Estradiol Promotes Breast Cancer Cell Migration via Recruitment and Activation of Neutrophils

Breast cancer is the most common type of cancer affecting women in the Western world, and metastasis is the main cause of death among breast cancer patients (1). Estrogen exposure plays a key role in breast cancer initiation and growth (2, 3). Two of three breast cancers express the estrogen receptor (ER), and antiestrogen therapy is a cornerstone of the medical treatment of these patients (4). However, this long-term therapy only reduces the risk of recurrence by 30% to 50% (5). Novel therapeutic targets of estrogen-dependent breast cancer progression are therefore warranted.

The importance of immune cells in tumor growth and metastasis has been described in numerous cancer types, including breast cancer (6, 7). Several types of immune cells express the ER, and estrogen affects the expression of inflammatory mediators in neutrophils and macrophages (8, 9). We have shown that estrogen increases the influx of macrophages into breast cancers and induces a protumorigenic phenotype (M2) in these cells (10). Neutrophils also play an important role in tumor growth and metastasis (7, 11). Neutrophils represent 40% to 75% of all white blood cells, and, similar to macrophages, neutrophils adopt a protumorigenic phenotype (N2) or antitumorigenic phenotype (N1) in response to various cytokines (11–13).

TGFβ1 is one of the most potent chemoattractants for neutrophils (11). TGFβ1 induces recruitment and N2 polarization of neutrophils, and increases the number of circulating low-density N2 neutrophils in several types of cancer, including breast cancer (11, 14, 15). TGFβ1 is a multifunctional cytokine that has dual roles in tumorigenesis: in normal tissues and early stages of cancer, it arrests cell proliferation and induces apoptotic pathways, whereas in later stages, it promotes angiogenesis, growth, and tumor progression (16, 17). Cross-talk between TGFβ1 and estradiol (E₂) in breast cancer has been reported; for instance, TGFβ1 and E₂ together enrich cancer stem cell populations in breast tumors, leading to increased migration and drug and radiation resistance (18). Estradiol also increases TGFβ1 secretion in experimental breast cancers both in vitro and in vivo (19). Cell–cell interactions in tumor microenvironments can be enhanced by integrins, which play a major role in cancer cell proliferation and metastasis (20, 21). Expression of the lymphocyte function-associated antigen 1 integrin (LFA-1, α_L/β₂, CD11a/CD18) in neutrophils promotes intra- and extravasation and transendothelial migration (22, 23). As LFA-1 is involved in the first steps of cell adhesion, and early steps of cancer cell dissemination would be efficient to target therapeutically, we focused our investigation on this integrin (24).

It is clear that E₂, TGFβ1, and neutrophils are important mediators of cancer progression and metastasis. However, the relationships among these three elements in the breast cancer metastasis process remain unaddressed. In the present study, E₂ increased the recruitment of neutrophils to the tumor-invasive margin in vivo in two different mouse breast cancer models. In addition, E₂ treatment promoted N2 polarization of neutrophils by inducing overexpression of the integrin LFA-1. These results were confirmed in vitro using mammosphere cultures. In a zebrafish metastasis model, nonmetastatic ER⁺ breast cancer cells became metastatic in the presence of neutrophils, and E₂ exposure further increased the dissemination of cancer cells via TGFβ1 and LFA-1 overexpression. Furthermore, in samples from breast cancer patients, extracellular levels of TGFβ1 were significantly increased, suggesting that this factor is a relevant target in human breast cancer.

Our data provide new insights into E₂-dependent mechanisms of breast cancer progression. In addition, we also describe a novel low-cost protocol for paraffin embedding of mammospheres, which can be used to investigate molecular mechanisms and therapeutic targets.

The regional ethical review board of Linköping approved the study, which was performed in accordance with the Declaration of Helsinki. All subjects gave written informed consent. Women diagnosed with breast cancer (n = 12) underwent microdialysis before surgery. Characteristics of the cancers are shown in Supplementary Table S1. Before catheter insertion, 0.5 mL lidocaine (10 mg/mL) was administrated intracutaneously. One microdialysis catheter was inserted intratumorally into the breast cancer and another into adjacent normal breast tissue as previously described (M Dialysis; 71 cat. #P000127, 10-mm; refs. 25, 26). Catheters, connected to a pump (CMA 107; CMA Microdialysis AB), were perfused with 154 mmol/liter NaCl and 60 g/liter hydroxyethyl starch (Voluven; Fresenius Kabi AB; cat. #066287) at 0.5 μL/min. After 60-minute equilibration, the perfusate was stored at −70°C.

Mice were housed at Linköping University. Animal care and treatment conformed to regulatory standards. The institutional animal ethics committee at Linköping University approved this study. Female athymic mice (BALB/c-nu/nu) and FVB/N mice (6–8 weeks; Scanbur) were housed in ventilated cages with a light/dark cycle of 12/12 hours, with rodent chow and water ad libitum. Mice were anesthetized by intraperitoneal injection of ketamine/xylazine and oophorectomized prior to s.c. implantation with 3-mm pellets containing 17β-estradiol (0.18 mg/60-day release, cat. #NE-121; Innovative Research of America) that produce serum concentrations of 150 to 250 pmol/L or placebo (27).

One week after surgery, MCF-7 cells, 5 × 10⁶ in 200 μL PBS, were injected into the dorsal mammary fat pads. Because MCF-7 cells require estrogen for tumor formation and growth, a nonestrogen control group was not possible in this in vivo model. At similar tumor sizes, mice either continued with estradiol or fulvestrant (5 mg/mouse every 3 days, s.c.; AstraZeneca; cat. #099215), a pure antiestrogen, was added to estradiol treatment.

FVB/N mice were injected in the dorsal mammary fat pad with cells, 1 × 10⁶ in 200 μL PBS, derived from a transgenic mouse strain expressing polyoma middle T (PyMT) antigen under the control of the mouse mammary tumor virus long terminal repeat (28). These mice developed spontaneous adenocarcinomas of mammary epithelium by 8 to 10 weeks of age. Tumors from 10-week-old mice were excised, dissociated in a collagenase/dispase solution (100 mL PBS) with 25 mg collagenase (Sigma; cat. #C7657) and 250 mg dispase (Roche; cat. #04 942 078 001), and cultured until confluence. As shown by us, and others, these breast tumors express the ER during the early stage of cancer and decrease ER expression at later stages (29, 30). Harvesting tumor tissue at early stages results in estrogen-dependent cancer growth in the syngeneic recipient mouse, as previously described (27, 30). Tumors were established in oophorectomized mice with/without estrogen supplementation and with estrogen together with fulvestrant as described above.

ER⁺ cell lines MCF-7 and ZR-75-1, and the ER⁻ cell line MDA-MB-231, were purchased in 2009, 2016, and 2009, respectively, from the American Type Culture Collection (ATCC). Cells, authenticated by the ATCC, were upon receipt immediately expanded and stored in liquid nitrogen. A new aliquot was resuscitated for each experiment and never used for more than 6 months. MCF-7 and MDA-MB-231 were maintained in DMEM medium (Gibco; cat. #11880) and ZR-75-1 cells in DMEM/F12 (Gibco; cat. #11039) with 2 mmol/L glutamine (Gibco; cat. #25030), penicillin-G (50 IU/mL), streptomycin (50 mg/mL; Gibco; cat. #15070), and 10% and 5% FBS, respectively (Gibco; cat. #10270). Human venous neutrophils were freshly isolated using buffy coats from healthy female donor that were immediately processed upon receipt. The buffy coat was diluted 1/3 in cold DPBS 1X (Gibco; cat. #14200-067) with 0.1% heat-inactivated FBS (Invitrogen; cat. #16140-071) and 2 mmol/L EDTA (Invitrogen; cat. #AM9260G) and washed by centrifugation. Peripheral blood mononuclear cells were removed by density gradient centrifugation with Ficoll-Paque (GE Healthcare; cat. #17-1440-02), and most of the red blood cells were removed by density gradient with sterile 3% dextran T-500 (Amersham Pharmacia Biotech AB; cat. # 17-0320-02) in 0.9 % NaCl. The residual red blood cells were completely eliminated by hypotonic lysis. The neutrophils were cultured at 37°C and 5% CO₂ for 45 minutes in DMEM/F12 with 0.02% of BSA (Merck; cat. #1.12018.0025), 10 μg/mL apo-transferrin (Sigma; cat. #T2036), and 1 μg/mL insulin (Sigma; cat. #I5500). For TGFβ1, neutrophils were cultured for 12 hours in DMEM (Gibco) with 5% charcoal-filtered FBS (Gibco; cat. #12676-029) and 2 mmol/L glutamine (Gibco). Treatments were as follows: 10⁻⁹ mol/L E₂ (Sigma; cat. #2758), 10⁻⁶ mol/L fulvestrant (Sigma; cat. #I4409), recombinant human TGFβ1 (200 pg/mL; rhTGFβ1; Peprotech; cat. #100-21), and antibodies (5 μg/mL) to human TGFβ1 (Acris Antibodies; cat. #DM1047), human LFA-1 (Biolegend; cat. #301213), or isotype IgG (Biolegend; cat. #401408).

MDA-MB-231 cells were seeded at 3 × 10³ cells/well in Ultra Low Attachment 96-well plates (Corning Inc.; cat. #7007) in DMEM medium with 2% ECM gel (Sigma; cat. #E1270). MCF-7 were seeded at 1 × 10³ cells/well with 2.5% ECM gel in Mammocult medium (Stem Cell Technologies Inc.; cat. #05620). Plates were centrifuged, 1,000 × g for 5 minutes, and incubated at 37°C with 5% CO₂. Two days before neutrophil infiltration, mammospheres, sized 500 μm in diameter, culture medium was gradually changed to serum-free DMEM/F12 supplemented as above. Note that 1 × 10⁵ neutrophils and 1 × 10⁴ neutrophils were added to MCF-7 and MDA-MB-231 mammospheres, respectively, and treated for 5 days. Medium was collected and mammospheres were fixed in 4% paraformaldehyde (PFA). Fixed mammospheres were stained with Mayer's hematoxylin (Histolab Products AB; cat. #01820) during 5 minutes, washed with PBS, and rehydrated in ethanol (70% 2 × 30 minutes, 90% 2 × 30 minutes, 99.5% 3 × 30 minutes), tissue clear at 2 × 30 minutes (Histolab; cat. #14250), and liquid paraffin at 60°C 3 × 30 minutes in an eppendorf lid. Then, mammospheres were transferred to a warm metal mold at 60°C for 30 minutes, centered with a warm spoon, and cooled at room temperature for about 15 minutes to keep them in place and then placed at 60°C with more liquid paraffin and a plastic mold for another 30 minutes. After incubation, the mold was carefully placed at room temperature for 10 minutes and then in 4°C to solidify the paraffin.

Optimization for neutrophils/mammospheres concentration was performed ranging from 2 × 10⁴ to 1 × 10⁵ per mammosphere. For MCF-7 mammospheres, in E₂ group, the various concentration of infiltrating neutrophils per mammosphere was not significantly different. The experiments were set up using 1 × 10⁵ neutrophils/MCF-7 mammospheres/well because there was a lower variability for this concentration (Supplementary Fig. S2A). However, for MDA-MB-231 mammospheres, 1 × 10⁵ neutrophils/mammospheres/well resulted in structural damaged and reduced growth (Supplementary Fig. S2B); therefore, a concentration of 1 × 10⁴ neutrophils/mammospheres/well was used.

TGFβ1 and intercellular adhesion molecule 1 (ICAM-1) were analyzed with ELISAs (Quantikine TGFβ1 ELISA kit; R&D Systems; cat. #DB100B and ICAM-1 ELISA; BioVision; cat. #K7161-100) according to the manufacturer's instructions.

MDA-MB-231 mammospheres were infiltrated with neutrophils during 3 days. Invasion assay was performed by the provider's instructions (R&D Systems; cat. #3500-096-K). Images were taken using the inverted microscope Axio Vert.A1 with AxioCam ICm1 CCD camera (Zeiss) at days 0 and 5. Invaded area (mm²) was calculated by subtracting the area at day 0 to the total area calculated at day 5 by using ImageJ 1.50i software.

Neutrophils, fixed in 4% PFA, were exposed to anti–human LFA-1 (Biolegend; cat. #301213) and Alexa-Fluor 488 (Abcam; cat. #ab150113). Mouse tumor sections were stained with rat anti-mouse lymphocyte antigen 6 complex locus G6D (Ly6G) 1:400 (BD Pharmingen; cat. #551459), using rat on a mouse HRP Polymer kit (BioCare Medical; #RT517G), counterstained with Mayer's hematoxylin. Negative controls did not stain. For immunofluorescence, sections were stained with rat anti-mouse Ly6G 1:400 (BD Pharmingen), rabbit anti-mouse LFA-1 1:100 (Abcam; cat. #ab203336), anti-rabbit Alexa 546 1:200 (Life Technologies; cat. #A11010), and anti-rat Alexa 488 1:200 (Life Technologies; cat. #A21210), and mounted using SlowFade Gold antifade reagent with DAPI (Life Technologies; cat. #S36938). Images were acquired by Olympus BX43 microscope light/fluorescence microscope, excitation filters BP460-495 and BP530-550, using an Olympus DP72 CCD camera and analyzed using Olympus CellSens Imaging software. Mammospheres, fixed for 30 minutes with 4% PFA at room temperature, were stained with anti-human CD45 1:200 (Biolegend; cat. #9624-01) and anti-mouse Alexa-Fluor 488 1:500 (Abcam; cat. #ab150113), and mounted in SlowFade Gold antifade reagent with DAPI. Images were taken using Zeiss Axio Imager with LSM 700 upright confocal microscope, and the infiltration of neutrophils was analyzed using the ImageJ 1.50i software. Images of MCF-7 mammospheres were deconvolved for better visualization where specified with Huygens compute engine 15.10.1p6 using classic maximum likelihood estimation.

Cytospins of freshly isolated neutrophils were dried and stained with a SNABB-DIFF kit (LABEX Instrument AB; cat. #2115) following the manufacturer's instructions. For immunocytochemistry, cytospins of freshly isolated neutrophils were dried, fixed with cold acetone for 10 minutes at −20°C, and incubated with anti–human LFA-1 at 1:100 (Biolegend; cat. #301213) at 4°C overnight. The MACH Universal HRP-polymer detection system (Histolab; cat. #BC-BRI4012L) and the betazoid DAB Kit (Histolab; cat. #BC-BDB2004H) were used. Slides were mounted with Glycergel (Dako; cat. #C0563), and negative control did not show stain. For immunohistochemistry, MCF-7 and MDA-MB-231 mammospheres were stained with mouse anti-human ICAM-1/CD54 1:50 (Biolegend; cat. #353101), mouse anti-human ICAM-2/CD102 1:100 (Novus Biological; cat. #NBP2-00320), and rabbit anti-human ICAM-3/CD50 1:100 (Sino Biological; cat. #10333-R002-50), and counterstained with Mayer's hematoxylin (Histolab). For immunofluorescence, sections of PyMT mouse tumors were double stained with rat anti-mouse Ly6G 1:400 (BD Pharmingen; cat. #551459) and rabbit anti-mouse F4/80 1:25 (Abcam; cat. #ab111101) at 4°C overnight. Anti-rabbit Alexa 488 1:200 (Abcam; cat. #ab150077) and anti-rat Alexa 546 1:200 (Invitrogen; cat. #A11081) were used as secondary antibodies. Slides were mounted using SlowFade Gold antifade reagent with DAPI (Life Technologies; cat. #S36938). Images were acquired by Olympus BX43 microscope light/fluorescence microscope with excitation filters BP460-495 and BP530-550, using an Olympus DP72 CCD camera, and analyzed using Olympus CellSens Imaging software.

MCF-7 cells were cultivated during 3 days in DMEM (Gibco; cat. #11880) with 10% charcoal-filtered FBS (Gibco; cat. #12676-029), 50 IU/mL Penicillin-G, 50 mg/mL streptomycin (Gibco; cat. #15070), and 2 mmol/L glutamine (Gibco; cat. #25030) ± E₂ 10⁻⁹ mol/L (Sigma; cat. #2758). Conditioned media from MCF-7 cells were used for migration and survival assays. Human neutrophils were freshly isolated as described above and re-suspended in DMEM/F12 with 0.02% of BSA (Merck; cat. #1.12018.0025), 10 μg/mL apo-transferrin (Sigma; cat. #T2036), and 1 μg/mL insulin (Sigma; cat. #I5500). Migration assay was performed with a CytoSelect 96-well cell migration assay kit (Cell Biolabs; cat. #CBA-105) according to the manufacturer´s instructions. Note that 5 × 10⁴ neutrophils/well were placed in the upper chamber ± E₂ 10⁻⁹ mol/L and conditioned media in the lower chamber, and migrated cells were quantified after 24-hour incubation at 37°C by using the Spark 10M multimode microplate reader (Tecan Group Ltd.). For survival assay, 5 × 10⁵ neutrophils/well were cultured in MCF-7 cell–conditioned media ± E₂ in a 96-well plate. Living cells were quantified at 24 and 48 hours of culture with trypan blue viability exclusion and reported as percentage of survival. For retention study, 7 × 10⁵ neutrophils/mL were added to a monolayer culture of MCF-7 cells grown in DMEM (Gibco) with 10% charcoal-filtered FBS (Gibco), Penicillin-G (50 IU/mL), streptomycin (50 mg/mL; Gibco), and 2 mmol/L glutamine (Gibco) ± E₂ 10⁻⁹ mol/L (Sigma). Cocultures were incubated at 37°C, and concentration of neutrophils in culture supernatant was quantified at 24 and 48 hours of culture in a Bürker chamber. Number of retained neutrophils was calculated by subtracting the concentration of neutrophils in culture supernatant to the initial concentration of neutrophils added to MCF-7 cell cultures.

The institutional animal ethics committee at Linköping University approved all zebrafish experiments. Cells were treated with/without E₂ before injections. Cancer cells were labeled with Fast DiI oil red dye (ThermoFisher Scientific; cat. #1635639), as previously described (31). Human neutrophils were labeled with 6 μg/mL of DiB blue dye (Biotium; cat. #60036) and re-suspended in medium with/without E₂. Neutrophils were diluted 1:2 (50%) with cancer cells, and anti–human TGFβ1, LFA-1, and isotype control antibodies were added at 0.1 mg/mL with/without E₂ immediately before injection.

Transgenic Tg(fli1:EGFP)^y1 zebrafish embryos, with green fluorescent vessels, were raised in E3 medium with 1-phenyl-2-thiourea (PTU). Cells were implanted into the perivitelline space of 2-day old zebrafish embryos, as previously described (31), and incubated at 28°C in E3 medium with 0.2 mmol/L PTU. Cell migration to the tail region was evaluated 24 hours after injection, the embryos were anesthetized with 0.02% MS-222 (Sigma-Aldrich; cat. #E10521), and pictures were acquired using an Olympus BX43 light/fluorescence microscope (10×/0.30 magnification), excitation filters BP360-370, BP460-495, and BP530-550, using an Olympus DP72 CCD camera. Images were acquired with the Olympus CellSens.

For time-lapse experiments, zebrafish embryos were embedded in 0.5% agarose low gelling temperature (Sigma; cat. #A9045) with 0.02% MS-222 and placed in 35 mm glass bottom petri dishes (MatTek Corporation; cat. #P35G-1.5-10-C) with E3 embryo medium containing 0.2 mmol/L PTU and with/without E₂. Z-stack images were acquired every 6 and 15 minutes for tail region and injection site, respectively. A Zeiss Observer with LSM 700 inverted microscope equipped with incubator chamber was used for Supplementary Videos S1 to S3. A Nikon Eclipse TE2000-U inverted confocal microscope equipped with BioRad Radiance 2100 MP laser scanning system and 2-photon excitation MaiTai laser was used for Supplementary Video S4. Time-lapse images were acquired with the Zeiss ZEN software.

Mammosphere confocal imaging was performed as follows: image size (pixels) 1024 × 1024, 16-bit depth, averaging 4, EC Plan-Neofluar 10×/0.3 M27 objective, laser wavelength 488 and 405 nm, Alexa-fluor 488 excitation/emission (nm): 488/518, DAPI excitation/emission (nm): 405/435, binning mode 1 × 1, and PMT detector. Time-lapse Video 3 was acquired with LaserSharp2000 software, image size 1024 × 768, 8-bit depth, 10× objective magnification, 1.2 objective zoom, PMT detector, detection filters HQ450/80, E570LP, and HQ515/30. Time-lapse Supplementary Videos S1, S2, and S4 were acquired with Zeiss ZEN software, at 512 × 512 resolution, 8-bit depth, Plan-Apochromat 10×/0.45 M27 objective magnification, PMT detector, binning mode 1 × 1.

Lysates were loaded on a 4% to 15% polyacrylamide gel (BioRad; cat. #456-1083), transferred to PVDF membrane (BioRad; cat. #170-4156), and incubated with anti-human ERα (Dako; cat. #M7047) overnight at 4°C and then with anti-mouse immunoglobulins HRP (Dako; cat. #P0447). ECL detection kit (GE Healthcare; cat. #RPN2232) following the manufacturer's instructions was used.

Data are presented as mean ± SEM. Two-tailed Student t tests for paired and unpaired analyses, where appropriate, were used. A P value < 0.05 was considered statistically significant. Statistics were performed with Prism 6.0 (GraphPad software).

To elucidate whether E₂ affects the recruitment and polarization of neutrophils in ER⁺ breast cancer, we set up two different mouse tumor models of breast cancer. Murine PyMT tumors displayed E₂-dependent growth in immune-competent mouse that was inhibited by fulvestrant treatment (Fig. 1A). In PyMT tumors, E₂ induced a significant increase of neutrophils in the invasive margin compared with tumors grown without E₂, and this was reversed by fulvestrant (Fig. 1A). Furthermore, E₂-recruited neutrophils in the tumor site showed increased expression of LFA-1 integrin, whereas fulvestrant-treated tumors exhibited similar LFA-1 expression as control tumors (Fig. 1B). In human estrogen-dependent experimental breast cancers (MCF-7) established in nude mice, the pure antiestrogen fulvestrant significantly decreased E₂-induced tumor growth (Fig. 1C). Also, fulvestrant significantly reduced the E₂-induced recruitment of neutrophils (Fig. 1C) and significantly decreased the percentage of LFA-1⁺ neutrophils (Fig. 1D). The results were similar in tumors of comparable size from the various treatment groups as well as in tumors of different sizes, i.e., large control tumors versus smaller tumors from the treatment groups. However, it cannot be ruled out that in the larger tumors, other factor may contribute to the recruitment and polarization of the neutrophils. In Supplementary Fig. S1A, we show that the neutrophil marker Ly6G specifically stains murine neutrophils and not macrophages.

To analyze the mechanism underlying neutrophil recruitment in response to E₂, we set up a MCF-7 mammosphere breast cancer model in vitro. First, we developed a protocol for paraffin embedding of mammospheres to avoid agarose or clot embedding, as described in Materials and Methods. The isolated cells from blood that were used for all experiments indeed were neutrophils with their characteristic multi-lobulated nuclei and with various expression of LFA-1 (Supplementary Fig. S1B and S1C).

Some studies suggest that ER⁺ cells lose ER expression when cultured in mammospheres, but mammospheres maintained ERα expression and their growth increased significantly in response to E₂ (Fig. 2A). When mammospheres were established, human neutrophils and treatments were added. Infiltration of human neutrophils was significantly increased by E₂, which was reduced in the presence of fulvestrant (Fig. 2B). This was not dependent on the number of neutrophils that were added to the mammospheres, as similar results were found at three different concentrations of neutrophils (Supplementary Fig. S2A). Orthogonal projection in Fig. 2C reveals that infiltrated neutrophils remained in the outer layers of mammospheres, resembling what was observed in vivo. The E₂-induced migration of neutrophils was confirmed using the Boyden chamber; 37,863 ± 1,687 neutrophils migrated in the control versus 45,606 ± 977 in the E₂ group (P < 0.01). E₂ exposure and coculture increased the survival of neutrophils; after 48 hours, 18% ± 2% survived in single culture versus 35% ± 4% in coculture (P < 0.05). In the presence of E₂, 32% ± 1% survived in single culture versus 53% ± 1% in coculture (P < 0.01). In line with previous data of neutrophil survival in human in tissue in vivo (32), neutrophils were detected up to 5 days in three-dimensional (3D) mammosphere culture. The retention of neutrophils to cancer cells increased during E₂ exposure; 1.9 × 10⁵ ± 0.1 versus 3.1 × 10⁵ ± 0.09 in the E₂ group (P < 0.01).

TGFβ1 secretion in neutrophil-infiltrated MCF-7 mammospheres was significantly increased by E₂, but was significantly reduced by fulvestrant (Fig. 2D). The cancer cells were the main source of TGFβ1 (Fig. 2E). To evaluate whether neutrophil infiltration in response to E₂ was mediated by TGFβ1, we added anti–human TGFβ1 to the mammospheres. Significantly decreased E₂-induced infiltration of neutrophils was detected (Fig. 2F). Treatment with hrTGFβ1 induced infiltration of neutrophils similar to E₂ (Fig. 2G). The presence of neutrophils did not affect the proliferation of the mammospheres (data not shown). This was supported by mammosphere size, which was unchanged after addition of TGFβ1 antibody (778 ± 7 and 769 ± 7 μm, respectively).

Next, we investigated whether E₂ affects LFA-1 expression in neutrophils. Primary cultures of human neutrophils were exposed to E₂, which significantly increased LFA-1 expression, and this effect was reduced in the presence of fulvestrant (Fig. 3A). Human recombinant TGFβ1 treatment increased LFA-1 expression in a similar fashion to E₂ treatment (Fig. 3A). As further confirmation, anti–human TGFβ1 decreased LFA-1 expression (Fig. 3B), and significantly higher concentrations of secreted TGFβ1 were observed in E₂-exposed neutrophils (Fig. 3C). Furthermore, anti–human LFA-1 significantly decreased E₂-induced neutrophil infiltration into mammospheres (Fig. 3D), underscoring the key role of LFA-1 expression in neutrophil migration.

PyMT transgenic mice develop metastases, especially to the lungs, at late carcinoma stages at approximately 14 to 18 weeks of age, and at this stage, the tumors have lost their ER expression (28, 29, 33). Here, we harvested tumors from 10-week-old mice and injected the cells into the mammary fat pad in recipient syngeneic mice to achieve an ER-expressing breast cancer model in an immune competent mouse. However, these tumors would not metastasize and therefore could not be used for studies of mechanisms of E₂-dependent metastatic growth. For these reasons, we used a zebrafish model to evaluate whether the E₂-induced overexpression of LFA-1 in neutrophils, via TGFβ1, increased the dissemination of otherwise nonmetastatic ER⁺ breast cancer cells. As previously shown, E₂ exposure alone did not affect the dissemination of MCF-7 cells (10). However, the presence of neutrophils significantly increased dissemination of MCF-7 cells (Fig. 4A). The presence of E₂ further increased cancer cell dissemination (Fig. 4A). In addition, the number of fish where cancer cells were disseminated increased from 33% ± 3.7% to 56% ± 3.2% in the presence of neutrophils, P < 0.05, and to 88% ± 6.1% in the presence of neutrophils and E₂, P < 0.01. Anti–human TGFβ1 and anti–human LFA-1 significantly reduced the dissemination of MCF-7 cells, confirming the importance of TGFβ1 and LFA-1 for cancer cell dissemination capacity (Fig. 4B and C).

To corroborate the results obtained with MCF-7 cells, we evaluated another ER⁺ breast cancer cell line in the presence of neutrophils and E₂. Similar to the MCF-7 cells, anti–human TGFβ1 and anti–human LFA-1 antibodies inhibited the dissemination of ZR-75-1 cell injected together with neutrophils in the presence of E₂ (Fig. 4D). These data suggest that E₂-induced dissemination via LFA-1 might be conserved among ER⁺ breast cancer cells.

In the zebrafish dissemination experiments, we observed that a high proportion of migrated breast cancer cells were accompanied by neutrophils, especially in the presence of E_2. Time-lapse imaging in zebrafish showed that neutrophils increased the migration and invasion of ER⁺ breast cancer cells at the tumor injection site, and that neutrophils comigrated with cancer cells and promoted cancer cell intravasation (Fig. 5A and B; Supplementary Videos S1 and S2). Neutrophils intravasated together with cancer cells from the tumor primary site, helping to establish new metastatic niches and to extravasate circulatory cancer cells in distant sites of the fish (Fig. 5B; Supplementary Video S3). Anti–human LFA-1 blocked the neutrophil-induced migration and intravasation of cancer cells (Supplementary Video S4).

To evaluate whether E₂-induced breast cancer progression and metastasis via LFA-1 plays a role in ER⁻ breast cancer, ER⁻ MDA-MB-231 was investigated. Growth of MDA-MB-231 tumors in nude mice was independent on E₂ exposure. Tumor volume 3 weeks after cancer cell injection in mice was 18 5 ± 45 mm³ versus 173 ± 55 mm³, P = 0.87, ± E₂, respectively. No differences in neutrophil count in the invasive margin in tumors grown were detected ± E₂; 552 ± 101 versus 654 ± 64, respectively, P = 0.4.

Exposure to E₂ did not affect the invasive capability of these cells in the presence or absence of neutrophils, and no significant difference in the invasion of neutrophils into MDA-MB-231 mammospheres was detected (Fig. 6A and B). Estradiol did not affect TGFβ1 secretion or infiltration of neutrophils in the presence or absence of anti–human LFA-1 (Fig. 6C and D). In addition, the dissemination of MDA-MB-231 cells was not affected by the presence of neutrophils or E₂ (Fig. 6E). Higher concentrations of neutrophils resulted in structural damage to the mammospheres (Supplementary Fig. S2B).

Immunostaining of paraffin-embedded mammospheres showed differential expression of LFA-1 ligands. For instance, we observed high expression of ICAM-3 and no or very low expression of ICAM-1 and ICAM-2 in the nonmetastatic ER⁺ MCF-7 mammospheres (Supplementary Fig. S3A), indicating that neutrophil infiltration could be carried out in this model through cell interactions via LFA-1 and ICAM-3 binding. On the other hand, the metastatic ER⁻ MDA-MB-231 mammospheres showed high expression of ICAM-1 and ICAM-3 and weak expression of ICAM-2 (Supplementary Fig. S3A). In addition, ER⁻ MDA-MB-231 mammospheres produced higher quantities of soluble ICAM-1 than ER⁺ MCF-7 mammospheres (Supplementary Fig. S3B).

Given the key role of TGFβ1 in mediating E₂-induced breast cancer progression and dissemination in vitro and in vivo, it is important to know whether extracellular concentrations of TGFβ1 are affected in breast cancer patients. Therefore, we performed microdialysis for in vivo detection of TGFβ1 in women with ER⁺ breast cancer. We found that TGFβ1 in breast cancers was significantly higher compared with adjacent normal breast tissue, suggesting TGFβ1 as a possible target for human breast cancer treatments (Fig. 7).

In this article, we showed that estrogen exposure increased the secretion of TGFβ1, leading to increased accumulation of neutrophils in the invasive margin of ER⁺ breast cancers. In addition to increased neutrophil numbers, exposure to E₂ increased the expression of LFA-1 in neutrophils. Neutrophils increased the ability of nonmetastatic ER⁺ breast cancer cells to disseminate, and E₂ exposure further enhanced this process via increased expression of TGFβ1 and LFA-1. Our observation of higher extracellular levels of TGFβ1 in human breast cancers compared with normal adjacent breast tissue suggests that TGFβ-1 may be exploited therapeutically.

Exposure to E₂ plays a key role in breast cancer growth and metastasis (34). During the first 5 first years after diagnosis, ER⁺ breast cancer patients have a better prognosis than ER⁻ breast cancer patients. However, the prognoses of these two groups converge over time; five to ten years after diagnosis, there is no difference in recurrence rates, and beyond 10 years, the risk of recurrence and death is higher in the ER⁺ group (35, 36). About half of ER⁺ breast cancers respond to hormone therapy; however, 25% of patients who receive hormone therapy will relapse (5). Because more than two thirds of breast cancers express the ER, investigating mechanisms of ER⁺ breast cancer dissemination is key for designing novel therapeutics for these cancers.

Although the mechanisms of estrogen-mediated primary cancer growth are well studied, mechanisms behind estrogen-dependent breast cancer dissemination need to be elucidated. One major obstacle for this research is that murine mouse models of breast cancer that spontaneously metastasize lose ER expression during the dissemination process; the metastases are then ER⁻. There has been a general belief that this is also a natural cause of breast cancer progression in humans. Evidence shows that this is not the case; the majority of ER⁺ primary breast cancers maintain ER expression in metastases, and nearly a third gain ER expression in metastatic lesions (37). This underscores the importance of understanding estrogen-dependent mechanisms of breast cancer progression as a prerequisite for finding novel therapeutic targets.

The zebrafish metastasis model allows for such investigation (10, 31). By using this model, we showed here that neutrophils increased ER⁺ breast cancer cell dissemination, which is in agreement with recent published data (7). We also showed that nonmetastatic ER⁺ breast cancer cells became highly metastatic in the presence of neutrophils and that estrogen treatment may further enhance this dissemination capability. Aside from the effects of cell–cell interactions between neutrophils and breast cancer cells, neutrophils may also pave the way for distant metastases by infiltrating and creating premetastatic niches before cancer cells arrive (7). Neutrophils did not increase the dissemination ability of ER⁻ metastatic breast cancer cells in the presence or absence of E₂, suggesting that this mechanism may primarily be important in low-grade cancers, such as ER⁺ breast cancers.

Neutrophils possess dual roles in cancer progression depending on their phenotype or number in the tumor microenvironment (38, 39). Massive infiltration of neutrophils may elicit a direct cytotoxic effect by releasing hydrogen peroxide, leading to tumor regression, whereas a low-grade neutrophil gradient is tumor progressive (38, 39). This is in line with our present data showing that addition of a high number of neutrophils caused a disintegration of mammospheres in vitro.

According to previous reports, high expression of ICAM molecules in cancer tissues is related to poor prognosis and increased cancer malignancy (40, 41). These data support our results for highly metastatic and invasive ER⁻ MDA-MB-231 mammospheres, which exhibited high expression of ICAM-1 and -3, weak expression of ICAM-2, and secretion of ICAM-1 molecules. The nonmetastatic ER⁺ MCF-7 mammospheres showed high expression of ICAM-3 only. ICAM-1 molecules interact with the actin-containing cytoskeleton and α-actinin, inducing cytoskeletal rearrangement (42, 43). This activity could explain the neutrophil-independent invasion and dissemination of MDA-MB-231 cancer cells because their high expression of ICAM molecules promotes their increased motility and invasion.

TGFβ1 exerts a potent chemotactic effect in neutrophils (44, 45). Many cell types, including immune cells, produce TGFβ1, which is secreted in a latent form bound to various proteins (45). Events in the microenvironment, such as protease activities, release active TGFβ1 into the extracellular fluid where it becomes available for cell–cell interactions (19). A correlation between TGFβ1 and increased metastasis has previously been shown only by immunostaining human breast cancers. However, previous studies did not determine whether TGFβ1 in its bioactive form is elevated in human breast cancer. Here, we provide evidence that, indeed, extracellular concentrations of TGFβ1 are increased in human breast cancers, suggesting that TGFβ1 is a viable target for treatment.

In addition to its chemotactic effects, TGFβ1 may decrease neutrophil cytotoxicity and mediate a polarization from antitumor N1 neutrophils to protumor N2 neutrophils (11). In our study, LFA-1 integrin, which is associated with N2 polarization and prolonged survival of neutrophils (46), increased upon E₂ exposure via elevated TGFβ1. This LFA-1 overexpression in neutrophils also increased cell–cell interactions, which led to increased ER⁺ breast cancer cell dissemination. We confirmed the roles of LFA-1 and TGFβ1 in mediating E₂ effects by showing that inhibition of these proteins by antibodies successfully blocked cancer cell dissemination in the presence of E₂.

We have previously shown that E₂ does not affect the number of neutrophils in central tumor tissues and that the numbers of neutrophils in the centers of tumors are very low (10). Our present data suggest that neutrophils are preferentially localized in the invasive margin and that E₂ significantly affects the number of these neutrophils both in immune-competent and immune-deficient mouse models of breast cancer. Also, the presence of LFA-1–expressing neutrophils in the invasive edges of tumors significantly increased with E₂ treatment. This suggests that combined therapies that target cell–cell interactions in tumor microenvironments could be feasible in ER⁺ breast cancer.

Our data were in part generated using mammosphere cultures, and we described here a protocol for paraffin embedding of these spheres. 3D models such as tumor spheres or microtissue more closely mimic the tumor microenvironment than do monolayers of cancer cells. Previous groups have attempted diverse paraffin-embedding techniques with several disadvantages, including difficulties in obtaining accurate sections of spheres due to agarose/clot embedding (47–49). When agarose/clot embedding of mammospheres is performed, mammospheres do not stay in a single plane but sink to different deep layers. In order to analyze all mammospheres, multiple slices must be collected. Our protocol provides a new and improved method for paraffin embedding. With the method described herein, it is possible to analyze all paraffin-embedded mammospheres simultaneously in a single slice, reducing slide number and staining procedures.

In summary, neutrophils may respond to microenvironmental cues and acquire a protumorigenic phenotype. Our data provide insight into these mechanisms in breast cancer, showing that E₂ promotes breast cancer metastasis by enhancing cell–cell interactions between neutrophils and ER⁺ breast cancer cells. E₂ increased the expression of LFA-1 via TGFβ1, leading to cancer cell migration from the tumor primary site to distant sites, creating new metastatic niches. These interactions caused nonmetastatic cells to become highly metastatic. Our data are clinically relevant because we detected increased TGFβ1 in breast cancers of women. Our work contributes to the understanding of the immune-modulatory effects of E₂ in ER⁺ breast cancer that drive breast cancer metastasis and provides insights into potential therapeutic targets for disseminated hormone-dependent breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: C. Dabrosin

Development of methodology: G. Vazquez Rodriguez, A. Abrahamsson, C. Dabrosin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Vazquez Rodriguez, A. Abrahamsson, L.D.E. Jensen, C. Dabrosin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Vazquez Rodriguez, A. Abrahamsson, C. Dabrosin

Writing, review, and/or revision of the manuscript: G. Vazquez Rodriguez, A. Abrahamsson, C. Dabrosin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Vazquez Rodriguez, A. Abrahamsson, C. Dabrosin

Study supervision: C. Dabrosin

This work was supported by grants to C. Dabrosin from the Swedish Cancer Society (2015/309), the Swedish Research Council (2013-2457), LiU-Cancer, and Research Funds of Linköping University Hospital.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.