Estrogen Receptor β-Mediated Inhibition of Actin-Based Cell Migration Suppresses Metastasis of Inflammatory Breast Cancer

Inflammatory breast cancer (IBC) is a rare but highly aggressive subtype of breast cancer accounting for about 10% of breast cancer–associated deaths in the United States. It is characterized by rapid progression, local and distant metastases, and younger age of onset (1). Although application of multimodal therapy approach that includes systemic neoadjuvant chemotherapy, surgery and radiotherapy has significantly improved clinical outcomes, 5-year survival rates are still approximately 35% to 40% and much lower compared with other breast cancers (2). Poor prognosis is primarily due to the high metastatic potential of IBC tumors. One-third of patients have distant metastases at diagnosis, and about 40% of those with stage III disease at diagnosis that receive standard treatment, eventually experience distant recurrence (3). Despite the contribution of metastasis to the aggressive behavior of IBC, its biology is poorly understood. Because of the small number of tested samples and differences in the characteristics of patient groups, genomic studies have failed to discover distinct biological mechanisms and a specific gene expression signature in IBC (4). However, a number of preclinical studies have suggested changes in signaling pathways and molecules that regulate stemness, epithelial-to-mesenchymal transition (EMT), actin-based cell migration as well as inflammation and immune modulation including ALDH1, NFκB, EGFR, HER2 and Rho GTPases (5). Among the Rho GTPases that promote actin-based cell migration is the cytoskeleton remodeler RhoC (4). Although RhoC is upregulated in many IBC tumors and increases the invasiveness of IBC cells, it is considered an “undruggable” target and the pathways that activate RhoC are not fully defined (6, 7). As of today, no specific therapeutic target for mitigation of IBC has been identified and there are currently no FDA-approved targeted therapies that are specific for the disease (1, 8).

One notable biological difference between IBC and non-IBC is the higher frequency of estrogen receptor α (ERα)-negative status in IBC tumors. Approximately 60% to 70% of IBC tumors lose expression of ERα compared with about 25% of non-IBC cancers (8). Despite the lack of ERα expression from the majority of IBC tumors, estrogen has been proposed to affect these cancers through ERα-independent pathways (9). These may include the second ER subtype (ERβ) that mediates effects of estrogen in normal and malignant breast (10). However, the expression of ERβ in IBC tumors and its role in biology of the disease have not been previously studied. It is generally accepted from studies in non-IBC that ERβ follows a different expression pattern than ERα. It has also been postulated that it elicits, as a transcription factor, distinct transcriptional responses and opposes ERα when both receptors are expressed in the same tumor (10–13). While ERα mediates the proliferative responses and is upregulated in luminal tumors, ERβ declines in carcinoma in situ and further decreases in invasive lesions of non-IBC tumors (14, 15). Conversely, as we and others have shown, enforced expression of ERβ in mesenchymal-like non-IBC triple-negative breast cancer (TNBC) cells promotes epithelial transformation and decreases invasion by inhibiting drivers of EMT including TGFβ, EGFR, and repressors of E-cadherin, indicating the potential for ERβ to function as an antimetastatic factor (12, 16–21).

The proposed anti-invasive activity of ERβ in non-IBC settings prompted us to investigate whether the receptor is expressed in IBC and to test its impact on growth properties and metastatic potential. We report here the analysis of clinical specimens that reveals expression of ERβ in IBC tumors and correlation of its expression with reduced metastatic phenotype. This is further supported by findings in preclinical models of IBC where ERβ and its agonists are associated with decreased burden of distant metastases. We link the antimetastatic function of ERβ to its ability to adjust the cytoskeleton architecture to inhibit cell migration by acting on Rho GTPases that are known mediators of IBC metastasis. Our findings indicate a novel role for ERβ as a regulator of IBC metastasis and suggest that the receptor may prove useful as a prognostic marker for the disease as well as an attractive therapeutic target.

KPL4, SUM190, BCX-010, IBC3 and SUM149 cell lines were from The University of Texas MD Anderson Cancer Center (UTMDACC). FC-IBC02 cells were obtained from Fox Chase Cancer Center (Philadelphia, PA) and MCF7, MDA-MB-231, MCF10A, ZR-75–1, T47D, UACC732, BT549 and HEK293 cells were from ATCC. All IBC cell lines were cultured in HAM-F12 media (Corning) supplemented with 7% FBS (Sigma), 10 mmol/L HEPES (Thermo Fisher), 5 μg/mL insulin (Sigma), 1 μg/mL hydrocortisone (Sigma) at 37°C in 5% CO₂. Non-IBC cell lines were cultured in DMEM (Corning) supplemented with 10% FBS with the exception of MCF10A cells that were maintained in DMEM containing 10% FBS, 10 μg/mL insulin, 0.5 μg/mL hydrocortisone, 1 ng/mL cholera toxin (Sigma), and 10 ng/mL EGF (Sigma). Cells were also incubated in phenol red–free medium supplemented with 1%–2% dextran-coated charcoal-treated FBS (Thermo Fisher) for 48 to 72 hours prior to treatment with the ERβ agonist LY500307 (ApexBio), 17β-estradiol (E2) or DMSO (Sigma) in the same media. Cell lines were used in experiments within three passages since thawing, were tested negative for Mycoplasma, and authenticated by short tandem repeat (STR) profiling. For the CRISPR knockout of ERβ, cells were transfected with pX330 vector containing the sgRNA hERβ CRISPR-7: GGATTGACTGCAGTTGTAGG that was kindly provided by Dr. Li (George Washington University) with plasmid carrying puromycin resistance in a ratio 3:1. Following selection with puromycin, resistant clones were isolated and screened by sequencing and immunoblotting. For sequencing, genomic DNA was extracted with Genomic DNA Purification Kit (Thermo Fisher) and PCR amplified using the primers ATTATAATGACCTTTGTGCC and GGATATTCATGGTGGCTGTC to produce a 190-bp fragment flanking the region targeted by CRISPR gRNA. Amplicons were cloned in pJET1.2 vector (Thermo Fisher) and clones were sequenced. For stable expression of both firefly luciferase and GFP, lentivirus containing the pCDH-CMV-MCS-EF1α-copGFP was produced by transfection and used to infect cells and the GFP-positive population was enriched by cell sorting. For stable expression of ERβ, cells were infected with lentivirus containing the pLenti6/V5-FLAG-ERβ1 (FLAG at N-terminus) construct as described previously (16). Cells were also transiently transfected with the pIRESneo-ERβ, pIRESneo-ERβ2, pIRESneo-ERβ5, and pIRESneo-ERα plasmids as published previously (22). For siRNA knockdown, cells were transfected in 6-well plates with 60 nmol/L ON-TARGETplus SMARTpool human ESR2, GPR141, RhoC, ELMO1, or HER3 siRNA pools or nontargeting siRNA pool as control (Horizon Discovery) using Lipofectamine RNAiMAX (Thermo Fisher). The target sequences of siRNA pools are listed in Supplementary Table S1. For expression of GFP-RhoC, cells were transfected with the GFP-RhoC plasmid (Addgene) using an empty vector (pEGFP-C2) as control (23). For expression of ELMO1, the ELMO1-YFP plasmid was kindly provided by Dr. Jin (Laboratory of Immunogenetics, NIH, NIAID, Bethesda, MD) and the pReceiver-M15 empty vector was used as control (GeneCopoeia; ref. 24). Plasmids were transfected with Lipofectamine 2000 (Thermo Fisher) as previously published (16). All reagents are listed in Supplementary Table S2.

In wound-healing assay, 7.5 × 10⁴ cells were seeded in Culture-Insert 2 well in μ-Dish 35 mm (Ibidi) and allowed to form monolayer. The insert was then removed and the migration area in the wound was imaged and evaluated in an IncuCyte S3 Live-Cell Analysis System (Sartorius) constantly for up to 3 days. Invasion assay was performed as previously published (16). Cells (2–2.5 × 10⁴ for KPL4 and 3.5 × 10⁴ for FC-IBC02) were plated in 6.5 mm Transwell champers (Sigma). Six to 18 hours later, depending on the invasive potential of the cell line, the cells that had invaded through the filter and attached to its bottom surface were stained with crystal-violet, imaged, and counted in five independent fields/well using a Nikon Eclipse TE2000-U microscope equipped with a DS-FI3 Camera and NIS-Element Software (Nikon).

The 2 × 10³ cells were seeded in 96-well plates containing 100 μL/well medium. Twenty-four hours later, the AlamarBlue Cell Viability Reagent (Thermo Fisher) was added in 1/10th volume directly to the cells and fluorescence was recorded in different time points using a fluorescence-based plate reader.

Total RNA was isolated from Control and ERβKO cells using Qiagen RNeasy Mini Kit (Qiagen) and quantitated in NanoDrop 1000 spectrophotometer (Thermo Fisher). The quality of RNA was evaluated by capillary electrophoresis on an Agilent 2100 Bioanalyzer (Agilent Technologies). Microarray analysis was carried out in the University of Pennsylvania Molecular Profiling Facility using Affymetrix Human Clarion D expression arrays (Thermo Fisher).

Differentially expressed genes that were identified by microarray analysis using a stringent (q-value <0.15 and fold change >2.5) or relaxed threshold (fold change >1.5) were analyzed using the Ingenuity Pathway Analysis (IPA; Ingenuity Systems) and Metascape portal to identify enriched biological pathways.

A 3071-bp DNA fragment containing estrogen response elements (ERE) and located upstream of GPR141 TSS was PCR amplified from human genomic DNA with forward 5-ACTGAGTTTTGCGCTTGTCA-3 and reverse 5-CATTCATGTCTCCCCAAACC-3 primers. Amplicons were first cloned into pJET vector and then subcloned between XhoI and BglII sites into pGL4.10 (Promega). The pGL4.10 recombinant construct was used as template for amplification of primer-assisted mutagenesis fragments. One fragment lacked a 189-bp region centered the ERE and in the second fragment the ERE TGACCCAGCTGCCCT was mutagenized to TGcCtgAGCctggga by combining the forward primer with the reverse primers 5-AAGTGAATGTGTTCTAAGTTCCTCA-3 and 5-AATGCATGCAGAGCATGGAAATGCCTGAGCCTGGG-AGCTGTTGAAACAGAATCCTATTT-3. Amplicons were cloned into the EcoRV site of pGL4.10. Cells were transfected using Lipofectamine 2000 with the reporter plasmids and a Renilla luciferase construct to normalize for transfection efficiency. Relative light units were calculated as ratio of firefly luciferase to Renilla luciferase activity. Luciferase activity was measured with Dual Luciferase Reporter Assay (Promega) as described previously (25).

Cells were plated on coverslips and 24 hours later were fixed with 3.7% paraformaldehyde (Sigma) in PBS, permeabilized with 0.3% Triton X-100 (Sigma) in PBS and blocked with 10% (v/v) FBS in PBS. For focal adhesion staining, cells were incubated with anti-vinculin antibody (Abcam) overnight at 4°C and secondary antibody for 1 hour at room temperature. Actin filaments and nuclear DNA were stained with Alexa-Fluor 488–conjugated phalloidin (Abcam) for 1 hour and 300 nmol/L DAPI (Thermo Fisher), respectively. Fluorescent images were acquired using a Zeiss Observer.Z1 microscope through 60 × lens with an AxioCam MRm camera and ZEN 2.3 Black edition (Carl Zeiss) software. Images were analyzed for focal adhesion density using NIS Elements Software (Nikon).

RhoC activity was assessed with RhoC Activation Assay Kit (Cell Biolabs) according to manufacturer's instructions. Briefly, cells were incubated with lysis buffer (25 mmol/L HEPES, pH 7.5, 150 mmol/L NaCl, 1% NP-40, 10 mmol/L MgCl₂, 1 mmol/L EDTA, 2% glycerol) containing protease inhibitors for 20 minutes on ice and lysates were cleared by centrifugation for 10 minutes at 14,000 × g at 4°C. Lysates were incubated with Rhotekin RBD agarose beads with rotation at 4°C overnight and the beads were pelleted at 14,000 × g for 10 seconds. After proper washing, proteins bound to beads were dissociated through resuspension in 2 × SDS-PAGE sample buffer and analyzed by immunoblotting.

Cells were lysed in RIPA buffer (Thermo Fisher) including the following protease and phosphatase inhibitors; Complete Mini, EDTA free (Roche), Protease Inhibitor cocktail (Sigma), Halt Phosphatase Inhibitor cocktail (Thermo Fisher) and phosphatase inhibitor cocktails 2 (Sigma) and 3 (Sigma). Proteins were separated in SDS-PAGE and transferred onto nitrocellulose membrane (Thermo Fisher). Membranes were blocked for 1 hour with 5% dry milk in TBS containing 0.05% Tween-20 at room temperature. Membranes were probed with primary antibodies overnight at 4°C and proteins were visualized with ECL Detection Kit (GE Healthcare) as previously described (16). ERβ was immunoblotted with two rabbit polyclonal (Millipore, GeneTex) and the 14C8 antibody (GeneTex) and recombinant ERβ protein (Thermo Fisher) was loaded to ensure specificity. Antibodies against RhoC, ERα, pAKT1, AKT1, GST, ELMO1, ROCK1, HER3, and GAPDH were from Cell Signaling Technology. Anti-β-Actin antibody was from Abcam and anti-GPR141 was from Thermo Fisher. All antibodies are listed in Supplementary Table S2.

Cells were fixed in 10% formalin for 30 minutes, resuspended in 0.8% agarose, and cell suspension was applied in donut-shaped agarose mold that was paraffin embedded. Five-micron–thick paraffin sections were immunostained using a Leica BOND RXm automated platform with the Bond Polymer Refine Detection Kit (Leica). Briefly, following dewaxing and rehydration, sections were pretreated with the epitope retrieval BOND ER2 high pH buffer (Leica) for 20 minutes at 98°C. Endogenous peroxidase was inactivated with 3% H₂O₂ for 10 minutes and nonspecific tissue–antibody interactions were blocked with PowerVision IHC/ISH Super Blocking solution (Leica) for 30 minutes. Sections were then incubated with the ERβ 503 antibody in the same blocking solution at a concentration of 1/300 for 45 minutes at room temperature. Immunoreactivity was revealed with diaminobenzidine (DAB) chromogen reaction and slides were counterstained in hematoxylin, dehydrated, cleared, and permanently mounted with a resinous mounting medium (Thermo Fisher).

Protein–DNA complexes were crosslinked with formaldehyde solution (50 mmol/L HEPES-KOH, pH 7.5, 100 mmol/L NaCI, 1 mmol/L EDTA, 0.5 mmol/L EGTA, 11% Formaldehyde) for 10 minutes at room temperature. Cells were harvested and cell suspensions were centrifuged. For nuclei purification, cell pellets were washed in PBS, snap-frozen in liquid nitrogen, and consecutively incubated in ice-cold hypotonic buffer (20 mmol/L HEPES-KOH, pH 7.5, 20 mmol/L KCl, 1 mmol/L EDTA, 10% glycerol, 1 mmol/L EDTA, 1 mmol/L DTT, 0.1 mmol/L PMSF) including protease inhibitors for 10 minutes at 4^o C and lysis buffer (50 mmol/L HEPES-KOH, pH 7.5, 140 mmol/L NaCl, 1 mmol/L EDTA, 10% glycerol, 0.5% NP-40, 0.25% Triton X-100, 1 mmol/L DTT, 0.1 mmol/L PMSF) with protease inhibitors. Nuclei was washed in ice-cold wash buffer (10 mmol/L Tris-HCL, pH 8, 200 mmol/L NaCl, 1 mmol/L EDTA, 0.5 mmol/L EGTA, 1 mmol/L DTT, 0.1 mmol/L PMSF). Chromatin in nuclear extract was sheared by sonication, clarified by centrifugation, and the supernatant was used for immunoprecipitation. Chromatin-bound FLAG-tagged ERβ was incubated with a monoclonal anti-FLAG M2 antibody (Sigma) or mouse IgG (Santa Cruz Biotechnology) and precipitated with protein G magnetic beads (Thermo Fisher). Protein–DNA complexes were eluted, decrosslinked at 65°C overnight, and after RNase (0.2 mg/mL) and proteinase K (0.2 mg/mL) treatment, the DNA was purified and the DNA enrichment was measured by real-time PCR using promoter-specific primers (Supplementary Table S3).

Animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Pennsylvania. In the orthotopic model, 6-week-old NCG female mice (Charles River) were anesthetized and injected with 5 × 10⁵ tumor cells into the mammary fat pad. In the experiments with the ligand treatment, NCG mice were ovariectomized at 4 week of age (Charles River) and injected with 5 × 10⁵ tumor cells into the mammary fat pad three weeks later. Pellets containing LY500307 (4.9 mg/pellet, 49-day release) or vehicle tablets (Innovative Research of America) were implanted subcutaneously in the dorsal neck region of mice at the time of cell injection. Mice were weighed and mammary tumors were measured once a week until sacrifice. Tumor volume (mm³) was calculated as V = (L × W²)/2. Once tumors reach approximately 500 mm³, mice were anesthetized and tumors were resected. Mice were also injected intravenously with 5 × 10⁵ tumor cells. Animals were monitored for signs for morbidity, body weight loss, and tumor ulceration and euthanized at the endpoint by cervical dislocation according to IACUC guidelines. Primary tumors and organs were harvested for analysis.

For bioluminescence imaging (BLI), following anesthesia with 2.5% isoflurane, mice were injected retro-orbitally with 100 μL of 15 mg/mL Firefly d-Luciferin substrate (Gold Biotechnology). In vivo images of mice and ex vivo images of harvested lung tissues were acquired 5 minutes following injection with d-Luciferin using the IVIS Spectrum system (Caliper Life Sciences). Data were analyzed with the Living Image 3.0 software (Caliper Life Sciences) and expressed as total photon flux.

For histologic analysis, harvested lungs were perfused and fixed in 10% buffered formalin (Sigma) for 24 hours. Lungs were then processed and embedded in paraffin, sectioned, and stained with hematoxylin and eosin. Metastases were quantified microscopically by a pathologist in blinded fashion by counting the number of neoplastic cell aggregates (more than 5 cells large) in 15 randomly selected ×200 microscopic fields. Bones were fixed in 10% formalin at 4°C under constant agitation for 3 days, soaked in 30% sucrose in PBS at 4°C overnight, and embedded in OCT (Sigma). Five-micron–thick frozen sections were cut and examined using the Zeiss Observer.Z1 microscope.

In the first cohort of breast cancers, a total of 40 inflammatory breast carcinoma tissue samples were collected from 2008 to 2015. The majority of these patients underwent modified radical mastectomy and received neoadjuvant chemotherapy (Taxol/Adriamycin). Histologic types of the total 40 samples were defined into grade II (12) and III (26). Clinical stage was defined according to WHO classification criteria (2007). IHC staining for ERβ was performed on total of 40 cancer specimens. In the second cohort, patients with IBC (25) and non-IBC (58) tumors received similar neoadjuvant chemotherapy (anthracycline- and taxane-based chemotherapy) followed by surgery at MD Anderson Cancer Center. Patients with IBC also had a similar histologic classification (12% grade II and 88% grade III) with those with non-IBC (22.5% grade II and 77.5 grade III).

Briefly, sections were deparaffinized by two washes in xylene and rehydrated in a gradient ethanol series (2 × 100%, 95% and 70%). Antigen retrieval was performed in a sodium citrate buffer (pH 6.0) for 25 minutes in subboiling temperature followed by a 20-minute incubation on ice. Sections were blocked and permeabilized with a 0.1% NP-40/3% BSA solution for 10 minutes at room temperature followed by incubation with the 503 anti-ERβ antibody (1:50) at 4°C overnight (15, 26, 27). Stained sections were washed in TBST and incubated with secondary biotin-conjugated anti-chicken antibody (Thermo Fisher) for 30 minutes and ABC system was used to visualize staining (DAKO). Sections were counterstained with hematoxylin, dehydrated, and mounted. Staining was assessed by the IBC pathologist blinded to the origination of samples and patient outcome. Staining scores were generated with the German semi-quantitative scoring system. Each specimen was assigned a score based on intensity of nuclear staining (no staining = 0; weak staining = 1, moderate staining = 2, strong staining = 3) and proportion of stained cells (0% = 0, 1–24% = 1, 25–49% = 2, 50–74% = 3, 75–100% = 4). The final immunoreactive score was estimated by multiplying the intensity score with the score of the % of stained cells ranging from 0 (minimum score) to 12 (maximum score). For ERβ, we defined 0–8 as negative and 9–12 as positive. This is similar to previously published cutoffs (28, 29).

Group comparisons were conducted by Student t test or Mann–Whitney test, with the mean values and the SD calculated for each group. Kaplan–Meier survival curves were calculated using overall survival and metastasis-free survival time for each patient from each group, with the log-rank test used to identify significant differences between the groups. The Cox proportional hazards model was used to assess the effect of ERβ expression while controlling for age, number of nodes, and grade.

The data that support the findings of this study are available from the corresponding author upon reasonable request. Patient-related clinical data are included as Supplementary tables. Microarray data have been uploaded to GEO repository under the accession code GSE149852.

The decline of ERβ expression in invasive lesions on breast cancer and the association of the receptor with decreased migration of breast cancer cells have been viewed as evidence of antimetastatic function (13, 14, 17, 19–21). However, a clear demonstration of the antimetastatic activity of ERβ in patients with breast cancer is lacking. To this end, we focused on IBC because ERβ has not been previously studied in this highly metastatic carcinoma. Moreover, since patients with IBC have different status of metastasis at diagnosis and a proportion of those without initial metastases eventually show disease progression, we set out to investigate whether differences in the expression of ERβ may account for the different metastatic potential of IBC tumors. We analyzed 40 archived tumor samples that were derived from post-neoadjuvant chemotherapy surgical specimens of mastectomy from patients with IBC that belong to all clinical breast cancer subtypes (see Methods). The clinicopathologic characteristics of the patients are summarized in Supplementary Table S4. To determine the expression levels of ERβ, we performed IHC in IBC samples using the 503 antibody that is specific for both the human and mouse proteins. Its specificity for mouse ERβ was previously tested in the mammary gland of conditional ERβ knockout mice (27). Specificity for human ERβ was validated with various methods including preincubation of the antibody with excess of recombinant human ERβ protein (15, 26). To further ascertain its specificity for human ERβ, we performed immunocytochemistry in HEK293 cells transiently transfected with wild-type ERβ, ERα, or the ERβ splice variants ERβ2 (also known ERβcx) and ERβ5 using the empty vector–transfected cells as negative control. ERβ isoforms have been reported to elicit contrasting actions in non-IBC breast cancer cells and associated with clinical outcome in non-IBC (30, 31). As shown in Supplementary Fig. S1A and S1B, strong nuclear staining was detected in ERβ-expressing cells, no staining in empty vector- and ERα-transfected cells, and modest nuclear signal in ERβ2- and ERβ5-expressing cells, suggesting that the 503 antibody is specific for ERβ and appears to have higher affinity for the wild-type receptor compared with its splice variants. This finding indicates that ERβ staining in human IBC tissues is specific for ERβ and may, in some extent, reflect expression of splice variants. ERβ staining was not evaluated in 3 of 40 clinical specimens due to absence of tumor tissue. We observed strong nuclear staining in tumor cells in 11 of 37 IBC samples (29%; Fig. 1A). Of those tumors that were stained positive for ERβ, 90% were ERα and progesterone receptor (PR) positive and 45% were HER2 positive. To assess the clinical importance of ERβ, we analyzed the survival and metastasis follow-up information of the same patients. From the 37 patients with known ERβ expression, those with high ERβ levels in tumors were significantly associated with better metastasis-free survival (MFS) than the ones with low ERβ tumors (5-year MFS: 50.0% vs. 17.2%, P = 0.02, Fig. 1B). Similarly, patients with high expression levels of tumor ERβ showed improved overall survival (OS) compared with those with low ERβ tumors (P = 0.1; Supplementary Fig. S1C). Moreover, we observed inverse correlations between the expression of ERβ and adverse pathologic features including tumor size, grade, and number of lymph nodes (Supplementary Fig. S1D). Notably, given the small patient sample size, these associations were not statistically significant. However, these trends and the correlation of ERβ with better MFS hold that tumors that express higher levels of ERβ are associated with favorable prognosis. To corroborate the inverse association of ERβ with metastasis, we simultaneously evaluated the effect of ERβ expression, age and grade on MFS by using the Cox proportional hazards model. As shown in Supplementary Fig. S1E, ERβ positivity remained an independent favorable prognostic factor for MFS (HR = 0.88, P = 0.025). Moreover, we analyzed a dataset with available gene expression and clinical information from 25 patients with IBC and 58 patients with non-IBC tumors that received similar treatment using an ERβ specific probe (211119_at; Supplementary Table S5; ref. 8). Although no significant difference in the expression of ERβ was observed between IBC and non-IBC tumors, high tumor mRNA levels of the receptor were significantly associated with longer OS in patients with IBC, but not in those with non-IBC tumors (Fig. 1C–E). Analysis using Cox regression model confirmed the ERβ as favorable prognostic factor for OS in patients with IBC (HR = 0.22, P = 0.0053; Supplementary Fig. S1F). To investigate whether the observed lack of association between ERβ expression and clinical outcome in non-IBC occurs only in this available cohort of patients, we analyzed publicly available datasets [The Cancer Genome Atlas (TCGA), WANG, and TRANSBIG] with non-IBC patients. As shown in Supplementary Fig. S2, ERβ is associated with better survival only in ERα-positive and HER2-positive patients from the TCGA dataset. In contrast, no significant associations were identified in WANG and TRANSBIG datasets as well as in TNBC patients from the TCGA dataset. Taken together, these results indicate that IBC tumors with high ERβ expression exhibit lower propensity to metastasize compared with ones with low levels of the receptor, implying a potentially protective role of ERβ against disease progression and its associated mortality.

To corroborate the inverse clinical association of ERβ with metastasis, we analyzed preclinical models of IBC. We, and others, previously associated the induction of ERβ expression in mesenchymal-like non-IBC TNBC cells with epithelial transformation and decreased invasion (12, 16–18, 20). In this study, we assessed for the first time the expression and function of ERβ in a panel of human IBC cell lines. Consistent with the presence of ERβ protein in IBC tumors (Fig. 1), we observed expression of endogenous receptor in IBC cell lines (Fig. 2A). We specifically detected significant expression in triple-negative (FC-IBC02, SUM149) and HER2-positive (UACC732, KPL4) IBC cell lines and low levels in BCX-010, IBC3, and SUM190 cells (Fig. 2A; refs. 32–34). We initially observed an inverse association between the endogenous amount of ERβ and the migratory morphology of IBC cell lines. FC-IBC02 and KPL4 cells with the highest ERβ levels look epithelial-like with polygonal scheme, whereas SUM190 and BCX-010 cells with the lowest expression display a morphology with membrane protrusions and decreased adhesion (Fig. 2B). To confirm this association, we generated clones of KPL4 cells with CRISPR/CAS9-mediated knockout of ERβ (ERβKO). As shown in Supplementary Fig. S3A, sequencing analysis of two ERβKO clones revealed insertions or deletions close to the protospacer adjacent motif (PAM) that led to premature stop codon near the beginning of the open reading frame and knockout of the targeted gene that was confirmed by the absence of the protein in immunoblots (Fig. 2C and D). Unlike ERβ, the low, yet detectable levels of ERα in these IBC cells did not decrease by the same CRISPR/CAS9 system (Supplementary Fig. S3B; ref. 33). Moreover, we stably reexpressed ERβ in one of the ERβ knockout clones (ERβKO + ERβ; Fig. 2D). We examined the morphology of cells in colonies of the same size to ensure equal plating density among cell cultures with different ERβ levels. Cells were also kept in FBS-containing media, suggesting that ERβ in control cells may be activated by the estrogen that is present in serum. In line with our initial observation, we have seen decreased adhesion and enlarged protrusions in two different clones of ERβ knockout IBC cells, whereas reexpression of the receptor in knockout cells rescued the more compact and cobblestone morphology of control cells (Fig. 2E). Similar to CRISPR-mediated knockout, knockdown of ERβ with small interfering RNA (siRNA) resulted in a more migratory phenotype in both KPL4 and FC-IBC02 cells (Fig. 2F; Supplementary Fig. S3C). In contrast, knockdown of ERβ in cultures with FBS-containing media or its activation by the selective agonist LY500307 in estrogen-depleted media did not significantly alter the morphology of SUM149 cells that express reduced endogenous ERβ levels compared with KPL4 and FC-IBC02 cells and display a mixed epithelial and mesenchymal phenotype (Supplementary Fig. S3D; ref. 35).

Guided by the inverse association of ERβ with the migratory morphology of IBC cells, we next examined whether the receptor can inhibit migration and decrease the invasiveness of the same cells when they grow in FBS-containing media. We initially analyzed wound healing as an indicator of cell migration in incuCyte live cell imaging system. As shown in Fig. 3A, knockout of ERβ enhanced migration in KPL4 cells as it is indicated by the faster wound closure in ERβ knockout cells, whereas reexpression of the receptor in knockout cells restored their collective motion to the speed of control cells. This effect was not due to higher proliferation of ERβ knockout cells as demonstrated by the results of a proliferation assay (Supplementary Fig. S4A). Moreover, overexpression of ERβ in control cells further decreased migration, underscoring the crucial role of ERβ expression in controlling migration in IBC cells (Supplementary Fig. S4B). Consistent with the increased migration in the absence of ERβ, a significantly higher number of ERβKO IBC cells invaded the membrane in Transwell chambers compared with ERβ-proficient cells (Fig. 3B). Conversely, activation of the receptor by its selective agonist LY500307 decreased the invasiveness of ERβ proficient but not ERβ knockout cells in estrogen-depleted media, corroborating both the anti-invasive function of ERβ and the specificity of the ligand (Fig. 3C; Supplementary Fig. S4C). Similar to KPL4 cells that express low levels of ERα, downregulation of ERβ enhanced invasion in ERα-negative FC-IBC02 cells (Fig. 3D; Supplementary Fig. S4D). Although ERβ may elicit its antimigratory activity in KPL4 cells by heterodimerizing with ERα, our findings suggest that ERβ can inhibit invasion in IBC cells irrespective of the ERα status (36). In contrast, and consistent with its effects on cell morphology, knockdown of the receptor in SUM149 cells altered migration and invasion at significantly less extent compared with KPL4 and FC-IBC02 cells (Supplementary Fig. S4E and S4F).

We next examined whether the antimigratory activity of ERβ had an impact on metastasis of IBC tumors in vivo. For this, we stably infected control, ERβKO and ERβKO + ERβ cells with lentivirus containing firefly luciferase and GFP-reported genes and injected the cells either orthotopically in the mammary fat pad or by tail vein in immunodeficient mice. In the lung colonization model, mice were analyzed by in vivo BLI and their harvested lungs by ex vivo BLI and histology four weeks after the inoculation of cells. As shown in Fig. 4A,–C; Supplementary Fig. S5A, mice that they were injected with cells lacking ERβ had a higher number of micrometastases in lungs compared with mice injected with control or ERβ knockout cells in which ERβ was reintroduced. In orthotopic model, mice and selected organs were evaluated four weeks following resection of primary tumors. Similar to the lung colonization model, mice with orthotopically-injected ERβ knockout tumors had a higher number of micrometastases in lungs and bones compared with the mice bearing ERβ-proficient tumors (control and ERβKO + ERβ) as it is demonstrated by the intensity of BLI signal and the abundance of GFP-positive cells in bone morrow (Fig. 4D,–F). Assessment of the primary tumor volume indicated that the higher metastatic burden in mice with ERβKO tumors was not due to their higher growth rate (Supplementary Fig. S5B). To ascertain the effect of activating endogenous ERβ on metastasis, we treated ovariectomized immunodeficient mice with orthotopically injected IBC tumors with the ERβ agonist LY500307 by using implanted pellets (5 mg/kg/day) for 7 weeks. As shown in Fig. 4G, harvested lungs from mice that were treated with LY500307 displayed reduced BLI signal compared with those that were exposed to vehicle pellets indicative of decreased number of metastases further supporting the repressive effect of ERβ on metastasis.

We next set out to explore the mechanism that is employed by ERβ to elicit its antimetastatic activity. We focused on pathways that regulate migration and factors that are known to be specifically altered in IBC. We carried out microarray analysis to identify differentially expressed genes between control and ERβ knockout KPL4 cells that grow in regular media. Using an absolute fold change of 1.5 cut off and q-value of ≤0.1, we identified 482 altered genes with 284 upregulated and 198 downregulated (Fig. 5A; Supplementary Table S6). Analysis of the differentially regulated genes using the Metascape portal showed enrichment of several biological processes. Among the highest ranking ones was the regulation of cell migration that is linked to the observed antimigratory effect of ERβ and the GO term that is related to response to steroid hormone that results from the knockout of ERβ (Fig. 5B). Heat map analysis of the genes annotated to cell migration revealed well known regulators of EMT and invasion (Fig. 5C). To acquire a more comprehensive picture of the altered pathways in ERβ knockout cells, we used a stringent threshold in selecting differentially regulated genes. Applying ≥2.5 as cut off for fold change the GO term that is associated with regulation of actin cytoskeleton reorganization was enriched (Supplementary Fig. S6A). Genes in this ontology group included both Rho GTPases (RhoD, RAC3) that drive actin-associated cell migration and factors (ARHGDIB) that regulate their activity (Supplementary Fig. S6B; refs. 6, 7, 37). Upregulation of genes that are involved in this type of cell migration was previously described as IBC-specific molecular alteration (4).

Increased formation of stress fibers through polymerization of actin (F-actin formation) and transmission of force through focal adhesions are common features of motile cells (38). Given the motile phenotype of ERβ knockout cells and the enrichment of genes that are involved in actin reorganization, we examined whether the receptor represses the formation of actin stress fibers and decreases the number of focal adhesions in IBC cells. As shown in Fig. 5D and E, increased formation of actin stress fibers and focal adhesions was observed in ERβ knockout cells as it is indicated by the immunofluorescence staining of polymerized actin with FITC-conjugated phalloidin and vinculin, a member of focal adhesion complexes. In contrast, reexpression of the receptor in ERβ knockout cells restored the cytoskeletal structure seen in control cells. Analysis of the images revealed that enrichment of ERβ knockout cells in actin filaments and membrane protrusions is accompanied by a significant enlargement in size (Fig. 5D). Because Rho GTPases promote both formation of actin stress fibers and focal adhesions and RhoC has been reported to increase the invasiveness of IBC cells, we examined whether ERβ affects the expression and activity of cytoskeleton remodelers in these cells (6, 7). Using a RhoC activation assay, we initially observed accumulation of active (GTP-loaded) RhoC in ERβ knockout cells (Fig. 5F). Analysis of mRNA levels indicated that elevated RhoC activity is not due to increased transcription in these cells (Supplementary Fig. S7A). In contrast to knockout of ERβ, treatment of IBC cells with the ERβ agonist LY500307 in estrogen-depleted media decreased the active form of RhoC (Fig. 5G). This effect was not due to ERα because LY500307 did not alter the expression of this receptor in IBC cells (Supplementary Fig. S7B; ref. 33). To evaluate the contribution of RhoC to the migratory phenotype of ERβ knockout cells, we knocked down RhoC in these cells using siRNA. Downregulation of RhoC reversed the increased invasion of ERβ knockout cells at a significant degree, suggesting that RhoC inhibition is essential for the antimigratory activity of the receptor (Fig. 5H).

We next examined how ERβ regulates the activity of RhoC. RhoC cycles between an inactive GDP-bound and an active GTP-bound state. Guanine nucleotide exchange factors (GEF) catalyze the exchange of GDP for GTP, while GTPase-activating proteins (GAP) promote GTP hydrolysis (39). In addition, AKT1 phosphorylation and activation of RhoC is associated with actin reorganization and invasion in IBC cells (40). Rho GTPases are also activated downstream of G-protein–coupled receptors (GPCR; ref. 41). GPCRs activate Rho GTPases by binding through the heterotrimeric G proteins to and activating GEFs (41). They can also transactivate growth factor receptors or recruit scaffolding proteins leading to activation of downstream effectors including AKT1 (41). One of the complexes that serves as GEF and is activated downstream of GPCRs is the ELMO1/DOCK. ELMO1 interacts with both heterotrimeric G proteins and Rho GTPases and is required for GPCR-mediated migration of breast cancer cells (24). Our microarray analysis revealed upregulation of ELMO1 and the member of rhodopsin family of GPCRs GPR141 in ERβ knockout cells (Supplementary Table S6), which was confirmed by qPCR (Fig. 6A and B). The repressive effect of ERβ was specific for GPR141 because knockout of the receptor did not alter the mRNA of GPR30, another GPCR that is expressed in IBC tumors and is associated with metastasis in breast cancer (Fig. 6A; ref. 9). Moreover, coimmunoprecipitation experiments showed interaction between GPR141, ELMO1, and RhoC, suggesting that ELMO1 operates downstream of GPR141 in IBC cells (Supplementary Fig. S7C). As in KPL4 cells, knockdown of ERβ with siRNA significantly increased the mRNA of GPR141 and ELMO1 in FC-IBC02 cells (Fig. 6C). Similar to mRNA, the protein levels of GPR141 and ELMO1 were also inversely correlated with that of ERβ in IBC cells (Fig. 6D). Moreover, the downstream effector of Rho ROCK1 and the phosphorylated form of AKT1 were upregulated in absence of ERβ, suggesting that AKT1 may be activated downstream of GPR141 contributing to RhoC activation in ERβKO cells (Fig. 6D; Supplementary Fig. S7D). Similar to the expression of ERβ, its activation by the agonist LY500307 decreased the levels of GPR141 in control cells in estrogen-depleted media (Supplementary Fig. S7E). On further examination of the involvement of ELMO1 and GPR141 on the effects of ERβ in IBC cells, we observed that their downregulation by siRNA reversed the migratory morphology of ERβ knockout cells and substantially decreased invasion (Fig. 6E and F). Treatment of the same cells with pertussis toxin (PTX) that inhibits GPCRs by disrupting their interaction with heterotrimeric G proteins, also significantly reduced the number of invading cells (Fig. 6G). In contrast, overexpression of ELMO1 in control cells promoted the formation of membrane protrusions and increased their invasive potential (Fig. 6H). Consistent with the inhibitory effect on invasion, knockdown of GPR141 completely reversed the activation of RhoC and AKT1 in ERβ knockout cells. Inhibition of GPR141 with PTX had a similar effect on RhoC activity, but it did not alter the phosphorylated AKT1, suggesting that AKT1, although it functions downstream of GPR141, is not direct effector of heterotrimeric G proteins (Fig. 6I). Downregulation of ELMO1 also decreased the activity of RhoC in ERβ knockout cells, albeit to a less extend than GPR141 (Fig. 6I; ref. 42). Conversely, upregulation of ELMO1 in control cells dramatically increased the activity of RhoC (Fig. 6I). In contrast, knockdown of HER3, which can be transactivated by GPCRs and is overexpressed in IBC tumors, did not substantially reduce the activity of RhoC in ERβ knockout cells (Fig. 6I). Taken together, these findings underscore the importance of GPR141 and ELMO1 in migration of IBC cells and their involvement in the antimigratory effect of ERβ.

Genomic analysis of the GPR141 locus revealed that GPR141 and ELMO1 are located near each other on chromosome 7 and in opposite orientation. This finding strengthens the possibility that their expression is regulated by direct binding of ERβ to nearby regulatory elements (Fig. 7A). To investigate this, we carried out chromatin immunoprecipitation (ChIP) experiments to examine whether ERβ interacts with sites that contain EREs within the promoter and close to the first exon of GPR141 and ELMO1 (Fig. 7A). We also analyzed sites from the promoter of ACTB that do not contain ERE. Because of lack of validated ChIP-specific ERβ antibodies, we compared the DNA-ERβ binding in regular media-growing ERβKO and ERβKO + ERβ cells that stably express FLAG-tagged ERβ using an anti-FLAG antibody for immunoprecipitation. As shown in Fig. 7B, a strong binding of ERβ to ERE-containing sites of GPR141 and ELMO1 was observed in cells expressing ERβ. To corroborate the interaction of ERβ with the promoters of GPR141 and ELMO1 and demonstrate the capacity of activating ligands to induce the DNA binding activity of the receptor in IBC cells, we assessed the association of ERβ with the same regulatory elements in ERβKO + ERβ cells following estrogen depletion and treatment with different concentrations of the ERβ agonist LY500307 or estrogen. We included in the same assay a site that was previously found to be exclusively bound by ERβ (43). We observed a substantial increase in the binding of ERβ to GPR141-associated site following treatment with LY500307 or estrogen that was similar to the binding to the previously validated region (Fig. 7C). We also saw increased association of ERβ with the regulatory element of ELMO1 with the higher concentration of LY500307; however, this was lower compared with that of GPR141-related site (Fig. 7C). To assess the functionality of the identified sites, we cloned a region containing the wild-type or mutated versions (base substitution and deletion) of the ERβ binding site of GPR141 into a luciferase reporter construct and assessed the activity in luciferase reporter assays. We observed that knockout of ERβ increases the expression of luciferase that is driven by the promoter with wild-type but no mutated ERE confirming the contribution of this regulatory region to the repressive effect of ERβ on GPR141 expression (Fig. 7D).

Given that GPR141 and ELMO1 were not previously studied in IBC, we examined their expression and function in IBC cells. We initially assessed the expression of each protein in a panel of breast cancer cell lines including five IBC and five non-IBC luminal and TNBC cell lines. We observed selective expression of ELMO1 in IBC because four out of five IBC cell lines presented with significant levels of ELMO1 and only one out of five non-IBC cell lines had comparable expression. GPR141 appeared to have a similar expression pattern with that of ELMO1 (Fig. 7E). Similar to protein, the mRNA levels of ELMO1 and GPR141 were higher in IBC cell lines compared with the non-IBC cells (Fig. 7F). Interestingly, ERβ displayed an opposite mRNA expression pattern with higher levels in non-IBC cell lines proposing an inverse association between its expression and those of ELMO1 and GPR141 (Supplementary Fig. S8A). This is further supported by the fact that SUM190 cells that have the lowest protein expression of ERβ and the most migratory phenotype among the IBC cell lines (Fig. 2A and B) express the highest mRNA and protein levels of ELMO1 and GPR141 (Fig. 7E and F). It is also consistent with the increase in the expression of GPR141 and ELMO1 following knockout of ERβ in IBC cells (Fig. 6A–D). Unlike GPR141 and ELMO1, RhoC mRNA levels did not substantially differ between IBC and non-IBC cells (Supplementary Fig. S8B). It is also interesting to note that SUM149 cells whose migration was slightly affected by ERβ downregulation was the only IBC cell line that did not express ELMO1 and GPR141 further supporting the involvement of both proteins in the antimigratory mechanism of ERβ action in IBC cells (Fig. 7E and F; Supplementary Fig. S4E and S4F). Because of the absence of GPR141 and ELMO1 in these cells, we examined whether ERβ can act directly on RhoC. Although immunoblotting showed that SUM149 cells indeed express higher levels of RhoC than FC-IBC02 and KPL4 cells, downregulation of ERβ was not able to increase the mRNA levels of RhoC in these cells replicating the effect of ERβ in KPL4 cells (Fig. 7G; Supplementary Figs. S7A and S8C).

Although the functional characterization of ERβ has improved our understanding about the mechanism of ER action and the regulation of estrogen signaling in normal and malignant tissues, controversy still surrounds its clinical importance in breast cancer. The progressive decline of ERβ in invasive lesions combined with its anti-invasive and prodifferentiative effects in preclinical models of breast cancer, lend support for a role as suppressor of both breast cancer progression and metastasis (12, 16–19). However, most of these associations were observed during the analysis of breast cancer cells with transfected ERβ and a convincing link between the loss of ERβ function and increased metastasis in breast cancer has been lacking. Besides, other publications have challenged the tumor-repressive properties of ERβ (44). To clarify the exact role of ERβ in metastasis, we focused on inflammatory breast cancer, a highly aggressive carcinoma with high rate of metastasis and poor prognosis (3). Because lack of targeted therapy is recognized as one of the causes of poor prognosis, and IBC tumors were not previously characterized for the presence and function of ERβ, we analyzed clinical specimens from patients that underwent mastectomy (2). We observed significant levels of ERβ in about 30% of tumors, and in support of our initial hypothesis, interrogation of clinical information revealed associations between high expression of ERβ and less metastatic potential. The same correlation is recapitulated in our xenograft models of IBC where the occurrence of distant metastases increased by the loss of ERβ function but declined by treatment with a selective ERβ agonist. It is likely that the antimetastatic activity of ERβ reflects its ability to inhibit single-cell dissemination or the hybrid E/M phenotype that is implicated in collective cell migration that enables the formation of highly metastatic clustered lymphatic emboli (45). Analysis of preclinical models of IBC supports both mechanistic explanations since ERβ was found to promote epithelial differentiation and decrease the invasiveness of IBC cells.

The increased metastasis of IBC has been associated with the overexpression of genes that are involved in actin-based cell migration including members of the Rho family of GTPases (4). Rho GTPases play an important role in cytoskeletal reprogramming, which, by increasing cell motility, controls morphogenesis and cancer cell movement during metastasis. Rho GTPases bind to and activate ROCK1 that in turn induces actin polymerization and actin-myosin contraction. Overexpression of RhoA and ROCK1 is implicated in breast cancer metastasis and upregulation of RhoC is believed to drive metastasis in IBC tumors (6, 46). Our experiments in IBC cell lines demonstrate strong repressive effects of ERβ on RhoC activity. Examination of upstream signaling that includes the direct activators GEFs and GPCRs that activate GEFs identified two novel direct target genes of ERβ, the GPCR GPR141 and the GEF-interacting protein ELMO1 (Fig. 7H; refs. 24, 41). The selective expression of these genes in IBC cells and their strong proinvasive activity in the same tissues underscore their importance both for the biology of IBC metastasis and as potential targets for therapeutic intervention. The possibility of targeting these molecules is strengthened by the fact that GPR141 belongs to the rhodopsin family of GPCRs that respond to high variety of natural ligands and small molecule inhibitors (47) and ELMO1 functions downstream of GPCRs (24).

We believe that our study helps to clarify the confusion that surrounds the presence and function of ERβ in breast cancer (48–50). First, our findings connect the loss of function of endogenously expressed ERβ with the increased metastasis in breast cancer. Therefore, our study stands out from previous investigations that have associated the artificially introduced receptor in breast cancer cells with decreased invasion (13, 17, 19–21). Second, our results demonstrate the capacity of ERβ and its agonists to exert antimetastatic actions in the highly aggressive IBC tumors and provide with mechanistic insights as to how the receptor signals in IBC cells. Furthermore, our clinical investigations provide with associations that support a protective effect of ERβ against disease progression. In light of these associations and because of the lack of biomarkers to identify patients with IBC who benefit from standard therapy, our research proposes ERβ as a potential predictor of clinical outcome (3). It is also of equal importance that the expression of ERβ in IBC tumors combined with the antimetastatic effects of ERβ agonists highlight the possibility of using ERβ ligands as a novel treatment to combat IBC metastasis and its associated mortality. On the other hand, the association of high ERβ expression with favorable outcome raises the question as to whether the receptor can be activated by endogenous ligands or act in a ligand-independent manner as it was previously suggested (10). Although this may be valid when the receptor is expressed at high levels, low expression of ERβ that occurs in the majority of tumors may require a potent and selective agonist to enhance its antimetastatic activity. Another question of equal significance stems from the variable expression of ERβ in IBC tissues, which suggests that there are tumors that can retain and others that lose the ability to express functional ERβ. Therefore, it is of paramount importance to identify what signaling pathways in tumor cells and the tumor microenvironment allow the expression of ERβ and why this occurs. Answers to these questions will be useful to establish ERβ as an important factor in both the prognosis and treatment of breast cancer metastasis.

No disclosures were reported.

C. Thomas: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. I.V. Karagounis: Formal analysis, methodology. R.K. Srivastava: Formal analysis. N. Vrettos: Formal analysis, methodology. F. Nikolos: Formal analysis. N. Francois: Formal analysis. M. Huang: Formal analysis. S. Gong: Formal analysis. Q. Long: Formal analysis. S. Kumar: Formal analysis. C. Koumenis: Resources. S. Krishnamurthy: Formal analysis, methodology. N.T. Ueno: Resources, formal analysis. R. Chakrabarti: Formal analysis, methodology. A. Maity: Resources, funding acquisition.

We thank Nektaria Leli, Hann-Hsiang Chao, Jayashree Karar, Sarah Hagan, Ioannis Verginadis, (University of Pennsylvania, Philadelphia, PA), Igor Bado (Baylor College of Medicine, Houston, TX), and Anders Strom (University of Houston, Houston, TX) for technical assistance. We thank James Monslow (University of Pennsylvania, Philadelphia, PA) for assistance with the analysis of immunofluorescence images. We also thank Jan-Ake Gustafsson (University of Houston, Houston, TX) for kindly providing the ERβ 503 antibody. We would also like to acknowledge the Penn Vet Comparative Pathology Core and Penn Center for Musculoskeletal Disorders Histology Core for embedding, sectioning and slide evaluation. Similar, we thank the Penn Cytomics and Cell Sorting Laboratory for assistance with cell sorting. We also thank the Penn Molecular Profiling Facility for microarray services and John Tobias at Penn Genomic Analysis Core for bioinformatics support. We finally thank Jie Willey, Anita Wood, Lily Villareal and Kenichi Harano of Morgan Welch Inflammatory Breast Cancer Research Program and Clinic at the UTMDACC for supporting the analysis of human samples. This work was supported by grants from the Pennsylvania Breast Cancer Coalition and NCI-R01 (R01CA237200 to C. Thomas) and NCI-R01CA182747 (to A. Maity And C. Koumenis).

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.