Interrogating Estrogen Signaling Pathways in Human ER-Positive Breast Cancer Cells Forming Bone Metastases in Mice

Cell Lines

American Type Culture Collection (ATCC, Manassas, VA) human breast cancer cell lines (MCF-7, T47D, and ZR-75-1), were cultured in E₂-replete Dulbecco's modified Eagle's medium (DMEM; Invitrogen, Carlsbad, CA) or RPMI-1640 (Invitrogen), as per ATCC's recommendation, containing 10% heat-inactivated fetal bovine serum (Atlanta Biologicals, Flowery Branch, GA), and 1% penicillin/streptomycin (Thermo Fisher, Waltham, MA) in 37 °C and 5% CO₂ in a humidified atmosphere. Bone tropic ER-negative MDA-MB-231 (MDA-SA) cells were generously provided by Dr. Theresa Guise (24, 46-48). BMET-derived ER+ human tumor cell lines (38-2M and 43-4M) were isolated from in vivo BMET in E₂-supplemented mice inoculated (intracardiac, IC) with ER+ MCF-7-derived tumor cells, as previously described (28). Two MCF-7-derived clonal lines encoding full length ERα proteins (and 46 kD variants) harboring 1 of 2 activating ESR1 mutations (MCF-7-Y537S CL3, MCF-7-D538G CL4) were generated from MCF-7luc cells (Cambridge Bioscience, Cambridge, UK) as previously described using CRISPR-Cas9 genome editing (49). Authentication of all human tumor cell lines were verified as previously described (47, 49).

Western Blot Analysis

Proteins were isolated from whole cell lysates of cells propagated in normal media, as described above; quantified by Bradford assay (Bio-Rad, Hercules, CA); and analyzed by Western blots with confirmation of even protein loading as previously described (28, 50). Blots were probed with primary rabbit antibodies against ERα (#8644, Cell Signaling Technology [CST], Danvers, MA; RRID:AB_2617128) followed by HRP-conjugated anti-rabbit secondary antibody (#7074, CST; RRID:AB_2099233) (28); or with primary mouse antibody against ERβ (#GTX70182, GeneTex, Irvine, CA; AB_370375) validated in doxycycline-inducible MDA-MB-231-ERβ cells (51), followed by HRP-conjugated anti-mouse secondary antibody (#7076, CST; RRID:AB_330924); or with primary goat antibody against GPER (#AF5534, R&D Systems, Minneapolis, MN; AB_2112482) followed by HRP-conjugated anti-goat secondary antibody (#R21459, Life Technologies, Carlsbad, CA; RRID:AB_2556529). Blots were visualized by chemiluminescence of SuperSignal West Femto ECL substrate (ThermoFisher). An antibody against β-actin (#4967, CST; RRID:AB_330288) was used as a loading control. Most blots are representative of ≥ 3 separate experiments.

Parathyroid Hormone–Related Protein Assay

For analysis of the tumor-secreted osteolytic factor PTHrP, cells were plated in 24-well tissue culture plates at a density of 1.3 × 10⁵ cells/well, maintained in E₂-deplete media (phenol red-free DMEM [Invitrogen], 10% charcoal-stripped fetal bovine serum [Valley Biomedical, Winchester, VA], 1% penicillin/streptomycin [Thermo Fisher], and 200 mM L-glutamine [Sigma-Aldrich, St. Louis, MO]) for 4 days (2 days for ER-negative MDA-SA cells) followed by treatments with TGFβ (5 ng/mL), E₂ (10⁻⁷ or 10⁻⁸ M, as indicated) and/or selective estrogen receptor-specific agonists, dosed according to their relative binding affinity (Fig. 1), for 24, 52, or 72 hours, as indicated, to optimize detection, depending on cell line. Cell numbers remained statistically unchanged from day 0 for each cell line, and between cell lines, after 4 days in E₂-deplete media. To assess the effects of various inhibitors on PTHrP secretion, cells were treated for 1 hour prior to addition of indicated ER agonists and/or TGFβ with nystatin (50 ug/mL; Alfa Aesar, Haverhill, MA) (52), rapamycin (1 nM; CST), or a p38 inhibitor SB202190 (10 μM; Selleckchem, Houston, TX) (52). Conditioned media, stored at −80 °C after addition of protease inhibitors (Sigma-Aldrich) was assayed for secreted PTHrP using a commercial immunoradiometric assay (DSL-8100, Beckman Coulter, Brea, CA: AB_3086791). A lack of treatment effect on cell number during the incubation periods was verified using a commercial MTT assay (ATCC).

Animal Studies

All animal protocols were approved by the Institutional Animal Care and Use Committee at The University of Arizona in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Female Foxn1^nu athymic nude outbred mice (Envigo, Indianapolis, IN) were housed in plastic cages in laminar flow isolated hoods with access to water and autoclaved mouse chow ad libitum (NIH-31 Modified diet, Envigo) (24, 25, 28, 47, 48). For one set of BMET experiments, 5-week-old mice (n = 10-13/group) were inoculated (1 × 10⁵ cells) via the left cardiac ventricle (IC) as previously described (24, 28, 48) with either: (i) human ER-negative MDA-MB-231 cells without estrogen supplementation; or (ii) individual ER+ human breast cancer cell lines (ATCC MCF-7, T47D, or ZR-75-1; or BMET-derived MCF-7 cells [38-2M, 43-4M]), 3 days post placement of 60-day extended release 17β-estradiol (E₂) pellets (0.05, 0.10, 0.36, or 0.72 mg, as indicated [Innovative Research of America, Sarasota, FL]) (24, 28).

In another set of experiments, 5-week-old mice were IC inoculated with MCF-7 BMET-derived ER+ 43-4M cells (1 × 10⁵), with start of E₂ and/or E₄ treatments 4 days prior to inoculation (28). Treatment groups (n = 7-12/group) included: (i) E₂ alone (0.72 mg 60-day extended release pellets [Innovative Research of America, Sarasota, FL]); (ii) E₂ pellets plus vehicle (100 μL peanut oil/d; #P2144, Sigma-Aldrich); (iii) E₄ (4 mg/kg/d) (#SML1523, Sigma-Aldrich; in peanut oil); or (iv) E₂ pellet and E₄ (4 mg/kg/d), with daily dosing of E₄ or vehicle via gavage for the first 7 days, followed by a 5x/week regimen without consecutive nontreatment days until end of experiment. The daily oral 4 mg/kg E₄ dosing strategy using peanut oil as a vehicle was the same as that used by Gerard et al to demonstrate in vivo stimulatory effects of 1 to 10 mg/kg/d E₄ on orthotopic MCF-7 tumor growth (and on normal mammary epithelium) (53) when used in isolation, and partial inhibition of E₂ driven growth when used in combination with E₄ (41). E₂ pellet dose used here (0.72 mg pellet), which yielded a maximal induction of ER + BMET in the E₂ dose-response experiments, was similar to that used by Gerard et al to support in vivo MCF-7 orthotopic tumor growth (1.7 mg pellet) (41). Because all reported outcomes for E₂ alone were unchanged by addition of vehicle, data for these groups were combined for analyses. In a third set of experiments, 4- to 5-week-old mice were IC inoculated (4 × 10⁵ cells) with either: (i) a 1:1 mix of MCF-7 clones harboring clinically common activating ESR1 mutations (MCF-7-Y537S CL3 and MCF-7-D538G CL4) without estrogen supplementation; or (ii) “parental” ESR1-wild-type MCF-7luc cells with, or without, E₂ treatment (0.72 mg 60-day pellets) beginning 4 days prior to inoculation, as indicated. At the end of all BMET experiments, mice were euthanized, hind limbs isolated for histologic analyses, and, in some cases (E₂/E₄ experiment), the fourth and ninth inguinal mammary glands were collected (54), and organ weights (eg, uteri) were recorded.

To assess bone dissemination, additional mice (vs untreated controls; n = 3-5/group) were either pelleted with 0.72 mg E_2, or treated by daily oral administration with 4 mg/kg E₄ for 3 days prior to IC inoculation with 8 × 10⁵ 43-4M cells freshly labeled with Vybrant DiD (Thermo Fisher), as per the manufacturer's instructions, with fluorescent staining remaining detectable for up to 7 days in culture, as previously described (24, 28). Daily E₄ administration continued until 3 days postinoculation when mice were euthanized and cells were isolated from the proximal (25%) tibia and imaged for enumeration of DiD-positive cells using the Cy5 filter of a Keyence BZ-X700 fluorescent microscope (Keyence Corporation of America, Itasca, IL), as previously described (24, 28). Results are reported as DiD+ cells/10⁶ bone marrow cells for each tibia. Similarly isolated bone marrow cells from tumor-naïve mice (n = 3) were included as negative controls and cultured 43-4M cells 3 days post DiD-labeling were used as positive controls.

Bone Metastatic Breast Cancer Tumor and Mammary Gland Histology

Hind limbs were removed at end of experiments, fixed, decalcified, and paraffin-embedded for histologic analyses of midsagittal (depth of 400-500 μM) sections (5-6 μM thick), as previously described (24, 28). Sections from age-matched tumor-free mice were stained with hematoxylin and eosin (H&E) to assess effects of E₄ vs E₂ (vs control) on bone histology and osteoblast numbers (24, 25, 28). Hematoxylin-stained mononuclear cuboidal osteoblasts lining trabecular bone surfaces and tartrate-resistant acid phosphatase (TRAP)-stained multi-nucleated osteoclasts were identified in a blinded fashion 0.25 mm below the proximal tibial growth plate and reported as cell number per mm of bone surface (BS), as previously described (24, 25, 28, 55, 56). Epithelial ER+ breast cancer tumors were immunohistochemically identified at end of all experiments and their size measured in hind leg bones using primary antibodies to pan-cytokeratin (#Z0622, Agilent Dako, Santa Clara, CA; AB_2650434), as previously described (24, 28). Of note, prior immunohistochemical experiments have confirmed the continued expression of ERa in the bone metastatic cells (24, 28). The fourth and ninth inguinal mammary glands were fixed, defatted, and stained with carmine alum as previously described (54) for examination of ductal area and terminal end buds (TEB) in mice treated with E₂ and/or E₄ vs controls.

Bone Imaging

Radiographs of mouse hind limbs (Faxitron UltraFocus 1000, Faxitron Bioptics, Tucson, AZ) in ER+ or ER-negative breast cancer cell-inoculated mice were obtained weekly to assess osteolytic lesion formation, which was analyzed as previously described in a blinded fashion by 3 independent investigators using ImageJ software (NIH) (24, 28, 47), with osteolytic BMET incidence and/or total hind limb radiographic lesion area reported per mouse, including animals without osteolytic lesions. Of note, no osteoblastic lesions were detected. Because E₂ can induce osteolytic osteosarcoma formation in nude mice (25), osteolytic breast cancer ER+ BMETs in each hind limb bone were verified by correlating radiographic lesions with cytokeratin-positive human tumors at end of experiments (24, 57). Areal bone mineral density (aBMD) of the proximal femur, a tumor-free area, was assessed by dual-energy x-ray absorptiometry (DXA; Faxitron UltraFocus 1000) in age-matched mouse hind limbs (with or without [naïve control] inoculated tumor cells) at end of experiments.

Statistical Analysis

Data are reported as mean ± SEM, with statistical significance of two-sided P values defined as P ≤ .05. Statistical differences were determined using Prism 8.0 software (Graphpad, San Diego, CA) for t tests, one- or two-way analyses of variance (ANOVA) with post hoc testing (as indicated), and tests for log-rank.

Bone Tropic ER+ Human Breast Cancer Cells Express ERα, ERβ, and GPER

Western blot analysis was used to interrogate estrogen receptor expression in “bone tropic” human breast cancer cell lines that secrete E₂ (if ER+)-inducible and/or TGFß-inducible osteolytic PTHrP and form osteolytic BMET in vivo that are TGFß- and estrogen-dependent (ER+ MCF-7, or MCF-7 BMET-derived 43-4M, 38-2M cells) or estrogen-independent (ER-MDA-MB-231), as compared to “non–bone tropic” ER+ cell lines that do neither (T47D, ZR-75-1) (24, 28) (and Supplementary Fig. S1) (58). GPER protein was detected in all ER + cell lines, independent of bone tropic behavior, but not in bone tropic ER-negative MDA-MB-231 cells (Fig. 2A). ERβ protein, using an antibody validated for Western blot analyses (51), was only readily detected in ER+ cell lines that were bone tropic (MCF-7, 38-2M, 43-4M), with questionable expression in bone tropic ER-negative MDA-MB-231 cells (Fig. 2B). As expected, ERα protein was identified by Western blot analysis in all ER+ breast cancer cell lines, but not in ER-negative MBC-MB-231 cells (Fig. 2B). However, ERα protein levels were much higher in bone tropic cells, where multiple ERα isoforms, including ERα-46, were also readily detected (Fig. 2B, left panel).

Nuclear ERα Signaling Is Sufficient to Induce Secretion of a Tumoral Mediator of Osteolysis From Bone Tropic ER+ Human Breast Cancer Cells

Having identified all 3 estrogen receptors in bone tropic human ER+ breast cancer cells where E₂ (and TGFß) drive tumor-associated osteolysis in vivo and osteolytic PTHrP secretion in vitro (24, 28) (and Supplementary Fig S1) (58), PTHrP secretion (in control or TGFß stimulated ER+ bone tropic cells) was utilized as an in vitro screening bioassay for estrogen receptor-mediated osteolytic potential, querying a role for ERα vs other estrogen receptors by testing effects of receptor- and signaling pathway–specific ligands using affinity-normalized dosing (Fig. 1). Isolated treatment of 2 different BMET-derived MCF-7 cell lines (38-2M and 43-4M) with the ERα agonist propyl pyrazole triol (PPT), or nuclear ER agonist (and MISS antagonist) estetrol (E₄), induced PTHrP secretion 2-fold over the media control, similar to levels induced by E₂ (Fig. 3A). In contrast, neither an ERβ agonist (R-diarylpropionitrile, R-DPN), isolated nongenomic ER MISS agonist (pathway preferential estrogen, PaPE-1), nor GPER agonist (G1) significantly altered PTHrP relative to untreated controls (Fig. 3A). Analogous to E₂, ERα agonist PPT and nuclear-only ER agonist E₄ each also exhibited additive PTHrP secretory effects with TGFβ (Fig. 3B), while agonists for the other receptors (ERβ, ER MISS, and GPER) did not. When agonists lacking isolated effects were combined with agonists sufficient to induce PTHrP expression, no additional synergies were identified; combining either ERβ agonist R-DPN (Fig. 4A) or GPER agonist G1 (Fig. 4B) with ERα agonist PPT did not alter PPT-induced PTHrP secretion in control or TGFß-stimulated cells. Agonists for ERβ (R-DPN) or GPER (G1), which individually were without effect, also did not induce PTHrP secretion when combined in control or TGFβ-stimulated cells (data not shown). Stimulatory effects of the ER nuclear agonist/MISS antagonist E₄ on PTHrP were not altered when combined with GPER agonist G1 (Fig. 4C), while in TGFβ-stimulated cells, PTHrP secretion was significantly reduced by G1 addition, as compared to E₄ alone. In toto, these data suggest that in ER+ breast cancer cells forming E₂ and TGFß-dependent osteolytic BMET in vivo (24, 28), nuclear effects of cytoplasmic ERα are sufficient to mediate both E₂-inducible PTHrP secretion and additive effects with TGFβ, without involvement of other estrogen receptors or nongenomic ERα MISS.

Other ERα Cell Signaling Pathways Contribute to Inducible PTHrP Secretion From Bone Tropic ER+ Human Breast Cancer Cells

While nuclear/genomic ERα signaling appeared sufficient for PTHrP induction, additional studies were conducted to query possible interactions between genomic vs nongenomic ERα signaling pathways, and involvement of additional intracellular signaling pathways known to alter ERα signal transduction. As E₄ has been reported to antagonize E₂ stimulated ERα MISS in MCF-7 cells (40), effects of E₄ (10⁻⁶ M) and E₂ (10⁻⁸ M) in combination vs isolation were assessed using comparable maximal doses (10⁻⁶ M vs 10⁻⁸ M, respectively), consistent with the 100-fold lower affinity of E₄ (vs E₂) for ERα (Fig. 1) (40), and their differential dose-dependent effects in stimulating PTHrP secretion from 43-4M cells (Fig. 5A). Each induced constitutive and TGFβ-stimulated PTHrP secretion to a similar degree without any evidence of an inhibitory effect when used in combination (Fig. 5B), which would have been suggestive of E₄ antagonism of an E₂-mediated stimulatory ERα MISS effect on PTHrP secretion. Instead, PTHrP levels were higher with combined E₂ and E₄ treatment, reaching statistical significance in TGFβ-stimulated cells (Fig. 5B), suggestive of possible additive effects of E₂/E₄ nuclear ERα signaling (potentially less likely given maximal doses used) and/or E₄ blockade of an E₂-mediated inhibitory ERα MISS effect when combined with TGFß. Pretreatment with rapamycin, an inhibitor of E₂-inducible mTOR signaling in 43-4M cells (28) that is reported to facilitate both genomic and nongenomic ERα signaling (59, 60), completely blocked inducible PTHrP secretion in response to E₄ or E₂ alone, and significantly inhibited their combined effects (Fig. 5C). Inhibition of p38 MAPK, an E₂-activated signaling pathway in 43-4M cells (28, 61, 62) that is possibly ER MISS-regulated (59, 63) and can also enhance nuclear ER transcriptional activity (via ER phosphorylation and/or stimulation of nuclear ER coactivators) (64, 65), partially blocked PTHrP induction by both E₂ and by E₄ (Fig. 5D). Because membrane ERα signaling involves receptor localization in lipid rafts (61, 62), as does TGFß-induced MAPK signaling, which is also induced in 43-4M cells (28), the lipid-raft disruptor nystatin was also examined (52), causing a partial (∼50%) reduction in E₂- and TGFβ-induced PTHrP secretion, alone or in combination (Fig. 5E). In toto, PTHrP screening bioassay data were thus consistent with a necessary and sufficient role for nuclear ERα signaling, possibly involving mTOR and p38 MAPK activation, in stimulating osteolytic PTHrP secretion from bone tropic ER+ breast cancer cells in response to E₂ or E₄. However, the ability of ERα MISS to alter stimulatory ERα nuclear effects in complex ways could not be ruled out, as evidenced by: 1) the partial suppressive effects of lipid-raft disruption on E₂ signaling; 2) the stimulatory effects of an ER MISS antagonist, E₄, when combined with E₂; and 3) the inhibitory effects of an agonist (G1) for GPER (which is reported to interact with membrane ERα) (41) only when combined with E₄, with all of the above being most notable in the context of TGFβ signaling, which also likely involves caveolin-enriched lipid rafts (66).

ER MISS Is Required to Support Estrogen-dependent ER+ BMET Progression In Vivo

Comparable effects of E₄ (to E₂) in supporting MCF-7 orthotopic tumor growth in vivo, as reported by Gerard et al (41), and stimulating osteolytic PTHrP secretion (alone or in combination with TGFß), as demonstrated here, suggested that pharmacologic activation of genomic ER signaling alone by E₄ could also be sufficient to support estrogen-dependent osteolytic human ER+ tumoral progression within bone. To test this hypothesis, effects of E₂ vs E₄ were compared in the ER+ BMET model. As previously described (24, 28), osteolytic BMET formed robustly in 43-4M cell-inoculated mice treated with E₂ in terms of both osteolytic lesion incidence (Fig. 6A) and size (Fig. 6B), forming bone-destructive tumors with TRAP+ osteoclasts at leading edges (Fig. 7A-7C). However, most notably and contrary to anticipated outcomes, in mice treated with an E₄ dose (4 mg/kg) within the range (1-10 mg/kg) previously reported to support orthotopic MCF-7 tumor progression in vivo (41, 42), there was no radiological evidence of osteolytic BMET formation (Fig. 6A-6B). Histologically, while cytokeratin-positive tumor cells were detected in the hind limbs of 43% (3/7) of E₄-treated animals (Fig. 6C), histologic lesions were quite small (Fig. 6D) and, while bone-adjacent (Fig. 7D-7F), were without evidence of associated osteolytic activity. A difference in bone dissemination as a potential driver of reduced osteolytic BMET in E₄-treated mice was ruled out by detection of similar number of bone-disseminated DiD-labeled 43-4M cells in E₂- vs E₄-treated (or untreated) mice within 3 days of tumor cell inoculation (Supplementary Fig. S2) (58).

Reduced effects of E₄, a nuclear ER-specific agonist (vs E₂), suggested nongenomic ER MISS was important for estrogen-dependent ER+ BMET progression; thus, combined effects of E₂ with E₄, which can antagonize E₂-stimulated ER MISS, were examined. Consistent with a possible stimulatory role of ER MISS in E₂-treated mice, the incidence (Fig. 6A) and size (Fig. 6B) of osteolytic BMET lesions developing in 43-4M cell-inoculated mice treated with E₂ in combination with E₄ were lower than in E₂-only, as were the incidence (Fig. 6C) and size (Fig. 6D) of cytokeratin-positive histologic tumors in bone; however, these trends did not reach statistical significance.

Sufficiency of In Vivo E₄ Dosing for Mediating Pharmacologic Effects

To determine whether the minimal effect of E₄ on human ER+ osteolytic BMET progression was due to insufficient steroid bioavailability or dosing, pharmacologic effects of E₄ (+/− E₂) on estrogen-dependent tissues were examined in the tumor-inoculated mice. Consistent with anabolic bone effects of E₂ in the ER+ BMET model that have been well documented by our laboratory and are identical even with > 10-fold lower E₂ dosing (24, 25), BMD in tumor-free proximal femurs of E₂-treated mice was significantly increased (vs age-matched control mice, Fig. 8A). In E₄-treated mice, BMD also significantly increased, but to a lesser degree (+8% [E₄] vs +18% [E₂], P < .01) (Fig. 8A). In mice treated with E₂ in combination with E₄, increased BMD in tumor-free proximal femurs was not statistically different than E₂ alone (Fig. 8A). Differential pharmacologic effects of E₄ vs E₂ were also seen in uterine weights, which tended to increase slightly in E₂-treated mice (+/− E₄), while being significantly reduced by E₄ treatment alone in these ovary-intact mice (Fig. 8B). Mammary gland ductal branching and alveoli development in also appeared qualitatively reduced in the presence of E₄ (Supplementary Fig. S3) (58), alone or in combination with E₂ treatment, as compared to naïve or E₂-treated mice, consistent with prior reports (41).

Anabolic Effects of ER Signaling in the Tumor Microenvironment Are Not Required for ER+ Human Breast Cancer Progression in Bone

In normal age-matched mice treated for 6 weeks with the same E₂ dose as in the BMET model, histologic assessment of the proximal tibiae, a common site of human ER-negative or ER+ breast cancer BMET formation in mice (14, 24, 28, 47), reflected a marked increase in bone with (vs age-matched control mice), as has been previously described (25), that was not recapitulated in mice treated with 4 mg/kg/d E₄ (Fig. 9A), where changes were modest (Fig. 9A). Osteoblast (Ob) numbers in the tibial metaphysis doubled in response to E₂ treatment but remained unchanged in E₄-treated mice (Fig. 9B), consistent with prior reports of a possible role for ER MISS in mediating estrogen effects on bone-forming osteoblasts (44). Osteoclast (Oc) numbers on trabecular bone surfaces were unchanged by either treatment (Fig. 9C) in these ovary-intact mice, suggesting that ER MISS effects on bone anabolism could be key drivers of the more marked E₂ (vs E₄) effects on the tumor microenvironment.

Bone effects of E₂ across a range of doses supporting ER+ human breast cancer BMET progression in mice are uniform (not dose-dependent) (24, 25) and can increase ER-negative BMET size (but not incidence) (24), suggesting that bone effects of estrogen are tumor-promoting. Given the less marked bone effects of E₄ (vs E₂), an experiment was undertaken to determine whether isolated ERα signaling in tumor (and not bone) could still support ER+ BMET progression. BMET formation in mice inoculated with ER+ MCF-7 cells with ESR1 activating mutations vs ESR1 wild-type cells was examined in the absence of E₂ supplementation. When E₂ supplementation was not given (and associated bone effects were absent), MCF-7 cells harboring activating ESR1 mutations formed BMET (with an incidence similar to ESR1 wild-type cells in mice supplemented with E₂), while ESR1 wild-type MCF-7 cells did not (Supplementary Fig. S4) (58).