Interaction Proteomics Identifies ERbeta Association with Chromatin Repressive Complexes to Inhibit Cholesterol Biosynthesis and Exert An Oncosuppressive Role in Triple-negative Breast Cancer*

Triple-negative breast cancers (TNBCs) are characterized by low overall survival and poor response to therapy because of their aggressiveness and limited available treatments. We found a subset of TNBC expressing the master regulator estrogen receptor beta (ERβ1) and characterized the effect of this nuclear receptor in human TNBC cells by a multiomics approach. Results highlight transcriptome deregulation by ERβ involving association with chromatin remodeling complexes, including PRC1/2, and provide an explanation for its oncosuppressive effects on TNBC cells. Graphical Abstract Highlights ERβ inhibits cell growth, migration and clonogenicity in TNBC cells. In TNBC ERβ deregulates the transcriptome and cholesterol biosynthesis pathway. ERβ interacts with multiple chromatin remodeling complexes including PRC1/2. Triple-negative breast cancer (TNBC) is characterized by poor response to therapy and low overall patient survival. Recently, Estrogen Receptor beta (ERβ) has been found to be expressed in a fraction of TNBCs where, because of its oncosuppressive actions on the genome, it represents a potential therapeutic target, provided a better understanding of its actions in these tumors becomes available. To this end, the cell lines Hs 578T, MDA-MB-468 and HCC1806, representing the claudin-low, basal-like 1 and 2 TNBC molecular subtypes respectively, were engineered to express ERβ under the control of a Tetracycline-inducible promoter and used to investigate the effects of this transcription factor on gene activity. The antiproliferative effects of ERβ in these cells were confirmed by multiple functional approaches, including transcriptome profiling and global mapping of receptor binding sites in the genome, that revealed direct negative regulation by ERβ of genes, encoding for key components of cellular pathways associated to TNBC aggressiveness representing novel therapeutic targets such as angiogenesis, invasion, metastasis and cholesterol biosynthesis. Supporting these results, interaction proteomics by immunoprecipitation coupled to nano LC-MS/MS mass spectrometry revealed ERβ association with several potential nuclear protein partners, including key components of regulatory complexes known to control chromatin remodeling, transcriptional and post-transcriptional gene regulation and RNA splicing. Among these, ERβ association with the Polycomb Repressor Complexes 1 and 2 (PRC1/2), known for their central role in gene regulation in cancer cells, was confirmed in all three TNBC subtypes investigated, suggesting its occurrence independently from the cellular context. These results demonstrate a significant impact of ERβ in TNBC genome activity mediated by its cooperation with regulatory multiprotein chromatin remodeling complexes, providing novel ground to devise new strategies for the treatment of these diseases based on ligands affecting the activity of this nuclear receptor or some of its protein partners.


In Brief
Triple-negative breast cancers (TNBCs) are characterized by low overall survival and poor response to therapy because of their aggressiveness and limited available treatments. We found a subset of TNBC expressing the master regulator estrogen receptor beta (ER␤1) and characterized the effect of this nuclear receptor in human TNBC cells by a multiomics approach. Results highlight transcriptome deregulation by ER␤ involving association with chromatin remodeling complexes, including PRC1/2, and provide an explanation for its oncosuppressive effects on TNBC cells.

Graphical Abstract
onset, aggressive clinical phenotype, and poor prognosis compared with other BCs (1). Moreover, because of the limited number of therapies applicable for its treatment, the need for novel molecular targets identification becomes critical.
Estrogen receptors belong to the nuclear receptor superfamily and comprise two main members: Estrogen Receptor ␣ (ER␣) and Estrogen Receptor ␤ (ER␤). Although much is known about ER␣ and its proliferative effect on breast epithelium, the role of ER␤ is still debated and not fully understood (2,3). Nevertheless, according to different studies, ER␤ appears to be a tumor suppressor gene with higher expression in normal breast tissue when compared with the cancerous one, thus representing an appealing clinical target for anticancer treatment because of the possibility of its selective activation with ER␤-specific agonists (4,5). In the TNBC context, several studies, including a meta-analysis, reported receptor expression in a small fraction (15-20%) of patients and its presence was correlated to improved patient outcomes (6,7). It has been demonstrated that ER␤ exerts its action in TNBC by targeting genes involved in cell cycle, proliferation, death and development (8). In basal-like BC it regulates epithelial to mesenchymal transition (EMT) either by up-regulation of miR-200a/b/429 (9) or by EGFR downregulation (9). Moreover, the presence of liganded ER␤ in TNBC cells causes a reduction of proinflammatory cytokines production and an increased synthesis of cystatin superfamily members, whose expression generally correlates with better relapse-free survival in TNBC patients and is linked to a concomitant decrease of migration and invasion of TNBC cells (10). In any case, the molecular mechanisms underlying ER␤ tumor-suppressive effects in TNBC are still not fully characterized, thus being a crucial point in therapy response evaluation and further classification of patients who might benefit from therapies targeting this estrogen receptor subtype.
Here, we show that ER␤ is detectable in a fraction of primary TNBCs and that its expression in TNBC cells in vitro leads to reduced cell proliferation by the increase of G1 cell cycle phase. Transcriptome analysis combined with genomewide ER␤ binding sites mapping revealed the involvement of the receptor in cholesterol biosynthesis downregulation through its recruitment to regulatory elements of the gene encoding for sterol regulatory element-binding transcription factor 1 (SREBF1), an upstream regulator of cholesterol biosynthesis pathway. Interactional proteomics, performed to unveil the molecular bases of ER␤ action, revealed its nuclear association with protein complexes involved in several key biological events, such as DNA replication, transcription regulation, post-transcriptional mRNA expression, and small molecule biochemistry control. Multiple complexes, such as polycomb repressor complexes 1 and 2, known to be involved in negative epigenetic regulation of transcription by chromatin remodeling, were found to be a part of ER␤ interactome.
These data allow us to suggest an immediate contribution of ER␤ and its molecular partners in the downregulation of key pathways in TNBC, including those involved in cholesterol metabolism.

EXPERIMENTAL PROCEDURES
Tissue Microarray (TMA) Construction-A breast Tissue MicroArray (TMA) was constructed using 217 samples of triple-negative breast cancer collected from 2003 to 2013 and 5 normal breast tissues from the Pathology Unit of the National Cancer Institute Fondazione G. Pascale of Naples. Informed consent was obtained from all patients. All tumors and controls were reviewed according to WHO classification criteria, using standard tissue sections and appropriate immunohistochemical slides. TMA was built using the most representative neoplastic areas of each sample by semi-automated tissue arrayer (Galileo TMA) as described previously (11).
Immunohistochemistry (IHC) and TUNEL Assay-Formalin-Fixed Paraffin-Embedded (FFPE) sections were deparaffinized in an organic solvent (Bio-Clear, Clodia Laboratori, Chioggia, Italy), in order to remove the including agent and rehydrated following a normal descending alcohol scale. Then, the endogenous peroxidase was blocked with 3% hydrogen peroxide for 10 min. Antigenic unbleaching was conducted using 10x citrate buffer (0,01M) in a decloaking chamber at 110°C for 20 min. After that, the slides were cooled, washed in TBS buffer solution (Tris buffer saline)/Tween and protein blockade was performed (5% BSA in 1ϫ PBS). The slides of TMA were incubated with two different primary antibodies that recognize ER␤: PPG5/10 (1:15; GeneTex, Irvine, CA) and PPZ0506 (1:60; ThermoFisher Scientific, Waltham, MA) overnight at 4°C and were washed in TBS/Tween buffer. The binding of the primary antibody to the antigen was visualized by incubation with a secondary antibody (anti-mouse) associated with horseradish peroxidase molecules (HRP) by a dextran polymer for 30 min at 4°C and followed by washing in TBS/Tween buffer (2 steps of 5 min each). The peroxidase activity was visualized by the addition of a chromogenic substrate (DAB, 3,3Ј-Diaminobenzidine and 2,5-3% hydrogen peroxide). The reaction with peroxidase produces a visible brown precipitate at the antigenic site. The tissue sections were immersed in 0.02% hematoxylin for about 30 s, to contrast the cores and dehydrated following an ascending scale of alcohol clarified by a passage in Bio-Clear and mounted using a non-aqueous permanent medium. Finally, the prepared slides were interpreted using a standard light field optical microscope by two expert pathologists. For each core sample, at least five fields and more than 500 cells were analyzed. Using a semi-quantitative scoring system, under the microscope, the observer evaluated the intensity, extent and subcellular distribution of the marker, for which there are no standardized criteria for assessing the intensity of the reaction. For the definition and evaluation of the score both qualitative and quantitative parameters were considered. For the qualitative criteria, we considered the intensity of the reaction subdividing it into "mild," "moderate," and "intense." For the quantitative criteria, the percentage of positive tumor cells was considered. The following antibodies were used for immunohistochemistry assay: rabbit polyclonal C-terminal anti-ER␤ (PPG5/10, Thermo Fisher Sci- Cell Culture and Clone Generation-HCC1806 (CRL-2335), MDA-MB-468 (HTB-132) and Hs 578T (HTB-126) cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, VA). All experiments relative to cell lines were performed under conditions of exponential growth and each cell line was grown in the appropriate cell culture medium, according to manufacturer protocol, and kept in an incubator at 37°C in the presence of 5% CO 2 . HCC1806 cells (CRL-2335) were grown in phenol red-containing Roswell Park Memorial Institute (RPMI) 1640 Medium (Euroclone, Milano, Italy) complemented with HEPES (pH 7.3; Euroclone) to a final concentration of 2.383 g/L, D-Glucose (Lonza, Basel, Switzerland) to a final concentration of 4.5 g/L and Sodium Pyruvate (Euroclone) to a final concentration of 0.11 g/L. MDA-MB-468 (HTB-132) and Hs 578T (HTB-126) were maintained in phenol red-containing Dulbecco's modified Eagle Medium (DMEM) (Sigma-Aldrich, St Louis, MO). For all cell lines, medium was complemented with 10% FBS (GE Healthcare, Chicago, IL), 1ϫ Pen-Strep (Lonza), 2 mM L-Glutamine (Lonza) and 0.25 g/ml of Amphotericin B (Sigma-Aldrich). Medium for Hs 578T cell line was complemented with 0.01 mg/ml bovine insulin (Sigma-Aldrich) according to vendor's instructions. ER␤ expression was induced using doxycycline (Sigma-Aldrich). Cells were STR authenticated and periodically tested for the presence of mycoplasma contamination using ABM mycoplasma PCR detection kit. All experiments were performed in media complemented with non-steroid depleted FBS. The indicated cell lines were used to generate clones stably expressing full-length-3xFlag-ER␤ using the Clontech Lenti-X Tet-On Advanced Inducible Expression System (Takara-Clontech Europe, Göteborg, Sweden). First, the 3xFlag-ER␤ DNA fragment, amplified from CSII-ESR2 vector, kindly provided by Dr. P. Dotto (12), was cloned into a pLVX-Tight-Puro vector downstream to a tetracycline-inducible promoter. The cloning procedure was performed using Clontech In-Fusion HD Cloning Kit. Subsequently, two lentiviral particle types, one containing Tet-On Advanced and another ER␤-encoding RNAs, were produced using the Clontech Lenti-X Packaging Single Shots (VSV-G) according to manufacturer protocol. Selected TNBC cell lines were infected first with Tet-On Advanced lentiviral particles, encoding for a transcription factor that in the presence of doxycycline binds to a tightly regulated inducible promoter, and separate clones were produced. One Tet-On Advanced clone for each cell line was then chosen and transduced with 3xFlag-ER␤ particles. Single ER␤ clones were grown and tested for receptor protein expression on doxycycline (Sigma-Aldrich) induction by Western blotting. For each cell line, one ER␤ clone was chosen for further experiments.
Total Protein Extraction-Cells were seeded in 60-mm plates and grown in the presence of doxycycline or vehicle for 9 days. The following doxycycline concentrations were used to induce ER␤ expression: 0.035 g/ml, 1.0 g/ml and 2.0 g/ml for HCC1806, Hs 578T and MDA-MB-468-derived cell lines respectively. For total protein extraction, cells were harvested, washed twice with ice-cold PBS-EDTA (0.5 mM EDTA), lysed using RIPA buffer (50 mM Tris-HCl pH 7.6, 150 mM NaCl, 0.1% SDS, 0.5% C 24 H 39 NaO 4 , 1% NP-40, 2 mM EDTA, 50 mM NaF) for 15 min on ice and centrifuged at 13,000 rpm for 30 min at ϩ4°C. Resulting protein extracts were quantified using Bradford protein assay followed by analysis by SDS-PAGE and Western blotting.
Cell Proliferation and Cell Cycle Analysis-Cell proliferation of ER␤expressing clones was evaluated using the MTT (3-(4,5-Dimethylthiazol-2-yr)-2,5-Diphenyltetrazolium Bromide)-based colorimetric assay (M6494, Invitrogen, Carlsbad, CA), according to the manufacturer instructions. Cells were seeded in 96-well plates and exposed to doxycycline or vehicle for 12 days. The same experiments were simultaneously performed also on the parental Tet-On Advanced cells (see Supporting Information). Absorbance was measured by a microplate reader at 570 nm, and corresponding background values read at 620 nm were subtracted for each sample. Cell cycle assay was performed in both ER␤ and Tet-On Advanced TNBC cell lines. To this aim, cells, grown in the presence or absence of the aforementioned doxycycline concentrations for 9 days, were detached using 1ϫ trypsin-EDTA in PBS (Aurogene), washed twice with ice-cold PBS, fixed with 70% Ethanol (300 l of PBSϩEDTA 20 mM and 700 l of Ethanol 100%), and stored at Ϫ80°C for 1 h. Once fixed, cells were washed with ice-cold PBS, centrifuged, resuspended in RNaseAcontaining PBS (0,05 g/l) and incubated for 10 min at room temperature. Finally, cells were stained with Propidium Iodide (0,05 g/l; Sigma Aldrich) and analyzed using BD FACSVerse (Becton Dickinson, Franklin Lakes, NJ). For each sample, the scatterplots were analyzed using ModFit LT 5.0 software (Verity Software House). Functional validation of ER␤ effects and the RNA profiling experiments described below were performed using the above-indicated doxycycline concentrations.
Total RNA Extraction, Profiling, and Data Analysis-Total RNA was extracted from HCC1806, MDA-MB-468 and Hs 578T ER␤ clones grown for 9 days in the presence and absence of doxycycline, by using TRIzol reagent (Life Technologies, Carlsbad, CA) according to the manufacturer instructions. As a control for doxycycline effect on RNA expression, total RNA was extracted also from MDA-MB-468 Tet-On Advanced cell line, cultured at the same conditions as the corresponding ER␤ clone. Once extracted, RNA concentration and purity were assayed by NanoDrop™ 2000/2000c spectrophotometer (Thermo Fisher Scientific) whereas its integrity was evaluated using TapeStation 2200 (Agilent, Santa Clara, CA) instrument through RNA ScreenTape Assay. Then, RNA was treated with DNase, using TURBO™ DNase kit (Thermo Fisher Scientific), and 0. ). Bioinformatics analysis was performed as described by Tarallo et al. (13). In brief, quality check of the reads was done using FASTQC [https://www.bioinformatics.babraham.ac.uk/ projects/fastqc] and STAR (14) software was used for reads alignment on human genome (assembly version hg38, GeneCode version 29). The list of ER␤-modulated transcripts in MDA-MB-468 cells was compared with that of deregulated RNAs in response to doxycycline in Tet-On Advanced MDA-MB-468 cell line. Transcripts showing a fold-change Յ2 between the two conditions were filtered out from the list of ER␤-regulated RNAs. The final data set of receptor-influenced transcripts in MDA-MB-468 cells was composed of RNAs specifically modulated by ER␤, those showing an opposite behavior in Tet-On Advanced and ER␤ cells in the presence of doxycycline and RNAs with the same behavior and a fold-change Ͼ2 when comparing the two cell lines. Identification of alternative splicing events was performed as previously described (15).
Co-Immunoprecipitation-For immunoprecipitation, nuclear protein extracts, obtained from TNBC cell clones treated (ER␤ϩ) or not (ER␤-) with doxycycline (0.5, 2.0 or 1.0 g/ml for HCC1806, MDA-MB-468, and Hs 578T respectively) for 24 h, were incubated with 50 l of anti-rabbit Magnetic Beads (Invitrogen) conjugated with the appropriate antibody as described earlier (16). 4 g of rabbit polyclonal anti-ER␤ antibody (PA1-313, Thermo Fisher Scientific) were used for ER␤ immunoprecipitation from 4 mg of HCC1806 nuclear extract or 1 mg of MDA-MB-468 or Hs 578T extract. Nuclear protein extraction was performed as it was previously described (13,17).
Mass Spectrometry and Data Analysis-MDA-MB-468 and HCC1806 nuclear ER␤ϩ and ER␤-extracts were used for interaction proteomics experiments. In brief, ER␤ was immunoprecipitated in both ER␤ϩ and ER␤-cells (cultured with and without doxycycline induction respectively) as it is described in Co-Immunoprecipitation section using same amounts of nuclear protein extracts, with the only difference that after the last wash, beads were resuspended in 100 mM ammonium bicarbonate buffer. On beads trypsin digestion was carried out by 0.5 g trypsin (Promega, Madison, WI) addition to each replicate followed by samples incubation at 37°C overnight. Additionally, 0.2 g of trypsin was added to the samples the next day followed by incubation at 37 C°for another 2 h. Peptides were acidified with 1% trifluoroacetic acid, desalted, concentrated with C18 zip tips (Thermo Fisher Scientific) and eluted twice with 0.1% trifluoroacetic acid/50% acetonitrile before drying and solubilization in 7 l 0.1% formic acid for the following mass spectrometry analysis. Each peptide mixture was analyzed on an Easy nLC1000 nano-LC system connected to a quadrupole-Orbitrap mass spectrometer (QExactive Plus, ThermoElectron, Langenselbold, Germany) equipped with a nanoelectrospray ion source (EasySpray; Thermo Fisher Scientific) as described previously (18). The resulting MS raw files were submitted to the MaxQuant software (version 1.6.1.10) for protein identification and quantitation using the Andromeda search engine. MaxQuant search was done against the UniProt Human database (October 2017, containing 20,239 entries). Carbamidomethyl was set as a fixed modification and protein N-acetylation and methionine oxidation were set as variable modifications. First search peptide tolerance of 20 ppm and main search error 4.5 ppm were used. Trypsin without proline restriction enzyme option was used, with two allowed miscleavages.
Functional Analyses-The lists of differentially expressed transcripts were submitted to Ingenuity Pathway Software (IPA, Ingenuity System, www.ingenuity.com; QIAGEN, Hilden, Germany) and investigation of ER␤-modulated disease and disorder gene networks, cellular processes and canonical pathways was carried out. Circos plot was generated using GOPlot (19). Molecular type and function enrichment analysis of the ER␤-interacting protein sets was performed by IPA. Receptor partners were further associated with known protein complexes annotated in CORUM database (20) if at least two complex members were present in the protein data set. For functional interaction networks visualization, FunRich v3.0 was used (http://www.funrich.org).
Experimental Design and Statistical Rationale-Proliferation was evaluated at different time points: 0 days (start point), 3, 6, 9 and 12 days (end point) using six biological replicates for each condition tested. Cell cycle assay was performed in triplicate for each condition, 15,000 events were acquired. Colony formation and migration assays were performed in triplicate for each of the conditions tested. Onetailed Student's t test was applied in all cases to assess statistical significance (p value).
For total RNA profiling experiment, three independent biological replicates were prepared for each treatment. Differentially expressed RNAs (fold change cutoff 1.5 and adjusted p value (p-adjՅ0.05) were identified using DESeq2 (21). For alternative splicing events, an Inclusion Level cut-off Ն 0.1 and FDRՅ0.05 have been considered.
For interaction proteomics experiment, three biological replicates of each ER␤ϩ and ER␤-control samples were analyzed. After Max-Quant search against the UniProt Human database, the minimal uniqueϩrazor peptides number was set to 1, and the allowed FDR was 0.01 (1%) for peptide and protein identification. Statistical analysis was performed using a permutation t test that was applied to MaxQuant protein "Intensities" values to obtain statistically significant difference (FDRՅ0.05) between ER␤ϩ and ER␤-samples. ER␤-cells (cultured in the absence of doxycycline) were used as a control for ER␤ϩ cells (where ER␤ expression was induced by doxycycline addition to the culture medium), in order to avoid potential antibody cross-reactivity and strengthen specificity of the interactions. Proteins identified and supported by statistical analysis and showing a fold-change ratio greater than 1.5 when compared with the controls were considered as potential ER␤ interactors. The indicated cutoff was chosen based on the average value of the median distribution of the fold-changes, computed in two generated data sets.
In the case of IPA cellular processes and canonical pathways analysis, only networks, processes and pathways showing a p value Յ0.05 were considered statistically significant.

ER␤ Detection in Primary TNBC
Tissues-TNBCs are known to lack ER␣ expression, but several studies reported that in a fraction of patients the tumor expresses its counterpart ER␤. Among the five known ER␤ isoforms, only fulllength isoform ER␤1 has an intact C-terminal ligand-binding domain and exerts antiproliferative effects, whereas other variants are instead associated with early disease recurrence (22,23). Recently, two studies focusing on validation of anti-ER␤ antibody specificity demonstrated that some commercially available antibody might not be specific for ER␤ in different assays (2,3), in particular by immunohistochemistry (IHC), raising concerns about the conclusions of previous studies focusing on the role of this nuclear receptor in TNBCs. The monoclonal antibody PPZ0506, recognizing the N terminus of the protein, was demonstrated instead to specifically detect ER␤ by IHC ER␤ by IHC (2). Therefore, PPZ0506 was selected to evaluate ER␤ expression in primary TNBC biopsies. As shown in the left panel of Fig. 1A, this antibody indeed recognizes ER␤1 by immunocytochemistry in fixed HCC1806, showing a positive signal only when expression of this protein is induced in the cells by doxycycline (Fig. 1B). IHC analysis with PPZ0506 of tissue microarrays containing 217 TNBC tissue samples revealed that 27% of these tumors are positive for ER␤ that, as shown in the micrographs displayed in the right panels of Fig. 1A, is detectable mainly in the cell nucleus. Interestingly, nearly half of the PPZ0506ϩ tumors (15% of total; light blue bar) scored positive only with this antibody, whereas the remaining PPZ0506ϩ (12% of total; dark blue bar) were stained also by PPG5/10, an anti-ER␤ antibody directed against the C terminus of the protein, whereas 31% of tumors were positive only for PPG5/10 and the remaining 42% were negative for both antibodies ( Fig. 1A and data not shown). Immunohistochemistry staining was performed on sequential slices from the same fixed tissue sample, using each antibody separately.
ER␤ Expression in TNBC Reduces Cell Proliferation and Inhibits Cell Cycle Progression-To our knowledge, there are no available TNBC cell lines that express detectable fulllength ER␤1 levels. For this reason, its activity and functions are generally investigated in stably ER␤ expressing cell models, where receptor sequence is introduced by exogenous transfer of its cDNA. In order to study the role of ER␤1 (from now on ER␤) in TNBC, three inducible ER␤-expressing cell models, belonging to different breast cancer subtypes, were generated. In particular, the following cell lines were used: MDA-MB-468, HCC1806 and Hs 578T that belong to basallike 1, basal-like 2 and claudin-low subtypes, respectively. Clones were produced as described in the experimental procedure section, to obtain ER␤ expression level after 9 days of doxycycline treatment similar for all three cell lines (Fig. 1B).
In order to characterize the generated cell lines and to determine ER␤ effect on cell proliferation, MTT assay was applied for cells grown in the presence (ER␤ϩ) and absence (ER␤-) of selected doxycycline concentrations. We demonstrated that ER␤ϩ cells grow considerably slower than ER␤ones (Fig. 1C), whereas the parental Tet-On Advanced-expressing cell lines, used for generation of ER␤-expressing clones, did not show reduced proliferation on doxycycline exposure (supplemental Fig. S1A). Interestingly, the basal-like 1 cell line MDA-MB-468 demonstrated the highest degree of growth inhibition (about 70% at 12 th day) on ER␤ induction (Fig. 1C). As shown in Fig. 1C, growth inhibition by ER␤ shows slow kinetics, reaching significant values between days 6 -9 of Doxy treatment for Hs578T and days 9 -12 for MDA-MB-468 and HCC1806 cells.
To check whether this growth inhibition may be caused by a reduction of cell cycle kinetics because of ER␤ activity, cell cycle phase distribution analysis was performed before and after induction of ER␤ expression by doxycycline and results confirmed this hypothesis. We observed an accumulation of ER␤ϩ cells in G 1 phase, accompanied by a parallel reduction of S and G 2 /M cells in the cultures (Fig. 1D). It is worthwhile to note that MDA-MB-468 and Hs 578T cell lines exhibited a slight increase (p-valϽ0.05) of cell number in sub-G 0 cell cycle phase, indirectly indicating ER␤-mediated increase of programmed cell death. Again, cell cycle profiles of Tet-On Advanced-expressing cell lines were not influenced by doxycycline exposure (supplemental Fig. S1B). The results recapitulate the ability of Estrogen Receptor beta to inhibit cell proliferation by cell cycle arrest, indicating that these cellular models are suitable for the investigation of ER␤ role in TNBC cells.
ER␤  Table S1). Signaling pathway analysis revealed statistically significant deregulation of several pathways, among which estrogen and cholesterol biosynthesis pathways demonstrated the highest level of deregulation (z-score Ͻ-2). Another two pathways (Wnt/Caϩ and Sirtuin), instead, demonstrated a tendency to be up-regulated (z-score Ͼ1) (supplemental Fig. S2A). To evaluate our results, we compared our list of differentially expressed genes with the one published by Shanle et al., where the same TNBC cell line was used for inducible ER␤ clone generation (8). Despite different experimental settings (the use of estradiol-free medium for cell cultivation and shorter time of ER␤ induction), the two data sets revealed 942 commonly deregulated genes, among which 735 (78%) showed the same behavior (up or downregulation), indicating the presence of a high degree of correlation between the two studies. To further characterize ER␤ impact on TNBC transcriptome, we evaluated if the receptor expression induces alternative splicing events in this breast cancer subtype, like it does in case of hormone-responsive breast cancer (15). To this aim, we compared RNA profiles of MDA-MB-468 ER␤ϩ and ER␤-cells and identified a total of 440 alternative splicing events, among which the following ones were present: exon skipping, intron retention, the use of alternative 5Ј donor and 3Ј acceptor sites and mutually exclusive exons (supplemental Fig. S2B). Splicing pattern was like the one observed in MCF-7 cell line, with exon skipping being the most frequent event (supplemental Fig. S2B).
Comparison of differentially expressed genes in MDA-MB-468 (fold-change cut-off Ն1.3, FDRϽ0.05) with ER␤-regulated transcripts (FDRϽ0.05) in the other cell lines revealed that HCC1806 and Hs 578T are characterized by less marked changes of mRNA profiles on ER␤ expression ( Fig. 2A, supplemental Tables S2 and S3 for HCC1806 and Hs 578T respectively). A total number of 1820 (857 up and 963 downregulated) transcripts showed a common response to ER␤ in at least two cell lines with several of them, shown in Fig. 2B, displaying the same behavior (fold-change cut-off 1.2 , FDRϽ 0.05) in all three cell lines (supplemental Table S4). The fact that a higher number of differentially expressed transcripts in the presence of ER␤ were down and not up-regulated, reflects the well-known repressive effect of ER␤ on gene expression. Among commonly deregulated transcripts in all three cell lines, we found some known to be hallmarks of TNBC, such as IGFBP3, ID1, TM4SF1, TSPO, and ABAT. Interestingly, all the above-mentioned mRNAs were downregulated on ER␤ expression, with the notable exception of ABAT. Functional annotation by Ingenuity Comparative Analysis performed on differentially expressed transcripts, revealed that apart from regulation of proliferation and cell cycle progression, ER␤ controls the expression of genes involved in tumor progression, invasion, angiogenesis, cell death, apoptosis and metabolism of steroids and cholesterol (Fig. 2C). Comparison of influenced signaling pathways (supplemental Fig. S2C) corroborated the result obtained for MDA-MB-468 cell line alone (supplemental Fig. S2A) and confirmed that sirtuin, cholesterol and estrogen biosynthesis pathways are characterized by the same behavior in all three TNBC experimental models, indicating that their deregulation represents a common effect of ER␤ expression in TNBC cells. Importantly, all three branches of cholesterol biosynthesis pathway were downregulated (supplemental Fig. S2C). Analysis of genes known to participate in cholesterol biogenesis revealed downregulation also of SREBF1 gene, encoding for an upstream regulator of cholesterol signaling pathway (supplemental Fig.  S2D). Indeed, it encodes for a transcription factor driving the expression of both cholesterol biosynthesis and fatty acid synthesis genes, indicating a profound regulation of this signaling pathway by ER␤ (supplemental Fig. S2E). On the other hand, some genes displayed discordant response to ER␤ between the three cell lines, suggesting subtype-specific effects of the receptor (supplemental Table S4 and supplemental Fig. S3). Finally, considering the possibility that the activity of ER␤ may influence the global TNBC character of the cell, a deep learning-based framework subtype classification was performed before and after ER␤ induction with DeepCC that employs PAM50 for intrinsic subtype classification and whose classification performance has been trained on TCGA RNA-Seq data sets (24). This method, that represents a biological knowledge-based framework for cancer molecular subclassification, considers the expression patterns of a given tumoral cell or tissue to classify it according to defined parameters. The results obtained indicate that the presence of ER␤ caused changes leading to re-classification of all three cell lines, with a repositioning from a TNBC-associated subtype to one like Luminal A (probability score Ͼ 99%).
ER␤ Inhibits TNBC Cell Clonogenic and Migratory Properties-Functional analysis of the RNA-seq results suggested that a significant number of genes and signaling pathways downregulated by ER␤ in all three TNBC cell lines studied are involved in cell migration, invasion and viability. This was confirmed experimentally by measuring the clonogenic and migratory potential of these cells before and after ER␤ induction. The results obtained show that this is indeed the case, as the expression of the receptor caused a significant reduction of clonogenic potential (supplemental Fig. S4A) and migratory capability (supplemental Fig. S4B) in all three cells lines.
ER␤ Binds Promoter of Downregulated Genes Including SREBF1-As far as ER␤ is a transcription factor, able for DNA binding with further regulation of gene activity, we hypothesized that transcriptome downregulation, including transcripts belonging to cholesterol biosynthesis pathway, may be directly controlled by receptor binding to regulatory elements of its genes. To investigate it and to further characterize the functional role of ER␤ in TNBC, ChIP-Sequencing (ChIP-Seq) was performed with an anti-Flag antibody in MDA-MB-468 ER␤ϩ and ER␤-cells as control. Results showed the ability of ER␤ to interact with the TNBC cell genome and led us to identify a total number of 15843 binding sites (supplemental Table S5), as displayed in supplemental Fig. S5A. The density plot reported in supplemental Fig. S5B shows the signal distribution profile of ER␤-binding sites, indicating transcription start site-specific association of the receptor and highlighting the prevalence of its positioning in gene promoter regions. An analysis of overrepresented transcription factor binding motifs among the binding sites revealed, as expected, the highest accumulation of EREs (estrogen response elements) and ERE-like sequences. Enrichment for the following transcription factor binding sequences was observed: RARA, RARB, PPARG, ESRRA, ESRRB, nuclear receptors NR2F2, NR2F6, and NR4A1 (Supplemental Fig. S5C), that are known to recognize DNA motifs similar to ERE. Genome distribution analysis of the receptor binding sites showed preferential ER␤ positioning in gene introns (ϳ47%) and intergenic regions (ϳ42%) with around 6% sites mapping to gene promoters (supplemental Fig. S5D). We further analyzed if commonly deregulated transcripts in all tested TNBC cell lines represent primary ER␤-responsive genes, namely hold a receptor binding site within their promoter region. In this way, we found that out of 843 ER␤-responsive RNAs, 280 represent direct targets of this nuclear receptor, 68% binding the receptor via one or more EREs. IPA functional annotation analysis performed on this gene set corroborated the results previously obtained by functional assays and RNA profiling, again suggesting the involvement of this receptor in regulation of cell cycle, growth and proliferation, cell death and survival, metabolic processes (supplemental Fig. S5E). Here, many novel ER␤-influenced processes were present, among which molecular transport, cell signaling and interaction, indicating a possible role of the receptor in intercellular communication. Interestingly, among downregulated genes in all three TNBC cell lines, having ER␤ binding sites in their gene body, we found, TM4SF1 and ABAT, whose association with TNBC was previously described by others, as it was mentioned above. Interestingly, expression of several components of the cholesterol biosynthesis and signaling pathway present in the list of ER␤-regulated genes is inhibited by the receptor in all TNBC cell lines analyzed and, in several cases, the transcription unit harbors a receptor binding site. CYP51A1, DHCR7, PMVK, and SREBF1 genes show an ER␤ binding site in the promoter region (supplemental Fig. S5F and data not shown), whereas DHCR24 and LSS carry it in the second and third intron respectively. Sequence analysis of the ER␤ binding sites of all aforementioned genes revealed the presence of ERE motifs in DHCR7, LSS, PMVK, and SREBF1. Altogether, these data provide mechanistic evidence of cholesterol biosynthesis regulation by ER␤ in TNBC cells. Further, as the SREBF mRNA isoform SREBP1a, known to activate both fatty acid and triglyceride biosynthesis together with the cholesterol pathway (25,26), it has been shown to be highly expressed in proliferating cells, including particular cancer cells (27), confirming that direct downregulation of the corresponding gene transcription by ER␤ may underlie the anti-proliferative actions of the receptor in TNBC cells.
Mapping Nuclear ER␤ Interactome Reveals Its Association with Polycomb Repressor Complexes 1 and 2-To unveil molecular bases of ER␤ action in TNBC, and thus its impact on transcriptome deregulation, characterization of receptor interactome in MDA-MB-468 cell line was performed. To this aim, native nuclear protein complexes were purified by immunoprecipitation, analyzed by nano LC-MS/MS and mapped (Fig. 3A). Potential contaminants, such as keratins and immunoglobulins, were excluded from further analysis. Data analysis allowed us to identify 1023 potential ER␤ interactors in this cell line (supplemental Table S6). The following known receptor molecular partners were present among ER␤ interactors: NRIP1 (also called RIP140) (28), NCOA5 and NCOA6 (also known as CIA and RAP250 respectively (29,30), DNT-TIP2 (otherwise known as ERBP) (31), KAT5 (also referred to as TIP60) (32), mediator complex (33) and RBM39 (which alternative name is CAPER) (34). Molecular type classification of receptor-associated proteins revealed the prevalence of enzymes and transcription factors within ER␤ interactome (Fig. 3B). Functional enrichment analysis performed by IPA, elucidated the involvement of receptor partners in multiple molecular functions, both relevant to cancer and functional processes known to be fulfilled by this nuclear receptor, such as RNA post-transcriptional regulation, gene expression, DNA replication, cell cycle, cell death and survival and protein synthesis (Fig. 3B). Comparative analysis of the present ER␤ interactome with the one of hormone-responsive cell line MCF7 (35), revealed the significant similarity of the two data sets. To support the results obtained for ER␤ interactome in MDA-MB-468, we performed the same experiment in another TNBC cell line: the HCC1806. Here, we successfully identified 462 ER␤ molecular partners (supplemental Table S7), among which 174 were common for two cell lines analyzed (supplemental Fig. S6A). Classification of HCC1806 ER␤ interactome demonstrated enrichment of enzymes and transcription factors also in this data set (supplemental Fig. S6B). A comparison of molecular functions exerted by ER␤ partners in these two cell lines was basically the same (supplemental Fig. S6C), demonstrating the high similarity of ER␤ functional activity in both cell lines.
The ER␤ partner proteins identified here are likely to be involved in multiple pathways controlled by the receptor in TNBC cell nuclei, such as those controlling cell proliferation and migration. An upstream regulator analysis based on the gene expression changes detected in the three cell lines profiled here supported this possibility, as it highlighted the following receptor interactors: CTNNB1, a coactivator of CREB, the transcription factors MTA1, E2F6, STAT3, RUNX1, and EGR1 and the chromatin remodeling factors HDAC1, HDAC2, NCOA6, SMARCA2, and EZH2, all connected to transcription regulatory pathways known to control these cellular functions in cancer cells.
The generated data sets were further screened for the presence of protein complexes annotated in CORUM database (20), leading to identification of multicomponent assemblies associated to ER␤ in TNBC cells nuclei involved in DNA replication, transcription regulation (RNA polymerase and mediator complex), RNA splicing and post-transcriptional regu-  Table S8). These results let us suggest that association with ER␤ could drive these proteins to DNA regulatory elements (e.g. enhancers and promoters) or RNAs to exert cooperative regulation of transcriptional activity and post-transcriptional RNA processing. Functionality of the observed protein associations was confirmed by the fact that ER␤ expression induces alternative splicing events, as described above (supplemental Fig. S2B). Among ER␤ molecular partners, we identified several proteins belonging to different chromatin remodelling assemblies, involved in both transcriptional activation (COMPASS and SWI/SNF) and repression (NCOR1, LSD1, Polycomb Repressor Complexes 1 and 2 (PRC1/2)) (Fig. 3C). Interestingly, both transcriptional activator complexes identified are known to antagonize PRC1/2 function, whereas transcriptional repressor LSD1 interacts with PRC2 via the long non-coding RNA HOTAIR and functions through demethylation of the H3K4 histone mark of transcriptionally active chromatin, whereas at the same time PRC2 methylates H3K27 leading to chromatin condensation (36 -38). Moreover, the core PRC2 component EZH2 is a histone methyltransferase known as a master regulator of chromatin rearrangements, involved in cell survival, proliferation, epithelial to mesenchymal transition, invasion and drug resistance of cancer cells (39). EZH2 high expression was also found in a wide range of cancer types, such as lymphoma, sarcoma, breast (including TNBC), prostate, bladder, colon, lung, and pancreatic cancers, where it correlates with advanced disease stages and poor prognosis (40,41). Thus, EZH2 inhibitors represent potential drugs for anticancer therapy, some of them being currently validated in pre-clinical and clinical trials (39). On the other hand, PRC1 is required for stabilizing PRC2-introduced epigenetic silencing and has been found recruited to oncogenic active enhancers in breast cancer cells to regulate their activity (42).
Given the key roles of PRCs in breast cancer, we further focused on the interaction between ER␤ and PRC1/2 complexes. Components of PRC2 and of canonical and noncanonical PRC1 (cPRC1 and ncPRC1) complexes were found among ER␤ interactors in both cell lines. In particular, the core PRC2 components SUZ12, EED, and EZH2 and the accessory proteins JARID2, PHF1, and RBBP7 were present at least in one of the two data sets, together with the cPRC1 and ncPRC1-participating proteins RING1 and RNF2, common to both complexes, CBX4, CBX8, PHC2, and PHC3, cPRC1specific and proteins PCGF6, CBX8, MGA, E2F6, and WDR5, ncPRC1-specific (43). ER␤ co-immunoprecipitation assay was performed under the same experimental conditions in both cell lines and the results, reported in Fig. 4  468 and HCC1806 cell lines. Interestingly, ER␤ϩ samples were enriched in both PHF1 and JARID2 proteins, that according to Holoch et al. (44) represent PRC2.1 and PRC2.1specific components, leading us to conclude that ER␤ interacts with both PRC2 complex subtypes in both cell lines analyzed. The interaction was further confirmed in the same cell lines by immunoprecipitation of key PRC2 components EED, EZH2 and SUZ12 ( Fig. 4 and supplemental Fig. 6D). Similarly, ER␤-PRC1 association was validated by RING immunoprecipitation ( Fig. 4 and supplemental Fig. S6D). EED and RING1 association with ER␤ was also confirmed by coimmunoprecipitation ( Fig. 4 and supplemental Fig. S6D). Finally, PRC1 and PRC2.1, but not PRC2.2, association with ER␤ were detectable also in Hs 578T cells probably reflecting subtle differences between TNBC subtypes (supplemental Fig. S8). Finally, it is worth mentioning that among potential ER␤ interactors we found several proteins involved in biosynthetic processes, including the lamin B receptor (LBR), essential for cholesterol biosynthesis (45).
Finally, because our analyses indicate EZH2 as a potential regulator of gene expression changes via association with ER␤, we evaluated if the interaction of this protein with the receptor indeed occurs on the chromatin, leading to both proteins binding together to regulatory elements of ER␤-responsive genes. To this aim, ChIP-Western blotting and ChIP-qPCR were performed in ER␤and ER␤ϩ MDA-MB-468 cells. Results, shown in supplemental Fig. S9A, demonstrated that the association between the two proteins can be detected also on isolated chromatin. Further, ChIP-qPCR focusing on the ER␤ binding sites identified by ChIP-Seq on SREBF1, CYP51A1 and DHCR7 transcription units (supplemental Fig.  S5F) revealed co-recruitment of both factors to the first intron of SREBF1 and the promoter regions of CYP51A1 and DHCR7 (supplemental Fig. S9B). On the other hand, EZH2 was not found associated with the receptor in the SREBF1 promoter region, in correspondence of one of the strongest ER␤ binding sites mapped in this cell line. This could indicate that the strength of ER␤ binding to chromatin might be a factor influencing its ability to interact with other proteins or, alternatively, that the nature/composition of its interacting complex could affect the kinetics of receptor association to the genome. DISCUSSION TNBC is a heterogeneous group, characterized by the lack of ER␣, PR, and HER2/neo, high proliferative rate, low response to current therapies and elevated risk of recurrence (46). In this study, using validated antibodies, we demonstrated that full-length ER␤ is expressed in a sizeable fraction of TNBCs, in agreement with other studies showing that these tumors can carry this receptor subtype, where it exerts oncosuppressive activities (47,48). Thus, the presence of ER␤ expression may represent a potential advantage for TNBC patients, because ER␤ activation with agonist ligands could enhance its antiproliferative activity and, in early cancer stages, amplify the receptor anti-tumoral effects (49). A good example of such drug is the selective ER␤ agonist LY500307, currently tested in phase 2 clinical trials for the treatment of estradiol withdrawal-induced mood symptoms in women with post-perimenopausal depression, raising the possibility of its use also in TNBC patients carrying ER␤ϩ neoplasms. We investigated here the effects of ER␤ in TNBC cells by combining interactional proteomics and genomics, in order to functionally elucidate the molecular mechanism of receptor action. ER␤ was found to inhibit proliferation, migration and clonogenicity of TNBC cells belonging to three distinct subtypes by reducing cell cycle kinetics with an accumulation of cells in G 1 cell cycle phase, as suggested by previous studies (8,50). Interestingly, ER␤-induced growth-inhibition occurs in all TNBC cell lines tested, although with slightly different kinetics, suggesting that the oncosuppressive effects of the receptor may be independent from the cellular background and, more important, from the TNBC subtype.
Characterization of the gene expression changes induced by ER␤ led to the identification of genes whose expression is greatly influenced by the receptor in TNBC cells, including several genes known to be critical in TNBC biology, such as IGFBP3, ID1, NRP2, MAPK12, KCNN1, TM4SF1, TSPO, and ABAT, all of which (except ABAT) result to be downregulated by the receptor. IGFBP3 was the most significantly ER␤downregulated gene in all three TNBC cell lines (fold-change in the range of Ϫ3.69 to Ϫ9.69). Interestingly, its role in basal-like TNBC cells was previously reported, showing that targeting of oncogenic signaling exerted by this protein with SphK and EGFR kinase inhibitors enhances mouse survival and increased apoptosis (51). In another study, targeting IGFBP3-mediated DNA repair function was shown to enhance chemosensitivity of basal-like TNBC (52). On the other hand, a high level of ID1 expression has been associated with stemness and EMT transition, and this protein is known as a mediator of lung metastatic colonization of TNBCs, together with its closely related family member ID3 (53,54). Further, NRP2 was shown to be expressed preferentially in tumorinitiating cells, whereas MAPK12 activation is known to promote cancer development and progression by stimulation of cancer stem-like cell expansion. These two proteins were proposed as novel therapeutic targets in TNBC (55,56). Inhibition of KCNN4 expression by specific siRNAs significantly inhibited cell proliferation, migration and promoted apoptosis, whereas its enhanced expression was correlated to EMT in MDA-MB-231 cells (57), whereas high levels of TM4SF1 mRNA correlated with poor TNBC prognosis (58). The gene encoding TSPO, a protein generally associated with advanced breast cancer stages and expressed at higher levels in ER-negative compared with ER-positive tumors, was also found to be downregulated by ER␤ in all three TNBC cell lines studied here. Interestingly, TSPO overexpression significantly altered cell migration and combined treatment of TSPO li-gands with the antiglycolytic drug lonidamine lead to decreased viability of ER-negative breast cancer cell lines (59), whereas a synergistic inhibition in TNBC cell and tumor growth was achieved by combining these ligands with drugs targeting the cannabinoid receptor CB2R (60). Among the genes up-regulated by ER␤ in all three cell lines, we noticed ABAT, whose decreased expression was correlated with shortened recurrence-free survival in both ERϩ and ERbreast cancer patients and induced tumorigenic and metastatic advantages to basal-like breast cancer by activating the GABA-mediated Ca 2ϩ -NFAT1 axis (61). All these gene expression changes in response to ER␤ induction support the oncosuppressive role of this transcription factor in TNBC, marking its significance in TNBC biology. ER␤ expression causes also inhibition of all three branches of the cholesterol biosynthesis pathway, as demonstrated by a strong downregulation of the genes encoding most of the enzymes catalyzing different steps of this process, as summarized in supplemental Fig. S2E. Supporting this possibility, genome-wide mapping of ER␤ binding to MDA-MB-468 cell genome revealed that SREBF1 and other genes participating in cholesterol biosynthesis are directly regulated by receptor binding to their promoter and/or other regions (supplemental Table S5 and supplemental Fig. S5). More interestingly, activation of the cholesterol biosynthesis pathway was found relevant for BC responses to extracellular stimuli (62) and disruption of cholesterol-containing lipid rafts induced apoptosis, expression of WNT receptor LRP6, survivin and common apoptotic markers in TNBC cells (63). Moreover, TNBC cells treatment with natural dietary compounds exerting anti-cancer activity resulted in reduced accumulation of esterified cholesterol caused by decreased SREBP1, SREBP2, FASN and ACAT-1 levels (64). Higher expression of genes involved in cholesterol biosynthesis has been found associated with shorter relapsefree survival in basal-like breast cancer, whereas patientderived mammospheres exhibit increased de novo cholesterol synthesis and their formation is reduced by genetic or chemical inhibition of this pathway (65). Interestingly, another epigenetic target of ER␤ is represented by the histone deacetylases sirtuins, whose effects on lipid metabolism in normal and cancer cells are well-known. SIRT1 directly deacetylases SREBPs and thereby impacts on SREBP ubiquitination, protein stability and activity on its target genes. SIRT1 chemical activators inhibit SREBP target gene expression, leading to a decrease of lipid and cholesterol levels in mice liver (66). Moreover, small molecule activators of sirtuins are considered as promising therapeutic agents for the treatment of metabolic diseases (67). The molecular mechanisms mediating the above described effects of ER␤ on gene activity and, as a consequence, on TNBC cell functions find an explanation in the nature and composition of the nuclear interactome of this receptor identified by interaction proteomics (Fig. 3B-3C, supplemental Fig. S6 to S8 and supplemental Tables S6 and S7), in particular when combined with distribution of ER␤ cistrome in the TNBC cells genome.
Among potential ER␤ interactors, we identified a network of transcription factors and chromatin remodeling proteins involved in the regulation of proliferation, cell cycle and migration that the IPA upstream transcriptional regulator module revealed to be potentially involved in transcriptome reprogramming by ER␤-mediated in TNBC cells. These include, among others: the CREB/CTNNB1 complex, considered important for TNBC biology and response to therapy (68), whose association with FOXM1 modulates cancer stem cells phenotype of TNBC (69); the tumor suppressor CDKN2A that acts through interaction with cyclin-dependent kinases to inhibit cell cycle progression leading to growth inhibition (70); the transcription cofactor MTA1, that is known to bind ER␣ in hormone-responsive BC cells thereby inhibiting the ability of estradiol to stimulate estrogen receptor-mediated transcription (71); the oncogenic transcription factor STAT3, whose activity has been linked with cancer initiation, progression, metastasis, chemoresistance and immune evasion and plays a critical role in TNBC, where its inhibitors have shown efficacy in inhibiting tumor growth and metastasis (72); the core binding factor RUNX1, already shown to play an important role in BC (73). Interestingly, the interaction proteomics results described here comprise several known ER␤-partners present in the BioGRID database (ca 54%), that although identified in other cell types provide indirect confirmation of our findings.
One of the most interesting findings provided here is ER␤ association with key components of the multiprotein epigenetic complexes COMPASS, SWI/SNF, NCOR1/HDAC3, LSD1 and PRC1/2 (Fig. 3C, 5 and supplemental Fig. S6D and S8), all known to be involved in chromatin remodeling and transcriptional and post-transcriptional gene regulation in normal and transformed cells including TNBC. It was recently shown that the H3K4 histone methyltransferase MLL4, a component of one of the COMPASS complexes, regulates a cohort of oncogenes and pro-metastatic genes linked to BC, including TNBC, cell proliferation and invasion in association with the H3K27 demethylase UTX (also known as KDM6A) (74). Knock-down of either one of these two genes significantly decreased cell proliferation and invasiveness, both in vitro and in vivo, whereas their high expression was found to associate with poor prognosis. On the other hand, the BRG1 and BRM ATP-dependent chromatin remodeling enzymes, key catalytic subunits of the SWI/SNF complex, were found to be overexpressed in most primary BCs, including TNBCs, where inactivation of the corresponding genes affects tumor formation in vivo and cell proliferation in vitro (75). The importance of another two ER␤ partners -HDAC5 and complex LSD1 in TNBC progression was demonstrated by the fact that combinational treatment with HDAC5 and LSD1 inhibitors acted synergically to inhibit the growth of TNBC xenografts in mice (76).
Among the protein complexes identified in the ER␤ interactome mapped here, of particular importance appear the polycomb repressive complexes (PRC) 1 and 2, as these are well known to control key features of TNBCs, mediated by their activity on chromatin repression (42) resulting in increased cell proliferation, tumor invasiveness, poor differentiation and aggressiveness in BC, typical signatures of TNBCs (73)(74)(75)(77)(78)(79). The ability of EZH2 to interact with ER␤ to control gene transcription, namely to inhibit expression of cholesterol biosynthesis pathway genes, is supported by the finding that both proteins can be recruited to regulatory elements of at least three genes involved in cholesterol biosynthesis and downregulated in ER␤ϩ cells. When combined, these evidences point to the possibility that ER␤ association with regulatory factors may likely cause their repositioning within TNBC cell genome, leading to reprogramming of the cell fate program to a less aggressive phenotype.
It should be mentioned that exogenous protein expression in cell models do not allow to reproduce an exact expression range of the expressed product that matches the one present in tumor samples, so that validations in alternative experimental models are generally required. For this reason, the most relevant results obtained here will need now further investigation in in vivo in animal models, patient-derived xenografts and tumor biopsies. Another interesting direction of future research, guided by results obtained here on chromating repressive complex association with the receptor and inhibition of cholesterol biosynthesis, will concern the effects of other known ER␤ isoforms in TNBCs.
In conclusion, the results reported here provide for the first time a comprehensive catalogue of ER␤ effects in the three most representative TNBC subtypes, that helps elucidate the oncosuppressive actions of this nuclear receptor in this aggressive cancer type. Further, available genetic and biochemical data point out to several molecular targets potentially exploitable to interfere with specific features of TNBC cells.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE
The study protocol received approval by the Ethics Committee of the Istituto Nazionale Tumori 'Fondazione Giovanni Pascale' (protocol n.er CEI/393/15) before the beginning of the study, in accordance with The Code of Ethics of the Declaration of Helsinki, and informed consent was obtained from all patients involved.