Bone Marrow Stromal Cells Transcriptionally Repress ESR1 but Cannot Overcome Constitutive ESR1 Mutant Activity

Abstract

Estrogen receptor α (ER) is the target of endocrine therapies in ER-positive breast cancer (BC), but their therapeutic effectiveness diminishes with disease progression. Most metastatic BCs retain an ER-positive status, but ER expression levels are reduced. We asked how the bone tumor microenvironment (TME) regulates ER expression. We observed ESR1 mRNA and ER protein downregulation in BC cells treated with conditioned media (CM) from patient-derived, cancer-activated bone marrow stromal cells (BMSCs) and the BMSC cell line HS5. Decreases in ESR1 mRNA were attributed to decreases in nascent transcripts as well as decreased RNA polymerase II occupancy and H3K27Ac levels on the ESR1 promoter and/or distal enhancer (ENH1). Repression extended to neighboring genes of ESR1, including ARMT1 and SYNE1. Although ERK/MAPK signaling pathway can repress ER expression by other TME cell types, MAPK inhibition did not reverse decreases in ER expression by BMSC-CM. ESR1 mRNA and ER protein half-lives in MCF7 cells were unchanged by BMSC-CM treatment. Whereas ER phosphorylation was induced, ER activity was repressed by BMSC-CM as neither ER occupancy at known binding sites nor estrogen response element–luciferase activity was detected. BMSC-CM also repressed expression of ER target genes. In cells expressing the Y537S and D538G ESR1 mutations, BMSC-CM reduced ESR1, but expression of target genes PGR and TFF1 remained significantly elevated compared with that of control wild-type cells. These studies demonstrate that BMSCs can transcriptionally corepress ESR1 with neighboring genes and inhibit receptor activity, but the functional consequences of the BMSC TME can be limited by metastasis-associated ESR1 mutations.

Breast cancer (BC) is the second leading cause of cancer-related deaths in women in the United States (1). Although BC is a heterogeneous disease composed of many subtypes, 60% to 70% express estrogen receptor α (ER), with the incidence of ER-positive disease expected to increase (2). Endocrine therapies are prescribed to target ER and, despite their effectiveness, ER-positive disease constitutes the greatest number of BC-associated mortalities (3). Most recurrent and metastatic diseases maintain their ER-positive status. However, ∼70% of metastatic disease (skin, lymph node, bone, and lung) show reduced responsiveness to ER-targeted therapy (4–6).

ER expression in BC can be modulated by the tumor microenvironment (TME). Clinical studies show that metastatic tumors generally have lower levels of ER expression relative to patient-matched primary tumors. For example, in bone metastases, ESR1 mRNA levels are decreased relative to those in the primary tumors (7). Cejalvo et al. (8) further found ESR1 was commonly repressed in metastatic samples derived from multiple sites. ER protein expression was also decreased in locoregional metastasis in the lymph node (9). In vitro studies of ER-positive BC cells show that coculture with proinflammatory macrophages and BC-associated fibroblasts (CAFs) resulted in negative regulation of ESR1 via activation of ERK/MAPK and miRNA control, respectively (10, 11). Lang et al. (12) showed that multiple cell lines of different origins can downregulate ER protein and support cell growth. As there is a direct correlation between ER levels and its transcriptional function, it is plausible that TME cell types alter BC cell behavior via modulation of cellular ER levels (13).

In this study, we investigate how stromal cells of the bone microenvironment, a site where ER-positive BCs preferentially metastasize, regulate ER expression and transcriptional function (14). We show that bone marrow stromal cells (BMSCs) repress ESR1 mRNA and ER protein in ER-positive BC cells and that ER downregulation is primarily driven via transcriptional repression. In addition to ESR1, we observed corepression of several neighboring genes. Unlike macrophages and primary CAFs, this transcriptional repression is independent of MAPK activation and does not involve posttranscriptional or posttranslational regulation of ER but does involve loss of RNA polymerase II (Pol II) occupancy at the ESR1 promoter and a distal enhancer site (ENH1) combined with loss of the histone mark H3K27Ac on ENH1. The reduction in ESR1 was accompanied by inhibition of ER transcriptional activation. BMSCs also decreased ER and ESR1 mRNA expression in metastasis-associated ER mutants Y537S and D538G, but in contrast to cells expressing wild-type receptor, ER mutants remained constitutively active. These studies indicate that although multiple cell types in the TME generally act to reduce ESR1 mRNA and ER protein expression, they do so via independent pathways that may have a variable impact depending on the tumor cell context.

Materials and Methods

Cell culture

MCF7 (validated via short tandem repeat profiling analysis), MCF7 (Y537S nos. 1 and 2), MCF7 (D538G nos. 1 and 2), and HS5 (ATCC CRL-11882; Manassas, VA) cells were maintained in DMEM (Gibco, Paisley, Scotland, UK) with 10% fetal bovine serum (FBS; Hyclone, Logan, UT) and 1% 10,000 U/mL penicillin-streptomycin (PS; Gibco) at 37°C and 5% CO₂. Y537S and D538G mutant cell lines were provided by Dr. Steffi Oesterreich, University of Pittsburgh Cancer Institute, and validated as previously described (15). T47D cells were maintained in RPMI 1640 media (Corning, NY) with the same concentration of FBS and PS. BMSCs characterized as normal were derived from healthy bone marrow donors in accordance with the University of Wisconsin–Madison Institutional Review Board exemption project (protocol no. 2016-1299) as previously described (16). Normal BMSCs were maintained in phenol red–free minimum essential media α (Gibco) with 1% nonessential amino acids (NEAAs; Gibco), 2% l-glutamine (Gibco), 10% FBS, and 1% PS. Cancer-associated bone marrow stromal cells were isolated in accordance with the University of Wisconsin–Madison Institutional Review Board requirements (protocol no. HO07403). Samples were obtained with informed consent from multiple myeloma patients at University of Wisconsin Hospital and Clinics. Following depletion of CD138⁺ cells in the bone marrow samples with CD138⁺ beads (Miltenyi Biotec, Bergisch Gladbach, Germany), nonadherent cells were removed after two passages to obtain >90% purity of BMSCs. BMSCs were maintained in Opti-MEM media (Gibco) with 2% l-GlutaMAX (Gibco), 10% FBS, 1% NEAAs (Gibco), and 1% PS at 37°C and 5% CO₂. For experiments, cell lines and patient-derived BMSCs were used no longer than passage 20 and 3, respectively.

Conditioned media and treatments

For generating conditioned media (CM), 4 × 10⁵ BMSCs or MCF7/T47D control cells were first maintained in stripped serum media (SS media; Gibco; phenol red–free media with 10% charcoal-dextran stripped FBS, 1% PS, and 2% l-glutamine) for 24 hours. Cells were then washed twice with PBS, and SS media were replaced with serum-free, phenol red–free DMEM (Gibco), or Opti-MEM (for patient-derived BMSCs) media (Gibco) with 1% PS, 2% l-glutamine, 1% NEAAs, and 0.05% BSA (Sigma-Aldrich, St. Louis, MO). The serum-free media were conditioned for 48 hours. At collection, CM was filtered through a 0.2-µm surfactant-free cellulose acetate membrane syringe filter (Thermo Fisher Scientific, Waltham, MA). To prepare BC cells (MCF7 or T47D), 4 × 10⁵ cells were maintained in steroid-deprived conditions for 72 hours in six-well plates prior to coculture and/or hormone treatments. Steroid-deprived recipient cells were washed twice with 1× PBS and then treated with CM for the indicated time.

Studies involving hormone treatments used a vehicle control (0.1% ethanol; Decon Labs, King of Prussia, PA) or 10 nM 17β-estradiol (E2; Steraloids, Newport, RI) in either MCF7-CM or HS5-CM. For mitogen-activated protein kinase kinase (MEK) inhibitor studies, cells were maintained in serum-free media for 24 hours, as above, followed by 1-hour pretreatment with UO126 (Promega, Madison, WI) before CM treatments. For studies with actinomycin D (ActD), cells were pretreated with 2 μM ActD (Sigma-Aldrich) for 30 minutes prior to CM treatments. Experiments with cycloheximide (CHX; Sigma-Aldrich) involved a 1-hour pretreatment prior to CM treatments. Treatments, when applicable, were reapplied to CM on plated cells.

Quantitative RT-PCR

RNA isolation, reverse transcription, and quantitative RT-PCR (qRT-PCR) were performed as previously described (17). Each sample was run in duplicate and cycle threshold (Ct) values were averaged. RPLP0 and GAPDH were used as housekeeping genes for qRT-PCR reactions. A minimum of three biological replicates were analyzed. Primer sequences are available in an online repository (18). qRT-PCR was run on the Bio-Rad CFX Connect real-time detection system using iQ SYBR Green Supermix (Bio-Rad Laboratories, Hercules, CA). Relative RNA levels were calculated using the ΔΔCt method, and relative expression was calculated based on values normalized against MCF7-CM–treated samples (19). mRNA half-life was determined by log₂-transforming mRNA levels and calculating half-life based on the slope of the log₂-transformed values. Statistical significance was determined by comparing the ΔCt values and using a paired Student t test.

Reporter gene assay

MCF7 cells were transiently transfected using Lipofectamine 2000 (Thermo Fisher Scientific) with reporter constructs estrogen response element (ERE)-tk-luc and cytomegalovirus β-galactosidase (β-gal) (20). Cells were treated with CM or 10 nM E2 for 24 hours. Luciferase (Promega) and β-gal (Tropix, Waltham, MA) assays were conducted according to the manufacturer’s instructions. Luciferase values were normalized to β-gal activity. Fold change was calculated with normalized luminescence values relative to MCF7-CM. Statistical significance was determined by a paired Student t test on the luciferase/β-gal values.

Western blot analysis

Samples were prepared by lysing cells in 1× sample buffer [62.5 mM Tris (pH 6.8), 2% SDS, 10% glycerol, 5% β-mercaptoethanol). Protein levels were measured using the RC DC protein assay (Bio-Rad Laboratories). Equal amounts of protein were run on a 10% SDS-PAGE gel and transferred to polyvinylidene difluoride membranes. Blots were preblocked using 5% milk diluted in Tris-buffered saline with Tween 20 and subsequently incubated overnight with primary antibodies for ER (HC-20; Santa Cruz Biotechnology, Santa Cruz, CA; sc-543, 1:3000) (21), phosphorylated S118-ER (E91; Abcam, Cambridge, UK; catalog no. ab32396, 1:1000) (22), Hsp90 (H-114; Santa Cruz Biotechnology, sc-7947, 1:3000) (23), and β-actin (AC-15; Sigma-Aldrich, 1:10,000) (24). Following incubation with secondary antibodies (anti-rabbit; GE Healthcare, Chicago, IL, 1:3000 or anti-mouse) (25, 26), blots were exposed and imaged with a ChemiDoc imaging system (Bio-Rad Laboratories). Quantification of band intensity was conducted using ImageLab software (Bio-Rad Laboratories). Expression was normalized to Hsp90 or β-actin protein levels. Relative changes in protein levels were compared with that in MCF7-CM controls. Protein half-life was determined by log₂ transformation of quantified protein levels and calculating protein half-life based on the slope of the log₂-transformed values. Statistical significance was determined by comparing the normalized protein levels and using a paired Student t test.

Immunofluorescence

Cells were maintained in SS media for 72 hours in four-well chamber slides prior to CM treatments. Following treatments, cells were prepared for analysis as previously described (12). Antibodies used include ER (HC-20; Santa Cruz Biotechnology; sc543, 1:250) (21) and Alexa Fluor 594 anti-rabbit (A11012; Invitrogen, Waltham, MA; 1:100) (27). Cell nuclei were visualized with 4′,6-diamidino-2-phenylindole (DAPI; Invitrogen) staining. Imaging was done on the Nikon Eclipse Ti microscope using a 10× objective. Images were collected from the same imaging plane with ER and DAPI fluorescence exposure times matching across samples. Batch image analysis was conducted on all images using identical parameters using the open-source image analysis software JEX (28). Data were analyzed using custom R scripts. log transformed, and normalized to the median ER fluorescence observed from MCF7-CM.

Chromatin immunoprecipitation

Chromatin immunoprecipitation (ChIP) assays were performed as described in Helzer et al. (29). Antibodies used for immunoprecipitation include ER (HC-20; Santa Cruz Biotechnology, sc543, 1 µg per sample) (21), (Pol II; MMS-126R; Covance, Princeton, NJ, 4 µL per sample) (30), H3K27Ac (Abcam, catalog no. ab4729, 2 μg per sample) (31), and histone H3 (Abcam, catalog no. ab1791, 2 μg per sample) (32). Quantitative PCR was performed with 20 nM forward and reverse primers amplifying regions of interest. Primer sequences are available in an online repository with genomic locations (18). Relative fold change in occupancy was compared with percentage input of MCF7-CM controls. Three biological replicates were performed. Statistical significance was determined by comparing percentage input between samples and using a paired Student t test.

Data availability

Supplemental tables and figures are available in an online repository [https://doi.org/10.6084/m9.figshare.8479214 (18) and https://doi.org/10.6084/m9.figshare.8479217 (33)].

Results

Patient-derived cancer-associated BMSCs and a BMSC cell line decrease ER expression

ESR1 expression in bone metastases has been reported to decrease by approximately twofold in 7 of 11 patients relative to patient-matched primary tumors (7). Therefore, we asked whether patient-derived normal or cancer-associated BMSCs could regulate ESR1 expression in ER-positive BC cells. CM was collected from three samples each of normal and cancer-associated BMSCs under serum-free conditions. Serum-free conditions were used to ensure against the possibility of contamination with serum growth factors and to prevent possible BMSC production of estrogen (E2), which are known to downregulate ER expression (34, 35). Furthermore, MCF7 cells were maintained in SS media prior to CM treatment to further deprive cells of estrogen.

Following 24 hours of CM treatment, ESR1 mRNA expression was measured and analyzed relative to cells treated with MCF7-CM [Fig. 1(a)]. ESR1 expression in MCF7 cells treated with CM from the three cancer-associated BMSCs decreased by ∼60%. In contrast, ESR1 mRNA expression in MCF7 cells treated with CM from the three normal BMSCs was not statistically significantly different from MCF7 cells treated with MCF7-CM. To examine whether BMSC cell lines also elicited ESR1 repression, MCF7 cells were treated with CM from the human-transformed BMSC line HS5. Replicating the effects of the patient-derived, cancer-associated BMSCs, ESR1 expression decreased by ∼80% in MCF7 cells treated with HS5-CM [Fig. 1(b)]. ER protein expression was also assessed via Western blot and immunofluorescence analyses. ER protein expression similarly decreased by ∼80% in MCF7 cells treated with HS5-CM relative to cells treated with MCF7-CM [Fig. 1(c)]. Single-cell analysis of ER expression using immunofluorescence showed that HS5-CM caused a leftward shift indicative of decreased ER in the total cell population [Fig. 1(d)]. Similar results were observed in another ER-positive cell line, T47D (33). These results demonstrate that patient-derived, cancer-associated BMSCs decrease ER expression, and this effect can be modeled using the BMSC cell line HS5.

Figure 1.

(a) MCF7 cells were treated with CM from MCF7- and patient-derived BMSCs (N1 to N3, normal BMSCs; C1 to C3, cancer-associated BMSCs) for 24 h and subsequently analyzed for ESR1 mRNA expression via qRT-PCR. Gene expression was calculated relative to MCF7-CM following normalization to housekeeping genes. Error bars are representative of two to three independent replicates (n = 2 for N1, C1, C2, and C3; n = 3 for N2 and N3) ± SD. (b) MCF7 cells were treated with CM from MCF7 and HS5-CM for 24 h and analyzed for ESR1 mRNA expression as described in (a). Error bars are representative of five independent replicates ± SD. (c) Same conditions as in (b), with Western blot analysis measuring ER protein levels. Western blot is representative of five independent replicates. Protein levels were quantified by normalizing to Hsp90 levels and to MCF7-CM subsequently. Error bars are representative of five independent replicates ± SD. (d) MCF7 cells were treated the same as in (b), and single-cell immunofluorescence analysis was performed using a 10× objective measuring ER protein intensity between MCF7-CM–treated (n = 4872) and HS5-CM–treated (n = 5893) cells. The left panel shows a representative merged image of ER (pink) and DAPI (blue). The integrated intensities of ER expression were calculated, log transformed, and plotted as a density histogram with MCF7-CM (light blue) and HS5-CM (purple) treatment. *P < 0.05, ***P < 0.001 vs MCF7-CM.

BMSC-CM transcriptionally represses ESR1

ER expression is controlled at multiple levels of regulation, including transcriptional, posttranscriptional, and posttranslational (36–39). Therefore, we first asked how patient-derived BMSCs decreased ESR1 mRNA. To directly measure the transcription of the ESR1 gene, the effect of patient-derived BMSCs and HS5 cells on nascent, unspliced transcripts was evaluated. Using primers directed toward the intron–exon junction of ESR1 mRNA, nascent transcripts were measured in MCF7 cells following treatment with CM from cancer-associated BMSCs and compared with those treated with MCF7-CM. Following 24-hour CM treatment of cancer-associated BMSCs, nascent ESR1 transcripts decreased by a magnitude (∼60% to 80%) similar to that of mature ESR1 mRNA [Fig. 2(a)]. Time course analysis with HS5-CM resulted in a maximal decrease of nascent ESR1 transcripts by 2 hours, whereas mature transcripts lagged and reached equivalent levels by 4 hours. Both mature and nascent ESR1 transcripts levels remained repressed by ∼80% up to 24 hours [Fig. 2(b)]. Decreases in nascent ESR1 transcripts were also observed in T47D cells treated with HS5-CM (33).

Figure 2.

(a) Same conditions as Fig. 1(a). Nascent ESR1 mRNA transcripts were measured in MCF7 cells treated with CM from cancer-associated BMSCs (C1 to C3) via qRT-PCR. Gene expression was calculated relative to MCF7-CM following normalization to housekeeping genes. (b) MCF7 cells were treated with HS5-CM, and mature and nascent ESR1 mRNA transcripts were measured via qRT-PCR at several time points during 24 h. Following normalization, relative expression was calculated relative to the 0-h time point. (c) ChIP of Pol II in MCF7 cells treated with either MCF7-CM or HS5-CM for 0.5 or 2 h. Pol II occupancy was measured on the ESR1 promoter (+135) and enhancer (ENH1) regions (−150 kb) with locations specified relative to the transcription start site of each gene and calculated as percentage of input. (d) ChIP of H3K27Ac and total H3 histone in MCF7 cells treated with either MCF7-CM or HS5-CM for 0.5 or 4 h. H3K27Ac and H3 histone occupancy were measured on the same genomic sites as in (c). H3K27Ac and total H3 histone levels were calculated as percentage of input. Error bars are representative of three independent replicates (n = 2 for N1, C1, C2, and C3) ± SD. *P < 0.05, ***P < 0.001 vs MCF7-CM or 0-h time point.

ChIP analyses were performed for Pol II occupancy on ESR1 regulatory elements, including the ESR1 proximal promoter and an upstream enhancer (ENH1) previously shown to regulate ESR1 expression (36, 40, 41). HS5-CM treatment reduced Pol II occupancy in both regulatory regions as early as 0.5 hour and at the distal enhancer at 2 hours [Fig. 2(c)]. H3K27Ac, a chromatin mark associated with active gene transcription, was also measured. H3K27Ac decreased on ESR1 ENH1 in cells treated with HS5-CM for 0.5 hour relative to cells treated with MCF7-CM [Fig. 2(d)] but returned to control levels by 4 hours. There was a similar trend of decreased H3K27Ac on the ESR1 promoter, but the decrease did not reach the level of statistical significance. Total H3 levels were not different in cells treated with MCF7-CM or HS5-CM. Taken together, these data indicate that cancer-associated BMSCs can transcriptionally repress ESR1 expression. Furthermore, the decrease in ESR1 by transformed BMSCs is associated with a decrease in Pol II occupancy at both regulatory elements, along with transient loss of H3K27Ac at ESR1 ENH1.

Because HS5-CM impacted Pol II occupancy at ESR1 ENH1, we asked whether repression by BMSCs extended to neighboring genes. Specifically, we measured the expression of ZBTB2, RMND1, ARMT1, CCDC170, and SYNE1 (42). Repression of all five genes occurred in cells treated with HS5-CM, with the largest magnitude of repression centered on ESR1 following CM treatment from all BMSCs (Fig. 3). CM from patient-derived, cancer-associated BMSCs similarly corepressed ARMT1, ESR1, and SYNE1. There was variable expression of ZBTB2, RMND1, and CCDC170. These data suggest that BMSC-mediated repression is not limited to ESR1 but includes corepression of several neighboring genes.

Figure 3.

MCF7 cells were treated with MCF7-CM, HS5-CM, or cancer-associated BMSC-CM for 24 h. Gene expressions of ZBTB2, RMND1, ARMT1, CCDC170, ESR1, and SYNE1 were normalized relative to housekeeping genes and expression from cells treated with MCF7-CM subsequently. Fold change of gene expression in cells treated with cancer-associated BMSC-CM (C1 to C3) were averaged together. Error bars are representative of four to five independent replicates for HS5-CM–treated cells and two independent replicates for each cancer-associated BMSC (C1 to C3) ± SD. *P < 0.05, **P < 0.01, ***P < 0.001 vs MCF7-CM.

BMSC-CM regulation of ESR1 mRNA is independent of MAPK and posttranscriptional mechanisms

Previous studies showed that CAFs derived from breast tumors of the basal subtype can repress ESR1 mRNA via posttranscriptional mechanisms involving hyperactivation of the ERK/MAPK signaling pathway (10). Likewise, activated macrophages were shown to repress the ESR1 gene via the same pathway (11). Therefore, we asked whether HS5 cells regulate ER expression similar to both cell types via the ERK/MAPK signaling pathway. First, we inquired whether HS5-CM can activate the ERK/MAPK signaling pathway in MCF7 cells. Consistent with previous findings, HS5-CM similarly induced ERK/MAPK signaling based on induction of phosphorylated ERK (p-ERK) [Fig. 4(a)]. Pretreatment with the MEK inhibitor UO126 inhibited p-ERK induction but was not able to prevent downregulation of ESR1 mRNA or ER protein [Fig. 4(a) and 4(b)]. To further examine possible posttranscriptional regulation of ESR1, mRNA stability was tested by measuring changes in ESR1 mRNA levels following treatment with the Pol II inhibitor ActD in the presence of HS5-CM with MCF7-CM serving as a control. ESR1 mRNA levels decreased at a similar rate in cells treated with MCF7-CM or HS5-CM [Fig. 4(c)], with an approximate half-life of 3.6 and 3.9 hours, respectively [Fig. 4(d)]. Taken together, these data indicate that BMSCs can repress ESR1 mRNA via pathways that are distinct from other reported cell types in the TME.

Figure 4.

(a) MCF7 cells were placed in serum-free media for 24 h and subsequently pretreated with the MEK inhibitor UO126 (5 µM) for 1 h. Following pretreatment, cells were treated with MCF7-CM + dimethyl sulfoxide (DMSO), HS5-CM + DMSO, or HS5-CM + UO126 for 24 h. ER, p-ERK, total ERK, and Hsp90 protein expression were measured via Western blot analysis. Western blot is representative of three independent replicates. (b) ESR1 mRNA was measured via qRT-PCR in the same conditions as in (a). Following normalization to housekeeping genes, relative expression was calculated relative to MCF7-CM + DMSO. (c) MCF7 cells were pretreated with ActD (2 µM) for 1 h prior to MCF7 or HS5-CM treatment with ActD. Cells were treated with CM for up to 8 h at several time points. Relative expression was calculated following normalization to RPLP0 relative to the 0-h time point. (d) ESR1 mRNA levels were log₂ transformed, and mRNA half-life was calculated for both MCF7-CM– and HS5-CM–treated cells described in (c). Error bars are representative of three independent experiments ± SD. *P < 0.05 vs MCF7-CM.

BMSC-CM does not regulate ER protein stability

Posttranslational regulation of ER protein stability is another important mechanism governing cellular receptor levels, and studies in rodent models suggest that posttranslational regulation of ER may be disrupted following metastasis (43). We observed that HS5-CM decreased ER protein levels [Fig. 1(c)]. Therefore, we asked whether changes in ER protein levels could be independently regulated by active protein degradation. A CHX pulse-chase experiment was carried out in the presence of HS5-CM or MCF7-CM. In the presence of CHX, ER protein levels decreased at a steady rate in cells treated with MCF7-CM [Fig. 5(a) and 5(b)]. ER protein levels decreased at a similar rate in cells treated with HS5-CM. ER protein half-life was calculated at 3.4 and 4.7 hours in cells treated with MCF7-CM and HS5-CM, respectively [Fig. 5(c)]. The difference in ER protein half-life was not statistically significant. Given that decreases in ER protein lag behind changes in ESR1 mRNA and there was no significant change in the half-lives of ER protein or ESR1 mRNA, these data indicate that BMSC regulation of ER levels is distinct from mechanisms invoked by other TME cell types, as BMSCs do not regulate ER protein levels through posttranscriptional or posttranslational control.

Figure 5.

(a) MCF7 cells were pretreated with CHX (10 µg/mL) for 1 h. Cells were subsequently treated with MCF7-CM or HS5-CM with (+) or without (−) CHX and samples were collected at indicated time points. Western blots of ER with β-actin as a loading control are representative of three independent replicates. (b) ER protein expression was quantified relative to 0 h for CHX-treated samples following normalization to β-actin expression. (c) ER protein half-life was calculated following log₂ transformation of normalized protein intensities of CHX-treated samples. Error bars are representative of three independent experiments ± SD.

BMSC-CM has differential effects on ER transcriptional function

To measure ER transcriptional activity in the presence of HS5-CM, we investigated several events of the ER transactivation pathway, including ER phosphorylation at serine 118 (pS118-ER), ER–DNA binding, and ER target gene expression. pS118-ER is an important posttranslational modification required for maximum ER transcriptional activity following E2 and growth factor stimulation (44, 45). This ER phosphorylation event is rapid and occurs as early as 15 minutes of receptor activation, with maximal phosphorylation occurring at 30 minutes (29). MCF7 cells were treated with E2, MCF7-CM, or HS5-CM during a 4-hour period, and Western blot analysis was performed for pS118-ER. E2 treatment was included as a positive control. As expected, pS118-ER was detected at 30 minutes in cells treated with E2 and was elevated for the entire treatment period [Fig. 6(a)]. pS118-ER was not induced in cells treated with MCF7-CM. Similar to E2, pS118-ER was detected in 30 minutes in cells treated with HS5-CM. pS118-ER was also detected in T47D cells treated with HS5-CM for 30 minutes (33).

Figure 6.

(a) MCF7 cells were treated with E2 (10 nM), MCF7-CM + ethanol (EtOH), or HS5-CM + EtOH. Samples were collected at indicated time points and subsequently analyzed via Western blot for pS118-ER, total ER, and Hsp90. Western blots are representative of three independent replicates. (b) ChIP of ER was conducted on MCF7 cells with the same treatments as described in (a) for 30 min or 4 h. qRT-PCR was performed on known ER binding sites (ESR1, +1600; TFF1 promoter, −200; PGR, +95,559) with locations specified relative to the transcription start site of each gene, and ER occupancy was calculated as percentage of input. (c) MCF7 cells transfected with an ERE-tk-luc construct were treated with E2, MCF7-CM + EtOH, and HS5-CM + EtOH for 24 h. ERE activity was measured and normalized to β-gal activity, and fold change was calculated relative to MCF7-CM + EtOH subsequently. (d) SNAI1, PGR, TFF1, and GREB1 mRNA expression were measured via qRT-PCR in cells treated with MCF7 or HS5-CM for 24 h. Transcript values of these genes were normalized relative to that of housekeeping genes. Shown is the expression level calculated relative to MCF7-CM control samples. Expression levels were log transformed. Error bars are representative of three to five independent experiments ±SD. *P < 0.05, **P < 0.01 vs MCF7-CM.

Given that pS118-ER is associated with an active receptor and is preferentially bound at sites of active enhancers (29), we asked whether HS5-CM induces ER occupancy at known ER genomic binding sites using ChIP assays. The ESR1 promoter was examined because ER can autoregulate its own gene expression through this regulatory element (36). The TFF1 promoter and PGR enhancer were examined because Lang et al. (12) previously observed an increase in TFF1 and PGR gene expression in a direct coculture study with MCF7 and HS5 cells. As expected, ER occupancy was detected at these three sites in cells treated with E2 at 0.5 and 4 hours [Fig. 6(b)]. Surprisingly, although ER is phosphorylated in response to HS5-CM, this was not sufficient to promote ER occupancy at these sites. Extending this analysis, a consensus ERE reporter gene assay was performed. The ERE-tk-luc reporter is driven by a multimerized consensus ERE and thymidine kinase promoter. Consistent with ChIP data, HS5-CM was unable to support activation of a consensus ERE [Fig. 6(c)]. Endogenous ER target gene expression was also examined [Fig. 6(d)]. TFF1 expression increased with HS5-CM, but unlike results observed from direct coculture studies with HS5-CM, PGR expression was repressed. GREB1 and SNAI1 expression both decreased following HS5-CM treatment. Taken together, these data suggest that despite inducing ER phosphorylation, BMSCs cannot support ER transcriptional function in MCF7 cells.

Approximately 20% to 40% of metastatic BCs contain ESR1 mutations that can cause constitutive activation of ER [reviewed in (46)]. Because BMSCs repressed ESR1 and ER transcriptional activation, we investigated how BMSCs affect the expression and activity of a constitutively active receptor in cells containing the metastasis-associated Y537S and D538G ESR1 mutations. MCF7 cell clones with a knock-in of the Y537S and D538G ESR1 mutation and the parental cell line were treated with HS5-CM, and ER protein was assessed by Western blot analysis. Y537S and D538G clones express lower basal levels of ER protein and ESR1 mRNA relative to control wild-type cells. Treatment with HS5-CM induced a further loss of Y537S and D538G ER protein and ESR1 mRNA (Fig. 7). Upon examination of ER target genes, we observed a decrease in GREB1 and PGR expression upon HS5-CM treatment in cells expressing wild-type receptor, as in Fig. 4. However, although HS5-CM treatment appeared to diminish GREB1 and PGR expression in mutant-bearing cells, this effect was not consistent across cell clones. Despite these differences, GREB1 and PGR expression levels were elevated between ∼4- and 14-fold and between ∼13- and 52-fold higher in HS5-CM–treated cells expressing the Y537S and D538G ESR1 mutations than in control wild-type cells [Fig. 7(b)]. TFF1 expression was also activated in the parental cell line by HS5-CM, consistent with our previous findings but was not changed in Y537S or D538G mutant cell lines (12). These data indicate that although BMSC-CM has a selective effect on ER activity in cells expressing wild-type ER, this repression is not sufficient to overcome the activity of a constitutive Y537S or D538G ESR1 mutant receptor.

Figure 7.

(a) Two clones of MCF7 cells expressing the Y537S ESR1 mutation, D538G ESR1 mutation, and the parental MCF7 cell line were treated with either their respective CM or HS5-CM for 24 h. Western blots of ER with β-actin as a loading control are representative of three independent replicates. (b) ESR1, GREB1, TFF1, and PGR mRNA expression were measured via qRT-PCR. Expression levels were normalized to that of housekeeping genes. Shown is the normalized expression level relative to that of MCF7-CM–treated controls. Error bars are representative of three independent experiments ± SD. *P < 0.05, **P < 0.01, ***P < 0.001 and ^#P < 0.05, ^##P < 0.01 relative to respective cell lines treated with their own CM.

Discussion

The bone is the most common site of BC metastasis, and it has been shown that cellular levels of ER are decreased in metastatic samples relative to the primary tumor samples (8, 14). In this study, we demonstrated that patient-derived, cancer-associated and transformed BMSCs repress ESR1 mRNA expression, which could be accounted for by direct transcriptional repression. The subsequent decrease in ER protein levels is likely a result of a reduction in ESR1 mRNA synthesis because neither posttranscriptional nor posttranslational regulation of ESR1 mRNA or ER protein was observed, respectively. Furthermore, our results indicate that the transcriptional repression of ESR1 occurs in conjunction with repression of several neighboring genes and is independent of ERK/MAPK activation. This makes BMSC regulation of ESR1 distinct from mechanisms engaged by other TME cell types (e.g., macrophages and BC-associated fibroblasts) that repress ESR1. Evaluation of the ER signaling pathway showed that BMSC-CM induces ER phosphorylation but does not support recruitment of ER to known ER binding sites or to a consensus ERE. Expression of endogenous ER target genes PGR, GREB1, and SNAI1 was also decreased, consistent with a diminution of ER functional activity. However, in the context of the metastasis-associated Y537S or D538G ESR1 mutations, BMSCs were unable to repress constitutive ER target gene activity despite downregulation of ESR1 mRNA and ER protein expression. Taken together, our results show that BMSCs could act along with other TME cell types to diminish ER function by limiting ER synthesis and protein levels. Importantly, the engagement of multiple transcriptional mechanisms by other cell types in the TME to repress ESR1 highlights the importance of receptor synthesis as a vulnerability that is exploited as a general mechanism to diminish ER control in BC. This vulnerability is overcome by ESR1 mutations that render the receptor constitutively active.

Previous studies have shown that cellular ER levels are tightly controlled through multiple mechanisms that limit receptor activity (36, 37, 39). ER protein can be induced to be degraded by the 26S proteasome [reviewed in (39)]. Huang et al. (47) reported that HS5-CM can increase ER ubiquitination and reduce ER protein half-life. In contrast, our data indicate that loss of ER under our conditions does not involve decreases in ER protein half-life. We observed that the loss of ER protein occurred subsequent to decreases in ESR1 mRNA. We also showed that BMSC-CM downregulates both Y537S/D538G ER protein and ESR1 mRNA, reinforcing the notion that the regulation is occurring at the level of the ESR1 gene rather than the protein. The observed discrepancy in findings is likely a reflection of differences cell culture conditions, as our studies were performed in serum-free conditions. Others have shown that ESR1 mRNA can also be controlled posttranscriptionally via a miRNA-mediated reduction in ESR1 half-life (37, 38). Studies by Shah et al. (10) identified miR-221/222 in exosomes released from BC-activated fibroblasts, which were previously shown to downregulate ER expression upon overexpression (48). Our studies showed that ESR1 mRNA stability was comparable to previously reported half-lives of ESR1. These data indicate that BMSCs regulate ESR1 mRNA via a distinct mechanism than BC-associated fibroblasts. Despite these differences, these data cumulatively indicate that a host of factors derived from the TME repress ESR1 expression, utilizing multiple mechanisms with the goal of reducing ER control of BC behavior.

Further substantiating the notion that distinct mechanisms are used by the TME to regulate ESR1, we found that downregulation of ESR1 mRNA and ER protein by BMSCs is independent of ERK/MAPK activation. The involvement of the ERK/MAPK signaling pathway in the regulation of ER expression is well documented. Constitutive activation of MAPK via Raf signaling resulted in a significant loss of ER protein (49). Similarly, hyperactivation of MAPK via constitutively active HER2 results in transcriptional repression of ESR1 by histone deacetylases (50). Stossi et al. (11) previously showed that proinflammatory macrophages can repress ESR1 expression in an AP-1–dependent manner. Similarly, we observed that BMSC-CM treatment results in ERK/MAPK activation and induction of AP-1 transcriptional activity (data not shown). However, inhibition of MEK with UO126 did not interfere with BMSC-mediated repression of ER. These data are consistent with reports indicating that BMSCs induce multiple signaling pathways, and it is likely that the collective activity of multiple factors is necessary to repress ESR1 (47).

Our study shows that transcriptional repression of ESR1 is the primary mechanism contributing to the decrease in both ESR1 mRNA and ER protein, and the consequence of decreased ESR1 mRNA synthesis results in overall decreases in ER protein and activity. We observed a decrease in nascent ESR1 mRNA levels within 2 hours of exposure to HS5-CM. We detected rapid loss of Pol II occupancy in the ESR1 proximal promoter and ESR1 ENH1, which were previously shown to play a direct role in regulating ESR1 expression (40, 41, 51). Furthermore, we observed a decrease in H3K27Ac on ESR1 ENH1, implicating a significant role of this region in regulating ESR1 expression by BMSCs. Consistent with this notion is the finding that ESR1 and its neighboring genes are coregulated, specifically corepression of ESR1 with ARMT1 and SYNE1 in cells treated with CM from cancer-associated BMSCs and HS5 cells. Studies by Bailey et al. (40) and Dunning et al. (52) noted that coregulation of these genes was associated with regulatory regions that contained single-nucleotide polymorphisms that are positively correlated with BC risk. Furthermore, deletion of a region overlapping ESR1 ENH1 caused repression of ESR1 with RMND1, ARMT1, and CCDC170 (40). Our laboratory has also shown that this enhancer is involved in transcriptional repression of ESR1 with E2 and bortezomib treatment. Given the loss of an activated chromatin mark on ENH1, we speculate that BMSCs drive transcriptional repression of ESR1 by inactivation of the distal enhancer ENH1.

The ER activation pathway involves a series of events that include phosphorylation of the receptor, ER–DNA binding, recruitment of coregulators, and basal transcriptional machinery. When assessing the effects of ER downregulation on ER activity, ER downregulation has been implicated as a marker for both activation and repression of ER activity (34, 53). Another study observed a direct correlation between ER protein levels and ER transcriptional activity (13). With HS5-CM, we observed increases in pS118-ER, which is required for maximal ER transcriptional activity and is associated with active enhancer regions (29). However, at known ER binding sites derived from an E2-stimulated ER ChIP sequencing data set, no significant change in ER occupancy was detected following HS5-CM treatment. This is consistent with the finding that phosphorylation is not required for ER to bind DNA (29). BMSCs also could not activate an ERE-driven reporter and appeared to repress ER transcription based on decreases in PGR, GREB1, and SNAI1 expression. These data are reminiscent of the effects of ER antagonists, including tamoxifen and fulvestrant, which likewise induce pS118-ER but repress ER function (45, 54, 55).

In Y537S and D538G ESR1 mutant cell lines, BMSC-CM was not sufficient to overcome the constitutive activity of the receptor. Both mutations have been found to mediate resistance to ER antagonists in both metastatic ER-positive BC patients as well as in in vitro studies (15, 56, 57). Our results implicate that these ESR1 mutants can overcome the repressive effects of the TME on ER expression levels to affect BC cell behavior. TFF1 was one ER target gene that was unaffected by HS5-CM in all ESR1 mutant cells. TFF1 has been shown to be regulated by growth factors in MCF7 cells that required de novo protein synthesis of a mediator protein (58). Given no detectable ER occupancy on the TFF1 promoter and no effects on Y537S or D538G ER activation of TFF1 with BMSC-CM, it is likely TFF1 is regulated independently of ER in this system. These data show that BMSCs have a repressive effect on ER transcriptional activity, but this repression is limited to those cells expressing wild-type ER.

Multiple platforms have been used to study paracrine interactions between different cells within the TME. A study in our laboratory showed that direct coculture of HS5 cells with ER-positive BC cells decreased ER expression and increased cell growth. This is consistent with work by Brechbuhl et al. (59), which showed that MCF7 tumors mixed with HS5 cells in mice and direct coculture of both cells in vitro resulted in tamoxifen-resistant tumor and cell growth, respectively. In contrast, Huang et al. (47) treated MCF7 cells with HS5-CM and observed a decrease in cell growth. These contrasting outcomes could be explained by different methodologies. Direct coculture studies allow for analyzing the effects of bidirectional signaling between cancer cells and the stromal compartment, and it is possible that HS5-mediated cell growth requires communication between the two compartments. Several coculture studies observed BC cell growth using direct or indirect coculture methods that promote bidirectional signaling (12, 60, 61). The goal of this study was to understand the mechanistic underpinnings by which BMSCs regulate ER expression and activity. Therefore, we investigated the effects of unidirectional signaling from BMSCs using CM, which limits the analysis to signals directly emanating from BMSCs and, as such, permits mechanistic analysis. To further ensure that downstream effects are strictly derived from BMSCs, we used serum-free conditions for our experiments. A variety of factors potentially found in serum can downregulate ER expression and/or induce ER activation, which are eliminated in serum-free conditions (34, 35, 62, 63). Together with coculture studies that show increased cell growth and hormone therapy resistance, we posit that the transcriptional repression of ESR1 is a direct action of BMSCs on BC cells whereas other endpoints (e.g., cell growth) require more complex cell–cell interactions.

Clinical data indicate that most recurrent/metastatic diseases maintain an ER-positive status, indicating selective pressure for the continued expression of ER. Owing to the highly integrated role of ER in regulating cell growth and survival, it is hypothesized that a reduction of ER expression diminishes hormone-dependent control in BC cells in favor of TME-driven mechanisms. This would promote ER-independent mechanisms, such as signaling pathways, as primary regulators of BC cell behavior (59). Reversing TME-driven ER repression may offer the opportunity to restore hormone dependency and responsiveness to anti-ER targeting drugs. However, our data indicate that such a strategy would only be effective in cells expressing wild-type ER, as mutant receptors remained constitutively active in the presence of BMSC-CM. The shift toward ER-independent growth that occurs during BC progression involves both intrinsic and extrinsic mechanisms, and identifying mechanisms that affect ER control will aid in improving the prioritization of targeted therapies.

Glossary

Abbreviations:

ActD

actinomycin D

BC

breast cancer

BMSC

bone marrow stromal cell

CAF

cancer-associated fibroblast

ChIP

chromatin immunoprecipitation

CHX

cycloheximide

CM

conditioned media

Ct

cycle threshold

DAPI

4′,6-diamidino-2-phenylindole

E2

17β-estradiol

ENH1

ESR1 distal enhancer

ER

estrogen receptor α

ERE

estrogen response element

FBS

fetal bovine serum

NEAA

nonessential amino acid

p-ERK

phosphorylated ERK

Pol II

RNA polymerase II

PS

penicillin-streptomycin

pS118-ER

ER phosphorylation at serine 118

qRT-PCR

quantitative RT-PCR

SS media

stripped serum media

TME

tumor microenvironment

β-gal

β-galactosidase

Additional Information

Disclosure Summary: The authors have nothing to disclose.

Data Availability:

All data generated or analyzed during this study are included in this published article or in the data repositories listed in References.