Mutations in ESR1 that confer constitutive estrogen receptor alpha (ER) activity in the absence of ligand are acquired by ≥40% of metastatic breast cancers (MBC) resistant to adjuvant aromatase inhibitor (AI) therapy. To identify targetable vulnerabilities in MBC, we examined steroid hormone receptors and tumor-infiltrating immune cells in metastatic lesions with or without ER mutations. ER and progesterone receptor (PR) were significantly lower in metastases with wild-type (WT) ER compared with those with mutant ER, suggesting that metastases that evade AI therapy by mechanism(s) other than acquiring ER mutations lose dependency on ER and PR. Metastases with mutant ER had significantly higher T regulatory and Th cells, total macrophages, and programmed death ligand-1 (PD-L1)-positive immune-suppressive macrophages than those with WT ER. Breast cancer cells with CRISPR-Cas9–edited ER (D538G, Y537S, or WT) and patient-derived xenografts harboring mutant or WT ER revealed genes and proteins elevated in mutant ER cells, including androgen receptor (AR), chitinase-3-like protein 1 (CHI3L1), and IFN-stimulated genes (ISG). Targeting these proteins blunted the selective advantage of ER-mutant tumor cells to survive estrogen deprivation, anchorage independence, and invasion. Thus, patients with mutant ER MBC might respond to standard-of-care fulvestrant or other selective ER degraders when combined with AR or CHI3L1 inhibition, perhaps with the addition of immunotherapy.

Significance:

Targetable alterations in MBC, including AR, CHI3L1, and ISG, arise following estrogen-deprivation, and ER-mutant metastases may respond to immunotherapies due to elevated PD-L1+ macrophages.

See related article by Arnesen et al., p. 539

Two-thirds of all early (nonmetastatic) primary breast cancers express estrogen receptor alpha (ER). First-line adjuvant treatment for the majority of patients with ER+ disease is aromatase inhibitor (AI) therapy, which blocks conversion of androgens to estrogens by the enzyme aromatase and significantly extends time to recurrence (reviewed in ref. 1). However, all patients who recur will eventually develop AI-resistant metastatic breast cancer (MBC) and acquired point mutations in the ESR1 gene often emerge under the selective pressure of this estrogen-deprivation therapy. While other known mechanisms of endocrine therapy resistance exist, including amplification and fusions of ESR1 (2), point mutations in exon eight encoding the ER ligand-binding domain are the most prevalent. Indeed, ER mutations are detectable in circulating tumor cells (CTC) or metastatic lesions in up to 40% of women with AI-resistant MBC (reviewed in ref. 3). Specific mutations in this region result in amino acid alterations that confer constitutive receptor activity in the absence of estrogen (4–10) and consequent estrogen-independent gene regulation (11–16). In preclinical models, breast cancer cells with mutant ER exhibit increased proliferation in vitro (13, 14) and in vivo (11, 12) compared with cells with wild-type (WT) ER and have a greater propensity to metastasize with or without 17beta-estradiol (E2; refs. 11, 12).

Despite ER being constitutively active in the absence of E2, breast cancer cells harboring mutant ER retain sensitivity to E2. Fulvestrant (Fulv), the selective ER degrader that serves as the current second-line therapy for AI-resistant disease, binds to mutant ER at lower affinity than to WT ER (reviewed in ref. 17). The poor bioavailability of Fulv necessitates intramuscular injections and higher doses are problematic. Hence, combination strategies to lower the effective dose of Fulv or alternative therapies are desirable. Cyclin-dependent kinase-4/6 (CDK 4/6) inhibitors, such as palbociclib (Palbo), improve survival for patients with ER+ MBC and Palbo+Fulv slightly improved survival in patients harboring mutant ER when compared with Fulv alone; however, CDK 4/6 inhibitors are not without side effects (18). Other therapies being tested in combination with Fulv include the CDK7 inhibitor THZ1 that showed enhanced efficacy in preclinical models (12) and second-generation selective ER degraders, such as AZD9496 (19). However, there is clearly a need for novel therapeutic approaches for patients with AI refractory ER+ MBC.

Although both WT and mutant ER arise from ER+ breast cancer that escapes adjuvant AI therapy, it is possible that due to differing biological properties, alternative therapeutic strategies may be required to treat these two forms of AI refractory MBC (11, 20, 21). In the companion article, Arnesen and colleagues molecularly characterize WT versus mutant ER breast cancer cells, studying the most commonly occurring D538G and Y537S mutants associated with poor patient outcomes (4–7, 22, 23). Interestingly, concurrent analyses of new data from Arnesen and colleagues (22) and two other studies (12, 13) indicated that many genes differentially expressed by mutant ER breast cancer cells compared with WT are not E2 regulated. These results suggested that the selective advantages conferred by mutant ER are not solely attributable to constitutive ER activity. Because ER mutations represent an acquired adaptation that facilitates breast cancer survival and recurrence as metastatic disease in women on AI therapy, we postulated that the selective advantages conferred by mutant ER might become more evident by modeling the clinical situation in which the mutations arise (E2 deprivation). Therefore, we examined specific proteins from mutant-enhanced pathways identified by genomics in Arnesen and colleagues (22), under conditions that mimic the altered hormonal milieu that ensues upon AI therapy and anchorage independence, to simulate the selective pressure of treatment and early metastasis. This analysis led to identification of innate immune cell activating signatures that were confirmed by multiplex analysis of tumor-infiltrating immune cells in MBC specimens. We also hypothesized that the long-term E2 deprivation induced by AI therapy blocking conversion of androgens to estrogens might cause tumor cells to shift to reliance on androgens/androgen receptor (AR). The complement of steroid hormone receptor proteins including ER, progesterone receptor (PR), AR, and glucocorticoid receptor (GR), had not previously been examined together in MBC with or without ESR1 mutations. Our approach combining data from preclinical models and biopsies of MBC, identified proteins that confer survival advantages relevant to AI refractory MBC that may serve as vulnerabilities and lead to new therapeutic approaches for this fatal form of the disease.

MBC patient samples

Mutation analysis

Core needle biopsies were acquired from patients, who gave their informed written consent, with ER+/HER2 measurable or evaluable MBC without central nervous system disease enrolled in clinical trial NCT02953860 (COMIRB 16-1001). Median age of patients was 61 years (46–87); PS 1 (0–1); a median of 2 prior chemotherapy and 2 prior hormonal therapies for metastatic disease (including 7 with prior Fulvestrant), and 90% had visceral disease. An additional three biopsies were from NCT01597193 (COMIRB 12-0970). Formalin-fixed paraffin-embedded (FFPE) sections were analyzed for mutations in ESR1 exon 8 as well as 67 other gene hotspots frequently altered in cancer using a modified Archer VariantPlex Solid Tumor Assay through the CMOCO Laboratory (Department of Pathology, University of Colorado, Aurora, CO). Mutation mutual exclusivity analyses were performed for commonly occurring mutations using cBioPortal (cBioPortal, RRID:SCR_014555) for our dataset and data from the 2016 INSERM database and the 2020 Metastatic Breast Cancer Project database. The majority (67%) of biopsies were from the liver. The other sites included skin (3.7%) and the rest equally divided across lymph node, other soft tissue, breast, and bone (at 7.4% each). Invasive ductal carcinomas represented 63% of the biopsies, invasive lobular carcinoma, 18.5%, and invasive mammary carcinoma 14.8% and 3.7% unknown.

IHC

IHC to detect ER (Agilent catalog no. M7047, RRID:AB_2101946), PR (Agilent catalog no. M3568, RRID:AB_2252608), AR (Agilent catalog no. M3562, RRID:AB_2060174), GR (Cell Signaling Technology, catalog no. 3660, RRID:AB_11179215), Ki67 (Agilent catalog no. M7240, RRID:AB_2142367), cleaved caspase-3 (Cell Signaling Technology, catalog no. 9661, RRID:AB_2341188), and IFITM3 (GeneTex, catalog no. GTX115407, RRID:AB_11172546) was performed on FFPE sections of the MBC biopsies described above. Antigen retrieval and detection were optimized for each antibody. IHC-stained slides were scored for intensity and percent cells positive by breast pathologist S.B. Sams who was blinded to ER mutation status. Pearson correlation values were calculated between the different receptors in MBC biopsies using percent positive cells.

Multiplex panel for tumor-infiltrating immune cells

FFPE sections were stained for immune markers using multiplex Opal TSA technology (Akoya Biosciences) along with the Vectra 3 Automated Quantitative Pathology Imaging System. Tumor-infiltrating lymphocyte antibodies used were: CD4 (Agilent, catalog no. M7310, RRID:AB_2728838), Foxp3 (Abcam, catalog no. ab20034, RRID:AB_445284), CD8 (Agilent, catalog no. M7103, RRID:AB_2075537), CD20 (Abcam, catalog no. ab9475, RRID:AB_307267), and CD68 (Agilent, catalog no. GA60961-2, RRID:AB_2661840), and pan cytokeratin (Agilent, catalog no. M3515, RRID:AB_2132885) was used to identify tumor epithelium. Programmed death ligand-1 (PD-L1)/CD68 coimmunofluorescence antibodies used were: PD-L1 (Abcam, catalog no. ab228462, RRID:AB_2827816) and CD68 (Agilent, catalog no. GA60961-2, RRID:AB_2661840). Dapi (Akoya, catalog no. FP1490) was used as a counterstain for each core needle biopsy, and positive cells in three to five 669 μm × 500 μm fields were scored using InForm software (Perkin Elmer) using either a pixel or cell-based algorithm including both tissue and cell segmentation.

Cell culture

Cell culture conditions

Breast cancer cell lines (MCF7 and T-47D) were obtained from Kathryn Horwitz at the University of Colorado Anschutz Medical Campus (Aurora, CO). Cell lines were authenticated by Short Tandem Repeat DNA Profiling (Promega) and tested for Mycoplasma in the University of Colorado Cancer Center (UCCC) Cell Technologies shared resource (September 2020). WT and mutant ER CRISPR cell lines were created by Jay Gertz at the Huntsman Cancer Institute using CETCH-seq to endogenously flag-tag WT or mutant (Y537S or D538G) ER using Cas9 mediated homologous recombination (see companion paper and ref. 24). MCF7 cells were maintained in Minimum Essential Medium (MEM) with 5% FBS, 1% nonessential amino acids, 6.0 μg/mL insulin, and 1% penicillin/streptomycin. T-47D cells were cultured in RPMI1640 media supplemented with 5% FBS and 1% penicillin/streptomycin. To mimic conditions resulting from AI therapy, where the ER mutations arise in patients, cells were also cultured in phenol red–free media (MEM for MCF7 and RPMI1640 for T-47D) supplemented with 5% dextran-coated charcoal (DCC)-stripped FBS for 6 months to generate long-term estrogen-deprived (LTED) cells. When appropriate, cell lines were cultured for eight or fewer passages prior to experimentation.

Immunoblotting

Cells were grown in full media under attached or forced-suspension conditions (on poly 2-hydroxyethyl methacrylate, poly-HEMA plates) for 24, 48, or 72 hours prior to harvest. Cells were then lysed in RIPA buffer (150 nmol/L NaCl, 1% IPEGAL, 0.5% Na-Deoxycholate, 0.1% SDS, 50 mmol/L Tris, and 1.0 mmol/L EDTA) containing protease and phosphatase inhibitors. Whole-cell protein lysates were separated by 10%–12% SDS-PAGE gels, transferred to polyvinylidene difluoride membranes, blocked in 3% BSA in Tris-buffered saline-Tween, and incubated with antibody overnight at 4°C. Primary antibodies used include: AR (Millipore, catalog no. 06-680, RRID:AB_310214), Flag (Sigma-Aldrich, catalog no. F1804, RRID:AB_262044), ER (Thermo Fisher Scientific, catalog no. RM9101-S), GR (Cell Signaling Technology, catalog no. 3660, RRID:AB_11179215), PR (Agilent, catalog no. M3568, RRID:AB_2252608), Tubulin (Sigma-Aldrich, catalog no. T5168, RRID:AB_477579), IFITM3 (GeneTex, catalog no. GTX115407, RRID:AB_11172546), and GAPDH (Sigma-Aldrich, catalog no. G8795, RRID:AB_1078991). Incubation in secondary antibody was followed by detection using an Odyssey CLx Imager (Odyssey CLx, RRID:SCR_014579). Densitometry was performed using the ImageJ software (ImageJ, RRID:SCR_003070) and reported as a ratio normalized to respective loading control (tubulin or GAPDH).

IHC on breast cancer cell lines and patient-derived xenografts

Cells were plated as described above and then were pelleted, fixed in 10% formalin, and paraffin embedded by the University of Colorado Denver Tissue Biobanking and Processing Core. Slides were deparaffinized in xylene and ethanol and heat-induced epitope retrieval was performed. Primary antibodies used were: AR (Abnova, catalog no. MAB10053, RRID:AB_10903299), ER (Thermo Fisher Scientific, catalog no. RM9101-S), PR (Agilent, catalog no. M3568, RRID:AB_2252608), IFITM3 (GeneTex, catalog no. GTX115407, RRID:AB_11172546), MUC1 (Cell Signaling Technology, catalog no. 4538, RRID:AB_2148549), and CHI3L1 (Abcam, catalog no. ab77528, RRID:AB_2040911). Representative images were taken at 200×, 400×, or 1,000× magnification. Quantification of nuclear AR was completed using the Aperio microscope (Aperio ScanScope XT Leica, RRID:SCR_018457) and ImageScope (ImageScope, RRID:SCR_014311), to calculate percent nuclei positive and staining intensities of 0, 1+, 2+, or 3+. Data represent the nuclear intensity of all cells in each cell pellet (N > 100 cells/pellet). For CHI3L1, IFITM3, and MUC1, ImageJ software (ImageJ, RRID:SCR_003070) was used to determine the average number of cells with 3+ staining in three representative images taken at 400×.

Soft agar colony formation

Soft agar colony formation assays were used to assay for anchorage-independent survival in WT and mutant ER MCF7 and T-47D cells treated with vehicle (ethanol), 20 μmol/L seviteronel, or 20 μmol/L enzalutamide (Enza). Experiments were performed in 6-well plates containing the following layers: 0.5% bottom agar, 0.3% top agar containing 20,000cells/well, and covered with the appropriate media. Cells were grown for approximately 21 days with biweekly media changes then fixed and stained with nitro blue tetrazolium at the time of harvest. Experiments were performed in triplicate (N = 3) and the average colony number was quantified using ImageJ software (ImageJ, RRID:SCR_003070).

Invasion assays

Cells were plated at 50,000 cells per well in 1% FBS on the top of a Matrigel-coated invasion chamber with 8.0 μm holes (Corning) and normal growth media (with 5% FBS) was used as a chemoattractant in the bottom of each well. The following day 10 μg/mL anti-CHI3L1 blocking antibody (Millipore, catalog no. MABC196) or IgG control (Thermo Fisher Scientific, catalog no. 31903, RRID:AB_10959891) was added to the top of each well. Seventy-two hours posttreatment, cells were fixed in 10% formalin and stained with 1% crystal violet. Cell invasion was determined by analyzing the average area of crystal violet staining on the bottom of each insert using ImageJ software. Presented is the average area of two separate experiments (N = 3/experiment).

Identification of mutant-specific and AR-regulated genes

D538G and Y537S ER mutant–specific genes were identified by analyzing RNA sequencing (RNA-seq) data from WT and mutant ER T-47D and MCF7 cell lines generated by the Gertz lab (22), the Oesterreich lab (UPMC Hillman Cancer Center; ref. 13), and the Brown lab (Dana Farber Cancer Institute groups; ref. 12) using an adjusted P value cutoff of <0.05. Pathway analysis of mutant-specific genes identified through the multivariate/multilab RNA-seq analysis (22) were analyzed using Illumina BaseSpace Correlation Engine to identify pathways and molecular and cellular processes that are significantly associated with mutant-specific gene expression changes. Gene sets were ranked by log2 fold change and P value. Gene ontology enrichment P values were calculated by Illumina BaseSpace Correlation Engine. A list of AR-regulated genes was compiled from multiple sources including prostate and breast cancer, and cell lines derived from both types of cancer (25–27). This curated list of AR-regulated genes was overlapped with mutant-specific genes identified in our multi-lab RNA-seq analysis described above. Significant overlap between mutant specific genes and AR-regulated genes was identified using a hypergeometric P value cutoff of <0.05.

Animal studies

All animal experiments were performed in accordance with protocols approved by the University of Colorado Institutional Animal Care and Use Committee using humane procedures. Six-week-old ovariectomized NOD scid gamma (NSG) mice (IMSR, catalog no. JAX:005557, RRID:IMSR_JAX:005557) were implanted subcutaneously at base of neck with pellets containing cellulose or pellets containing 2 mg 17beta-E2 in cellulose (N = 10/group). Mice were injected via the tail vein with T-47DGFP/Luc cells containing either WT or D538G ER to generate four experimental groups of mice: cellulose pellets plus WT T-47D cells, cellulose pellets plus D538G T-47D cells, E2 pellets plus WT T-47D cells, and E2 pellets plus D538G T-47D cells (N = 5/group). Mice were imaged weekly by IVIS (Perkin Elmer IVIS Spectrum In-Vivo Imaging System, RRID:SCR_018621) beginning the day after tumor cell injection. Six weeks after cell injection, mice were imaged by IVIS, then were sacrificed, and excised lungs were imaged by IVIS ex vivo. Tumor burden at endpoint (6 weeks) is presented as total flux (photons/second) of luminescence from IVIS imaging. Patient-derived xenografts (PDX) were kindly provided by Alana Welm, Huntsman Institute and Carol Sartorius, University of Colorado (Aurora, CO). PDX were passaged in cycling female NSG mice supplemented with E2 with the exception of HCI-013-EI tumors passaged in unsupplemented mice.

Statistical analysis

Data are presented as the mean ± SEM. When comparing two groups of data, an unpaired Student t test was used. When heteroscedasticity was present or data were skewed, the nonparametric Mann–Whitney test was used. For statistical analysis between more than two groups, a one-way ANOVA or two-way ANOVA with a Tukey multiple comparison test was used. In the case of heteroscedasticity or skewed data, nonparametric Kruskal–Wallis or Friedman tests were used. When data were presented on a log-scale (Fig. 4) data analysis using parametric tests was conducted on log-transformed data to account for heteroscedasticity between groups. Statistical significance was defined as P < 0.05 and all tests were two sided. Analyses were done using GraphPad Prism (GraphPad Prism, RRID:SCR_002798, ver8.3). Power calculations for the animal study were conducted under the supervision of a collaborating biostatistician Kathleen Torkko. Animal numbers were calculated at 80% power to the expected difference at P < 0.05 (two-tailed).

Steroid hormone receptor status of metastases harboring ER mutations

Core needle biopsies from accessible metastases were collected from consented patients primarily enrolled in NCT02953860 and 3 from NCT01597193. All patients had ER+HER2 breast cancer at time of diagnosis. The steroid hormone receptor proteins, ER, PR, AR, and GR, had not previously been examined together in the same cases of MBC harboring WT versus mutant ER. FFPE sections from core needle biopsies were stained for ER, PR, AR, and GR (Fig. 1A and B), and Ki67 and cleaved caspase-3 (Supplementary Fig. S1A and S1B). IHC was scored by breast pathologist (S.B. Sams) and graphed according to ER mutation status. ER and PR protein were significantly higher in biopsies harboring mutant ER as compared with metastases with no detectable mutations in exon 8 (designated as WT ER; Fig. 1A). In contrast, AR and GR were not significantly different in biopsies with WT ER compared with those with mutant ER (Fig. 1A). In WT ER metastases, where ER and PR were often low, AR was frequently maintained (patients: 01-006, 01-010, 01-014, 01-022, 01-028, 02-006, 02-007; Fig. 1B) and was therefore not significantly different between WT and mutant biopsies. Interestingly, some mutant ER cancers had a substantially higher AR than PR (patients: 01-002, 01-004 and 01-027; Fig. 1B), and mutant ER metastases with high PR had low AR. Ki67 and cleaved caspase-3 were not significantly different between the two groups, although there was a trend toward higher Ki67 in metastases with mutant ER compared with WT ER (Supplementary Fig. S1A and S1B). We also examined correlations between hormone receptors and as would be expected because ER regulates PR, these two receptors are positively correlated. ER is correlated with AR as well. It is evident that while many ER mutant containing biopsies (red) are high for ER and PR, the ER WT (black) have predominately lost PR, but often retain AR expression. ER does not correlate with GR and PR and AR do not correlate (Supplementary Fig. S2A–S2D).

Figure 1.

Steroid hormone receptor expression in biopsies of MBC with WT or mutant ESR1. A, FFPE sections of core needle biopsies (N = 18 WT ER; N = 12 mutant ER) were stained by IHC for ER, PR, AR, and GR. Depicted are the mean scores (intensity × percent cells staining) ± SEM. Mann–Whitney tests were performed for each receptor stained. B, Representative images for all WT ER metastases (left) and all mutant ER metastases (right) stained for ER, PR, AR, and GR are shown at 400×. N.S., not significant.

Figure 1.

Steroid hormone receptor expression in biopsies of MBC with WT or mutant ESR1. A, FFPE sections of core needle biopsies (N = 18 WT ER; N = 12 mutant ER) were stained by IHC for ER, PR, AR, and GR. Depicted are the mean scores (intensity × percent cells staining) ± SEM. Mann–Whitney tests were performed for each receptor stained. B, Representative images for all WT ER metastases (left) and all mutant ER metastases (right) stained for ER, PR, AR, and GR are shown at 400×. N.S., not significant.

Close modal

Mutant ER breast cancer cells upregulate AR expression and activity

To study mutant ER expressed from its native locus in MCF7 and T-47D human breast cancer cell lines, FLAG-tagged ER-mutant models were engineered using CRISPR to introduce one allele of FLAG-tagged ESR1 containing the D538G or Y537S mutation or FLAG-tagged WT ESR1 (22). Immunoblot for FLAG confirmed CRISPR construct expression in two clones for each cell line (Supplementary Fig. S3A and S3B). Further experiments were conducted using two clones pooled for each line.

Steroid hormone receptor expression was assessed by immunoblot of WT, D538G, and Y537S MCF7 and T-47D cells grown in culture media containing full serum (containing steroid hormones). However, to mimic conditions resulting from AI therapy, where the ER mutations arise in patients, the cells were also cultured in phenol red–free media containing DCC-stripped serum for 6 months to generate LTED cells. While AR was slightly elevated in the MCF7 ER mutant compared with WT cells cultured in full serum, in the LTED conditions ER, PR, AR, and GR were all increased. Interestingly, AR and PR were dramatically increased in the LTED mutants compared with LTED WT (Fig. 2A). In T-47D LTED resulted in increased ER, PR AR, and GR compared with the cells cultured in full serum. AR and GR increased with LTED in both WT and mutants, indicating that WT cells that survive LTED can also upregulate AR and GR. PR did not increase as a result of LTED in the WT (Supplementary Fig. S4).

Figure 2.

AR protein and gene regulation differs in WT versus mutant ER–expressing MCF7 breast cancer cells and PDX, particularly after LTED. A, Western blot analysis of whole-cell lysates from WT, D538G, and Y537S MCF7 cells grown in full serum media (left) or charcoal-stripped serum containing media (LTED, right), which were probed for AR, ER, GR, PR-A/B, and tubulin. B, ER+ PDX tissue microarrays from Alana Welm, Huntsman Cancer Institute (HCI) and ER+ PDX from University of Colorado (UCD) with ESR1 mutation status indicated were stained for AR by IHC, with representative images at 100× (left) and 400× (right). For the estrogen-independent version of HCI-013, HCI-013-EI, images representative of primary tumor (PT) and lung metastasis (Met) are shown. C, Heatmap of known AR-regulated genes in MCF7 (left) and T-47D (right) depicting relative expression of genes displayed as a Z-score across WT and mutant cells after 5 days of growth in hormone-depleted media, followed by treatment with either 10 nmol/L E2 or vehicle (DMSO) for 8 hours. Significant overlap was determined using hypergeometric tests with a P value cutoff of <0.05: AR genes versus D538G upregulated genes, 5.5 × 10−05; AR genes versus D538G downregulated genes, 1.0 × 10−08; AR genes versus Y537S upregulated genes, 3.2 × 10−08; AR genes versus Y537S downregulated genes, 9.3 × 10−11.

Figure 2.

AR protein and gene regulation differs in WT versus mutant ER–expressing MCF7 breast cancer cells and PDX, particularly after LTED. A, Western blot analysis of whole-cell lysates from WT, D538G, and Y537S MCF7 cells grown in full serum media (left) or charcoal-stripped serum containing media (LTED, right), which were probed for AR, ER, GR, PR-A/B, and tubulin. B, ER+ PDX tissue microarrays from Alana Welm, Huntsman Cancer Institute (HCI) and ER+ PDX from University of Colorado (UCD) with ESR1 mutation status indicated were stained for AR by IHC, with representative images at 100× (left) and 400× (right). For the estrogen-independent version of HCI-013, HCI-013-EI, images representative of primary tumor (PT) and lung metastasis (Met) are shown. C, Heatmap of known AR-regulated genes in MCF7 (left) and T-47D (right) depicting relative expression of genes displayed as a Z-score across WT and mutant cells after 5 days of growth in hormone-depleted media, followed by treatment with either 10 nmol/L E2 or vehicle (DMSO) for 8 hours. Significant overlap was determined using hypergeometric tests with a P value cutoff of <0.05: AR genes versus D538G upregulated genes, 5.5 × 10−05; AR genes versus D538G downregulated genes, 1.0 × 10−08; AR genes versus Y537S upregulated genes, 3.2 × 10−08; AR genes versus Y537S downregulated genes, 9.3 × 10−11.

Close modal

Because AR was upregulated in ER mutant versus WT cells and by LTED, we evaluated AR by IHC in multiple ER+ PDX (Fig. 2B). Upon knowledge of the ER mutational status of the PDX, AR was found to be higher in PDX with ER mutations as compared with those with WT ER. Furthermore, the Y537S mutant HCI-013-EI (estrogen independent) PDX, which was derived from the parental HCI-013 PDX passaged in mice without E2 supplementation (to mimic the E2-deprivation therapy in patients with breast cancer), exhibited elevated AR expression compared with parental HCI-013 tumors passaged in E2-supplemented mice. Human AR was also maintained in spontaneous metastases from the HCI-013-EI primary tumors to lung (Fig. 2B).

Cross-referencing known AR target genes with genes differentially expressed in mutant ER breast cancer cells revealed statistically significant enrichment (hypergeometric test P values < 6 × 10–5) of AR target genes in mutant ER cells compared with WT (Fig. 2C; genes identified in Supplementary Table S1). In this experiment, cells were cultured in phenol red–free media containing DCC-stripped serum for 5 days, then supplemented ± E2 for 8 hours. AR target genes were upregulated in mutant ER cell lines with or without E2, indicating that the mutant-specific alterations in known AR target genes occurred in an E2-independent manner.

AR inhibition abrogates the selective advantage of mutant ER breast cancer cells for anchorage-independent survival

We previously reported that AR mRNA, protein, and activity were increased in triple-negative breast cancer (TNBC) cells that survive under anchorage-independent conditions as compared with adherent cells, and AR protected against caspase-mediated cell death (28). To determine whether AR also increased in anchorage-independent ER+ breast cancer lines, parental MCF7 cells were grown for 24–72 hours under adherent conditions (Att) or forced suspension conditions on poly-HEMA–coated plates (Susp), which prevent cell adherence. AR was elevated at 24, 48, and 72 hours of anchorage-independent growth as compared with adherent cells, while ER and GR were not (Supplementary Fig. S5A). PR was slightly elevated at 24 hours of anchorage-independent culture as compared with attached cells, but to a much lesser extent than AR. IHC on cell pellets harvested at 24, 48, and 72 hours demonstrated a striking increase in AR at each time point in the anchorage-independent cells compared with attached cells, while ER and PR were not altered (Supplementary Fig. S5B).

To determine whether AR protein increased in mutant ER cells under anchorage-independent conditions, AR was analyzed by IHC on FFPE sections of cell pellets generated from WT, D538G, and Y537S MCF7 cells grown in attached or forced suspension conditions for 24 hours. AR protein increased in all three MCF7 cell lines following suspension culture for 24 hours as compared with adherent cells and AR expression decreased with addition of the CYP17A1 lyase inhibitor seviteronel that also acts as an AR antagonist (Fig. 3A; ref. 29). Quantification showed that strong (3+) nuclear AR staining in D538G cells was 5-fold higher at baseline than in WT MCF7 cells (Fig. 3B). The number of nuclei strongly positive for AR was increased by 19-fold for WT, 7.5-fold for D538G, and 21-fold for Y537S MCF7 cells grown in suspension as compared with attached conditions. Sevi dramatically reduced the number of AR+ nuclei and the staining intensity (Fig. 3A and B). Similar results were observed for WT, D538G, and Y537S T-47D cells (Supplementary Fig. S6A and S6B). Furthermore, when grown on soft agar, D538G MCF7 cells, which at baseline expressed the highest level of AR, formed significantly more colonies than WT counterpart (1.6-fold higher; Fig. 3C and D). To determine whether increased AR was indicative of a dependency on AR for survival in anchorage-independent conditions, as in TNBC lines (28), MCF7 cells were treated with Sevi, which completely abolished growth on soft agar of all MCF7 lines. Because the AR inhibitor Enza showed clinical benefit in AR+ TNBC (as did bicalutamide; refs. 30, 31), we also tested Enza, which significantly inhibited D538G cell growth in soft agar. Importantly, Enza negated the survival advantage of D538G cells compared with WT MCF7 in soft agar. Similar results were seen in D538G T-47D cells, where both Sevi and Enza inhibited growth on soft agar (Supplementary Fig. S6C). WT and mutant ER MCF7 and T-47D cells generated in the Oesterreich lab (13) were also studied on two-dimensional (2D) standard culture as compared with three-dimensional (3D) culture using ultralow adhesion plates and 3 days of hormone deprivation (Supplementary Fig. S7). AR was higher in both D538G- and Y537S-mutant ER MCF7 lines compared with WT (in both 2D and 3D). While AR decreased in 3D compared with 2D in the WT cells, it remained elevated in 3D in the mutant ER lines, again demonstrating a potential role for AR in the ER mutants. Results in T-47D mutants versus WT were similar.

Figure 3.

AR increased in breast cancer cells grown in soft agar and AR inhibition abolished the selective advantage of mutant ER cells for anchorage-independent survival. A, WT, D538G, and Y537S mutant MCF7 cells were grown in full serum media under attached (Att) or suspension conditions (on poly-HEMA plates, Susp) and treated ± 20 μmol/L seviteronel (Sevi) for 24 hours. Cells were pelleted, FFPE, and AR IHC performed. Representative images are shown at 400×. B, Quantification of A with ImageScope software for percent positive nuclei for staining intensities of 0–3+ for all conditions attached (Att), suspended (Susp), and ± seviteronel. C, WT, D538G, and Y537S cells plated in 0.3% agar and treated with 20 μmol/L seviteronel, 20 μmol/L Enza or EtOH control and grown for 3 weeks with biweekly media and drug changes. Representative images of anchorage-independent growth are shown. D, Colony number quantified using ImageJ software for assay in C (N = 3). Mean ± SEM. One way ANOVA with Tukey multiple comparison test is depicted. Two-way ANOVA, interaction between cell line and treatment P = 0.003.

Figure 3.

AR increased in breast cancer cells grown in soft agar and AR inhibition abolished the selective advantage of mutant ER cells for anchorage-independent survival. A, WT, D538G, and Y537S mutant MCF7 cells were grown in full serum media under attached (Att) or suspension conditions (on poly-HEMA plates, Susp) and treated ± 20 μmol/L seviteronel (Sevi) for 24 hours. Cells were pelleted, FFPE, and AR IHC performed. Representative images are shown at 400×. B, Quantification of A with ImageScope software for percent positive nuclei for staining intensities of 0–3+ for all conditions attached (Att), suspended (Susp), and ± seviteronel. C, WT, D538G, and Y537S cells plated in 0.3% agar and treated with 20 μmol/L seviteronel, 20 μmol/L Enza or EtOH control and grown for 3 weeks with biweekly media and drug changes. Representative images of anchorage-independent growth are shown. D, Colony number quantified using ImageJ software for assay in C (N = 3). Mean ± SEM. One way ANOVA with Tukey multiple comparison test is depicted. Two-way ANOVA, interaction between cell line and treatment P = 0.003.

Close modal

To determine whether mutant ER cells have a selective advantage for anchorage-independent survival or metastatic outgrowth in vivo under conditions modeling the E2-depleted state, we introduced labeled WT or D538G T-47D cells, in which AR was elevated (Fig. 4A), by tail vein injection into ovariectomized mice that were supplemented with either cellulose control or E2 pellets. Whole mouse IVIS imaging at the end of study showed significantly higher metastatic burden in mice injected with D538G T-47D cells compared with those with WT cells, both with and without E2 (Fig. 4B and C). Ex vivo imaging of lungs completed immediately after necropsy also showed significantly higher metastatic burden in lungs harboring D538G cells as compared with those injected with WT T-47D with or without E2 (Fig. 4D and E). In non-E2-supplemented mice, WT tumor cells were nearly undetectable when imaged in vivo or ex vivo.

Figure 4.

Breast cancer cells harboring D538G ER mutation thrive at metastatic sites in an experimental metastasis model in estrogen-deprived mice, while WT ER cells do not. A, WT and D538G ER T-47DGFP/Luc cells were stained for AR and representative images are shown at 400×. B, The same lines were delivered via tail vein injection into 6-week-old ovariectomized NSG mice (N = 5/group) with or without E2 (cellulose) and 6 weeks after cell injection, mice were imaged by IVIS. C, Quantification of whole mouse luminescence signal (total flux). Mean ± SEM. Raw data are presented with P values from a Student unpaired two-tailed t test conducted on the log-transformed data due to heteroscedasticity. One-way ANOVA (P = 0.006). D, At end of study, mice were sacrificed, lungs excised, and imaged ex vivo by IVIS. E, Quantification of luminescence in lungs ex vivo. Mean ± SEM. Raw data are presented with P values from a Student unpaired two-tailed t test conducted on log-transformed data due to heteroscedasticity. One way ANOVA (P = 0.0003).

Figure 4.

Breast cancer cells harboring D538G ER mutation thrive at metastatic sites in an experimental metastasis model in estrogen-deprived mice, while WT ER cells do not. A, WT and D538G ER T-47DGFP/Luc cells were stained for AR and representative images are shown at 400×. B, The same lines were delivered via tail vein injection into 6-week-old ovariectomized NSG mice (N = 5/group) with or without E2 (cellulose) and 6 weeks after cell injection, mice were imaged by IVIS. C, Quantification of whole mouse luminescence signal (total flux). Mean ± SEM. Raw data are presented with P values from a Student unpaired two-tailed t test conducted on the log-transformed data due to heteroscedasticity. One-way ANOVA (P = 0.006). D, At end of study, mice were sacrificed, lungs excised, and imaged ex vivo by IVIS. E, Quantification of luminescence in lungs ex vivo. Mean ± SEM. Raw data are presented with P values from a Student unpaired two-tailed t test conducted on log-transformed data due to heteroscedasticity. One way ANOVA (P = 0.0003).

Close modal

Mutant ER breast cancer have elevated expression of prometastatic CHI3L1

Analysis of RNA-seq data from three independent laboratories using mutant ER cell lines generated by different methods showed that CHI3L1 (also known as YKL-40) was elevated in Y537S MCF7 and T-47D compared with WT (Supplementary Fig. S8A and S8B; ref. 22). CHI3L1 is a well-established androgen/AR-regulated gene in prostate cancer where it is detectable in tumor and patient serum and correlates with disease progression (32–34). Because we previously reported that CHI3L1 increased in TNBC lines under anchorage-independent conditions (28), we evaluated CHI3L1 protein in WT, D538G, and Y537S MCF7 cells in attached versus suspension culture for 24 hours (Fig. 5A). CHI3L1 was significantly higher in both D538G (5-fold) and Y537S (3-fold) MCF7 cells than WT; however, CHI3L1 was not further enhanced by 24 hours of anchorage-independent culture. Because CHI3L1 promotes invasion in other cancer types, we compared the invasive capacity of D538G and Y537S MCF7 to WT and found that D538G MCF7, which had the highest expression of CHI3L1, were significantly more invasive than WT MCF7 in Matrigel-coated Boyden chamber assays (Fig. 5B). Importantly, a CHI3L1 blocking antibody negated this significant invasive advantage of D538G MCF7 cells compared with WT, suggesting that mutant ER breast cancer cells can utilize CHI3L1 to promote invasion.

Figure 5.

MCF7 breast cancer cells with mutant ER express increased CHI3L1 and blocking CHI3L1 decreased mutant ER breast cancer cell invasion. A, IHC staining for CHI3L1 protein in WT, D538G, and Y537S MCF7 cells grown in full serum media under attached (Att) or suspension conditions (on poly-HEMA plates, Susp) for 24 hours. Representative images are shown at 1,000× (left) and quantification of percent cells strongly positive (3+) for CHI3L1 is shown (right). Mean ± SEM; N = 3 photos/pellet; one-way ANOVA with Tukey multiple comparison test is depicted. Two-way ANOVA, interaction between cell line and treatment was not significant. B, Invasion through Matrigel was determined for WT, D538G, and Y537S MCF7 cells grown in full serum media. Cells were treated with 10 μg/mL anti-CHI3L1 blocking antibody or IgG control 24 hours after plating and harvested 72 hours posttreatment. After crystal violet staining and imaging, the relative amount of invasion was determined using ImageJ software. Mean ± SEM of two separate experiments; N = 3. One-way ANOVA with Tukey multiple comparison test is depicted. Two-way ANOVA, interaction between cell line and treatment was not significant. C, CHI3L1 IHC was conducted on Y537S-mutant HCI-013 grown in E2-supplemented mice and HCI-013-EI grown in ovariectomized mice. Representative images are shown at 400× (left) and quantification of strongly positive (3+) cells was conducted using the ImageScope Software (right). Mean ± SEM; N = 6–7; Mann–Whitney test.

Figure 5.

MCF7 breast cancer cells with mutant ER express increased CHI3L1 and blocking CHI3L1 decreased mutant ER breast cancer cell invasion. A, IHC staining for CHI3L1 protein in WT, D538G, and Y537S MCF7 cells grown in full serum media under attached (Att) or suspension conditions (on poly-HEMA plates, Susp) for 24 hours. Representative images are shown at 1,000× (left) and quantification of percent cells strongly positive (3+) for CHI3L1 is shown (right). Mean ± SEM; N = 3 photos/pellet; one-way ANOVA with Tukey multiple comparison test is depicted. Two-way ANOVA, interaction between cell line and treatment was not significant. B, Invasion through Matrigel was determined for WT, D538G, and Y537S MCF7 cells grown in full serum media. Cells were treated with 10 μg/mL anti-CHI3L1 blocking antibody or IgG control 24 hours after plating and harvested 72 hours posttreatment. After crystal violet staining and imaging, the relative amount of invasion was determined using ImageJ software. Mean ± SEM of two separate experiments; N = 3. One-way ANOVA with Tukey multiple comparison test is depicted. Two-way ANOVA, interaction between cell line and treatment was not significant. C, CHI3L1 IHC was conducted on Y537S-mutant HCI-013 grown in E2-supplemented mice and HCI-013-EI grown in ovariectomized mice. Representative images are shown at 400× (left) and quantification of strongly positive (3+) cells was conducted using the ImageScope Software (right). Mean ± SEM; N = 6–7; Mann–Whitney test.

Close modal

We also examined CHI3L1 expression in HCI-013-EI (estrogen independent) PDX tumors, which harbor the Y537S mutant and were grown in the absence of E2, as compared with parental HCI-013 PDX tumors, which were grown in the presence of E2 (Fig. 5C). HCI-013-EI tumors had a >3-fold increase in CHI3L1 expression compared with HCI-013 tumors, demonstrating that CHI3L1 is enhanced by propagating this PDX under conditions similar to the selective pressure of E2 depletion that selects for ER mutations in patients with breast cancer.

Mutant ER breast cancer cells upregulated the immunomodulatory type-1 IFN pathway

To identify additional mutant ER-specific vulnerabilities, we conducted pathway analysis of mutant-specific genes that revealed innate immune, viral recognition, and type-1 IFN among the top pathways preferentially upregulated in mutant versus WT ER MCF7s (Fig. 6A,C; ref. 22). The IFN-induced transmembrane (IFITM) gene IFITM3 was consistently increased in D538G and Y537S MCF7 cells in data from multiple laboratories (Supplementary Fig. S8C). Both D538G and Y537S MCF7 cells demonstrated enhanced levels of a number of IFN-stimulated genes (ISG) such as OAS1/2, IFIT1/2/3/5, IFITM1/3, and ISG15 compared with WT MCF7 (Fig. 6B and C; Supplementary Fig. S8D–S8F), demonstrating robust activation of this pathway in mutant ER breast cancer cells as compared with WT. Because IFITM proteins have been reported to play a role in AI resistance (35, 36), we explored the expression of IFITM3 in WT, D538G, and Y537S MCF7 cells ± IFNβ stimulation, a key cytokine and activator of the type-1 IFN signaling pathway. Baseline IFITM3 levels were —2- to 4-fold higher in D538G and Y537S MCF7 cells compared with WT and were further enhanced by IFNβ (Fig. 6D). IHC quantification showed that IFITM3 was higher in D538G MCF7 than WT (Fig. 6E). Given that MUC1 drives IFITM1 overexpression in AI-resistant breast cancer cells (37) and MUC1 was also on the list of genes significantly elevated in cells containing mutant versus WT ER, we also examined MUC1 protein by IHC and found it was elevated in D538G MCF7 cells compared with WT (Fig. 6E). Finally, we performed IHC for IFITM3 in biopsies of 5 patients with MBC harboring the D538G ER mutation as compared with all of the WT biopsies presented in Fig. 1 and found IFITM3 protein was significantly upregulated in the ER mutant versus WT biopsies (Fig. 6F).

Figure 6.

Mutant ER breast cancer cells and specimens have increased type-1 IFN signaling pathway signature and increased IFITM3 protein compared with those with WT ER. A, Illumina BaseSpace Correlation Engine was used to identify mutant-specific pathways from MCF7 mutant-specific genes identified through the multivariate/multilab RNA-seq analysis (22). B and C, Heatmap depicts relative expression, displayed as a Z-score, of genes associated with Innate immunity (B) and type I IFN (C), identified by two different pathway analyses (David and ENRICHR) in ER WT and mutant MCF7 cells grown in hormone-depleted media without E2 for 3–5 days depending on the laboratory. D, WT, D538G, and Y537S MCF7 cells were treated with 1,000 units of IFNβ for 24 hours, whole-cell lysates were generated, and were analyzed by Western blot analysis for IFITM3. E, IHC for IFITM3 and MUC1 was performed on FFPE pellets generated from WT, D538G, and Y537S MCF7 cells grown in media containing full serum. Mean ± SEM; N = 3 photos/pellet. One-way ANOVA with Tukey multiple comparison test was conducted for each staining (right). Representative images are shown at 1,000× (left). F, FFPE sections of core needle biopsies (N = 14 WT ER; N = 5 D538G mutant ER) were stained by IHC for IFITM3. Depicted are the mean scores (intensity × percent cells staining) ± SEM, Student unpaired two-tailed t test.

Figure 6.

Mutant ER breast cancer cells and specimens have increased type-1 IFN signaling pathway signature and increased IFITM3 protein compared with those with WT ER. A, Illumina BaseSpace Correlation Engine was used to identify mutant-specific pathways from MCF7 mutant-specific genes identified through the multivariate/multilab RNA-seq analysis (22). B and C, Heatmap depicts relative expression, displayed as a Z-score, of genes associated with Innate immunity (B) and type I IFN (C), identified by two different pathway analyses (David and ENRICHR) in ER WT and mutant MCF7 cells grown in hormone-depleted media without E2 for 3–5 days depending on the laboratory. D, WT, D538G, and Y537S MCF7 cells were treated with 1,000 units of IFNβ for 24 hours, whole-cell lysates were generated, and were analyzed by Western blot analysis for IFITM3. E, IHC for IFITM3 and MUC1 was performed on FFPE pellets generated from WT, D538G, and Y537S MCF7 cells grown in media containing full serum. Mean ± SEM; N = 3 photos/pellet. One-way ANOVA with Tukey multiple comparison test was conducted for each staining (right). Representative images are shown at 1,000× (left). F, FFPE sections of core needle biopsies (N = 14 WT ER; N = 5 D538G mutant ER) were stained by IHC for IFITM3. Depicted are the mean scores (intensity × percent cells staining) ± SEM, Student unpaired two-tailed t test.

Close modal

Metastases harboring mutant ER have significantly more CD4+ T-cell and macrophage infiltrates

Given that mutant ER cell lines showed enhanced innate immune and IFN pathway signatures compared with WT ER, we tested biopsies of MBC with mutant versus WT ER for differences in tumor-infiltrating immune cells. A multiplex panel consisting of antibodies that distinguish T cells (FoxP3+CD4+ T regulatory cells; FoxP3CD4+ Th cells; or CD8+ T effector cells), macrophages (CD68+), B cells (CD20+), and tumor cells (cytokeratin+) was utilized. InForm Software analysis demonstrated that the number of T regulatory (CD4+FoxP3+) and Th cells (CD4+FoxP3) were significantly increased in mutant ER breast cancer metastases when compared with WT ER, although no difference in CD8+ T effector cells was observed (Fig. 7A). Interestingly, in both our data (Supplementary Fig. S9A) and data from two publicly available studies (Supplementary Fig. S9B and S9C), ESR1 mutations were mutually exclusive with TP53 mutations. Furthermore, patients with either TP53 or PIK3CA mutations contained significantly higher CD8+ cytotoxic T effector cells (Supplementary Fig. S10A and S10B). There was no significant change in CD20+ B cells in mutant ER metastases compared with WT (Fig. 7B). However, the number of CD68+ macrophages in mutant ER metastases was significantly increased when compared with biopsies with WT ER (Fig. 7C). Finally, PD-L1 expression was analyzed in the same biopsies due to the recent testing of PD-L1 inhibitors in ER+ breast cancer (Fig. 7D; refs. 38, 39) and FDA approval in TNBC (40). Interestingly, while the overall level of PD-L1 did not differ between mutant and WT ER biopsies, PD-L1+ macrophages were significantly higher in mutant ER biopsies compared with WT, suggesting that macrophages infiltrating mutant ER metastases are more immunosuppressive. Interestingly, this finding was more significant when examining macrophages intimately associated with the tumor cells as compared with those found predominately in the intratumoral stroma (Supplementary Fig. S11).

Figure 7.

CD4+ T cells and PD-L1+ macrophages were significantly higher in biopsies of MBC harboring mutant ER as compared with WT ER. Biopsies of the same metastases characterized in Fig. 1 (N = 14 with ER WT, N = 10 with mutant ER) were stained for tumor-infiltrating immune cells using Opal TSA technology (Akoya Biosciences), with the following colors indicating respective immune cell markers: CD4 (yellow), Foxp3 (green), CD8 (magenta), CD68 (orange), CD20 (red), and cytokeratin (cyan). Slides were scanned using Vectra 3 Automated Quantitative Pathology Imaging System (Perkin Elmer) technology and three to five representative 20× fields/tumor were analyzed using InForm software (Perkin Elmer) and a pixel-based algorithm for percent positivity. Mean percent positive cells ± SEM; Mann–Whitney test (top); representative images (20×; bottom) for: A, T regulatory cells (CD4+ FoxP3+), Th cells (CD4+ FoxP3), and cytotoxic T cells (CD8+); B and C, B cells (CD20+; B) and macrophages (CD68+; C). D, The same biopsies were stained for PD-L1 and CD68 using Opal TSA technology (Akoya Biosciences). Slides were scanned using a Vectra 3 Automated Quantitative Pathology Imaging System (Perkin Elmer), and up to five representative fields/tumor were analyzed for either total PD-L1 or dual expression of PD-L1 and CD68 using a cell phenotype-based algorithm for percent positive cells. Mean positive cells ± SEM; Mann–Whitney test.

Figure 7.

CD4+ T cells and PD-L1+ macrophages were significantly higher in biopsies of MBC harboring mutant ER as compared with WT ER. Biopsies of the same metastases characterized in Fig. 1 (N = 14 with ER WT, N = 10 with mutant ER) were stained for tumor-infiltrating immune cells using Opal TSA technology (Akoya Biosciences), with the following colors indicating respective immune cell markers: CD4 (yellow), Foxp3 (green), CD8 (magenta), CD68 (orange), CD20 (red), and cytokeratin (cyan). Slides were scanned using Vectra 3 Automated Quantitative Pathology Imaging System (Perkin Elmer) technology and three to five representative 20× fields/tumor were analyzed using InForm software (Perkin Elmer) and a pixel-based algorithm for percent positivity. Mean percent positive cells ± SEM; Mann–Whitney test (top); representative images (20×; bottom) for: A, T regulatory cells (CD4+ FoxP3+), Th cells (CD4+ FoxP3), and cytotoxic T cells (CD8+); B and C, B cells (CD20+; B) and macrophages (CD68+; C). D, The same biopsies were stained for PD-L1 and CD68 using Opal TSA technology (Akoya Biosciences). Slides were scanned using a Vectra 3 Automated Quantitative Pathology Imaging System (Perkin Elmer), and up to five representative fields/tumor were analyzed for either total PD-L1 or dual expression of PD-L1 and CD68 using a cell phenotype-based algorithm for percent positive cells. Mean positive cells ± SEM; Mann–Whitney test.

Close modal

Mutations in ESR1 that render ER constitutively active in the absence of ligand are a common mechanism of acquired resistance to AI therapy. While increased PR resulting from constitutively active mutant ER is well documented, the four main steroid hormone receptor proteins (ER, PR, AR, and GR) had not been evaluated together in MBC biopsies harboring WT versus mutant ER. While E2 is extremely low in postmenopausal women with breast cancer on AI, circulating and intratumoral androgens are increased (41). This, combined with the predominantly nuclear localization of AR in clinical samples (indicating ligand bound receptors), led us to postulate that long-term E2 deprivation via AI therapy, which blocks conversion of androgens to estrogens, might cause tumor cells to shift to reliance on androgens and AR. Indeed, in TNBC, where ER is absent, AR binds chromatin in a manner more similar to ER in ER+ breast cancer cells than to AR in prostate cancer cells, suggesting that AR can replace ER in some TNBC (42) and confer the “luminal AR” designation (43). Our analysis of steroid hormone receptors suggests that ER+ MBC can become resistant to AI treatment by either (i) acquiring mutations that render ER constitutively active (retaining robust ER and PR positivity) or (ii) by expressing little to no ER and PR in an apparent “subtype switch” to resemble TNBC, which often retain AR and/or GR.

Cross-referencing known AR-regulated genes with genes differentially expressed by mutant ER versus WT ER breast cancer cells revealed significant enrichment for AR-regulated genes. Although AR protein had not specifically been examined in WT versus mutant ER MBC, it was elevated in both primary and MBC resistant to AI therapy when compared with responsive disease (44–49). AR was often maintained or elevated in MBC compared with patient-matched primary tumors, whereas ER was decreased (50). Breast cancer patient CTCs have been found to be AR+ in several studies and AR+ CTC associate with bone metastases (51) and AR mRNA was higher in CTC from breast cancer PDX compared with primary tumors in mice (52). Furthermore, AR inhibition decreased metastatic burden of breast cancer cells from an ER+/AR+ PDX (50) and conversely, ligand-activated AR potentiated metastasis in ER+/AR+ breast cancer cell lines (53). These data, together with our analysis herein of hormone receptor status in clinical biopsies of AI-resistant MBC and other data presented in this study suggest that AR could be active in a subset of ER+ MBC, and might facilitate survival under conditions of E2 deprivation or anchorage-independent survival. Indeed, MCF7 and T-47D ER-mutant breast cancer lines, had higher ER, PR, and AR compared with WT, similar to results from the clinical biopsies of MBC. While all steroid receptors increased under LTED conditions, PR and AR were the most strikingly different in mutant compared with WT LTED cells. Historically, ER+ PDX were propagated in mice supplemented with E2 to increase take rate and growth rate for research studies regardless of ER mutational status. We now know that many of these ER+ PDX, derived from MBC in postmenopausal women on AI therapy (a long-term E2-deprived context) do harbor ER mutations. Placing these into mice supplemented with E2 could result in tumors that are “readdicted” to E2 and therefore do not accurately model the selective pressure of long-term E2 deprivation under which the mutations arise. We stained ER+ PDX for AR and found that PDX harboring ER mutations had higher AR than those with WT ER. Furthermore, in support of the theory that E2 deprivation enhances AR expression, the E2-independent HCI-013-EI PDX (with Y537S ER mutation) propagated in E2-deprived mice, expressed more AR protein than its parental counterpart propagated in mice supplemented with E2. Importantly, HCI-013-EI lung metastases originating from the orthotopic site still expressed AR. Thus, maintaining models of mutant ER disease under conditions similar to those in which they arise in the clinic (E2-deprived) may reveal important, clinically relevant molecular vulnerabilities.

In addition to E2 deprivation, we used anchorage independence as a model of early metastasis. Although AR significantly increased in both ER mutant and WT breast cancer cells surviving anchorage-independent conditions, antiandrogens abrogated the selective advantage of mutant ER cells for anchorage-independent survival. This is consistent with previous studies in TNBC cells that demonstrated increased AR in anchorage-independent cells, where it protected against apoptosis (28), rendering anchorage-independent cells highly sensitive to AR inhibition (54). Therefore, AR targeting may limit AI-resistant MBC recurrence, perhaps for both patients with either WT or mutant ER, because some WT ER biopsies retained AR but had low ER and PR.

CHI3L1, encoding chitinase 3-like 1, was preferentially expressed in breast cancer cells harboring mutant ER. CHI3L1/YKL-40 is linked to chronic inflammation and cancer and is upregulated in the mammary gland during involution, where it contributes to a protumorigenic wound healing like environment via enhanced macrophage infiltration (55). In prostate cancer, CHI3L1 is upregulated by androgens/AR (34), and in breast cancer, it is immune suppressive (56), and correlates with M2 macrophage infiltration (57), enhanced metastasis (58, 59), and poor prognosis (60). While CHI3L1 was noted previously as increased in mutant ER breast cancer cells (13), it was not further explored. In this study, CHI3L1 protein was higher in mutant ER MCF7 compared with WT. Because CHI3L1 is associated with decreased E-cadherin and enhanced MMP-9 activity in the involuting mammary gland (55), we investigated its role in invasion and found that the selective advantage of D538G mutant ER MCF7 cells for invasion was abolished by a CHI3L1 neutralizing antibody. While a gene signature suggestive of increased migration/invasive capacity was linked to mutant ER (12, 13), to our knowledge this is the first demonstration that a specific protein can be targeted in mutant ER cells to decrease invasiveness.

Pathway analysis of genes significantly differentially expressed by mutant versus WT ER in data from three studies demonstrated that genes in the type-1 IFN signaling and innate immune pathways were enhanced in mutant ER lines. In normal tissues, infectious pathogens are recognized by pattern or damage recognition receptors that activate type-1 IFN production. IFNα/β signal through IFN receptors and STAT1 activate transcription of ISGs to facilitate an innate immune response against the pathogen. The cyclic GMP-AMP synthase (cGAS) is a known stimulator of ISGs via the STING signaling pathway that can be upregulated in cancer cells in response to cytoplasmic DNA or mitochondrial DNA (reviewed in ref. 61). We focused on the ISGs IFITM3 and M1 because they were elevated in mutant ER cells compared with WT in our mRNA profiling, correlated with AR positivity in TNBC (26), elevated in AI-resistant breast cancer cells (36), induced in therapy-resistant prostate cancer models (62, 63), and upregulated by MUC1 in breast cancer cells (64). MUC1 also emerged as a mutant ER-specific gene increased in mutant ER cells at the protein level. We confirmed that both D538G and Y537S MCF7 cells have elevated IFITM3 protein at baseline that was further elevated by IFNβ. Importantly, clinical biopsies harboring the D538G mutation had significantly more IFITM3 protein compared with metastatic lesions with WT ER.

Given the enhanced expression of the type-1 IFN pathway protein IFITM3 in mutant ER MBC specimens, we sought to determine whether there was a difference in immune cell infiltrate in WT versus mutant ER biopsies. We found significantly elevated Th and T regulatory cells, as well as enhanced macrophage infiltration in MBC with mutant compared WT ER. Androgens stimulated alternative activation (M2 protumor polarization) of macrophages in an AR-dependent manner in an allergic asthma model (65). Furthermore, CHI3L1 stimulated by inflammation leads to M2 macrophage polarization during allergic reactions (66). Thus, enhanced macrophage infiltration in mutant ER MBC may be the result of elevated innate immune signaling pathways/proteins like type-1 IFN, active AR, and/or CHI3L1.

Although ER+ breast cancer is more “immunologically cold” than TNBC, the recent efficacy of anti-PD-L1 antibody in high-risk patients with ER+ disease (38), confirms that some ER+ breast cancer may respond to immunotherapy. Indeed, we found that PD-L1–positive macrophages were significantly higher in mutant ER metastatic lesions than those with WT ER. Interestingly, the other two genes most commonly mutated in breast cancer, TP53 and PIK3CA, did not correlate with altered macrophages, but did correlate with increased CD8+ cytotoxic T cells. These data suggest that enhanced immune suppression via PD-L1+ macrophages may be uniquely acquired by ER-mutant MBC. Because expression of PD-L1 in the tumor microenvironment, primarily its expression on myeloid cells, predicts breast cancer response to checkpoint inhibitors (40), future studies will explore checkpoint inhibition in WT versus mutant ER MBC.

Evidence presented here, and in Arnesen and colleagues (22), suggest that mutant ER containing cells may be poised to survive the stress of E2 deprivation and anchorage independence by the constitutive activity of mutant ER and by changes in non-ER-regulated genes. Our studies in breast cancer cells, PDX, and patient biopsies revealed proteins/pathways that may serve as targetable vulnerabilities in ER-mutant disease, including AR, CHI3L1, and ISGs. On the other hand, metastases without ESR1 exon 8 mutations expressed significantly less ER and PR than those with mutant ER, but clearly adapt to the same selective pressure of E2 deprivation (AI therapy) and the metastatic cascade. Adaptations may occur through shared targetable mechanisms like activated AR, or unique mechanisms that remain to be determined via preclinical modeling under clinically relevant conditions that mimic the selective pressures under which they evolve in patients. The differences in hormone receptor protein expression and tumor-infiltrating immune cells in biopsies with ESR1 mutations versus ESR1 WT suggest that AI-resistant MBC should be examined in light of the different pathways and adaptations that lead to resistance, which in turn result in new vulnerabilities to guide novel therapeutic strategies.

M.M. Williams reports grants from NIH during the conduct of the study. S. Arnesen reports grants from Department of Defense during the conduct of the study. S. Oesterreich reports grants from NCI during the conduct of the study. J. Gertz reports grants from Department of Defense during the conduct of the study and grants from Zenopharm outside the submitted work. No disclosures were reported by the other authors.

M.M. Williams: Conceptualization, data curation, formal analysis, investigation, writing-original draft, project administration, writing-review and editing. N.S. Spoelstra: Data curation, formal analysis, methodology, writing-review and editing. S. Arnesen: Data curation, formal analysis, writing-review and editing. K.I. O'Neill: Data curation, writing-review and editing. J.L. Christenson: Data curation, writing-review and editing. J. Reese: Data curation, formal analysis, methodology, writing-review and editing. K.C. Torkko: Formal analysis, methodology, writing-review and editing, statistical analysis. A. Goodspeed: Data curation, formal analysis, bioinformatics expertise. E. Rosas: Data curation. T. Hanamura: Data curation. S.B. Sams: Data curation, formal analysis, pathology analysis. Z. Li: Resources, writing-review and editing. S. Oesterreich: Resources, writing-review and editing. R.B. Riggins: Resources, writing-review and editing. B.M. Jacobsen: Conceptualization, data curation, formal analysis, writing-original draft, writing-review and editing. A. Elias: Resources, funding acquisition, visualization, writing-review and editing. J. Gertz: Conceptualization, funding acquisition, visualization, writing-review and editing. J.K. Richer: Conceptualization, supervision, funding acquisition, visualization, writing-original draft, project administration, writing-review and editing.

The authors acknowledge the use of the University of Colorado Cancer Center/NIH/NCI Cancer Core Support Grant P30 CA046934 for Tissue Biobanking and Histology, Molecular Pathology, Animal Imaging, Biostatistics and Bioinformatics, and Cell Technologies shared resources. We thank Adrie Van Bokhoven, Kathryn Zolman, and Kurtis Davies in particular for their contributions. We would also like to thank the patients enrolled in NCT02953860 (COMIRB 16-1001) and patient advocates Karen Raines Hunt, Jane Perlmutter, and Vickie Tosher. This investigation was supported by DOD BCRP W81XWH-16-1-0422 BC151357 and BC151357P1 (to J. Gertz and J.K. Richer); DOD BCRP BC120183 W81XWH-13-1-0090 (to J.K. Richer and A. Elias); NIH R01CA187733 (to J.K. Richer), NIH R01CA221303 (to S. Oesterreich), T32CA190216 and NIH NRSA F32 CA239436 (to M.M. Williams) and DOD W81XWH-17-1-0615 BC161497 (to R.B. Riggins). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.

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.

1.
van Hellemond
IEG
,
Geurts
SME
,
Tjan-Heijnen
VCG
. 
Current status of extended adjuvant endocrine therapy in early stage breast cancer
.
Curr Treat Options Oncol
2018
;
19
:
26
.
2.
Hartmaier
RJ
,
Trabucco
SE
,
Priedigkeit
N
,
Chung
JH
,
Parachoniak
CA
,
Vanden Borre
P
, et al
Recurrent hyperactive ESR1 fusion proteins in endocrine therapy-resistant breast cancer
.
Ann Oncol
2018
;
29
:
872
80
.
3.
Jeselsohn
R
,
De Angelis
C
,
Brown
M
,
Schiff
R
. 
The evolving role of the estrogen receptor mutations in endocrine therapy-resistant breast cancer
.
Curr Oncol Rep
2017
;
19
:
35
.
4.
Karnik
PS
,
Kulkarni
S
,
Liu
XP
,
Budd
GT
,
Bukowski
RM
. 
Estrogen receptor mutations in tamoxifen-resistant breast cancer
.
Cancer Res
1994
;
54
:
349
53
.
5.
Li
S
,
Shen
D
,
Shao
J
,
Crowder
R
,
Liu
W
,
Prat
A
, et al
Endocrine-therapy-resistant ESR1 variants revealed by genomic characterization of breast-cancer-derived xenografts
.
Cell Rep
2013
;
4
:
1116
30
.
6.
Toy
W
,
Shen
Y
,
Won
H
,
Green
B
,
Sakr
RA
,
Will
M
, et al
ESR1 ligand-binding domain mutations in hormone-resistant breast cancer
.
Nat Genet
2013
;
45
:
1439
45
.
7.
Robinson
DR
,
Wu
YM
,
Vats
P
,
Su
F
,
Lonigro
RJ
,
Cao
X
, et al
Activating ESR1 mutations in hormone-resistant metastatic breast cancer
.
Nat Genet
2013
;
45
:
1446
51
.
8.
Barone
I
,
Iacopetta
D
,
Covington
KR
,
Cui
Y
,
Tsimelzon
A
,
Beyer
A
, et al
Phosphorylation of the mutant K303R estrogen receptor alpha at serine 305 affects aromatase inhibitor sensitivity
.
Oncogene
2010
;
29
:
2404
14
.
9.
Herynk
MH
,
Parra
I
,
Cui
Y
,
Beyer
A
,
Wu
MF
,
Hilsenbeck
SG
, et al
Association between the estrogen receptor alpha A908G mutation and outcomes in invasive breast cancer
.
Clin Cancer Res
2007
;
13
:
3235
43
.
10.
Reinert
T
,
Saad
ED
,
Barrios
CH
,
Bines
J
. 
Clinical implications of ESR1 mutations in hormone receptor-positive advanced breast cancer
.
Front Oncol
2017
;
7
:
26
.
11.
Yu
L
,
Wang
L
,
Mao
C
,
Duraki
D
,
Kim
JE
,
Huang
R
, et al
Estrogen-independent Myc overexpression confers endocrine therapy resistance on breast cancer cells expressing ERalphaY537S and ERalphaD538G mutations
.
Cancer Lett
2019
;
442
:
373
82
.
12.
Jeselsohn
R
,
Bergholz
JS
,
Pun
M
,
Cornwell
M
,
Liu
W
,
Nardone
A
, et al
Allele-specific chromatin recruitment and therapeutic vulnerabilities of ESR1 activating mutations
.
Cancer Cell
2018
;
33
:
173
86.e5
.
13.
Bahreini
A
,
Li
Z
,
Wang
P
,
Levine
KM
,
Tasdemir
N
,
Cao
L
, et al
Mutation site and context dependent effects of ESR1 mutation in genome-edited breast cancer cell models
.
Breast Cancer Res
2017
;
19
:
60
.
14.
Martin
LA
,
Ribas
R
,
Simigdala
N
,
Schuster
E
,
Pancholi
S
,
Tenev
T
, et al
Discovery of naturally occurring ESR1 mutations in breast cancer cell lines modelling endocrine resistance
.
Nat Commun
2017
;
8
:
1865
.
15.
Fuqua
SA
,
Fitzgerald
SD
,
Chamness
GC
,
Tandon
AK
,
McDonnell
DP
,
Nawaz
Z
, et al
Variant human breast tumor estrogen receptor with constitutive transcriptional activity
.
Cancer Res
1991
;
51
:
105
9
.
16.
McGuire
WL
,
Chamness
GC
,
Fuqua
SA
. 
Estrogen receptor variants in clinical breast cancer
.
Mol Endocrinol
1991
;
5
:
1571
7
.
17.
Katzenellenbogen
JA
,
Mayne
CG
,
Katzenellenbogen
BS
,
Greene
GL
,
Chandarlapaty
S
. 
Structural underpinnings of oestrogen receptor mutations in endocrine therapy resistance
.
Nat Rev Cancer
2018
;
18
:
377
88
.
18.
Turner
NC
,
Slamon
DJ
,
Ro
J
,
Bondarenko
I
,
Im
SA
,
Masuda
N
, et al
Overall survival with palbociclib and fulvestrant in advanced breast cancer
.
N Engl J Med
2018
;
379
:
1926
36
.
19.
Hamilton
EP
,
Patel
MR
,
Armstrong
AC
,
Baird
RD
,
Jhaveri
K
,
Hoch
M
, et al
A first-in-human study of the new oral selective estrogen receptor degrader AZD9496 for ER(+)/HER2(-) advanced breast cancer
.
Clin Cancer Res
2018
;
24
:
3510
8
.
20.
Li
Z
,
Levine
KM
,
Bahreini
A
,
Wang
P
,
Chu
D
,
Park
BH
, et al
Upregulation of IRS1 enhances IGF1 response in Y537S and D538G ESR1 mutant breast cancer cells
.
Endocrinology
2018
;
159
:
285
96
.
21.
Gelsomino
L
,
Gu
G
,
Rechoum
Y
,
Beyer
AR
,
Pejerrey
SM
,
Tsimelzon
A
, et al
Erratum to: ESR1 mutations affect anti-proliferative responses to tamoxifen through enhanced cross-talk with IGF signaling
.
Breast Cancer Res Treat
2017
;
163
:
639
40
.
22.
Arnesen
S
,
Blanchard
Z
,
Williams
MM
,
Berrett
KC
,
Li
Z
,
Oesterreich
S
, et al
Estrogen receptor alpha mutations in breast cancer cells cause gene expression changes through constant activity and secondary effects
.
Cancer Res
2021
;
81
:
539
51
.
23.
Chandarlapaty
S
,
Chen
D
,
He
W
,
Sung
P
,
Samoila
A
,
You
D
, et al
Prevalence of ESR1 mutations in cell-free DNA and outcomes in metastatic breast cancer: a secondary analysis of the BOLERO-2 clinical trial
.
JAMA Oncol
2016
;
2
:
1310
5
.
24.
Savic
D
,
Partridge
EC
,
Newberry
KM
,
Smith
SB
,
Meadows
SK
,
Roberts
BS
, et al
CETCh-seq: CRISPR epitope tagging ChIP-seq of DNA-binding proteins
.
Genome Res
2015
;
25
:
1581
9
.
25.
Cuzick
J
,
Swanson
GP
,
Fisher
G
,
Brothman
AR
,
Berney
DM
,
Reid
JE
, et al
Prognostic value of an RNA expression signature derived from cell cycle proliferation genes in patients with prostate cancer: a retrospective study
.
Lancet Oncol
2011
;
12
:
245
55
.
26.
Doane
AS
,
Danso
M
,
Lal
P
,
Donaton
M
,
Zhang
L
,
Hudis
C
, et al
An estrogen receptor-negative breast cancer subset characterized by a hormonally regulated transcriptional program and response to androgen
.
Oncogene
2006
;
25
:
3994
4008
.
27.
Jin
HJ
,
Kim
J
,
Yu
J
. 
Androgen receptor genomic regulation
.
Transl Androl Urol
2013
;
2
:
157
77
.
28.
Barton
VN
,
Christenson
JL
,
Gordon
MA
,
Greene
LI
,
Rogers
TJ
,
Butterfield
K
, et al
Androgen receptor supports an anchorage-independent, cancer stem cell-like population in triple-negative breast cancer
.
Cancer Res
2017
;
77
:
3455
66
.
29.
Norris
JD
,
Ellison
SJ
,
Baker
JG
,
Stagg
DB
,
Wardell
SE
,
Park
S
, et al
Androgen receptor antagonism drives cytochrome P450 17A1 inhibitor efficacy in prostate cancer
.
J Clin Invest
2017
;
127
:
2326
38
.
30.
Gucalp
A
,
Tolaney
S
,
Isakoff
SJ
,
Ingle
JN
,
Liu
MC
,
Carey
LA
, et al
Phase II trial of bicalutamide in patients with androgen receptor-positive, estrogen receptor-negative metastatic breast cancer
.
Clin Cancer Res
2013
;
19
:
5505
12
.
31.
Traina
TA
,
Miller
K
,
Yardley
DA
,
Eakle
J
,
Schwartzberg
LS
,
O'Shaughnessy
J
, et al
Enzalutamide for the treatment of androgen receptor-expressing triple-negative breast cancer
.
J Clin Oncol
2018
;
36
:
884
90
.
32.
Brasso
K
,
Christensen
IJ
,
Johansen
JS
,
Teisner
B
,
Garnero
P
,
Price
PA
, et al
Prognostic value of PINP, bone alkaline phosphatase, CTX-I, and YKL-40 in patients with metastatic prostate carcinoma
.
Prostate
2006
;
66
:
503
13
.
33.
Johansen
JS
,
Brasso
K
,
Iversen
P
,
Teisner
B
,
Garnero
P
,
Price
PA
, et al
Changes of biochemical markers of bone turnover and YKL-40 following hormonal treatment for metastatic prostate cancer are related to survival
.
Clin Cancer Res
2007
;
13
:
3244
9
.
34.
Jeet
V
,
Tevz
G
,
Lehman
M
,
Hollier
B
,
Nelson
C
. 
Elevated YKL40 is associated with advanced prostate cancer (PCa) and positively regulates invasion and migration of PCa cells
.
Endocr Relat Cancer
2014
;
21
:
723
37
.
35.
Ogony
J
,
Choi
HJ
,
Lui
A
,
Cristofanilli
M
,
Lewis-Wambi
J
. 
Interferon-induced transmembrane protein 1 (IFITM1) overexpression enhances the aggressive phenotype of SUM149 inflammatory breast cancer cells in a signal transducer and activator of transcription 2 (STAT2)-dependent manner
.
Breast Cancer Res
2016
;
18
:
25
.
36.
Choi
HJ
,
Lui
A
,
Ogony
J
,
Jan
R
,
Sims
PJ
,
Lewis-Wambi
J
. 
Targeting interferon response genes sensitizes aromatase inhibitor resistant breast cancer cells to estrogen-induced cell death
.
Breast Cancer Res
2015
;
17
:
6
.
37.
Escher
TE
,
Lui
AJ
,
Geanes
ES
,
Walter
KR
,
Tawfik
O
,
Hagan
CR
, et al
Interaction between MUC1 and STAT1 drives IFITM1 overexpression in aromatase inhibitor-resistant breast cancer cells and mediates estrogen-induced apoptosis
.
Mol Cancer Res
2019
;
17
:
1180
94
.
38.
Nanda
R
,
Liu
MC
,
Yau
C
,
Asare
S
,
Hylton
N
,
Veer
LVt
, et al
Pembrolizumab plus standard neoadjuvant therapy for high-risk breast cancer (BC): results from I-SPY 2
.
J Clin Oncol
35
:
15s
, 
2017
(
suppl; abstr 506
).
39.
Kyte
JA
,
Andresen
NK
,
Russnes
HG
,
Fretland
SO
,
Falk
RS
,
Lingjaerde
OC
, et al
ICON: a randomized phase IIb study evaluating immunogenic chemotherapy combined with ipilimumab and nivolumab in patients with metastatic hormone receptor positive breast cancer
.
J Transl Med
2020
;
18
:
269
.
40.
Schmid
P
,
Adams
S
,
Rugo
HS
,
Schneeweiss
A
,
Barrios
CH
,
Iwata
H
, et al
Atezolizumab and Nab-Paclitaxel in advanced triple-negative breast cancer
.
N Engl J Med
2018
;
379
:
2108
21
.
41.
Gallicchio
L
,
Macdonald
R
,
Wood
B
,
Rushovich
E
,
Helzlsouer
KJ
. 
Androgens and musculoskeletal symptoms among breast cancer patients on aromatase inhibitor therapy
.
Breast Cancer Res Treat
2011
;
130
:
569
77
.
42.
Robinson
JL
,
Macarthur
S
,
Ross-Innes
CS
,
Tilley
WD
,
Neal
DE
,
Mills
IG
, et al
Androgen receptor driven transcription in molecular apocrine breast cancer is mediated by FoxA1
.
EMBO J
2011
;
30
:
3019
27
.
43.
Lehmann
BD
,
Bauer
JA
,
Chen
X
,
Sanders
ME
,
Chakravarthy
AB
,
Shyr
Y
, et al
Identification of human triple-negative breast cancer subtypes and preclinical models for selection of targeted therapies
.
J Clin Invest
2011
;
121
:
2750
67
.
44.
Dees
EC
,
Carey
LA
. 
Improving endocrine therapy for breast cancer: it's not that simple
.
J Clin Oncol
31
:
2s
, 
2013
(
suppl; abstr 171–3
).
45.
Giuliano
M
,
Schiff
R
,
Osborne
CK
,
Trivedi
MV
. 
Biological mechanisms and clinical implications of endocrine resistance in breast cancer
.
The Breast
2011
;
20
:
S42
9
.
46.
Harvell
DME
,
Spoelstra
NS
,
Singh
M
,
McManaman
JL
,
Finlayson
C
,
Phang
T
, et al
Molecular signatures of neoadjuvant endocrine therapy for breast cancer: characteristics of response or intrinsic resistance
.
Breast Cancer Res Treat
2008
;
112
:
475
88
.
47.
Harvell
DME
,
Richer
JK
,
Singh
M
,
Spoelstra
N
,
Finlayson
C
,
Borges
VF
, et al
Estrogen regulated gene expression in response to neoadjuvant endocrine therapy of breast cancers: tamoxifen agonist effects dominate in the presence of an aromatase inhibitor
.
Breast Cancer Res Treat
2008
;
112
:
489
501
.
48.
De Amicis
F
,
Thirugnansampanthan
J
,
Cui
Y
,
Selever
J
,
Beyer
A
,
Parra
I
, et al
Androgen receptor overexpression induces tamoxifen resistance in human breast cancer cells
.
Breast Cancer Res Treat
2010
;
121
:
1
11
.
49.
Cochrane
DR
,
Bernales
S
,
Jacobsen
BM
,
Cittelly
DM
,
Howe
EN
,
D'Amato
NC
, et al
Role of the androgen receptor in breast cancer and preclinical analysis of enzalutamide
.
Breast Cancer Res
2014
;
16
:
R7
.
50.
D'Amato
NC
,
Gordon
MA
,
Babbs
B
,
Spoelstra
NS
,
Carson Butterfield
KT
,
Torkko
KC
, et al
Cooperative dynamics of AR and ER activity in breast cancer
.
Mol Cancer Res
2016
;
14
:
1054
67
.
51.
Aceto
N
,
Bardia
A
,
Wittner
BS
,
Donaldson
MC
,
O'Keefe
R
,
Engstrom
A
, et al
AR expression in breast cancer CTCs associates with bone metastases
.
Mol Cancer Res
2018
;
16
:
720
7
.
52.
Lawson
DA
,
Bhakta
NR
,
Kessenbrock
K
,
Prummel
KD
,
Yu
Y
,
Takai
K
, et al
Single-cell analysis reveals a stem-cell program in human metastatic breast cancer cells
.
Nature
2015
;
526
:
131
5
.
53.
Feng
J
,
Li
L
,
Zhang
N
,
Liu
J
,
Zhang
L
,
Gao
H
, et al
Androgen and AR contribute to breast cancer development and metastasis: an insight of mechanisms
.
Oncogene
2017
;
36
:
2775
90
.
54.
Barton
VN
,
D'Amato
NC
,
Gordon
MA
,
Lind
HT
,
Spoelstra
NS
,
Babbs
BL
, et al
Multiple molecular subtypes of triple-negative breast cancer critically rely on androgen receptor and respond to enzalutamide in vivo
.
Mol Cancer Ther
2015
;
14
:
769
78
.
55.
Hughes
K
,
Wickenden
JA
,
Allen
JE
,
Watson
CJ
. 
Conditional deletion of Stat3 in mammary epithelium impairs the acute phase response and modulates immune cell numbers during post-lactational regression
.
J Pathol
2012
;
227
:
106
17
.
56.
Cohen
N
,
Shani
O
,
Raz
Y
,
Sharon
Y
,
Hoffman
D
,
Abramovitz
L
, et al
Fibroblasts drive an immunosuppressive and growth-promoting microenvironment in breast cancer via secretion of Chitinase 3-like 1
.
Oncogene
2017
;
36
:
4457
68
.
57.
Steenbrugge
J
,
Breyne
K
,
Demeyere
K
,
De Wever
O
,
Sanders
NN
,
Van Den Broeck
W
, et al
Anti-inflammatory signaling by mammary tumor cells mediates prometastatic macrophage polarization in an innovative intraductal mouse model for triple-negative breast cancer
.
J Exp Clin Cancer Res
2018
;
37
:
191
.
58.
Chen
Y
,
Zhang
S
,
Wang
Q
,
Zhang
X
. 
Tumor-recruited M2 macrophages promote gastric and breast cancer metastasis via M2 macrophage-secreted CHI3L1 protein
.
J Hematol Oncol
2017
;
10
:
36
.
59.
Ma
B
,
Herzog
EL
,
Lee
CG
,
Peng
X
,
Lee
CM
,
Chen
X
, et al
Role of chitinase 3-like-1 and semaphorin 7a in pulmonary melanoma metastasis
.
Cancer Res
2015
;
75
:
487
96
.
60.
Wan
G
,
Xiang
L
,
Sun
X
,
Wang
X
,
Li
H
,
Ge
W
, et al
Elevated YKL-40 expression is associated with a poor prognosis in breast cancer patients
.
Oncotarget
2017
;
8
:
5382
91
.
61.
Loo
TM
,
Miyata
K
,
Tanaka
Y
,
Takahashi
A
. 
Cellular senescence and senescence-associated secretory phenotype via the cGAS-STING signaling pathway in cancer
.
Cancer Sci
2020
;
111
:
304
11
.
62.
Borghesan
M
,
Fafian-Labora
J
,
Eleftheriadou
O
,
Carpintero-Fernandez
P
,
Paez-Ribes
M
,
Vizcay-Barrena
G
, et al
Small extracellular vesicles are key regulators of non-cell autonomous intercellular communication in senescence via the interferon protein IFITM3
.
Cell Rep
2019
;
27
:
3956
71.e6
.
63.
Civenni
G
,
Bosotti
R
,
Timpanaro
A
,
Vazquez
R
,
Merulla
J
,
Pandit
S
, et al
Epigenetic control of mitochondrial fission enables self-renewal of stem-like tumor cells in human prostate cancer
.
Cell Metab
2019
;
30
:
303
18
.
64.
Provance
OK
,
Lewis-Wambi
J
. 
Deciphering the role of interferon alpha signaling and microenvironment crosstalk in inflammatory breast cancer
.
Breast Cancer Res
2019
;
21
:
59
.
65.
Becerra-Diaz
M
,
Strickland
AB
,
Keselman
A
,
Heller
NM
. 
Androgen and androgen receptor as enhancers of M2 macrophage polarization in allergic lung inflammation
.
J Immunol
2018
;
201
:
2923
33
.
66.
Kwak
EJ
,
Hong
JY
,
Kim
MN
,
Kim
SY
,
Kim
SH
,
Park
CO
, et al
Chitinase 3-like 1 drives allergic skin inflammation via Th2 immunity and M2 macrophage activation
.
Clin Exp Allergy
2019
;
49
:
1464
74
.