Contrasting activities of estrogen receptor beta isoforms in triple negative breast cancer.

As shown in Figure 1a , human ERβ1 is the longest isoform (530-amino acids) and has a domain structure closely related to ERα. ERβ1 contains domains A/B, C, D, E and F, which include the DNA-binding region C, activation functions 1 and 2 in the A/B and E regions, respectively, and the ligand binding domain E. ERβ2 (495 amino acids), ERβ4 (481 amino acids) and ERβ5 (472 amino acids) are truncated forms, shortened and altered in their C-terminal regions due to alternative splicing of the ESR2 gene [ 25 , 26 ], ERβ6 (440 amino acids) is missing the segment between amino acids 317 and 408. Our analysis of TCGA clinical data [ 27 ] showed that ERβ1 and ERβ6 RNAs were present in less than 10% of breast tumors of all subtypes ( Figure 1b ). ERβ4 RNA was present in about 60% of breast cancers, but the mRNA level was very low ( Figure 1c ). ERβ2 and ERβ5 RNAs were the predominant RNA forms, and were present in 80-95% of breast tumors. The mRNA levels of ERβ2 and ERβ5 were also higher than that of the other isoforms, with ERβ5 being the most abundantly expressed isoform. To explore the possible prognostic value of ERβ2 and ERβ5, we examined disease-free survival (DFS) and overall survival (OS) probabilities. Log rank analyses of the clinical data indicated an association of high ERβ2 with decreased DFS and OS in TNBC that was significant for the latter ( Figure 1d ). By contrast, we found no significant association for ERβ5 and DFS or OS in TNBC ( Figure 1e ). Further analyses of clinical data for ERβ isoform expression and TNBC tumor stage revealed that the median expression level of ERβ2 increased slightly in TNBC tumor stage II and III vs. stage I, whereas the median level of ERβ5 expression was similar in stages I, II, and III ( Supplementary Figure 1 ). We also did Kaplan Meier analyses and found no significant association of ERβ2 with DFS or OS in all breast cancers (n=1,000 cases, TCGA). By contrast, our Kaplan Meier analyses of HR+ breast cancers (n=753 cases, TCGA) showed that high expression of ERβ2 or ERβ5 was associated with an overall better DFS (but short of statistical significance p=0.08) but not OS ( Supplementary Figure S2 ). This effect of ERβ2 and ERβ5 in HR+ breast cancer is opposite to that shown in Figure 1d and 1e for TNBC. We next examined the expression of ERβ isoforms in several TNBC cell lines using RT-PCR. ERβ2 and ERβ5 were found in all cell lines, with highest expression in MDA-MB-468 and BT-549 cells ( Figure 1f ). ERβ1, ERβ4, and ERβ6 were either not detectable or were present at very low levels in these cell lines and are therefore not shown. Western blot analysis after subcellular fractionation showed ERβ2 and ERβ5 proteins to be almost exclusively in the nuclear fraction, whereas β-actin was largely cytoplasmic, as expected ( Figure 1g ). Based on mRNA expression ( Figure 1f ), we expected that ERβ5 would be the more abundant protein in the cells; however, Western blot analyses in Figure 1g showed that ERβ2 levels were at least 5-fold higher than ERβ5. Using a Dox-inducible lentiviral expression system ( Supplementary Figure S3 ), we observed upregulated expression of ERβ2 and ERβ5 proteins ( Figure 1h ), and cotreatment with the proteasome inhibitor MG132 further increased the level of ERβ5 detected, suggesting that ERβ5 might be susceptible to proteasomal degradation in cells. In contrast, we did not observe any difference in ERβ2 protein level with MG132 treatment, indicating that ERβ2 protein is relatively more stable. To examine the effect of ERβ2 and ERβ5 on cellular activities, we downregulated their expression using siRNA. We could not design siRNA specifically targeting either ERβ5 or ERβ2 because there are only 4 unique amino acids at the far C-terminus end of ERβ5 compared to ERβ2. Therefore, we designed and used siRNA targeting both ERβ2 and ERβ5. The efficiency and specificity of ERβ2/β5 siRNA knockdown were confirmed by RT-PCR and Western blot. As shown in Figure 2a , upon siRNA treatment, the levels of ERβ2 and ERβ5 mRNA were reduced, as was the level of the proteins ( Figure 2b ). Downregulation of ERβ2 and ERβ5 led to significant reduction in the rate of cell proliferation ( Figure 2c ). Migration of MDA-MB-231 and BT549 cells was also reduced ( Figure 2d ), as was cell invasiveness ( Figure 2e ). To delineate the individual effects of ERβ2 and ERβ5 in TNBC, we overexpressed ERβ2 or ERβ5 using Dox-inducible lentiviral expression. Upon Dox induction, ERβ2 mRNA was elevated 13-fold and ERβ5 mRNA by 35-fold ( Supplementary Figure S4a , b ), resulting in significant increase in cell proliferation as well as migration and invasion ( Figure 3a – 3f ). These findings were opposite to but consistent with the findings with siRNA knockdown of ERβ2 and ERβ5, implying that both of these ERβ isoforms enhance cellular properties associated with tumor growth and aggressiveness. In contrast to what we observed with overexpression of ERβ2 or ERβ5, overexpression of ERβ1 using the Dox-inducible lentiviral system ( Figure 4a , Supplementary Figure S4c ) was found to suppress the proliferation, migration, and invasion of the TNBC cells ( Figure 4b – 4d ). We further confirmed the suppressive effect of ERβ1 on proliferation, migration, and invasion in MDA-MB-231 cells and in other TNBC cells (BT549 and DT28) using adenoviral infection as another means to elevate ERβ1 levels ( Supplementary Figure S5 ). ERβ1, the full-length isoform, is known to bind ligands [ 28 , 29 ]; so, to further delineate the mechanism underlying the ERβ1 mediated growth inhibitory effects, we examined the effect of ERβ1 on cells upon treatment with specific ligands. As seen in Figure 4b , increasing the level of ERβ1 in TNBC cells (Dox+veh) reduced cell proliferation, but treatment with estradiol or the ERβ-selective ligands, LY500307 (LY) and chloroindazole (CLI) [ 21 , 30 , 29 , 31 ], had no additional effect on proliferation ( Figure 4b ). In contrast, treatment with ligands further enhanced the inhibitory effect of ERβ1 on cell migration and invasion ( Figure 4c , 4d ), with LY500307 and CLI being more efficacious than E2 at the concentrations studied. Cystatins are natural inhibitors of cysteine proteases and are known to suppress cancer cell migration, invasion and metastasis [ 32 – 34 ]. Therefore, we examined the effects of ERβ1 and ligands on cystatins expression. Overexpression of ERβ1 significantly upregulated the expression of cystatin (CST) 1, 2, 4, and 5, and treatment with LY500307 and CLI, as well as E2, further increased the expression of these cystatins ( Figure 4e ). Epithelial cadherin (E-cadherin) plays a crucial role in cell-cell adhesion and is considered a tumor suppressor [ 35 – 38 ]. Of note, upregulation of ERβ1 greatly elevated E-cadherin protein level ( Figure 4a ), and stimulation with the ERβ-selective agonist ligands LY and CLI, as well as E2, further elevated E-cadherin protein and coordinately reduced cellular survivin ( Figure 4f and 4g ), considered to be a pro-oncogenic protein [ 39 ]. In cells with Dox-inducible ERβ2 or ERβ5, E-cadherin was very low, and no effect of the ERβ-selective agonist ligands, LY or CLI was observed (not shown). Survivin (BIRC5), a well-known factor for cell survival and cancer progression [ 39 ], is upregulated in primary breast tumors compared to normal breast ( Figure 5a ), and has the highest expression in the TNBC subtype ( Figure 5b ). Notably, siRNA mediated knockdown of ERβ2/ERβ5 reduced the level of survivin mRNA and protein to about half ( Figure 5c , 5d ), and Dox-inducible upregulation of ERβ2 or ERβ5 increased the cellular level of survivin protein ( Figure 5e , 5f ). In contrast, adenoviral ( Figure 5g ) or lentiviral ( Figure 4f ) overexpression of ERβ1 reduced survivin. These findings suggest that the differential regulation of the malignant behavior of TNBC by ERβ isoforms might involve survivin. To examine the effect of survivin in TNBC cells, we knocked down survivin using siRNA, reducing survivin/BIRC5 mRNA by 90% ( Figure 6a ) and survivin protein to half of that in non-specific siRNA (NS)-treated cells ( Figure 6b ). Downregulation of survivin/BIRC5 greatly reduced cell proliferation ( Figure 6c ) and also suppressed cell migration and invasion ( Figure 6d , 6c ).

MDA-MB-231 and BT549 cell lines were obtained from the American Type Culture Collection (Manassas, VA). MDA-MB-231 cells were maintained in DMEM (Sigma, St Louis, MO) supplemented with 5% fetal bovine serum (FBS), and 100 units/mL each of penicillin and streptomycin (Thermo Fisher) [ 19 ]. BT549 cells were maintained in RPMI 1640 supplemented with 10% FBS and 100 units/mL of penicillin and streptomycin. DT28 TNBC cells were derived from a primary TNBC basal human breast tumor and were kindly provided by Dr. Dorraya El-Ashry and were maintained in cell culture as described [ 20 ]. Cells were grown in phenol red-free media plus 5% charcoal-dextran-treated FBS for at least 5 days when depletion of estrogen was required. 17β-Estradiol (E2) was from Sigma-Aldrich. LY500307 was from Cayman Chemical Company, and chloroindazole (CLI) was prepared as described [ 21 ]. A doxycycline-inducible system was used for constructing ERβ1, ERβ2 and ERβ5 expressing cell lines. EGFP-Luciferase containing MDA-MB-231 cells were infected with 100 MOI of pLV[Exp]-CMV>Tts/rtTA/Hygro lentivirus and pLV[Exp]-Pur-TRE>hESR2[ NM_001437.2 , ERβ1]/HA, pLV[Exp]-Pur-TRE>hESR2[NM_00140275.1, ERβ2]/HA or pLV[Exp]-Pur-TRE>hESR2 [ NM_001214903.1 , ERβ5]/HA lentiviral particles in the presence of polybrene (8 μg/ml). At 48 h after transduction, the infected cells were selected using both 300 μg/ml hygromycin B and 2 μg/ml puromycin. After 2 weeks of selection, the clones were assayed for expression of the individual ERβ isoforms. For transient overexpression of ERβ1, cells were infected with either control adenovirus expressing β-galactosidase (AdGal) or with adenovirus expressing ERβ1 (AdERβ1) as described [ 11 , 17 , 15 ] for 8 h before excess virus was removed with media change. The concentration of adenoviral particles used for infection was 100 MOI for MDA-MB-231 cells and 25 MOI for BT549 and DT28 cells. After 48 h of transduction, cells were trypsinized and reseeded in 6-well plates for RNA collection, in 24-well plates for migration/invasion assays, or in 96-well plates for proliferation assays. ERβ2/ERβ5 siRNA (Dharmacon) was transfected into cells using Dharmafect Reagent 4 (T-2001, Dharmacon) at a concentration of 50 nM according to the manufacturer’s instructions. Of the two siRNAs tested for ERβ2 and ERβ5. the more efficient siRNA was chosen for further experiments. The sequences of the siRNA are as follows: ERβ2/ERβ5 siRNA Sense: 5′ AGGCATGCGAGGGCAGAAATT 3′, Antisense: 5′ UUUCUGCCCUCGCAUGCCUUU 3′. Non-targeting siRNA (NS) Pool 1 (D-001206-13, Dharmacon) was used as the control siRNA. WST-1 assay (Roche, Basel, Switzerland) was used to quantify cell viability as described [ 15 ]. Absorbance was measured at 450 nm using a VICTOR X5 PerkinElmer 2030 Multilabel Plate Reader. All assays were performed in triplicate and analyzed using Graph Pad Prism 8.0. Cell migration was assayed using 8.0 μm PET membrane 24-well tissue culture plates (VWR). Cell invasion was assayed using matrigel-coated transwell cell culture inserts (8 μm pore size) (BD Bioscience). Briefly, 22,000-25,000 cells were resuspended in 200 μL serum-free medium and seeded into the upper chamber of cell culture inserts. The lower wells of the chamber were filled with 800 μL medium supplemented with 10% FBS as a chemoattractant. Following 48 h of incubation, the membranes were excised and fixed with 100% methanol, stained with crystal violet, and washed with 1X PBS. Noninvasive cells in the upper chamber were removed by wiping with a cotton swab. The migrated/invaded cells at the bottom surface of the membranes were photographed at 40x magnification and quantified using Image J software, as previously described [ 22 ]. Total RNA was prepared using TRIzol (Thermo Fisher) and reverse transcribed using MMTV reverse transcriptase (New England BioLabs) according to the manufacturer’s recommendations. Real-time PCR was performed using SYBRgreen PCR Master Mix (Quantabio) as described [ 22 , 15 ]. Relative mRNA levels of genes were normalized to the housekeeping gene 36B4, and fold-change calculated relative to control. Results are the average ± SD from at least two independent experiments carried out in triplicate. Primer sequences for the genes studied were obtained from the Harvard Primer Bank, and are listed in Supplementary Table 1 . Whole cell extracts were prepared using lysis buffer containing 20 mM Tris, 150 mM NaCl, 1% NP-40, 1% SDS, 5% glycerol, and the Complete Mini protease inhibitor cocktail tablet (Roche). Cytoplasmic and nuclear fractions were isolated using CE buffer (HEPES [10 mM] pH 7.9, KCI [10 mM], EDTA [0.1 mM], NP-40 0.3% (added just before use) and 1X protease inhibitors (added just before use)) and NE buffer (HEPES [20 mM] pH 7.9, NaCl [0.4 M], EDTA [1 mM], glycerol 25%, 1X protease inhibitors (added just before use)) exactly as detailed previously [ 19 ]. Proteins were separated on 4–12% SDS-PAGE gels, and transferred to nitrocellulose membranes. The membrane was dried for 1 h to bind proteins tightly to the membrane and blocked with Intercept (PBS) Blocking Buffers (LI-COR) before incubation with primary antibody. Western blotting used antibodies against ERβ isoforms (MC-10, Thermo Fisher 14-9336-82), ERβ5 (Bio-Rad, Clone 5/25, MCA4676GA), survivin (Cell Signaling Technologies), E-cadherin (Cell Signaling Technologies), and β-actin (Sigma-Aldrich) as an internal loading control. Both IRDye 800 CW goat anti-rabbit secondary antibody (LI-COR, 926-32211) and IRDye 680 CW goat anti-mouse secondary antibody (LI-COR, 926-68070) were diluted (1:5000) for incubation with the blots. Quantification of protein bands used the LI-COR Odyssey CLx Imaging System and LI-COR analysis software. For analysis of clinical tumor datasets, RNA expression data for the ERβ isoforms were downloaded from Splicing Variants Database (TSVdb), an online tool for TCGA splicing variant analysis [ 23 , 24 ]. The TCGA patient intrinsic subtype, clinical stage, disease free survival (DFS) and overall survival (OS) data were downloaded through the Genomic Data Commons Data Portal ( https://portal.gdc.cancer.gov/ ). Welch’s t-test was used to compare ERβ isoform expression between the different breast cancer subtypes and tumor stage. The cutoff values for ERβ isoforms were defined by the upper quartiles. Kaplan-Meier survival analysis and log-rank test were performed to plot survival curves and compare disease-free survival (DFS) and overall survival (OS) times between the different groups. Statistical analysis used SPSS for Windows (IBM SPSS Statistics 24.0) with p < 0.05 considered to be statistically significant. For the in vitro cell studies, statistics were calculated using unpaired two-tailed Student’s t-test and GraphPad Prism 8.0 software. Significance was designated as * for p<0.05, ** for p<0.01, *** for p<0.001, and **** for p<0.0001.

Our work reveals the contrasting biological effects of ERβ isoforms in triple negative breast cancers (TNBCs) and the effects of ERβ-selective ligands. Notably, among the isoforms, ERβ1, the full-length ligand-binding form, was the least abundant form in human breast tumors and in the TNBC cell lines we examined. ERβ5 and ERβ2 were the most abundant forms. ERβ4 and ERβ6 were present at very low, barely detectable levels in the cell lines examined. The TNBC cell lines studied here are all basal breast cancers, representing the largest subgroup within TNBC. By up or down regulating the cellular levels of ERβ2 or ERβ5 vs. ERβ1, we found that these proteins elicited distinctly different effects ( Figure 7 ). While increasing ERβ1 reduced cell proliferation, migration, and invasion, increasing ERβ2 or ERβ5 enhanced these activities, and knockdown of endogenous ERβ2/β5 suppressed these cellular activities. These differences were mirrored in the effects of the ERβs on the RNA and protein expression of survivin and E-cadherin. As might be expected, knockdown of ERβ2/β5 suppressed cell aggressiveness and reduced survivin. Likewise, overexpression of ERβ1 decreased survivin, and increased expression of several cystatins, cysteine protease inhibitors that are associated with suppression of cancer progression [ 32 – 34 ]. Upregulation of ERβ1 also increased expression of the tumor suppressor, E-cadherin [ 36 , 35 , 37 , 38 ]. Our findings and those of others showing that cellular levels of cystatins [ 18 ] and E-cadherin are upregulated by ERβ1 alone, and that this is further augmented by treatment with E2, LY500307 and CLI, supports prior reports that ERβ1 has some basal ligand-independent activity but is also able to bind and respond to E2 and other estrogen and antiestrogen ligands with good affinity [ 15 , 16 , 18 ]. ERβ2 and ERβ5 have C-terminal portions of their ligand binding domains truncated and altered due to alternative splicing and are reported to have little or no ability to bind ligands [ 25 , 40 – 42 ]. Thus, our findings indicate that ERβ2 and ERβ5 promote properties associated with tumor progression, whereas ERβ1 has contrasting cancer suppressive effects, which in TNBC are enhanced further by treatment with ERβ3-selective agonist ligands ( Figure 7 ). The oncogenic activity of ERβ3 has been observed in glioblastoma, where ERβ5 was found to be associated with glioblastoma progression [ 43 ]. Of interest, our analyses of human tumor database information showed high ERβ2 to correlate with a worse DFS and OS. By contrast, although we found that experimentally upregulating ERβ5 in TNBC cells increased oncogenic cell properties of proliferation, migration and invasion, and increased levels of survivin, the level of ERβ5 in clinical triple negative tumor samples did not correlate with clinical outcome at a statistically significant level. Notably, although ERβ5 mRNA level was the highest of all the ERβ isoforms we analyzed, we found that ERβ5 protein levels were considerably lower than those of ERβ2. This is most likely the consequence of its more efficient proteasomal degradation, since we observed that treatment with proteasome inhibitor MG132 increased the cellular ERβ5 protein level. TNBCs lack three of the most important nuclear receptors in breast cancer, ERα, progesterone receptor, and HER2. Consequently, there has been great interest in determining whether other nuclear receptors might be critical regulators of TNBC cell behavior, and possibly serve as prognostic markers or therapeutic targets for treatment. In this regard, Conzen has proposed that use of a GR antagonist along with chemotherapy might be an optimal approach for treatment of metastatic high risk TNBC [ 3 ]. Likewise, in TNBCs that express AR, it has been suggested that AR antagonists might be beneficial in reducing cancer progression [ 4 , 44 , 45 ]. Our findings likewise suggest a role that ERβ might be playing in TNBC. ERβ2 and ERβ5 are associated with a more aggressive phenotype and their knockdown reduced TNBC migration and invasion. By contrast, ERβ1 promoted a more indolent and less aggressive phenotype, consistent with reports by others [ 18 , 46 , 47 ]. Indeed, ERβ1 has also been shown to suppress the pro-proliferative effects of ERα in ERα-positive breast cancers [ 15 , 16 ], and to have an important impact on TNBC metabolism and cholesterol homeostasis [ 47 ]. Although our studies here did not examine effects of ERβ isoforms in hormone receptor (HR)-positive breast cancer cells, our Kaplan Meier analyses of clinical HR+ breast tumors suggest that high expression of ERβ2 and ERβ5 are associated with overall improved DFS, implying different impacts of these ERβ isoforms dependent on the presence or absence or ERα. The effects of the ERβ isoforms on the behavior of TNBC cells reflect the changes we observed in levels of survivin, E-cadherin, and cystatins, with high survivin and low E-cadherin associated with poor disease-free and overall patient survival. Unlike AR and GR which are ligand regulated, the predominant isoforms of ERβ, ERβ2 and ERβ5. do not appear to be ligand regulated. Given that in TNBC the levels of ERβ1, the only isoform capable of ligand binding, are low, it is likely that ER ligands acting through ERβ might have only limited effects directly on TNBC tumor cells. However, ERβ, including the ligand-binding ERβ1 isoform, is expressed in a number of normal tissues that comprise a part of the microenvironment that hosts both primary and metastatic TNBCs [ 14 ]. Hence, by working through ERβ1 expressed in these tumor tissue microenvironments, ER ligands might have important antiproliferative, anti-invasive, and antimetastatic effects. Such tumor and microenvironment interactions have been noted in a few model systems of breast, as well as other cancers, and seem to involve inflammatory processes and immune modulation [ 48 ]. It is worth noting that in other inflammatory-driven pathological conditions (e.g., multiple sclerosis, endometriosis, vascular inflammation), ERβ ligands can have very strong immunomodulatory effects that appear to provide therapeutic benefit [ 49 – 51 , 31 ]. It remains an open question whether such ERβ-driven anti-inflammatory effects of ER ligands might prove efficacious in affecting TNBCs. It is now well appreciated that TNBCs are heterogeneous in their properties and that several subtypes are encompassed within the designation TNBC [ 52 , 53 ]. Our findings imply that understanding the absolute amounts and the relative ratios of the different ERβ isoforms might have prognostic and therapeutic relevance, and could enable better selection of optimal approaches for treatment of this often aggressive form of breast cancer.

Triple negative breast cancers (TNBC) represent a heterogeneous subtype of breast cancer that is considered the most difficult to treat and the subtype most likely to recur at early times after initial treatment [ 1 , 2 ]. TNBCs lack the three most analyzed ligand-regulated receptors (estrogen receptor α (ER), progesterone receptor (PgR), and HER2) that have proven to be useful therapeutic targets in other breast cancer subtypes, because agonist and antagonist ligands can be used to regulate receptor activity. Clearly, there is a need for new therapeutic targets for TNBCs, and in this regard, there are ongoing studies of the roles for other ligand-regulated receptors, such as androgen receptor and glucocorticoid receptor, in TNBC [ 3 , 4 ]. Because ERβ, a receptor structurally related to ERα, is present along with ERα in the normal breast [ 5 , 6 ], we wished to delineate the isoforms of ERβ present in TNBCs and assess their ligand dependent and independent activities, with the aim of evaluating their possible roles as prognostic biomarkers or therapeutic targets for this subtype of breast cancer. ERβ from human and rat was first identified in the 1990’s [ 7 , 8 ] and shown to be structurally related to ERα [ 9 ], and it was reported to be present in ovary [ 10 ], brain, and in the normal breast, as well as other tissues [ 11 ]. Much of the previous ERβ literature has been focused on ERβ1. the longest form of ERβ, and is confounded by employing some widely used ERβ antibodies that may not have been sufficiently specific for ERβ nor been able to identify the different isoforms of ERβ [ 12 – 14 ]. To clarify the endogenous expression status of ERβ, we first examined the expression of ERβ isoforms using clinical database information and also monitored ERβ isoforms in different TNBC cell lines. Then we assayed ERβ protein using antibodies shown to be selective for ERβ and that could distinguish among the different ERβ isoforms. Although ERβ1, the longest form of ERβ, has been shown to suppress the activity of ERα in ER-positive breast cancers [ 15 – 17 ] and to suppress the growth of TNBC [ 18 ], the presence and activities of other known ERβ isoforms in TNBC remain open to further investigation. In the present study, we show that ERβ2 and ERβ5 are the most abundant isoforms in TNBC and that they have effects on proliferation, migration, invasion, and gene regulation opposite to that of ERβ1. Our findings highlight the contrasting effects of specific human ERβ isoforms on malignant features of TNBC cells, with only ERβ1 showing response that can be modulated by ERβ agonist ligands.