-based treatment options to be better tailored to RESTless tumors compared to tumors that contain REST. In order to further investigate loss of REST in breast cancer we analyzed REST protein levels in different breast cancer cell lines. MCF-7 cells were used to represent hormone responsive breast cancers while MDA-MB-231 cells represent TNBC. REST expression was found in both MCF-7 and MDA-MB-231 with MCF-7 cells expressing higher levels of endogenous REST. Additionally, REST was effectively knocked down in both cell lines using siRNAs targeting REST (Additional file 6).

We studied the regulation of gene expression in MCF-7 cells to investigate how loss of REST affected tumorigenic cell signaling pathways. First, REST was knocked down in cultured cells using specific siRNAs and 24 h post siRNA transfection, the cells were treated with vehicle (PBS), estradiol (10 nM), progesterone (1uM), or both. Comparing with siControl, the siRNAs targeting REST were efficient at significantly downregulating REST (p < 0.05) at the protein level and mRNA level (Fig. 2A and B). Aliquots of RNA from the samples were subjected to RNA sequencing in order to determine gene expression changes that are influenced by the loss of REST (GSE173857). REST knock down, treatment with estradiol and/or progesterone created distinct changes in the gene expression profile of the effected samples as seen by the principal component projection (PCA) of their expression (Fig. 2C). An agglomerative hierarchical clustering of the significantly differentially expressed genes between different treatment groups illustrates the distinct subsets of genes expressed under these conditions (Additional file 7). Ingenuity Pathway Analysis (Qiagen, IPA) was used to further understand tumorigenic pathways affected by the loss of REST. Table 1 shows the top fifteen genes that were significantly upregulated upon REST knockdown in the absence of hormonal treatment. Additional file 7 shows all these genes lie in Sub cluster 1. Table 2 and Additional file 7 (Sub cluster 4) show a large proportion of these genes are significantly altered by estradiol treatment. IPA predicted that the REST pathway was significantly inhibited (activation z-score -2.415, p-value 5.76E-09) in all of the MCF-7 siRNA treated groups (Fig. 2D). REST downregulation, indicated in green, leads to its target genes being overexpressed (red) (Fig. 2D). Additionally, IPA predicted that estrogen receptor signaling was significantly activated (activation z-score 2.433, p-value 0.0176) after REST knockdown even in the vehicle treated group (Fig. 2E). Investigation of the genes affected showed a large number of them are known to play a role in breast cancer (Additional file 8). Additionally, a highly significant increase in the number of genes upregulated by estrogen was observed in MCF-7 cells that lost REST expression (Additional file 8). These data suggest that knockdown of REST alters the cells’ response to sex steroid hormones, a fundamental aspect of breast cancer pathophysiology and may impact a patient’s response to available treatments.

Analysis of RNAseq data from MCF-7 cells using IPA revealed that many of the aberrantly expressed REST target genes could contribute to disease progression. We prioritized a smaller number of genes that may be playing a role in this process. We selected genes that were known as ‘REST targets’ and novel genes that were potential REST targets based on available ChIP sequencing data sets (Additional file 1). We performed RT-qPCR using the siREST transfected MCF-7 samples. Expression of the steroid hormone receptors, estrogen receptor (ESR1) and progesterone receptor (PGR), remained stable when comparing control and REST knockdown samples. Known REST target genes including, AP3B2, CPLX1, CHGA, and DISP2 all show upregulation in the knockdown samples (Fig. 3 and Additional file 7 Sub cluster 1). This result confirms a sufficient knockdown was achieved in the MCF-7 cells. Genes downstream of ERα signaling that were upregulated included MMP24, GABBR2, SYP, STMN3, SNAP25, FGF12 and BSN (Fig. 3 and Additional file 7 Sub cluster 1). Upregulation of these genes changed depending on the hormone treatment present indicating variable responses to hormones could occur in the absence of REST.

Additionally, IPA predicted genes involved in cell migration, cell–cell adhesion, and extracellular matrix organization including CEMIP and MMP24 were significantly upregulated upon REST knockdown. CEMIP and MMP24 have been previously shown to be upregulated in breast cancer patient samples and cell lines [20, 47, 48]. However, transcriptional regulation of these genes in breast cancer is poorly understood. We further investigated how REST regulates the expression of these two genes with reported roles in breast cancer metastasis. First, we looked at the relationship of REST and CEMIP expression in different breast cancer cell lines, including MDA-MB-468, T-47D, HCC1937 + , HCC1937-, HCC1937, MCF-7, and MDA-MB-231. All cell lines showed high REST expression and no expression of CEMIP except for MDA-MB-231 (Fig. 4A). Here low REST expression was observed with higher levels of CEMIP. Upon REST knockdown, using siRNA, CEMIP expression was increased in MCF-7 and in MDA-MB-231 (Fig. 4A). We picked high REST expressing MCF-7 cells and low REST expressing MDA-MB-231 cells to further characterize REST’s regulation of CEMIP and MMP24. Using siRNAs specific to REST, we confirmed CEMIP was significantly upregulated upon REST knockdown with a 2.46 fold change (p = 0.011). Additionally, MMP24 was significantly upregulated (p = 0.0001) with a fold change of 44.6 upon REST knockdown (Fig. 4B). These results were also observed in RNA-sequencing data (CEMIP, FC = 1.58, FDR = 0.06; MMP24, FC = 20.33, FDR = 1.04E-82). REST knockdown was repeated in MDA-MB-231 cells with a 40% decrease in REST expression (p = 0.01). CEMIP showed a fold change of 7.35 (p = 0.0053) and MMP24 showed a fold change of 55.2 (p = 0.0027). Moreover, MDA-MB-231 had higher expression levels and CEMIP expression increased when the knockdown was extended to 48 h (Fig. 4C and D). MDA-MB-231 cells already express lower levels of REST protein compared to MCF-7 cells and this may contribute to the higher target gene expression levels (Fig. 4A and Additional file 6).

Based on our findings that loss of REST and high CEMIP levels are positively correlated, we next wanted to see if REST binds within the CEMIP gene located on chromosome 15. REST binding sites, known as RE1 sites, were found in several locations of the CEMIP gene (Additional file 9). Site 1 (-1501) is located upstream of the transcription start site (+ 1) in the promoter region. Site 2 is located in intron 1 (+ 1022) and site 3 is located at + 1442 within the first intron (Fig. 5A). Interestingly, site 2 falls within a CpG island in the first intron of CEMIP. Kuscu et al. showed this CpG island (+ 525 to + 1059) is methylated in nonaggressive or minimally expressing CEMIP cell lines, including MCF-7. However, in aggressive breast cancer cell lines, MDA-MB-231, this CpG island is shown to be demethylated [49]. Given the role of REST in the epigenetic regulation of a multitude of its target genes, we sought to confirm the association of REST with its binding consensus sequence using semi-quantitative ChIP-PCR analysis in MCF-7 cells transfected with siControl or siREST. Cross-linked chromatin from the cells was immunoprecipitated using anti-Histone H3 (positive control), anti-REST, anti- RNA polymerase II, anti-H3K27Me3, and anti-IgG (control) antibodies. Three primer pairs for CEMIP were designed for semi-quantitative PCR as seen in Additional file 3. ChIP- PCR results indicated that REST was not associated with the RE1 sequence site 1 (-1501) in the control or in siREST knockdown MCF-7 cells (Fig. 5B). Site 2 (+ 1022) and site 3 (+ 1442) showed REST binding to the RE1 sequences in the control non-targeting siRNA treated samples (Fig. 5B). Additionally, the enrichment of these two RE1 sites decreased in the chromatin immunoprecipitants after REST knockdown (Fig. 5B), indicating that the interaction was REST specific. Additionally, REST knockdown appears to change binding of RNA polymerase II and H3K27Me3 in the siREST samples (Fig. 5B). However, future experiments using ChIP qPCR are required to confirm these epigenetic/ transcriptional changes upon REST knockdown as well as in patient samples with loss of REST expression. Taken together, these data confirm REST binding within the CEMIP gene and loss of REST binding upon siREST knockdown. Our finding that REST associated sequence elements overlap with the CpG island within the CEMIP locus will have important implications on the epigenetic regulation of this gene by REST and in tumor metastasis.

While the loss of REST has previously been shown to increase MMP24 expression in breast cancer cells, transcriptional regulation of MMP24 by REST has not been studied so far. We wanted to confirm that REST directly regulates MMP24 in breast cancer cells. Sequence analysis showed several putative RE1 sites in the MMP24 gene locus on chromosome 20 (Additional file 10). Site 1 (-329) is located in the promoter region of MMP24 while site 2 (+ 804) and site 3 (+ 5653) are located within the first intron (Fig. 6A). To confirm the association of REST with its binding consensus sequence, we performed semi-quantitative ChIP-PCR analysis in MCF-7 cells transfected with siControl or siREST. Cross-linked chromatin from the cells was immunoprecipitated using anti-histone H3 (positive control), anti-REST, anti-RNA polymerase II, anti-H3K27Me3, and anti-IgG (control) antibodies. Three primer pairs for MMP24 were designed for semi-quantitative PCR as seen in Additional file 3. RE1 site 1 (-329) and site 3 (+ 5653) were not bound by REST. Site 2 showed REST was bound and this interaction was decreased by the REST knockdown (Fig. 6B). These results suggest that the loss of REST may alter the epigenetic regulation of MMP24, a matrix metalloproteinase with putative roles in breast cancer progression and metastasis.