Interaction Between MUC1 and STAT1 Drives IFITM1 Overexpression in Aromatase Inhibitor–Resistant Breast Cancer Cells and Mediates Estrogen-Induced Apoptosis

An estimated 70% of breast cancers express the estrogen receptor alpha (ERα), the progesterone receptor (PgR), or both, and estrogen stimulation of these receptors is a significant factor in the development and growth of breast cancer (1). In the context of ER-positive breast cancer, disrupting estrogen activity using aromatase inhibitors (AI) represents the primary and most effective approach to therapy. AIs inhibit the growth of ER-positive breast cancer by blocking the aromatase enzyme, which regulates the conversion of androgen to estrogen in peripheral tissues, thus resulting in an almost complete loss of circulating estrogen (2). Despite the effectiveness of these agents, however, 30% to 40% of patients eventually develop resistance to AI treatments. Clinical recurrence of breast cancer after AI treatment usually occurs within 3 years of the conclusion of therapy (3, 4). A better understanding of the mechanisms of AI resistance may contribute to the development of new therapeutic strategies and aid in the search for new therapeutic targets and agents.

Mucin 1 (MUC1) is a transmembrane heterodimeric glycoprotein that is localized to the apical border of epithelial cells and is aberrantly overexpressed in approximately 90% of breast cancers (5–8). The MUC1 N-terminal (MUC1-N) subunit is the heavily glycosylated mucin component of the heterodimer and it supports the mucous layer, functions in cell adhesion and immune suppression, and aids in the trapping of microbes (9). The MUC1 C-terminal (MUC1-C) subunit functions as an oncoprotein by promoting signaling pathways that control cell proliferation and survival (10, 11). Of relevance to breast cancer, MUC1-C is known to associate with ERα upon stimulation by 17β-estradiol (E₂; ref. 12). MUC1-C binds directly to the ERα DNA-binding domain and stabilizes ERα by blocking its ubiquitination and degradation. MUC1-C also enhances ERα promoter occupancy, increases recruitment of coactivators, and stimulates ERα-mediated transcription (12). In contrast, loss of MUC1 confers inhibition of ERα translational activity and cell proliferation (13). Clinically, MUC1 expression correlates with ERα levels in breast tumors and its overexpression is associated with resistance to tamoxifen therapy (6, 12, 14). Its role in AI resistance, however, is not known.

MUC1 expression has previously been shown to be upregulated by IFNγ in cell lines derived from breast carcinomas (15). Most recently, we published that the IFNα signaling pathway is hyperactivated in AI-resistant breast cancer cells and that these resistant cells constitutively overexpress several IFNα-stimulated genes (ISG), which promote an aggressive phenotype (16). Of note, we found that IFN-induced transmembrane protein 1 (IFITM1), a critical downstream target of the IFNα signaling pathway, was overexpressed in AI-resistant cells and AI-resistant breast tumors. Targeting IFITM1 expression induced cell death in the resistant cells in vitro and in vivo (16, 17). Of particular interest was the observation that estrogen treatment suppressed IFITM1 expression in AI-resistant MCF-7:5C cells, which was associated with cell death (16). IFITM1 expression is regulated by IFNs through activation of the JAK/STAT signaling pathway. Previous studies have reported that MUC1 can interact with STAT1 protein in breast cancer cells to maintain activating phosphorylation and that this interaction is associated with poor overall survival (18). We hypothesize that MUC1 augments IFITM1 expression in AI-resistant MCF-7:5C cells and that crosstalk between MUC1/IFITM1 promotes an aggressive phenotype in the resistant cells.

In this study, we investigated the interaction between MUC1 and IFITM1 in AI-resistant MCF-7:5C cells and AI-sensitive T-47D and MCF-7 cells. We examined whether estrogen treatment or inhibition of JAK/STAT signaling can disrupt this crosstalk and thus contribute to death in the resistant cells. We found that siRNA knockdown of MUC1 markedly reduced IFITM1 expression in AI-resistant MCF-7:5C cells, which was associated with cell death, and that treatment with E₂ further enhanced this phenotype. In addition, we found that MUC1, STAT1, and STAT2 directly bound to the IFITM1 promoter induce its expression through modulation of JAK/STAT1 signaling in the resistant cells. Finally, using tissue microarray of 94 ER⁺ breast tumors and in silico analysis of 4,151 patient samples in publicly available databases, we demonstrated that high MUC1 and IFITM1 expression correlate with poor patient outcome and overall survival. Together, these findings demonstrate a critical role for MUC1/IFITM1 crosstalk in promoting the aggressiveness of AI-resistant breast cancer and they suggest that inhibiting MUC1 or IFITM1 expression either with low-dose estrogen therapy or JAK/STAT inhibition has potential as a viable treatment option for AI-resistant breast cancer.

The MCF-7 cell line was obtained from Dr. V. Craig Jordan (University of Texas MD Anderson Cancer Center, Houston, TX) and maintained in RPMI1640 medium supplemented as described previously (19). The long-term estrogen-deprived human breast cancer cell line; MCF-7:5C was cloned from parental MCF-7 cells following long-term (>12 months) culture in estrogen-free medium composed of phenol red–free RPMI1640 supplemented with 10% FBS treated three times with dextran-coated charcoal (Thermo Fisher Scientific, catalog no. 7440-44-0; ref. 19). The T-47DA:18 (20) cell line (hereafter referred to as T-47D) was derived from T-47D (21) cells originally obtained from ATCC and cultured in the same media as the MCF-7 cells. All cell lines were cultured at 37°C under 5% CO₂.

Cells were seeded in 6-well plates and allowed to acclimatize overnight. Following 24-hour treatment as indicated with 1 nmol E₂ (Sigma, catalog no. E8875), 72-hour treatment with siMUC1 (Santa Cruz Biotechnology, catalog no. SC-37266), or 48-hour treatment with ruxolitinib/Jakafi (Rux; Cayman Chemical, catalog no. 11609) cells were harvested. Total cell lysate was separated by gel electrophoresis and transferred to polyvinylidene difluoride membrane as described previously (17, 22). Target proteins were detected using either anti-MUC1 (Santa Cruz Biotechnology, catalog no. SC-7313), anti-ERα (Santa Cruz Biotechnology, catalog no. SC-544), anti-P-STAT1 (Santa Cruz Biotechnology, catalog no. SC-8394), anti-P-STAT2 (Cell Signaling Technology, catalog no. 88410S), anti-STAT1 (Santa Cruz Biotechnology, catalog no. SC-464), anti-STAT2 (Santa Cruz Biotechnology, catalog no. SC-514193), anti-PARP-1 (Santa Cruz Biotechnology, catalog no. SC-8007), anti-IFITM1 (Santa Cruz Biotechnology, catalog no. SC-374026), or anti-β-actin (Cell Signaling Technology, catalog no. 3700S) antibodies. The appropriate horseradish peroxidase (HRP)-conjugated secondary antibody (Cell Signaling Technology, catalog no. 7076S and catalog no. 7074S) was applied and the positive bands were detected on autoradiography film as described previously (17, 22).

Cells were seeded in 6-well plates and allowed to acclimatize overnight. Following 24-hour treatment with 1 nmol E₂ or siMUC1, RNA was isolated using the RNeasy Mini kit (Qiagen, catalog no. 74104). First-strand cDNA synthesis was performed from 3 μg total RNA using M-MLV Reverse Transcriptase (Invitrogen, catalog no. 28025-013) on a Bio-Rad MyCycler. RT-PCR was conducted using the ViiA 7 Real-Time PCR system (Applied Biosystems) and SYBR Green Reagent (Applied Biosystems, catalog no. 4367659) with 25 pmol primers specific for human MUC1 (sense: 5′-ACCTACCATCCTATGAGCGAG-3′; antisense: 5′-GGTTTGTGTAAGAGAGGCTGC-3′), IFITM1 (sense: 5′-GGATTTCGGCTTGTCCCGAG-3′; antisense: 5′- CCATGTGGAAGGGAGGGCTC-3′), ERα (sense: 5′-AAGAGGGTGCCAGGCTTTGT-3′; antisense: 5′-CAGGATCTCTAGCCAGGCACAT-3′), STAT1 (sense: 5′- CCGCCATGTTTACAGCAGAT-3′; antisense: 5′-GTCCCCTAGGACCTCCTCAT -3′), and STAT2 (sense: 5′-GCAGCACCATTTGCGGAA -3′; antisense: 5′-ACAGGTGTTTCGAGAACTGGC-3′). PUM1 was used as the internal control (sense: 5′-TCACCGAGGCCCCTCTGAACCCTA-3′; antisense: 5′-GGCAGTAATCTCCTTCTGCATCC T-3′). Relative mRNA expression level was determined as the ratio of the signal intensity to that of PUM1 using the formula: 2^−ΔC_t. When cells were treated, fold change in gene expression was normalized to PUM1 and then compared with the untreated value for that cell line using the formula: 2^−ΔΔC_t.

IF was performed as described previously (17). Briefly, cells were seeded in 2-well slides and allowed to acclimatize overnight. Following 24-hour treatment with 1 nmol E₂, cells were fixed with methanol. Because of the use of mouse antibodies on mouse tissue, blocking and antibody dilution were performed using the Mouse on Mouse (MOM) Kit (Vector Labs, catalog no. FMK-2201) following manufacturer's instructions. Sections were stained using antibodies against anti-MUC1 (Santa Cruz Biotechnology, catalog no. SC-7313), anti-ERα (Santa Cruz Biotechnology, catalog no. SC-544). Secondary antibodies were FITC (Santa Cruz Biotechnology, catalog no. SC-2359) or Texas Red (Santa Cruz Biotechnology, catalog no. SC-2781) conjugated. Slides were visualized on a Leica TCS SPE confocal microscope in the Confocal Imaging Core at The University of Kansas Medical Center (Kansas City, KS). Images were collected and analyzed using the Leica LAS AF Lite software (Leica Biosystems).

Cells were transiently transfected with siRNA for MUC1 (Santa Cruz Biotechnology, catalog no. sc-37266) or a scrambled negative control (Santa Cruz Biotechnology, catalog no. sc-37007). The MUC1 and control siRNAs were pools of three target-specific 20–25 nt siRNAs. Cells were seeded the night before transfection and allowed to reach 60% confluence by the time of transfection. Twenty nanomoles of each siRNA was introduced using Lipofectamine 2000 (Invitrogen, catalog no. 11668-027) in OptiMEM Reduced-Serum Medium (Gibco, catalog no. 11058-021) according to the manufacturer's instructions. After overnight incubation, the transfection mixture was replaced with normal culture medium, containing E₂ only where indicated.

Cells were assayed for viability and proliferation in 24-well plates in triplicate in estrogen-free medium, supplemented with E₂ only where indicated for 72 hours. After treatment, cells were counted by Trypan blue (Sigma, catalog no. T8154) exclusion direct cell counts.

Cells were seeded in 2-well slides and allowed to acclimatize overnight. After the indicated treatment conditions, TUNEL staining was conducted using the Click-iT Plus TUNEL Assay Kit (Invitrogen, catalog no. C10618) following manufacturer's instructions. Tumor samples were deparaffinized by clearance in xylene, rehydrated through graded ethanol series, and then subjected to TUNEL staining as above. The average TUNEL intensity was quantified using the red color channel on ImageJ software for a minimum of three images.

For IFITM1 promoter assays, 0.8 μg of plasmid DNA and the pRL CMV Renilla vector were used as described previously (22). For analysis of IFITM1 promoter activity, the pGL3 plasmid with the first 750 nucleotides of the IFITM1 promoter inserted (pGL3-IFITM1 [−750/−1]), was used (22). The pGL3-Basic-IRES was a kind gift from Joshua Mendell (Addgene, catalog no. 64784; ref. 23). After 24 hours, transfection reagent was replaced with normal cell culture media containing Rux where indicated. Luciferase and Renilla activities were measured 24 hours later using the Dual-Luciferase Reporter Assay Kit (Promega, catalog no. E1910) according to the manufacturer's instructions on a BioTek Synergy 4 microplate reader using the Gen 5 data analysis software (BioTek Instruments).

For coimmunoprecipitation (coIP) experiments, cell lysates were collected in RIPA buffer (Thermo Scientific, catalog no. 89901) with protease inhibitor cocktail (Roche Diagnostics, catalog no. 11836-153-001) and phosphatase inhibitor (Sigma, catalog no. P0044), and sonicated on ice. Cell lysates containing equivalent protein concentrations (5 μg) were precleared with 50:50 Protein A/G–coated magnetic beads (Invitrogen, catalog no. 10001D/catalog no. 10003D) then incubated overnight at 4°C with 2 μg appropriate antibody or control IgG. 50:50 Protein A/G–coated magnetic beads were then added for the final 1 hour of incubation time. Immune complexes were washed three times with PBS, resuspended in Laemmli sample buffer containing dithiothreitol and β-mercaptoethanol (Invitrogen, catalog no. NP0007), boiled for 5 minutes, and subjected to Western blotting analysis.

Chromatin immunoprecipitation (ChIP) was performed using the ChIP-IT Express Kit (Active Motif, catalog no. 53008) according to the manufacturer's instructions using sonication as the method for chromatin shearing. Lysates were immunoprecipitated (IP) overnight (18 hours) with the following antibodies anti-STAT1 (Santa Cruz Biotechnology, catalog no. SC-464), anti-STAT2 (Santa Cruz Biotechnology, catalog no. SC-514193), anti-MUC1 (Santa Cruz Biotechnology, catalog no. SC-7313) or an equal amount of mouse (Santa Cruz Biotechnology, catalog no. SC-2762) or rabbit IgG (Santa Cruz Biotechnology, catalog no. SC-2027). Resulting DNA was analyzed using qPCR as described previously (24), and data are represented as a percentage of input DNA.

Eight- to 10-week-old virgin female athymic nude (nu/nu) or NOD-SCID IL2Rgamma^null (NSG) mice were purchased from Jackson Laboratories and used for in vivo experiments. Animal experiments were conducted following protocols approved by the University of Kansas School of Medicine Animal Care and Use (ACUP#: 2016-2341). Cells were suspended in 50:50 PBS/Matrigel (Corning, catalog no. 354262) and bilaterally injected into fourth mammary fat pads as described previously (25). A total of 3 × 10⁶ cells were delivered per injection in a volume of 100 μL. Tumor volume was calculated weekly as described previously (25). When tumors reached a mean volume of 0.20 cm³, groups of 5–15 mice were randomly assigned to treatment groups. Where indicated, mice were administered 50 μg/kg body weight of ruxolitinib (LC Laboratories, catalog no. R6688) suspended in methylcellulose (Acros Organics, catalog no. 9004-67-5) by oral gavage every other day. Experiments were terminated after 43 days due to gastrointestinal side-effects in the treatment group. When using MCF-7 and MCF-7:5C cells for estrogen experiments, capsules containing 1:4 β-estradiol/cholesterol, which produce a mean serum estradiol level of approximately 80 pg/mL, were implanted subcutaneously as described previously (26).

IHC staining was performed after tissue deparaffinization by clearance in xylene and hydration through graded ethanol series as described previously (17). Sections were stained using primary human antibodies targeted against anti-IFITM1 (Santa Cruz Biotechnology, catalog no. SC-374026), anti-MUC1 (Santa Cruz Biotechnology, catalog no. SC-7313), anti-P-STAT1 (Santa Cruz Biotechnology, catalog no. SC-8394), and horseradish peroxidase–conjugated biotinylated secondary antibodies (Cell Signaling Technology, catalog nos. 7076S and 7074S). Immunoperoxidase signal was produced using 3,3′-Diaminobenzidine (DAB; Vector Laboratories, catalog no. SK-4100) and amplified using the Vectastain Elite ABC Kit (Vector Laboratories, catalog no. PK-4000). Tissue sections were counter stained using hematoxylin and mounted in xylene. Slides were imaged on a Nikon Eclipse 80i Upright Microscope in the Imaging Core of The University of Kansas Medical Center (Kansas City, KS).

A tumor microarray of 104 patient samples as described previously (17) was analyzed for MUC1 and IFITM1 expression and scored by staining intensity by three independent analysts blinded to the patient details. The MUC1 and IFITM1 antibodies were validated using samples of human colorectal cancer as described previously. The MUC1 antibody was also validated using normal human breast tissues samples taken from reduction mammoplasties at KUMC (Supplementary Fig. S7A). Clinicopathologic data including age, race, clinical stage, and Her2 staining were published previously (17). Two cores from each tumor were analyzed and averaged in an attempt to account for the impact of tumor heterogeneity on protein expression. MUC1 staining intensity was quantified manually on a scale of 0–3 where 0 means no staining, 1+ is faint staining, 2+ is moderate staining and 3+ is strong staining (Supplementary Fig. S7B). Cores were scored separately for IFITM1 and MUC1 expression by three independent individuals before accessing patient medical records. Any discrepancies were resolved by group consensus. Demonstration of IFITM1 staining intensity and final distribution of IFITM1 expression was published previously on the same scale (17). For Kaplan–Meier plotting, a score of 0 to 1 was labeled as negative (−) and 2 to 3 as positive (+). Histopathologic parameters were extracted from patient pathology records. Clinical parameters including disease recurrence and survival data were obtained from electronic medical records for 94 of 104 patients.

We previously reported the development of an AI-resistant breast cancer cell line, MCF-7:5C, which was derived from parental MCF-7 cells following long-term estrogen deprivation (19, 27). A unique phenotype of MCF-7:5C cells is that they express high levels of functional ERα but lack PgR and undergo apoptosis in the presence of 17β-estradiol (E₂; refs. 27, 28). Our study begins by examining the effect of E₂ treatment on MUC1 expression in AI-sensitive T-47D and MCF-7 cells and AI-resistant MCF-7:5C cells. First, we measured baseline expression of MUC1 and ERα protein and mRNA in all three cell lines and we found that MUC1 protein and mRNA levels were highest in T-47D cells compared with MCF-7 and MCF-7:5C cells; however, ERα protein and mRNA levels were highest in AI-resistant MCF-7:5C cells (Fig. 1A). Next, we confirmed that estrogen differentially regulates MUC1 expression in AI-sensitive T-47D and MCF-7 cells compared with AI-resistant MCF-7:5C cells. Estrogen starvation causes upregulation of ERα in the AI-sensitive cell lines and subsequent estrogen treatment (1 nmol/L E₂) significantly increased MUC1 protein and mRNA level in MCF-7 and T-47D cells within 24 hours; however, in AI-resistant MCF-7:5C cells, estrogen treatment significantly reduced MUC1 protein and mRNA level within 12 hours (Fig. 1B; Supplementary Fig. S1A). Notably, in MCF-7:5C cells, MUC1 levels remained reduced up to 72 hours, which was followed by increased PARP cleavage, and reduced P-STAT1 and IFITM1 expression (Supplementary Fig. S1C). MUC1 downregulation in AI-resistant MCF-7:5C cells was confirmed with immunofluorescence (Fig. 1C). For comparison, similar experiments were performed with the pure antiestrogen ICI 182,780 and we found that ICI 182,780 decreased MUC1 levels in T-47D and MCF-7 cells, but increased MUC1 expression in AI-resistant MCF-7:5C cells (Fig. 1B), further suggesting differential hormonal regulation of MUC1 in AI-sensitive versus AI-resistant breast cancer cells.

To test whether MUC1 expression held any functional significance on the aggression of AI-resistant cells, we first knocked down MUC1 with three unique MUC1-specific siRNAs (Supplementary Fig. S1B) and a combined pool of all three siRNAs (Fig. 2A). Trypan blue direct cell counting confirmed that E₂ treatment significantly enhanced the growth of MCF-7 and T-47D cells, but reduced the growth of AI-resistant MCF-7:5C cells (Fig. 2B), as previously reported (27). Notably, loss of MUC1 resulted in lower cell proliferation in AI-resistant MCF-7:5C cells; however, it did not alter the growth of AI-sensitive MCF-7 and T-47D cells (Fig. 2C). To further investigate the effects of MUC1 loss on AI-resistant MCF-7:5C cells we combined MUC1 knockdown with estrogen treatment. Phase contrast images demonstrated an increase in dead, floating cells (Fig. 2D) and TUNEL staining verified induction of apoptosis most significantly in the MUC1 knockdown plus estrogen-treated cells compared with siMUC1 or estrogen treatment alone (Fig. 2E).

We previously reported that IFITM1, an IFN-stimulated gene, is overexpressed in AI-resistant MCF-7:5C cells and that its overexpression promotes an aggressive phenotype. (16, 17) We therefore examined whether MUC1 crosstalk with JAK/STAT signaling plays a role in regulating IFITM1 expression in MCF-7:5C cells. In Fig. 3A and D, we demonstrate that knockdown of MUC1 with siRNA significantly reduced IFITM1 and P-STAT1 expression in AI-resistant MCF-7:5C cells with a less pronounced effect seen in AI-sensitive T-47D cells. Notably, the AI-sensitive MCF-7 cell line does not express P-STAT1, P-STAT2, or IFITM1. Next, we targeted JAK/STAT signaling using the JAK1/2 inhibitor, ruxolitinib, and found reduced P-STAT1 and IFITM1 expression at the protein and mRNA level in MCF-7:5C and T-47D cells (Fig. 3B and D). Notably, E₂ treatment also reduced P-STAT2 and IFITM1 expression in both cell lines (Fig. 3C). Physical binding of MUC1 with P-STAT1 and P-STAT2 was confirmed by coimmunoprecipitation using ERα as a positive control. Under normal conditions (no treatment), MUC1 is bound to P-STAT1 and P-STAT2 in AI-resistant MCF-7:5C cells, but only P-STAT2 in T-47D cells. Inhibition of STAT phosphorylation with ruxolitinib markedly reduced this association in both cell lines (Fig. 3E). In addition, we confirmed the role of MUC1 in IFITM1 expression using the MUC1 inhibitor, GO-201, in T-47D, MCF-7, and MCF-7:5C cells (Supplementary Fig. S1D). These data support a role for MUC1 upstream of IFITM1 expression. Notably, inhibition of JAK/STAT signaling with ruxolitinib did not significantly enhance the antiproliferative effect of either ICI 182,780 or estrogen in the AI-sensitive or AI-resistant cells, respectively (Supplementary Fig. S2).

Because inhibition of JAK/STAT signaling with ruxolitinib reduced IFITM1 expression in MCF-7:5C and T-47D cells at the mRNA level, we next assessed the effect of ruxolitinib on the IFITM1 promoter. We performed ChIP assay to assess the binding of MUC1, STAT1, and STAT2, to the IFN-stimulated response element (ISRE) located at the -1 position in the IFITM1 promoter. Baseline levels showed greater binding of MUC1, STAT1, and STAT2 to the IFITM1 promoter in AI-resistant MCF-7:5C and AI-sensitive T-47D cells compared with MCF-7 cells (Supplementary Fig. S3B). Upon treatment with ruxolitinib, binding of MUC1, STAT1, and STAT2 was significantly decreased compared with controls in both MCF-7:5C and T-47D cells indicating a role for these proteins in promoting IFITM1 expression (Fig. 3F). Dual luciferase reporter assay supported these findings (Supplementary Fig. S3A). Inhibition of IFITM1 expression had no effect on upstream MUC1 expression (Supplementary Fig. S3C). These studies found a more prominent role for P-STAT2 in driving IFITM1 expression in AI-sensitive T-47D cells, while indicating that IFITM1 expression in AI-resistant MCF-7:5C cells is more dependent on P-STAT1. Overall, these data validate the importance of MUC1 in transcriptional regulation of IFITM1 expression.

Because our in vitro data indicated that estrogen (E₂) treatment significantly reduced MUC1 and IFITM1 expression in AI-resistant MCF-7:5C cells (Fig. 3C) and that this reduction was associated with cell death (Fig. 2), we next determined the effect of E₂ treatment on AI-resistant tumor growth and MUC1 and IFITM1 expression in vivo. For this experiment, AI-sensitive MCF-7 cells and AI-resistant MCF-7:5C cells (3 × 10⁶) were bilaterally injected into the fourth mammary fat pad of NSG mice. As expected, treatment with E₂ significantly enhanced the growth of MCF-7 tumors; however, it markedly reduced the growth of AI-resistant MCF-7:5C tumors (Fig. 4A and B). TUNEL staining of MCF-7:5C tumors revealed that the reduction in tumor growth was due to an increase in apoptosis (Fig. 4C). Similar experiments were performed using athymic nude mice and the results were comparable with those seen with NSG mice (Supplementary Fig. S4A–S4C). Of importance, IHC staining of MCF-7:5C tumors revealed high expression of MUC1, P-STAT1, and IFITM1 under control conditions (-E₂); however, treatment with E₂ completely reduced the expression of all three proteins (Fig. 4D). A similar finding was observed for MCF-7:5C tumors generated in Nude mice (Supplementary Fig. S5A and S5B).

Because our in vitro data revealed that P-STAT1 and IFITM1 expression were markedly elevated in AI-resistant MCF-7:5C cells compared with parental MCF-7 cells and that treatment with the JAK inhibitor ruxolitinib markedly reduced P-STAT1 and IFITM1 expression in these cells (Fig. 5A), we next determined whether ruxolitinib has therapeutic benefits in vivo. MCF-7 and MCF-7:5C cells were bilaterally injected into the fourth mammary fat pad of NSG mice and once the tumors reached approximately 200 mm³ in size the mice were randomized to either the control group (-Rux) or the ruxolitinib-treated group (50 mg/kg). Our data showed that ruxolitinib treatment significantly reduced the growth of AI-resistant MCF-7:5C tumors within the first 2 weeks (Fig. 5B); however, it did not alter the growth of MCF-7 tumors (Fig. 5B) or T-47D tumors (data not shown) throughout the duration of the study. TUNEL staining showed increased apoptosis after ruxolitinib treatment in MCF-7:5C tumors (Fig. 5C), but not in MCF-7 tumors (Supplementary Fig. S6) and this increase in cell death was associated with a reduction in P-STAT1 and IFITM1 expression, as confirmed by Western blot analysis (Fig. 5D) and IHC staining (Fig. 5E).

Normal breast tissue from 6 reduction mammoplasties and tumor samples from a tumor microarray of 94 patients with ER⁺ breast cancer were analyzed by IHC and scored according to staining intensity (Supplementary Fig. S7). χ² analysis confirmed a strong correlation between MUC1 and IFITM1 expression (Fig. 6A). We have previously reported an association between recurrence-free survival and overall survival based on IFITM1 expression alone in this patient cohort (17). Here, MUC1 intensity alone did not predict patient outcomes (Fig. 6B and C); however, positive coexpression of both MUC1 and IFITM1 significantly correlated with recurrence of disease and overall survival (Fig. 6D and E). We sought to verify this correlation in a larger cohort of breast cancers and mined data from BCCRC xenograft, METABRIC, TCGA, and TCGA 2015 datasets. We found that MUC1 and IFITM1 were coexpressed in 13%–27% of breast tumors (Fig. 7A) and that their coexpression was significantly associated with decreased recurrence-free and overall survival (Fig. 7B). These findings suggest that the presence of MUC1 and IFITM1 in breast tumors confers a poor prognosis for patients.

While AIs remain the standard of care for patients with ER-positive breast cancer, approximately 30% of patients will develop resistance to these agents. Mechanisms of resistance to AIs have been extensively studied, yet few novel, targetable regulators of AI resistance have been discovered and used clinically. Resistance can occur through ER-independent signaling, upregulated growth factor signaling, and changes to cell-cycle regulators and apoptotic mechanisms (29). We have previously reported that JAK/STAT signaling drives IFITM1 expression in our AI-resistant model, which directly promotes the aggressive phenotype of these cells. We found that targeting IFITM1 causes the death of AI-resistant cells in vivo and in vitro (16, 17). In this study, we demonstrate for the first time that MUC1 augments the underlying JAK/STAT signaling, enhances IFITM1 expression, and promotes an aggressive phenotype in AI-resistant breast cancer cells. Clinically, we found that coexpression of MUC1 and IFITM1 positively correlated with aggressive disease and resistance to therapy. Notably, in vivo studies using NSG mice demonstrated that treatment with ruxolitinib, a JAK/STAT inhibitor, and estrogen (E₂) effectively reduced the growth of AI-resistant breast tumors with no significant effect seen for AI-sensitive tumors. Overall, these findings suggest that targeting MUC1 and the JAK/STAT signaling pathway simultaneously might be a novel strategy to treat a subset of patients with AI-resistant ER⁺ breast cancer.

We first investigated the correlation between MUC1 and IFITM1 in promoting breast cancer aggressiveness. MUC1, a known oncoprotein, is overexpressed in many types of cancer, including breast (30). Its C-terminus is a coactivator in many tumor-promoting signaling pathways in both ER-positive and -negative breast cancers (31, 32). Here, we elucidate a role for MUC1 in augmenting IFNα-mediated JAK/STAT signaling in a subset of AI-resistant tumors. Following constitutive activation of the type 1 IFNα receptor (IFNAR) by IFNα, MUC1 binds to STAT1 to maintain phosphorylation and regulate the expression of IFN-stimulated genes (ISG), including IFITM1 (18). Previously, our laboratory has shown that IFITM1 drives the aggressive phenotype of AI-resistant breast cancer (17). The IFITM1 promoter contains both IFN-stimulated response elements (ISRE) and STAT response elements. We show for the first time through in vitro (Fig. 3) and in vivo (Fig. 5) studies that MUC1 interacts with STAT1/2 to drive expression of IFITM1 and promote an aggressive phenotype in AI-resistant cells. Specifically, we found that MUC1 enhances STAT binding to the IFITM1 promoter through maintenance of STAT activation, which is consistent with reports of MUC1 cooperativity with STAT1 signaling (ref. 18; see visual overview). In addition, MUC1 may interact with chromatin remodelers and recruit coactivators to cause upregulation of IFITM1 expression, as it has a known role in enhancing recruitment of CBP, EzH2, c-Src, and DNMT (33–36). Future research should investigate the role of MUC1 in regulating SWI/SNF complexes, which are known to regulate ISG expression, including IFITM1 (37). Downregulation of MUC1 abrogates IFITM1 protein and mRNA expression (Fig. 3). MUC1 also associates with the IFITM1 ISRE promoter element in conjunction with STAT1/2 (Fig. 3), playing a role in constitutive activation of IFITM1 to maintain aggressiveness and AI resistance. IFITM1 expression causes tumor progression by downregulating the CDK inhibitor p21 (17). p21 increases the expression of proteins involved in cell progression and proliferation, like PCNA, Ki67, cyclin D1 and E, pRb, and E2F1 (38). The key role of MUC1 in regulating IFITM1 is of primary importance in breast cancer as high coexpression of MUC1 and IFITM1 together may predict patient outcomes, demonstrated through tumor microarray analysis of patients with ER-positive breast cancer (Fig. 6). Not only does IFITM1 expression predict overall outcome (17), as previously reported, but high expression of MUC1 and IFITM1 was associated with disease recurrence and poor overall survival in large patient datasets, emphasizing the opportunity to use MUC1 and IFITM1 as biomarkers of aggressive disease. Overexpression of MUC1 and IFITM1 was also examined through data mining and MUC1 and IFITM1 were found to co-occur in 13% to 27% of breast cancers from four distinct databases (Fig. 7). This resistance-driving crosstalk led us to further probe the interaction between MUC1 and IFITM1 and investigate strategies for targeting these tumors in the clinic.

Mechanistically, we uncovered a role for MUC1 in inducing IFITM1 expression and found that disrupting this interaction leads to cell death (Visual Overview). Notably, we observed differential regulation of MUC1 by estrogen in AI-resistant cells compared to AI-sensitive cells. In AI-sensitive cells, ERα is only active upon E₂ treatment, which then upregulates MUC1 expression. MUC1 is influenced by estrogen treatment and ERα expression due to estrogen response elements (ERE) within its promoter. In addition, MUC1 can directly bind ERα which prevents its ubiquitination and degradation (30). This interaction is enhanced in the presence of estrogen and promotes the binding of ERα with its coactivators (12). Loss of MUC1 confers inhibition of ERα translational activity and cell proliferation (13). In contrast, treatment of AI-resistant cells and tumors with E₂ results in MUC1 suppression and induction of death (Figs. 1 and 4). This dichotomy may tie back into initial mechanisms of AI resistance.

It is well-established that estrogen promotes breast cancer development and progression; however, preclinical data from our laboratory (16, 25, 27) and other investigators have previously shown that estrogen is also capable of inducing cell death in ER-positive breast cancers that have acquired resistance to estrogen deprivation (i.e., AI-resistant breast cancer). Clinically, low-dose estrogen therapy (6 mg/daily E₂) has also been shown to be effective in treating a subset of postmenopausal women (30%–40%) with ER-positive AI-resistant breast cancer (39, 40). The ability of estrogen to induce death in AI-resistant cells is partly attributed to activation of the mitochondrial death pathway. However, our current study highlights a novel pathway in which estrogen also downregulates MUC1 expression, which reduces IFITM1 expression in AI-resistant cells. The loss of MUC1 and IFITM1 leads to cell death in AI-resistant MCF-7:5C cells but not AI-sensitive MCF-7 and T47D cells. AI-resistant cells can rely on ligand-independent signaling where ERα is constitutively active and promotes upregulation of MUC1. Estrogen may interfere with ERα stability, which in turn may abrogate the transcription and/or stabilization of MUC1 and its rate of turnover. Alternatively, it is possible that E₂ treatment may disrupt ERα-coactivator binding at the MUC1 promoter thereby reducing MUC1 and consequently IFITM1 expression in the resistant cells (41). MUC1 can also be regulated by other nuclear hormone receptors, including PgR and the androgen receptor (AR), which may be inhibited by activation of ERα (42–44). Uncovering this mechanism allowed us to isolate MUC1 and consequently, IFITM1, as targets of E₂-induced cell death in AI-resistant cells. MUC1 can also be constitutively overexpressed in ER-negative breast cancer through amplification of the MUC1 gene and is associated with patient prognosis (31, 45). Recently, high MUC1 expression has been associated with targetable CD274/PDL1 gene activation, PD-L1 expression, immune evasion and epithelial to mesenchymal transition in ER-negative breast cancer (32).

In addition to our in vitro data, we also investigated the in vivo effects of estrogen treatment on AI resistance and IFITM1 expression. Estrogen acts paradoxically in AI-resistant cells and instead of promoting their growth, induces death. We observed this effect in AI-resistant MCF-7:5C cells upon treatment with E₂ (Fig. 2D and E). Perhaps this occurs because of the downregulation of MUC1 and IFITM1 levels after E₂ treatment (Fig. 3A), which leads to an increase in p21, a decrease in phospho-p21, and cell death (17). Significantly, E₂ treatment caused tumor regression and downregulation of these proteins in vivo, exclusively in the AI-resistant cell line (Fig. 4).

As an alternative therapy to estrogen treatment or potentially as a combination therapy, we investigated the effects of ruxolitinib on MUC1 and IFITM1 expression and AI resistance. Ruxolitinib is a small-molecule inhibitor of JAK1 and JAK2 and so widely dampens the JAK/STAT signaling pathway. Ruxolitinib is currently under clinical investigation in combination treatment with trastuzumab in metastatic HER2⁺ breast cancer (32), with chemotherapy in triple-negative inflammatory breast cancer (46), and with pembrolizumab in metastatic triple-negative breast cancer (TNBC; ref. 47). However, a phase II clinical trial of ruxolitinib in patients with metastatic TNBC demonstrated that patients treatment was ineffective in this subpopulation of patients as a single agent (48). These patients were selected on the basis of high P-STAT3 expression; alternatively, more successful results may have been achieved by selecting patients with high MUC1 and IFITM1 expression. The dichotomy of treatment efficacy between HER2⁺ and triple-negative breast cancers and our research in AI-resistant breast cancer cells may be due to the effects of ruxolitinib on downregulating IFITM1, a linchpin in maintaining AI resistance, specific to ER-positive breast cancer. Our data indicate that ruxolitinib, as a single agent or in combination with other therapies (i.e., estrogen), could be effective in treating patients with ER-positive AI-resistant or IFITM1-expressing breast cancer. This could be due to the downregulation of MUC1 and IFITM1 expression upon ruxolitinib treatment that occurs both in vitro and in vivo (Figs. 3C–F and 5). We should note that the pure antiestrogen, fulvestrant (ICI 182,780), which is known to degrade ER, is currently a second-line therapy for ER-positive breast cancers that fail tamoxifen and/or AI treatment. ICI 182,780 profoundly inhibited the proliferation of AI-sensitive MCF-7 and T47D cells; however, its inhibitory effect was not augmented when combined with ruxolitinib. Similarly, ruxolitinib alone blocks IFITM1 expression and significantly inhibited the proliferation of AI-resistant MCF-7:5C cells; however, the inhibitory effect of ruxolitinib was not further enhanced when combined with fulvestrant (ICI 182,780). Although ruxolitinib blocks IFITM1 expression in the AI-sensitive T-47D cells, the JAK/STAT inhibitor did not significantly inhibit cell proliferation in this model, suggesting that AI sensitivity and ER expression is a main determinant of ruxolitinib sensitivity. These data suggest that the clinical benefit of ruxolitinib may be limited to high IFITM1-expressing AI-resistant breast tumors.

Finally, we demonstrated that MUC1 knockdown in combination with estrogen treatment causes the most effective increase in cell death in AI-resistant cells. We also examined the effects of a MUC1 inhibitor, GO-201 (45, 49), on IFITM1 expression and observed decreases in IFITM1 expression in vitro (Supplementary Fig. S1D). Currently, a MUC1 third-generation inhibitor, GO-203, is undergoing phase I/II clinical trials for patients with acute myeloid leukemia (45, 49). However, in vivo experiments within our laboratory have shown limited response to GO-201 and GO-201 demonstrates a cytotoxic effect on normal human fibroblasts (50). In addition to small-molecule inhibitors, MUC1 vaccines are also being investigated for their efficacy in multiple tumor types (51, 52). Our research elucidated two novel mechanisms to target IFITM1 expression both in vitro and in vivo through treatment with estrogen or ruxolitinib. The combination of these therapies will be explored in future research.

In conclusion, this study demonstrated that MUC1 is a key regulator of JAK/STAT signaling and IFITM1 transcription, which promotes AI-resistant breast cancer cell aggression and survival. MUC1 and IFITM1 coexpression could be used as a biomarker to isolate patients who may successfully be treated with estrogen, Rux, or combination therapy. Future studies are needed to determine the exact role that IFITM1 has in tumorigenesis and how combination therapies can effectively treat AI-resistant patients. Our data suggests that MUC1 and IFITM1 crosstalk may be a marker of aggressive breast cancer and that current therapies can be adapted to target resistant breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: T.E. Escher, A.J. Lui, E.S. Geanes, O. Tawfik, C.R. Hagan, J. Lewis-Wambi

Development of methodology: T.E. Escher, A.J. Lui, E.S. Geanes, O. Tawfik, C.R. Hagan, J. Lewis-Wambi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.E. Escher, A.J. Lui, E.S. Geanes, K.R. Walter, O. Tawfik, J. Lewis-Wambi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.E. Escher, A.J. Lui, E.S. Geanes, C.R. Hagan, J. Lewis-Wambi

Writing, review, and/or revision of the manuscript: T.E. Escher, A.J. Lui, E.S. Geanes, K.R. Walter, O. Tawfik, C.R. Hagan, J. Lewis-Wambi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.E. Escher, E.S. Geanes

Study supervision: J. Lewis-Wambi

The authors would like to thank the KUMC Confocal Imaging Facility and Imaging Core. The confocal core is supported, in part, by NIH/NIGMS COBRE grant P20GM104936. We acknowledge support from the University of Kansas (KU) Cancer Center's Biospecimen Repository Core Facility staff for helping obtain human specimens and performing histological work. The authors also acknowledge support from the KU Cancer Center's Support Grant (P30 CA168524). This study was supported, in part, by grants from the Department of Defense (W81XWH-12-1-0139, to J. Lewis-Wambi), the National Cancer Institute (K01CA120051, to J. Lewis-Wambi), start-up funds from the University of Kansas Medical Center (KUMC; to J. Lewis-Wambi), the KUMC Biomedical Research Training Program (BRTP, to A.J. Lui), the National Cancer Institute (1F30CA203160-01, to A.J. Lui), the American Medical Association (AMA) Foundation (to A.J. Lui), the University of Kansas Cancer Center (CCSG grants P30 CA168524-0, to J. Lewis-Wembi, C.R. Hagan, and K.R. Walter), and the University of New Mexico Comprehensive Cancer Center (GMaP grant 3P30CA118100-12S2, to J. Lewis-Wembi). We acknowledge the Flow Cytometry Core Laboratory at the University of Kansas Medical Center, which is sponsored, in part, by the NIH/NIGMS COBRE grant P30 GM10332.

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.