An AIB1 isoform rewires glucocorticoid receptor signaling to promote TNBC progression

Summary

The role of glucocorticoid receptor (GR) signaling in triple-negative breast cancer (TNBC) progression remains poorly defined. Here, we describe a GR-dependent mechanism driving TNBC invasion, mediated by the presence of a subpopulation of cancer cells that express an N-terminal truncated splice isoform of the nuclear receptor coactivator AIB1. Invasion was driven through direct contact of this subpopulation with neighboring cancer cells, and suppressed by GR antagonists or depletion of GR. Crosstalk between the AIB1 isoform-expressing cells and full-length AIB1-expressing cells triggered enhanced GR activation, GR signaling, and distinct patterns of AIB1 genomic engagement. Notably, GR signaling selectively activated pathways driven by Myc in the AIB1 isoform-expressing population, and the reduction of Myc reduced invasion. These findings identify the emergence of an AIB1 isoform-expressing subpopulation as a key mechanism driving progression in TNBC and suggest sensitivity to GR-targeted therapies.

Graphical abstract

Highlights

• •An AIB1 splice isoform rewires GR signaling to drive collective invasion in TNBC
• •AIB1 isoform-expressing TNBC cells are selectively sensitized to glucocorticoids
• •AIB1 isoform presence is a potential biomarker for GR antagonist therapy in TNBC

Molecular biology; Cell biology; Cancer

Introduction

Triple-negative breast cancer (TNBC) lacks estrogen receptor (ER), progesterone receptor (PR), and HER2 amplification. Advanced-stage TNBC frequently metastasizes to the lungs, liver, and central nervous system and is associated with poor prognosis and a median overall survival of less than 2 years.^1,2,3,4 Despite its lethality, the molecular drivers of TNBC invasion and metastasis remain incompletely understood.

The glucocorticoid receptor (GR) is expressed in some patients with TNBC and is associated with worse overall survival.^5,6 GR activity has been linked to the upregulation of pro-survival and epithelial-to-mesenchymal transition (EMT) genes, as well as the activation of oncogenic transcriptional programs.^5,7,8,9 Recent studies have defined stress-responsive signaling mechanisms that converge on GR to promote TNBC progression. For example, phosphorylation of GR integrates glucocorticoid and hypoxia pathways through interaction with the transcription coregulator PELP1 and HIF, driving the induction of Breast tumor kinase (Brk) and metastatic behavior.^10,11,12 These findings highlight that GR can function as a nodal sensor of both hormonal and cellular stress cues, amplifying pro-invasive transcriptional programs in aggressive TNBC. Additionally, elevated circulating glucocorticoids have been shown to promote metastasis in TNBC models through the increased expression of kinase ROR1.¹³ However, the cellular context and co-regulatory mechanisms through which GR promotes tumor progression remain poorly defined.

We previously identified an mRNA splice isoform of the nuclear receptor coactivator AIB1/SRC-3/NCOA3, termed AIB1Δ4, which lacks exon 4 and the corresponding N-terminal 223 amino acids encompassing the bHLH-PAS domain.¹⁴ This isoform was initially identified through the RT-PCR analysis of total RNA from MCF-7 human breast cancer cells, which overexpress AIB1.¹⁵ Alternative splicing of exon 4 leads to a new translation start site on exon 7 in frame with the full-length AIB1 isoform.¹⁶ AIB1 is frequently amplified or overexpressed in breast and other cancers^{15,17,18,19,20} and drives oncogenic progression,^21,22,23,24 including in TNBC.²⁵ The AIB1Δ4 isoform retains the nuclear receptor domain (NRD) containing LXXLL motifs required for interaction with the GR,²⁶ but its truncated N-terminus results in the loss of interactions with tumor suppressors and in increased efficacy as a nuclear receptor coactivator.^14,16,27,28 The percentage of TNBCs that express the AIB1Δ4 isoform is unknown, but it is barely detectable in normal mammary epithelium and upregulated in breast cancers, including TNBC.^14,29 Its contribution to breast cancer progression is supported by the emergence of TNBC in transgenic mice by the expression of the AIB1Δ4 isoform.³⁰ Using CRISPR editing, we generated cells that exclusively express AIB1Δ4 and demonstrated that this subpopulation can increase primary tumor size and invasion and metastasis of surrounding parental TNBC cells in vitro and in vivo, despite being largely indolent themselves. We designated these indolent AIB1Δ4-expressing cells as “enablers” of the invasion of the surrounding parental cells.²⁹ The increased invasion of the parental cells promoted by the enabler cells in vitro was further enhanced by a synthetic glucocorticoid dexamethasone (dex), which can activate the GR directly or through GR heterodimerization.³¹

Here, we demonstrate that GR antagonists can inhibit the ability of cells that express only AIB1Δ4 (DCISΔ4) to promote the invasion of parental DCIS cells in 3D invasion assays in vitro and in zebrafish models of extravasation. MCFDCIS (DCIS) triple-negative ductal carcinoma in situ cells are a derivative of the MCF10A premalignant line established by Miller et al. that is the most widely used model of human early-stage breast cancer progression.^32,33 The MCFDCIS model was derived from xenograft outgrowths of MCF10AT lesions and represents a basal-like, ER-/PR-/HER2-line that harbors a PI3K gain-of-function mutation and expresses signaling molecules associated with malignant progression.³⁴ MCFDCIS cells can give rise to DCIS lesions in xenografts that contain precursor populations with metastatic potential. Human triple-negative DCIS is detected in about 5%–10% of cases, and triple-negative DCIS lesions are thought to rapidly progress to high-grade DCIS and invasive breast cancer,^34,35 underscoring the importance of understanding mechanisms driving early invasion in this subtype. The inhibition of DCIS invasion we observe is primarily through direct effects on GR signaling in the AIB1Δ4-expressing population. Consistently, GR inhibition in mice reduces the primary tumor growth of mixed DCIS/DCISΔ4 xenografts in immunocompromised models.

To determine if the GR dependence also occurred in mouse models with an intact immune system, we identified a previously unknown AIB1 N-terminal truncated murine homolog, AIB1Δ5, which can similarly drive tumor progression in mouse allograft models. GR inhibition suppresses invasion pathways in a syngeneic 67NR/67NRΔ5 model with an intact immune system, corroborating the conserved function of this isoform.

Mechanistically, we show that AIB1Δ4-expressing cells exhibit distinct GR-driven transcriptional responses compared to full-length AIB1-expressing cells, and that these responses are amplified upon cell-cell crosstalk. CUT&RUN profiling reveals AIB1 isoform-specific differences in GR and AIB1 chromatin engagement, along with distinct patterns of chromatin accessibility. Importantly, Myc emerges as a critical downstream effector of GR signaling in AIB1Δ4-expressing cells following crosstalk with full-length AIB1-expressing parental cells.

Together, these findings identify a glucocorticoid-responsive subpopulation of TNBC cells defined by AIB1 isoform expression and uncover a targetable GR-Myc axis that drives invasion through tumor cell crosstalk.

Results

Glucocorticoid receptor inhibition decreases invasion driven by AIB1Δ4-expressing subpopulations

The invasion and metastasis of DCIS parental cells is significantly increased by the presence of a small fraction of DCIS cells that only express the AIB1Δ4 splice isoform (DCISΔ4 cells) (Figure 1A) and can be enhanced by dex, a synthetic glucocorticoid.^29,36 To investigate the role of GR activation and signaling in this effect, we utilized a 3D invasion assay. Fluorescently labeled DCIS and DCISΔ4 cells were cocultured at a ratio of 4:1 (DCIS:DCISΔ4) in the presence of GR antagonists and embedded in a mixture of collagen I and matrigel. Both the GR/PR/androgen receptor (AR) antagonist mifepristone (RU486) and the more selective GR antagonist relacorilant (rela) significantly decreased the invasive area of spheres compared to control-treated cells (Figures 1B and 1C), indicating GR activity is rate-limiting for the increased invasion of the mixed population of DCIS cells. To confirm that the investigational drug rela was inhibiting GR transcriptional activity, GR-response genes were analyzed via RT-qPCR (Figures S1A and S1B). Similar to RU486, rela treatment resulted in a significant decrease in the gene expression of KLF9 and fibronectin (FN1), known glucocorticoid target genes.¹³

Figure 1.

GR inhibition reduces DCISΔ4 cell-driven invasion

(A) Schematic representation of how a subpopulation of DCISΔ4 (green) cells drives DCIS parental (red) cells to invade and metastasize through cell-cell crosstalk.

(B) Representative images of a 4:1 (DCIS:DCISΔ4) mix 3D sphere invasion in 80% collagen I and 20% matrigel treated with either vehicle control, 0.1 μM RU486, or 0.1 μM relacorilant (Rela). White arrows point to invasive DCIS cells. Scale bars, 200 μm.

(C) Quantification of the invasive area of spheres (n = 12 spheres) in each condition from B. Median plotted as a solid line. ∗∗p < 0.01; ∗∗∗∗p < 0.0001. p value was calculated using two-tailed unpaired Student’s t-tests.

(D) Schematic representation of the experiment where a 4:1 mixture of DCIS-RFP:DCISΔ4-GFP cells was injected into the circulation of zebrafish embryos via the duct of Cuvier. Fish swam in water treated with vehicle control or 0.1 μM Rela, and DCIS extravasation was scored. Representative image of a zebrafish embryo with RFP labeled DCIS cells extravasating into the tissue of the tail region (white arrows).

(E) Scoring of zebrafish embryos as having DCIS cells extravasating into the tissue in vehicle control or 0.1 μM Rela treatment groups. Data represent n = 37 control treated fish and n = 39 Rela treated fish. ∗p < 0.05. p value calculated using Fisher’s exact test.

(F) Schematic representation of mouse in vivo experimental design. One million DCIS cells at a 4:1 mix of DCIS parental and DCISΔ4, respectively, were injected bilaterally into the mammary fat pads (MFP) of 6-week-old female NOD/SCID mice. Rela (30 mg/kg) or vehicle treatment was started on Day 7 and was administered every Monday, Wednesday, and Friday via oral gavage. Primary tumors were removed on Day 34. Treatments resumed on Day 42. All mice were euthanized on Day 62, and tissues were collected. n = 6 mice per group.

(G) Tumor size in the control treated (n = 11) and rela treated (n = 12) groups. Data are represented as mean ± SD. ∗p < 0.05. p value was calculated using two-tailed unpaired Student’s t test.

(H) Ingenuity pathway analysis of RNA-seq from tumors of control and rela treated groups (n = 5 control and n = 5 rela tumors). Comparison is rela treated tumors versus control treated tumors.

Because FN1 is downstream of GR signaling and has been implicated in breast cancer invasion,^37,38 we inhibited FN1 using an RGDS peptide and evaluated DCIS cell invasion in coculture with DCISΔ4 cells. FN1 inhibition did not reduce parental cell invasion compared to the control peptide (Figure S1C). Inhibition of another GR target, Eph receptor signaling,^39,40 also did not reduce the invasive area of DCIS cells in coculture with DCISΔ4 cells (Figure S1D). Thus, neither of these pathways is a key mediator of the glucocorticoid-dependent enabling phenotype.

To determine whether the GR activity was exerted through the DCISΔ4 or the parental DCIS responder cells, shRNAs were used to knockdown GR in both cell types. DCISΔ4 GR knockdown cells did not proliferate (Figure S1E), indicating these cells had become dependent on GR for viability. Consistent with this, DCISΔ4 cells treated with RU486 or rela also had significantly reduced cell proliferation (Figure S1F, right). In contrast, knockdown of GR in DCIS parental cells (Figures S1G and S1H) did not affect their proliferation (Figure S1F, left) nor invasive potential when cultured alone in 3D (Figures S1I and S1J). The knockdown GR DCIS cells retained their ability to invade when cocultured with DCISΔ4 cells (Figures S1K and S1L). Taken together, these data demonstrated that GR was dispensable for the parental responder cells to invade and implied that the main impact of GR is exerted through the DCISΔ4 enabler cells.

To determine if GR modulation could influence the invasion of mixed DCIS/DCISΔ4 cell populations in vivo, we first examined the transgenic zebrafish (Tg(kdrl: GRCFP)zn1) as a model of cell extravasation. This zebrafish model has green reef coral fluorescent protein (GRCFP) expressed in the vascular endothelia under the control of a VEGFR2 promoter.⁴¹ Fluorescently labeled DCIS and DCISΔ4 cells were mixed at a 4:1 ratio, respectively, and injected into the duct of Cuvier of the zebrafish embryos, and rela or vehicle was added to the water (Figure 1D). Tissue-invasion of DCIS cells was then scored in the tail region of the zebrafish, quantitating the number of RFP-labeled parental cells that had extravasated from the fluorescently marked vessels. Extravasation of the DCIS cells in zebrafish was significantly decreased by rela treatment compared to treatment controls (Figure 1E).

To test the effect of GR inhibition on invasion in a mammalian immune-compromised model, DCIS/DCISΔ4 mixed cell populations were injected bilaterally into the mammary fat pad of female NOD/SCID mice (Figure 1F). Primary tumors were significantly smaller in the rela treated group compared to control (Figure 1G), indicating a reversal of the increased tumor size observed with DCIS mixed cells compared to DCIS parental cells alone in this model.²⁹ The inhibitory effect of GR antagonism on mixed DCIS/DCISΔ4 tumors likely reflects the disruption of GR-driven crosstalk between the two populations, as indicated later in discussion in Figure 5. RNA-seq analysis of tumors revealed the downregulation of several cancer-related pathways in the rela-treated group compared to controls (Figure 1H). Notably, pathways such as WNT/β-catenin signaling, Epithelial Adherens Junction Signaling, and Molecular Mechanisms of Cancer showed significantly negative activation z-scores, indicating the suppression of pathway activation.

Figure 5.

Crosstalk between parental DCIS and DCISΔ4 expressing cells significantly enhances the glucocorticoid response

(A) Schematic representation of an unmixed RNA-seq experimental design. A 4:1 mix of fluorescently labeled DCIS:DCISΔ4 cells, respectively, was cocultured for 48 h prior to treatment with either vehicle control, 5 nM dex or 5 nM dex plus 500 nM RU486. After 24 h of treatment, cell populations were separated via FACS, and RNA was extracted for RNA-seq. Data from these samples is called “unmixed.” n = 3 biological replicates for each cell line and condition.

(B) Total number of differentially expressed genes from RNA-seq between dex versus dex plus RU486 treatment in the indicated cells. All differentially expressed genes had an FDR ≤0.05. Data represent n = 3 biological replicates for each cell line and treatment condition.

(C) Overlap of differentially expressed genes (RNA-seq) between dex treated versus dex plus RU486-treated DCISΔ4 and DCISΔ4 unmixed cells. All differentially expressed genes had an FDR ≤0.05. Data represent n = 3 biological replicates for each cell line and treatment condition.

(D) Ingenuity pathway analysis (IPA) upstream regulators of RNA-seq differentially expressed genes from dex versus dex plus RU486 treated DCISΔ4 unmixed cells. Red lines are at -log₁₀(0.05) and Z score −1, 1.

(E) Number of GFP+ DCISΔ4 unmixed cells from FACS after treatment with either vehicle control, 5 nM dex, or 5 nM dex plus 500 nM RU486. Data are represented as mean ± SD. Data represent n = 3 biological replicates from each condition. ∗∗p < 0.005. p value was calculated using two-tailed unpaired Student’s t test.

(F) Proliferation of DCISΔ4 siRNA control (DCISΔ4-siCtrl) and DCISΔ4 siRNA Myc (DCISΔ4-siMyc-1 and DCISΔ4-siMyc-2) cells measured by ECIS. Data are represented as mean ± SD. ∗p < 0.05. p-value was calculated using two-tailed unpaired Student’s t test. Data represent n = 2 biological replicates for DCISΔ4-siCtrl and n = 3 biological replicates for each siMyc cell line.

(G) Representative images of a 4:1 (DCIS:DCISΔ4) mix 3D sphere invasion in 80% collagen I and 20% matrigel with control siRNA (DCISΔ4-siCtrl) or siRNA Myc DCISΔ4 cells (DCISΔ4-siMyc). Scale bars, 200 μm.

(H) Quantification of the invasive area of DCIS cells (n = 10 spheres) in each condition from G. Invasive area was compared to that of DCISΔ4-siCtrl mixed with DCIS cells. ns, non-significant. Median plotted as a solid line. ∗p < 0.05. p value was calculated using two-tailed unpaired Student’s t-tests.

From the 3D and in vivo models, it is clear that the effect of the AIB1Δ4 DCIS cells on the invasion of the surrounding cells can be highly responsive to glucocorticoid, at least in the immune-compromised zebrafish and mouse models.

Development of syngeneic models to examine the effects of AIB1 isoform subpopulations: Discovery and characterization of mouse AIB1Δ5, a homolog of human AIB1Δ4

Given the known pivotal role of glucocorticoids in immune cell biology,⁴² we needed to develop syngeneic mouse models where we could examine the invasive potential of cells expressing an N-terminal truncated isoform of AIB1 and the impact of GR signaling in an animal with an intact immune system. To avoid immune rejection, we needed to use cells that would express a mouse homolog of AIB1Δ4, although such a murine homolog had not been previously described. Through analyzing RNA sequencing data in a number of mouse cell lines, we discovered recurring reads skipping exon 5 in mouse Ncoa3 (AIB1) transcripts (Figure 2A, blue arrow). PCR with cDNA from the mammary gland carcinoma cell line E0771 using primers in exon 4 and exon 6 confirmed a major band of 350 bp corresponding to full-length AIB1 and a lower band that corresponds to the predicted 101 bp shorter fragment size resulting from skipping of exon 5 (Figure S2A). Sanger sequencing of the DNA fragment confirmed the presence of a continuous sequence of a fusion between the coding of exons 4 and 6 (Figure S2B), verifying exon 5 exclusion in the mRNA isoform, herein referred to as AIB1Δ5.

Figure 2.

Mouse AIB1Δ5 is a homolog of human AIB1Δ4

(A) Representative Sashimi plot of RNA sequencing reads skipping exon 5 (blue arrow) in mouse E0771 cells.

(B) Expression level of the GR-regulated gene Fkbp5 in E0771 empty vector (E0771-EV) and E0771-AIB1Δ5 overexpressing (E0771-Δ5) cells treated with either 10 nM dex or 1 μM rela for 48 h (n = 3 technical replicates per condition). Data are represented as mean ± SD. ∗p < 0.05. ∗∗∗p < 0.001. p value was calculated using two-tailed unpaired Student’s t-tests.

(C) Relative level of AIB1Δ5 mRNA in indicated E0771 samples, normalized to full-length AIB1. Normal lung samples are from control female C57BL/6 mice without E0771 tumor cell injection. Data represent n = 4 biological replicates for E0771 cells grown in culture, n = 3 biological replicates for tumor samples, n = 3 biological replicates for lung metastasis, and n = 3 biological replicates for tumor naive lung tissue. Data are represented as mean ± SD. ∗p < 0.05. p value was calculated using two-tailed unpaired Student’s t test.

(D) Expression level of AIB1 and AIB1Δ5 mRNA using RT-qPCR in the indicated samples. Tumor and immune cells were dissociated from primary tumor and spleen tissues and FACS sorted for CD45⁻, CD45⁺, or CD45+/CD3+ cells prior to RNA extraction and RT-qPCR. Median plotted as a solid line. Data represent n = 4 biological replicates for E0771 cells grown in culture, n = 8 tumor samples (n = 4 mice), and n = 3 mouse spleens. ∗∗p < 0.01. ∗∗∗p < 0.001. ∗∗∗∗p < 0.0001. p value was calculated using two-tailed unpaired Student’s t test.

Mouse AIB1Δ5 cDNA produces a protein product similar in size to human AIB1Δ4 (Figure S2C). The exclusion of exon 5 results in an in-frame stop codon in exon 6, requiring the usage of an alternative translation initiation site to generate the protein isoform. Mass spectrometry confirmed the presence of an N-terminal acetyl group on peptide 235-MLEEGEDLQCCMICVAR, indicated by a 43 Da mass shift in the predicted peptide mass (Figure S2D). N-terminal acetylation occurs on most eukaryotic proteins,⁴³ providing support that methionine 235 represents the translation start site of AIB1Δ5. Human AIB1Δ4 translation starts at M224, eleven amino acids upstream of mouse AIB1Δ5 (Figure S2E). Thus, both isoforms have a comparable deletion of the N-terminal bHLH-PAS domain.

One of the predominant features of human AIB1Δ4 is its increased potency of the transcriptional activation of nuclear receptor transcription compared to full-length AIB1¹⁴ due to the loss of N-terminal interactions with co-repressors such as ANCO1.^27,28 Mouse AIB1Δ5 behaved similarly as a more potent PR coactivator (Figure S2F). RT-qPCR of RNA from cells that were treated with dex or with rela confirmed the GR coactivator ability of AIB1Δ5 in overexpressing E0771 cells. Rela treatment induced a significant decrease in the mRNA expression of the GR-regulated gene Fkbp5 in both E0771 empty vector (E0771-EV) and E0771 AIB1Δ5 overexpression (E0771-Δ5) cells and a significant increase in the expression of Fkbp5 in the E0771-Δ5 compared to E0771-EV cells after dex treatment (Figure 2B). Thus, mouse AIB1Δ5 is structurally and functionally analogous to human AIB1Δ4.

Human AIB1Δ4 mRNA increases expression during malignant progression toward tumor invasion and metastasis.^14,29 Therefore, we evaluated AIB1Δ5 mRNA expression in E0771 cells during tumor progression. E0771 cells were injected subcutaneously into female C57BL/6 mice and, following tumor growth, primary tumors and lung metastases were removed for analysis. AIB1Δ5 mRNA levels in E0771 primary tumors and lung metastases were higher than in E0771 cells grown in culture (Figure 2C). In fact, the lung metastases samples showed a 5-fold increase compared to E0771 cells and a 2-fold increase compared to the primary tumor. One caveat in the interpretation of this data relates to the fact that immune cells are also known to express AIB1/SRC-3.⁴⁴ To evaluate the cell type-specific expression of AIB1Δ5, primary tumors were sorted for CD45, and AIB1 and AIB1Δ5 mRNA levels were assessed in CD45-positive and CD45-negative cell populations. AIB1Δ5 significantly increased in CD45-negative cells that comprise E0771 cells grown into a tumor compared to E0771 cells in culture (Figure 2D). Interestingly, full-length AIB1 expression decreased in tumor-infiltrating CD45-positive cells compared to CD45-positive cells isolated from the spleen, whereas AIB1Δ5 expression increased. AIB1 is known to control immune cell proliferation,⁴⁴ and the shift in the relative expression of AIB1/AIB1Δ5 in tumor-derived immune cells could be a reflection of the immune response in the microenvironment.

AIB1Δ5 promotes cancer progression in vivo and is sensitive to glucocorticoid receptor inhibition

To evaluate AIB1Δ5’s ability to enable invasion and metastasis, AIB1Δ5 was overexpressed in E0771 cells (Figures S3A and S3B). Empty vector alone (E0771-EV), AIB1Δ5 alone (E0771-Δ5), or a 4:1 mixture of E0771-EV to E0771-Δ5 (E0771-Mix) cells were injected bilaterally into the mammary fat pad of female C57BL/6 mice (Figure 3A). Mice underwent survival surgery to remove the primary tumors, and all mice were euthanized on Day 49 for the analysis of lung metastasis. Ex vivo tumor weight was measured after survival surgeries, and there was a significant increase in tumor weight in the E0771-Mix group compared to E0771-EV (Figure 3B). Analysis of the expression level of AIB1Δ5 in these tumors showed a significantly higher level of AIB1Δ5 mRNA in the E0771-Δ5 group, indicating that the expression of the isoform was maintained in vivo (Figure S3C). Mice in the E0771-Mix group died at a higher rate than the other groups, likely from large, locally invasive recurrent tumors (Figure 3C). Analysis of lung metastasis via hematoxylin and eosin (H&E) staining showed a significantly higher number of mice with lung metastasis in the E0771-Mix group compared to E0771-Δ5 or E0771-EV (Figures 3D and 3E). RNA-seq of primary tumors revealed glucocorticoid and dexamethasone as significant upstream regulators of pathways in E0771-Mix tumors compared to E0771-EV tumors (Figure 3F). Dex was also a significant upstream regulator of pathways for E0771-Mix tumors compared to E0771-Δ5 tumors (Figure S3D).

Figure 3.

A subpopulation of mouse AIB1Δ5 cells promotes overall invasion and is responsive to GR antagonism

(A) Schematic representation of mouse in vivo experimental design. E0771 cells overexpressing either empty vehicle (EV) control, AIB1Δ5 (E0771-Δ5), or a 4:1 mix of E0771-EV and E0771-Δ5 (E0771-Mix), respectively, were injected bilaterally into the mammary fat pads of 6-week-old female C57BL/6 mice. Primary tumors were removed on Days 28 and 31. Any surviving mice were euthanized on Day 49, and lungs were analyzed for metastasis via H&E staining.

(B) Tumor weight of primary tumors following removal on Days 28 and 31. Data represent n = 17 tumors (n = 9 mice) in the EV group, n = 10 tumors (n = 5 mice) in the Mix group, and n = 8 tumors (n = 4 mice) in the Δ5 group. Data are represented as mean ± SD. ∗p < 0.05. p value was calculated using two-tailed unpaired Student’s t test.

(C) Overall survival of mice in the three indicated groups. Death during primary tumor removal was due to blood loss from large, vascularized tumors. Two mice in the mixed group died during surgery, and one mouse in each of the other groups died during surgery. The remainder of the mice were either found dead or had to be euthanized. All surviving mice were euthanized on Day 49. ∗∗p < 0.01. p value calculated using the log-rank (Mantel-Cox) test. Data represent n = 9 mice in the EV group, n = 6 mice in the Mix group, and n = 5 mice in the Δ5 group.

(D) Representative H&E staining of lungs from E0771-EV and E0771-Mix injected female C57BL/6 mice. Scale bars, 500 μm.

(E) Number of mice in each group with lung metastasis as evident by H&E staining. ∗p < 0.05. p value calculated using the chi-square test.

(F) Ingenuity pathway analysis upstream regulators of primary tumor RNA-seq differentially expressed genes from E0771-Mix versus E0771-EV. Red lines are at -log₁₀(0.05) and Z score −1, 1. Data represent n = 4 tumors from each group.

(G) Representative images of indicated 67NR cells in a 3D sphere invasion embedded in 50% collagen I and 50% matrigel. 67NR-Mix cells are a 4:1 mix of 67NR parental and 67NRΔ5 cells, respectively. 0.1 μM rela treatment was used to inhibit GR. Scale bars, 500 μm.

(H) Quantification of the invasive area of spheres (n = 8) in each condition from G. Median plotted as solid line. ∗p < 0.05. ∗∗∗∗p < 0.0001. p value was calculated using two-tailed unpaired Student’s t-tests.

67NR cells were next used as an additional TNBC model to test GR inhibition against the background of an intact immune system. 67NR cells are a less aggressive TNBC cell line that is able to form primary tumors in vivo but does not metastasize.⁴⁵ A CRISPR strategy we have described previously²⁹ was used to make 67NR derivative cells that exclusively express AIB1Δ5. In the CRISPR screen, we identified clone 32 that only expressed AIB1Δ5 mRNA (Figure S3E). AIB1Δ5 protein expression was confirmed in a Western blot of extracts from clone 32 and HEK 293T cells overexpressing AIB1Δ5 (Figure S3F) and 67NR parental cells (Figure S3G). In a 3D invasion experiment, a 4:1 mixture of 67NR parental and clone 32 (67NRΔ5), respectively, was more invasive than the parental or 67NRΔ5 cells alone (Figures 3G and 3H). Importantly, treatment with rela significantly reduced 3D invasion of these cocultured cells.

Using a similar experimental design for 67NR cells to the DCIS in vivo experiment in Figure 1F, we observed that there was no significant difference in ex vivo tumor weight at the time of survival surgery between the control and rela treated groups (Figures S3H and S3I). However, gene expression analysis of primary tumors showed the downregulation of invasion pathways, such as FAK Signaling and Cell surface interactions at the vascular wall, in the rela treated tumors (Figure S3J).

Taken together, the 3D and in vivo data with E0771 and 67NR cells show that, similar to the human AIB1Δ4 phenotype, the expression of AIB1Δ5 in a subpopulation of cancer cells can enhance overall invasion that is dependent on glucocorticoid signaling.

Transcriptional response to glucocorticoid varies between AIB1 isoform-expressing cells

The above data suggested that the activation of GR specifically in the AIB1 N-terminally truncated isoform expressing cells (AIB1Δ4/AIB1Δ5) leads to the progression of TNBC through enhanced cell invasion. We therefore hypothesized that the response to glucocorticoid must be different in AIB1 isoform-expressing cells compared to parental cells.

AIB1Δ4, as well as full-length AIB1, are coactivators for the GR and enhance receptor-mediated transcription (Figure 4A). In fact, similar to AIB1Δ4’s more potent effect on PR- and ER-induced transcription,¹⁴ AIB1Δ4 elicited a stronger GR transcriptional response than full-length AIB1. DCIS and DCISΔ4 cells express similar endogenous levels of GR protein (Figure 4B) as well as GR (NR3C1) mRNA and lower levels of other nuclear receptors from RNA-seq analysis (Figure S4A). Transcriptome profiling of differentially expressed genes between dex versus dex plus RU486 treatment in each cell line revealed 493 differentially expressed genes in the DCIS line and 883 differentially expressed genes in the DCISΔ4 line, with an overlap of 330 genes (Figure 4C). The uniquely activated and uniquely repressed glucocorticoid-regulated genes in each cell line (Figure 4D) indicated that the AIB1 isoforms control markedly different gene sets in response to glucocorticoids. It is worth pointing out that the differences in response to glucocorticoids were not simply due to the differences in the baseline transcriptomes (Figures S4B and S4C) because the baseline expression of the uniquely glucocorticoid-regulated genes was similar (Figure S4C, diagonal line). For example, VIPR1 is one of the top uniquely upregulated genes in the DCISΔ4 cell line following glucocorticoid treatment (Figure S4D) but has similar expression levels at baseline. In fact, the majority of genes that are differentially expressed at baseline respond similarly to glucocorticoids (Figures S4E and S4F).

Figure 4.

Glucocorticoid transcriptional responses are altered by the switch to AIBΔ4 expression in TNBC cells

(A) Luciferase assay using MMTV promoter-driven firefly luciferase. HEK 293T cells were transfected with an MMTV promoter-driven firefly construct, a human GR expression vector, and either a pcDNA3 empty vector, pcDNA3-AIB1, or pcDNA3-AIB1Δ4 expression vector. Cells were treated with either vehicle control or 10 nM dex. Luciferase activity was measured 24 h after transfection. Dot blot shows AIB1 protein levels in the three samples treated with dex. Below the dot blot are the protein expression levels normalized to pcDNA3-empty vector. Data are represented as mean ± SD (n = 3). RLU = relative light unit. ∗∗∗∗p < 0.0001. p value was calculated using two-tailed unpaired Student’s t-tests.

(B) Western blot for GR in DCIS and DCISΔ4 cells after overnight treatment with vehicle, 5 nM dex, or 5 nM dex plus 500 nM RU486.

(C) Differentially expressed genes (RNA-seq) of dex (5 nM) treated versus dex (5 nM) plus RU486 (500 nM) treated DCIS or DCISΔ4 cells. All differentially expressed genes had an FDR ≤0.05. Data represent n = 3 biological replicates for each cell line and treatment condition.

(D) Bar graph depicting differentially expressed genes in B that are either commonly or uniquely activated or repressed in the DCIS and DCISΔ4 cells.

(E) Western blot for GR in DCIS and DCISΔ4 cells after 1 h treatment with vehicle, 5 nM dex or 5 nM dex plus 500 nM RU486. Cells were cultured in media without serum and hydrocortisone for 4 h prior to drug treatment in full serum.

(F) Peak overlap from a CUT&RUN experiment with a GR antibody in DCIS and DCISΔ4 cells. Cells were treated for 1 h with dex prior to conducting the CUT&RUN experiment. Data represent n = 3 biological replicates for each cell line. Peaks were called with SEACR in stringent mode and MACS3 with a q-value cutoff of ≤0.05. Consensus peaks were those called in at least 2 of the 3 replicates.

(G and H) Categorization of significantly enriched motifs from the GR CUT&RUN experiment in DCIS (E) and DCISΔ4 (F) cells. bZIP = basic leucine zipper; bHLH = basic-helix-loop-helix; HMG = high-mobility group; Zf = zinc finger; NR = nuclear receptor.

(I) Heatmap representation of ATAC-seq peaks found in DCIS and DCISΔ4 cells. Cells were treated for 1 h with dex prior to the ATAC-seq experiment. Overlap peaks were called in both cell lines (n = 87,356). DCIS enriched peaks were only called in the DCIS parental cell line (n = 5,459), and DCISΔ4 enriched peaks were only called in the DCISΔ4 cell line (n = 60,715). Data represent n = 3 biological replicates for each cell line. Peaks were called with MACS3 with a q-value cutoff of ≤0.01. Consensus peaks were those called in at least 2 of the 3 replicates.

(J) GR CUT&RUN reads (RPKM) in DCIS and DCISΔ4 cells plotted at ATAC-seq peak loci from G.

To evaluate GR chromatin engagement in the two cell lines, we first confirmed that the short dex treatment used for CUT&RUN did not alter GR protein expression or phosphorylation status. Western blot analysis after 1 h of dex treatment showed comparable GR protein levels and molecular weights in DCIS and DCISΔ4 cells (Figure 4E), indicating no differential protein stabilization or modification. We then performed CUT&RUN with a GR antibody to map GR genomic binding. The glucocorticoid response element (GRE) was the top-enriched motif in both cell lines (Figures S5A and S5B). However, over 2000 more loci were found in the DCISΔ4 cells (Figure 4F). There was also a modest increase in GR binding to homeobox motifs in DCISΔ4 cells when compared to DCIS cells (Figures 4G and 4H). Annotating GR CUT&RUN peaks in the DCISΔ4 cells to the nearest gene revealed 32 genes that are upregulated and 45 genes that are downregulated, respectively, after glucocorticoid treatment (Figure S5C). The top transcription factor target of the upregulated genes was NR3C1, and the top transcription factor target for the downregulated genes was NFKB1 (Figures S5D and S5E). NFKB1 is a known regulator of inflammation, and GR can bind to and inhibit NFKB activity.⁴⁶

Further genomic analysis with ATAC-seq (Assay for Transposase-Accessible Chromatin using sequencing) revealed more open chromatin regions in the DCISΔ4 cells (Figures 4I and S6A) with a slight decrease in open chromatin surrounding promoter regions and an increase surrounding introns and intergenic regions (Figures S6B and S6C). Interestingly, the open chromatin regions were overlapping with cell line-specific GR engagement sites (Figure 4J).

GR overlap with histone marks was next evaluated in both cell lines. In the DCIS parental cells, 19.9% of GR-annotated genes also had H3K4me3 and H3K27ac marks that delineate transcriptionally active chromatin region⁴⁷ (Figure S6D). Pathway analysis of these genes revealed several cancer-related pathways (Figure S6E). There was considerably more overlap with GR and the histone marks in the DCISΔ4 cells. 35.2% of GR annotated genes overlapped with the transcriptional activation marks H3K4me3 and H3K27ac (Figure S6F). Pathway analysis revealed multiple signaling pathways as significantly enriched from this gene set (Figure S6G). This initial genomic analysis indicates that the sole expression of the AIB1Δ4 isoform in the DCIS cells profoundly sensitizes the cells to GR signaling, causing increased open chromatin regions with GR engagement and significant and increased alterations in both induced and repressed gene expression.

Cellular crosstalk enhances the DCISΔ4 GR response and activates Myc pathways

We have shown previously that the ability of AIBΔ4-expressing cells to enhance invasion occurs only when the DCIS and DCISΔ4 cells are in direct contact.²⁹ Therefore, we wanted to understand how this cellular crosstalk affected the glucocorticoid transcriptional response. Fluorescently labeled DCIS and DCISΔ4 cells were cocultured for 48 h and then treated with either vehicle, dex, or dex plus RU486 for 24 h followed by flow cytometry to separate the two cell lines, now referred to as “unmixed” (Figure 5A). The total number of glucocorticoid-regulated genes was similar between DCIS cultured alone and DCIS unmixed cells (493 genes and 621 genes, respectively) (Figure 5B). However, the DCISΔ4 unmixed cells had a substantial increase in the total number of glucocorticoid-regulated genes (2700) compared to the DCISΔ4 cells cultured alone (883) (Figure 5B). Multidimensional scaling (MDS) analysis also revealed that the most distinct population was the DCISΔ4 unmixed cells (Figure S7A).

Comparison of glucocorticoid-regulated genes from cells cultured alone versus the unmixed populations demonstrated significant overlap but also new distinct genes now being regulated after cell-cell crosstalk, both for DCIS parental and DCISΔ4 cells (Figures S7B and 5C). The lower overlap in DCISΔ4 cells is due to the large number of new genes (2251) that are regulated in these cells by glucocorticoids after they have been exposed to the parental DCIS cells. The parental cells had a more modest increase, with only 416 new glucocorticoid-regulated genes after exposure to the subpopulation of DCISΔ4 cells. These data suggest that the DCISΔ4 cells become sensitized to glucocorticoid signaling after exposure to the parental cells, and this in turn allows them to enable the invasion of the parental population.

Ingenuity pathway analysis (IPA) of the upstream regulators in the DCISΔ4 unmixed cells of glucocorticoid-regulated genes showed significant involvement of Myc-regulated pathways (Figure 5D) that was not seen in the parental DCIS cells (Figure S7C). Myc pathways are highly involved in cellular metabolism and growth responses,⁴⁸ and this was reflected in the metabolic and proliferative pathways activated by glucocorticoids in the DCISΔ4 unmixed cells (Figure S7D). For example, the EIF2 signaling pathway was the most significantly upregulated pathway in these cells, which is involved in protein synthesis.⁴⁹ In contrast, the most activated pathways in the DCIS unmixed parental are directly related to known glucocorticoid effects, such as pulmonary fibrosis (Figure S7E). Supporting the pathway analysis findings, we saw an increase in DCISΔ4 cell proliferation in the presence of dex only when mixed with DCIS parental cells (Figures 5E and S7F).

The above data suggested that Myc is an important molecule regulated by the GR in the DCISΔ4 cells, specifically after cell-cell crosstalk with parental cells. Consistent with this, we found that Myc knockdown in DCISΔ4 cells (DCISΔ4-siMyc) (Figure S7G) caused a significant decrease in cell proliferation (Figure 5F). Furthermore, a 3D invasion assay with DCIS parental cells cocultured with DCISΔ4-siMyc cells in the presence of dex showed a significant decrease in the invasive area of the parental cells (Figures 5G and 5H), demonstrating Myc’s involvement downstream of GR in the enabling phenotype. Of note, the transcription factor Myc had a positive Z Score in the IPA upstream regulator analysis when comparing E0771-Mix tumors versus E0771-Δ5 tumors (Figure S3D). Although this change did not reach significance, taken together, these data indicate that Myc is likely playing a major role in the GR effects on AIB1 isoform-expressing cells in both human and murine TNBC cells.

The AIB1Δ4 cistrome is dramatically altered by cellular crosstalk

The distinct GR cistromes revealed above suggested that AIB1 and AIB1Δ4 would also have different engagement profiles induced by glucocorticoids. A CUT&RUN performed with an AIB1 antibody after dex treatment in the DCIS and DCISΔ4 lines revealed that there was only an overlap of 44 peaks, with the majority of AIB1 or AIB1Δ4 peaks being unique to each cell line (Figure 6A). The top significantly enriched motifs also differed between DCIS parental and DCISΔ4 cells (Figures S8A and 6B). Notably, the GRE was in the top ten sites for AIB1Δ4 but not full length AIB1. A comparison of overall motifs showed a small increase in homeobox motifs in the DCISΔ4 line and larger changes in the bHLH and ETS motifs, as well as an increase in nuclear receptor motif binding for AIB1Δ4 (Figures S8B and S8C). The AIB1 and AIB1Δ4 peaks were less enriched relative to the cell line-specific open chromatin regions revealed by ATAC-seq (Figure S8D). These observations showed that the AIB1Δ4 engagement pattern as a result of glucocorticoid stimulation is markedly distinct from that of full-length AIB1. Notably, GR and AIB1Δ4 have similarly enriched GRE engagement in DCISΔ4 cells.

Figure 6.

AIB1Δ4 chromatin engagement is affected by cell-cell crosstalk

(A) Heatmap representation of peaks called from a CUT&RUN experiment with an AIB1 antibody in DCIS and DCISΔ4 cells. Cells were treated for 1 h with dex prior to conducting the CUT&RUN experiment. Data represent n = 3 biological replicates for each cell line. Peaks were called with SEACR in relaxed mode and MACS3 with a p value cutoff of ≤0.005. Consensus peaks were those called in at least 2 of the 3 replicates.

(B) Top ten significantly enriched motifs found in the DCISΔ4 cell line from AIB1 CUT&RUN.

(C and D) Peak overlap from a CUT&RUN experiment with an AIB1 antibody in DCISΔ4 and DCISΔ4 unmixed cells (C) and DCIS and DCIS unmixed cells (D). Cells were treated for 1 h with dex prior to conducting the CUT&RUN experiment. Data represent n = 3 biological replicates for each cell line. Peaks were called with SEACR in relaxed mode and MACS3 with a p value cutoff of ≤0.005. Consensus peaks were those called in at least 2 of the 3 replicates.

(E) Top six significantly enriched motifs found in the DCISΔ4 unmixed cell line from AIB1 CUT&RUN.

(F) Top human Enrichr ChEA 2022 terms for the 467 DCISΔ4 unmixed upregulated genes with dex treatment that also have AIB1 binding in DCISΔ4 unmixed cells (from Figure S8H). The ChEA dataset is from target genes of transcription factors from published ChIP-chip, ChIP-seq, and other transcription factor binding site profiling studies.

A CUT&RUN experiment with “unmixed” cells was run to determine how AIB1 chromatin engagement is affected by cell-cell crosstalk. Similar to the large increase in DCISΔ4 glucocorticoid-regulated genes after crosstalk with DCIS parental cells (Figure 5C), there is a substantial increase in AIB1 chromatin engagement in DCISΔ4 unmixed cells (Figure 6C). AIB1 chromatin engagement is also increased in the DCIS unmixed cells, but not to the same extent (Figure 6D). The top significantly enriched AIB1 motif in the DCISΔ4 unmixed cells was GRE (Figure 6E), whereas TEAD3 was the top enriched AIB1 motif in DCIS unmixed cells (Figure S8E). Notably, c-Myc was the sixth most significantly enriched motif in DCISΔ4 unmixed cells. The majority of AIB1 was found to be located in introns or intergenic regions in both cell lines (Figures S8F and S8G).

Annotating AIB1 CUT&RUN peaks to the nearest gene in the DCISΔ4 unmixed cells reveled 467 genes with AIB1 engagement that are upregulated with glucocorticoid after cell-cell crosstalk (Figure S8H). The top transcription factor target of the upregulated genes was Myc (top 4), followed by NR3C1 (Figure 6F).

Together, these data demonstrate the substantial change in AIB1Δ4 chromatin engagement after cell-cell crosstalk with parental AIB1-expressing cells, with an enrichment at known GR and Myc target genes.

Discussion

Glucocorticoids have been implicated in breast cancer progression and metastasis,^13,50,51 yet the molecular context that defines GR dependence across tumor subtypes remains poorly understood. In this study, we identify a subpopulation of early-stage TNBC cells expressing the AIB1Δ4 isoform as uniquely sensitized to glucocorticoid signaling. These cells not only respond strongly to GR activation but also drive the glucocorticoid-driven invasion of surrounding, otherwise less aggressive, tumor cells. Glucocorticoid transcriptional programs are further amplified by cell-cell crosstalk, which dramatically reshapes glucocorticoid-responsive gene expression patterns in both DCISΔ4 and parental cell populations. Together, these findings suggest that the expression of a coactivator isoform can rewire the cellular response to systemic hormones, with profound implications for early TNBC progression. Our data also show that a specific splice variant is conserved across species and creates N-terminally truncated isoforms of AIB1 that are upregulated during the progression of TNBC. While splicing is known to be impacted during tumorigenesis,⁵² further investigation of the specific AIB1 splicing pathways involved will be relevant for future studies.

At the mRNA expression level, DCISΔ4 cells exhibit a significantly expanded set of glucocorticoid-regulated genes compared to parental DCIS cells, including both up- and down-regulated targets. These transcriptional differences correspond with altered GR genomic binding patterns, suggesting that AIB1Δ4 modifies GR cistrome engagement. AIB1Δ4-expressing cells also show distinct patterns of coactivator chromatin association, potentially due to the loss of the N-terminal bHLH-PAS domain, which mediates interactions with tumor suppressors such as ANCO1.^16,27,28 The inability of AIB1Δ4 to bind ANCO1 may enhance its coactivator promiscuity and facilitate preferential stabilization at distinct GR-bound loci, contributing to the divergent transcriptional programs observed. Although both AIB1 isoforms display diffuse chromatin engagement, likely reflecting their interaction with multiple transcription factor families (e.g., GR, AP-1, TEADs, E2Fs),^53,54,55 their differential binding patterns appear central to the isoform-specific transcriptional responses we observe. Future work using proteomics approaches such as RIME could further define the composition of GR-AIB1 isoform complexes and shed more light on their functional divergence.

The dramatic transcriptional shifts observed in DCISΔ4 cells after coculture with parental cells suggest that intercellular crosstalk plays a key role in enhancing glucocorticoid sensitivity. This may involve paracrine signaling or cell-cell adhesion-mediated pathways, particularly at the invasive front where DCISΔ4 and parental cells are in close proximity. Notably, this crosstalk results in the selective activation of Myc pathways in DCISΔ4 cells. Given that the Myc promoter harbors GREs,^56,57,58 and Myc is a well-established driver of proliferation and metabolism,⁵⁹ its induction may underpin the glucocorticoid-dependent invasive phenotype. The dependence on Myc could also explain why GR knockdown, but not GR antagonism, results in the loss of viability of DCISΔ4 cells. We speculate that GR antagonists, such as relacorilant or RU486, permit occupancy of the receptor on chromatin in a repressive conformation that is not functionally equivalent to GR depletion.

The mechanism by which crosstalk between AIB1Δ4-expressing and parental DCIS cells enhances GR-dependent transcription and Myc activation needs further study. While our findings demonstrate that intercellular communication is necessary for the enabling phenotype, the underlying signals could involve either direct cell-cell contact or soluble factors such as metabolites, cytokines, or extracellular vesicles. The “unmixed” experiments were performed after prior coculture, and therefore cannot distinguish juxtacrine from paracrine signaling. Prior work showed that physical proximity is required for enabling,²⁹ but additional layers of communication may further amplify GR-Myc pathway activation. Future studies, such as transwell separation, conditioned-media transfer, metabolomic profiling, or extracellular vesicle isolation, will help define the precise mechanisms by which AIB1Δ4-driven crosstalk enhances glucocorticoid signaling and invasion.

The original observation of the invasion/enabling of cancer cells expressing different AIB1 isoforms was made in human cancer cells in immune compromised models.²⁹ To evaluate whether this held up in an immunocompetent system, we identified a murine homolog of AIB1Δ4, which we named AIB1Δ5. This isoform results from alternative splicing that excludes exon 5 and produces a truncated protein analogous in size and function to human AIB1Δ4. AIB1Δ5 promotes invasion and metastasis in mouse TNBC models, and its expression increases during tumor progression. We posit that AIB1Δ5 enhances invasion and metastasis through cell-cell crosstalk similar to what we observe with human AIB1Δ4-expressing TNBC cells.²⁹ Importantly, GR inhibition with relacorilant significantly reduces invasion pathways in AIB1Δ5-expressing syngeneic models, mirroring findings with human cancer cells.

The identification of the murine homolog AIB1Δ5 allowed us to also examine the effects of isoform expression in immune cells from the murine model. The high expression levels of AIB1Δ5 in CD45-positive immune cells are of particular interest and suggest a possible role of this splice variant beyond its effects on tumor cells. Examining the role of AIB1Δ5/AIB1Δ4 in immune cell subtypes will be an interesting area of future study since AIB1/SRC-3 has a demonstrated role in regulatory T cells⁶⁰ and also in lymphoma.⁴⁴

The translational implications of our findings are significant. Using Basescope technology and isoform-specific primers, we can detect AIB1Δ4 expression in a subset of primary human TNBC tumors (Figure S9), raising the possibility that this isoform could serve as a predictive biomarker for response to GR antagonists. Relacorilant, which is currently in clinical trials for ovarian and other cancers,^61,62,63 may have particular efficacy in patients whose tumors harbor AIB1Δ4-expressing subpopulations. It will be important to determine whether AIB1Δ4 is similarly expressed and functional in other cancer types where GR signaling plays a role, and whether its presence predicts sensitivity to GR-targeted therapies.

In summary, our study identifies an AIB1 isoform-specific, GR-dependent mechanism of invasion in TNBC and reveals a layer of complexity in tumor cell crosstalk and steroid hormone response. These findings not only provide insight into early metastatic programming but also open the door for AIB1 isoform-guided precision therapies in GR-positive cancers, especially as a method of delaying early-stage TNBC progression.

Limitations of the study

Although AIB1Δ4-associated gene signatures have been reported,²⁹ clinical datasets capturing both AIB1 isoform expression and GR-targeted therapy outcomes are not currently available. Future studies leveraging such datasets will be required to determine whether AIB1Δ4 expression predicts therapeutic response.

The pathway analysis and Myc knockdown experiments support the conclusion that Myc functions as a key downstream effector of GR signaling in AIB1Δ4-expressing cells; however, these findings are largely correlative. The current study demonstrates an association between GR activation, increased Myc pathway activity, and the enabling phenotype, but we have not directly shown that GR-Myc interactions are causally required for invasion. How GR/Myc crosstalk mechanistically enhances enabling will be the subject of further studies.

Finally, although the enabling phenotype requires crosstalk between AIB1Δ4-expressing and parental cells, our experiments do not distinguish whether this interaction is mediated primarily by direct contact or by soluble factors such as cytokines, metabolites, or extracellular vesicles. Follow-up studies using compartmentalized or media-exchange systems could help to define the nature of this intercellular communication.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Anna Riegel, ariege01@georgetown.edu.

Materials availability

This study did not generate new unique reagents or materials.

Data and code availability

• •RNA-seq, ATAC-seq, and CUT&RUN data have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject PRJNA1327441 and were released to the public.
• •The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD034909 and https://doi.org/10.6019/PXD034909.
• •This article does not report original code.
• •Any additional information required to reanalyze the data reported in this article is available from the lead contact upon request.

Acknowledgments

We would like to thank Dr. J. Torchia for kindly providing the mouse pCMX-Ncoa3 expression vector. We would also like to thank Georgetown’s Animal Models Shared Resource – Zebrafish, Animal Models Rodent Shared Resource, Flow Cytometry & Cell Sorting Shared Resource, and Mass Spectrometry and Analytical Pharmacology Shared Resource for all of their help with experiments. This work was funded with support from the National Institutes of Health, USA: T32-CA009686 to RR, MEM and AJK, F30-CA250307 to MEM, R01-CA205632, R21-CA226542, R21-CA296504 to ATR, R01-CA231291 to AW. The project described used the Tissue Culture & Biobanking, Flow Cytometry & Cell Sorting, Animal Model, and Mass Spectrometry and Analytical Pharmacology, which are partially supported by award number P30-CA051008 (PI: L. Weiner) from the NCI. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or NCI.

Author contributions

A.J.K.: conceptualization, formal analysis, investigation, methodology, visualization, writing – original draft, and writing – review and editing. G.M.S.: conceptualization, formal analysis, investigation, methodology, supervision, and writing – review and editing. M.E.M: formal analysis, methodology, and writing – review and editing. R.R.: investigation and writing – review and editing. S.P.: investigation and writing – review and editing. M.O.S.: data curation and writing – review and editing. C.W.: investigation, methodology, and writing – review and editing. J.M.: investigation, methodology, and writing – review and editing. E.G.: investigation, methodology, and writing – review and editing. A.W.: conceptualization, funding acquisition, methodology, supervision, and writing – review and editing. A.T.R.: conceptualization, funding acquisition, methodology, supervision, writing – original draft, and writing – review and editing.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
NCOA3/SRC3 CUTANA CUT&RUN Antibody	EpiCypher	13-2013; RRID: NA
Glucocorticoid Receptor (D6H2L) Rabbit Monoclonal Antibody	Cell Signaling Technology	12041; RRID: AB_2631286
SRC-3 (5E11) Rabbit Monoclonal Antibody	Cell Signaling Technology	2126; RRID: AB_823642
GAPDH (14C10) Rabbit Monoclonal Antibody	Cell Signaling Technology	2118; RRID: AB_561053
Bacterial and virus strains
One Shot Stbl3 Chemically Competent E. coli	Thermo Fisher Scientific	C737303
Biological samples
Patient-derived xenograft (PDX) HCI-001	DeRose et al.64	https://cancertools.org/pdx-models/hci-001-pdx-162069/
Chemicals, peptides, and recombinant proteins
Eph inhibitor 1	MedChemExpress	HY-114199
RGDS peptide	Millipore Sigma	A9041
RGES peptide	AnaSpec	AS-62527
Relacorilant (in vivo)	Corcept Therapeutics	NA
Relacorilant (in vitro)	Targetmol Chemicals Inc.	NA
Mifepristone	Millipore Sigma	475838
Lipofectamine RNAiMAX Transfection Reagent	Thermo Fisher Scientific	13778100
Dexamethasone	Millipore Sigma	D4902
Puromycin	Thermo Fisher Scientific	A11138-03
FuGENE6	Promega	E2692
PEG-it Virus Precipitation	System Biosciences	LV825A-1
Rat Collagen Type I	Millipore Sigma	08-115
Matrigel	Fisher Scientific	CB-40230
Critical commercial assays
CUT&RUN Assay Kit	Cell Signaling Technology	86652
DNA Purification Buffers and Spin Columns	Cell Signaling Technology	14209
NEBNext Ultra II DNA Library Prep Kit for Illumina	New England Biolabs	E7645
ATAC-Seq Kit	Active Motif	53150
Deposited data
RNA-seq	NCBI Sequencing Read Archive (SRA)	PRJNA1327441
ATAC-seq	NCBI Sequencing Read Archive (SRA)	PRJNA1327441
CUT&RUN	NCBI Sequencing Read Archive (SRA)	PRJNA1327441
ChIP-seq	Gene Expression Omnibus	GSE139367
MS Proteomics	PoteomeXchange Consortium	PXD034909 and https://doi.org/10.6019/PXD034909
Experimental models: Cell lines
Human: MCF10DCIS.com	Dr. Susette Mueller (Georgetown University)	RRID:CVCL_5552
Human: HEK 293T	ATCC	RRID:CVCL_0063
Mouse: E0771	ATCC	RRID:CVCL_GR23
Mouse: 67NR	Karmanos Cancer Institute	NA
Experimental models: Organisms/strains
Mouse: NOD.Cg-Prkdcscid/J	The Jackson Laboratory	IMSR_JAX:001303
Mouse: C57BL/6J	The Jackson Laboratory	IMSR_JAX:000664
Mouse: BALB/cJ	The Jackson Laboratory	IMSR_JAX:000651
Zebrafish: Tg(kdrl:GRCFP)zn1	Georgetown’s Animal Models Shared Resource – Zebrafish core facility	NA
Oligonucleotides
Mission siRNA Universal Negative Control #1	Sigma-Aldrich	SIC001-1NMOL
Predesigned siRNA for MYC 1	Sigma-Aldrich	NM_002467; SASI_Hs01_00222676
Predesigned siRNA for MYC 2	Sigma-Aldrich	NM_002467, SASI_Hs01_00222680
Mouse Ncoa3 sgRNA: CACCGTTCTTCAGAACGCGGAATGC	Integrated DNA Technologies (IDT)	NA
Recombinant DNA
pLKO.1 Human NR3C1 shRNA	Millipore Sigma	TRCN0000245004
pLKO.1 Human NR3C1 shRNA	Millipore Sigma	TRCN0000245007
LentiCRISPR v2 puro	Addgene	98290
psPAX2 D64V	Addgene	63586
pCMX-Ncoa3 (mouse)	Dr. J. Torchia	NA
Software and algorithms
Qiagen Ingenuity Pathway Analysis (IPA)	QIAGEN	https://digitalinsights.qiagen.com/products-overview/discovery-insights-portfolio/analysis-and-visualization/qiagen-ipa/
GraphPad Prism (version 8.3.0)	Graphpad	https://www.graphpad.com/
R (version 4.3.1)	The R Foundation for Statistical Computing	https://www.r-project.org/
BioRender	BioRender.com	https://www.biorender.com/
TrimGalore (version 0.6.6)	Felix Krueger	https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/
Bowtie2 (version 2.3.5.1)	Langmead et al.65	https://github.com/BenLangmead/bowtie2
Picard (version 2.18.14-199 SNAPSHOT)	Broad Institute	https://github.com/broadinstitute/picard
Samtools (version 1.12)	Li et al.66	https://samtools.sourceforge.net/
SEACR (version X)	Meers et al.67	https://github.com/FredHutch/SEACR
MACS3 (version X)	Zhang et al.68	https://macs3-project.github.io/MACS/
DiffBind (version 3.18.0)	Bioconductor	https://bioconductor.org/packages/release/bioc/html/DiffBind.html
HOMER (version 3.4)	Heinz et al.69	http://homer.ucsd.edu/homer/
deepTools (version 3.5.1)	Ramírez et al.70	https://deeptools.readthedocs.io/en/develop/
HISAT2 (version 2.2.1)	Kim et al.71	https://github.com/DaehwanKimLab/hisat2
featureCounts (version 2.0.2)	Subread	https://subread.sourceforge.net/
edgeR (version 3.42.4)	Bioconductor	https://bioconductor.org/packages/release/bioc/html/edgeR.html

Experimental model and study participant details

Cell lines

Human female breast epithelial line MCF10DCIS.com (hereafter DCIS) cells and CRISPR-engineered DCISΔ4 cells were cultured in DMEM/F12 (1:1) medium (Gibco, 11039-021) supplemented with 5% horse serum, 20 μg/mL epidermal growth factor (EGF), 100 μg/mL hydrocortisone, 10 μg/mL insulin, and 100 ng/mL cholera toxin. HEK293T cells were used for lentiviral production. Murine female mammary cancer cell lines E0771 and 67NR were maintained in DMEM (Gibco, 11995-065) supplemented with 10% fetal bovine serum (FBS).

Cells were maintained at 37°C with 5% CO₂ and routinely tested for mycoplasma contamination.

67NR cells were obtained from Karmanos Cancer Institute. MCF10DCIS.com cells were originally described by Miller et al.³² and obtained from Dr. Susette Mueller. HEK293T and E0771 were obtained from ATCC.

Animal models

All mouse experiments were approved by the Georgetown University Animal Care and Use Committee (2016-1138).

Xenograft studies: Female NOD/SCID mice (Mus musculus) aged 6-8 weeks were injected orthotopically with DCIS/DCISΔ4 mixed populations.

Syngeneic models: Female C57BL/6 or female BALB/c mice (Mus musculus) aged 6-8 weeks were injected orthotopically with E0771, E0771Δ5, or E0771/E0771Δ5 mixed populations or 67NR/67NRΔ5 mixed populations.

Mice were randomized prior to treatment, and investigators were not blinded to treatment groups.

Mice were housed under standard conditions with a 12-hour light/dark cycle and ad libitum access to food and water.

All zebrafish experiments were conducted in accordance with institutional animal care guidelines.

Zebrafish (Tg(kdrl:GRCFP)zn1) embryos were maintained under standard conditions and injected at 48 hours postfertilization with labeled DCIS cells as described below.

Human samples

Human TNBC patient-derived xenograft (PDX) tissue (HCI-001) was processed for BaseScope analysis.

Method details

3D spheroid invasion assays

Cell aggregates were formed in 81-well agarose molds⁷² in full culture media for 24 hours and then embedded in a mixture of 80% collagen I (Millipore Sigma, 08-115) and 20% matrigel (Fisher Scientific, CB-40230) to form 3D spheres. Images were taken using the Olympus IX-71 Inverted Epifluorescence Microscope 24-72 hours after embedding. Cells protruding from sphere boundaries into the collagen and matrigel were defined as invading. Invasion area was quantified with ImageJ. Indicated drugs were added at the time of cell aggregation. The Eph inhibitor 1 was purchased from MedChemExpress (HY-114199), The RGDS (Arginine-Glycine-Aspartic Acid-Serine) peptide was purchased from Millipore Sigma (A9041) and the control RGES (Arginine-Glycine-Glutamic Acid-Serine) peptide was purchased from AnaSpec (AS-62527).

Zebrafish extravasation assay

The zebrafish experiment was done with the help of Georgetown’s Animal Models Shared Resource – Zebrafish core facility. Fifty mixed DCIS-RFP:DCISΔ4-GFP (4:1) cells were injected directly into the circulation via the duct of Cuvier of Tg(kdrl:GRCFP)zn1 zebrafish embryos 48 hours after fertilization. Vehicle control or 0.1 μM relacorilant (Corcept Therapeutics) was added to the zebrafish water for systemic administration. DCIS cell extravasation was then scored in the tail region in live, tricaine anesthetized embryos 72 hours after injection.

Mutagenesis

Mouse pCMX-Ncoa3 expression vector was kindly provided by Dr. J. Torchia. Mutagenesis to create the AIB1Δ5 expression vector was done using Agilent’s QuikChange II XL Site-Directed Mutagenesis Kit according to the manufacturer’s protocol. Mutagenic primers: Δ5_fwd: CGGCAAATAAAAGAACAAGGCACTGGATGGTTTCCTG; Δ5_rev: CAGGAAACCATCCAGTGCCTTGTTCTTTTATTTGCCG.

Luciferase reporter assay

Glucocorticoid receptor

HEK 293T cells were plated in a 6-well plate. The next day, cells were transfected with 1 μg MMTV promoter-driven firefly construct, along with 100 ng human GR expression vector and 3 μg of either pcDNA3 empty vector, pcDNA3-AIB1 or pcDNA3-AIB1Δ4 expression vector in the presence of FuGENE 6 Transfection Reagent (Promega Corporation, E2692). Cells were treated with either ethanol vehicle control or 10 nM of dexamethasone (Millipore Sigma, D4902). Luciferase activity was measured 24 hours after transfection with a Synergy H4 plate reader.

Progesterone receptor

HEK 293T cells were plated in a 6-well plate. The next day, cells were transfected with 1 μg PRE promoter-drive firefly contruct, along with 100 ng PR expression vector and 3 μg of either pCMX empty vector, pCMX-AIB1 (mouse) or pCMX-AIB1Δ5 expression vector in the presence of FuGENE 6 Transfection Reagent (Promega Corporation, E2692). Cells were treated with either ethanol vehicle control or 10 nM of R5020. Luciferase activity was measured 24 hours after transfection with a Synergy H4 plate reader.

Dot blot

Cells were lysed in 1X NP40 lysis buffer with protease inhibitor cocktail (Sigma-Aldrich) for 30 minutes on ice. Samples were centrifuged at max speed at 4°C for 5 minutes and supernatants were collected. Protein concentration was determined using the Bradford protein assay (Bio-Rad, 5000001). Protein was loaded onto a PVDF membrane and dried for 1 hour at room temperature. Membrane was then blocked with 5% milk in PBST for 1 hour at room temperature followed by incubation with AIB1/SRC-3 primary antibody (Cell Signaling Technology, 2126) in 2% milk PBST for 1 hour at room temperature. The membrane was then washed 3X with PBST and incubated with HRP-conjugated secondary antibody for 1 hour at room temperature. Following 3X washes with PBST, protein was detected with Immobilon Western Chemiluminescent HRP Substrate (Millipore Sigma).

Generating CRISPR constructs and AIB1Δ5 single cell clones

Single guide RNA (sgRNA) (5’-TTCTTCAGAACGCGGAATGC) was designed around mouse exon 5. Oligos were suspended to 100 μM in dH2O and then oligo pairs were annealed at 95°C for 5 minutes and then brought to room temperature slowly. Annealed oligos were diluted 1:200 with dH2O. 2 μg of LentiCRISPR v2 puro (Addgene, 98290) was digested overnight with BsmBI at 55°C. The digested plasmid was then gel purified. 50 ng of digested plasmid was ligated with 1 μL of diluted oligo duplex using T4 DNA ligase (NEB, M0202S). One Shot Stbl3 Chemically Competent E. coli (ThermoFisher, C737303) were transformed with 1 μL of ligation product. Bacteria colonies were then miniprepped and Sanger sequenced with U6-forward primer for evaluation of insertion.

To make lentivirus, HEK 293T cells were transfected using FuGENE 6 (Promega Corporation, E2692) following manufacturer’s instructions. The CRISPR construct was mixed with psPAX2-D64V (Addgene, 63586) to make integrase-deficient lentivirus and pVSV-G envelope plasmid prior to transfection. Viral containing media was collected 48 hours and 72 hours after transfection. PEG-it Virus Precipitation (SBI, LV825A-1) was used to concentrate lentiviral particles.

67NR cells (Karmanos Cancer Institute) were infected with the lentiviral CRISPR construct using 5 μg/mL polybrene (Sigma-Aldrich). Cells were then selected with 5 μg/mL puromycin (Fisher Scientific, A11138-03) two days after infection. Infected cells were FACS single cell sorted into 96-well plates. Once confluent, clones were transferred to a well of a 6-well culture plate. From there, one half of the cells from each well were froze and the remaining half were used for analysis of expression of AIB1Δ5 either via RT-qPCR or Western blot.

Real-time electric cell impedance sensing (ECIS)

Cell proliferation was monitored using E-plates from xCELLigence, Agilent according to the manufacturer’s protocol. Briefly, 5000 cells were plated in each well of an E-plate with indicated drugs, when applicable. N = 2-3 replicates per condition. Cell impedance was measured for indicated time. Mean and standard deviation were plotted using GraphPad Prism 8.

In vivo mouse experiments

DCIS cells with relacorilant treatment

A total of one million cells were injected bilaterally into the mammary fat pad of 6-8 week old female NOD/SCID mice, n=6 mice per treatment group. All mice were injected with 80% DCIS parental + 20% DCISΔ4 mixed cells. Relacorilant (Corcept Therapeutics), 30 mg/kg, or vehicle control treatment started one week after cells were injected. Drug was administered via oral gavage every Monday, Wednesday and Friday. Relacorilant was prepared in 10% DMSO, 0.1% Tween 80, 0.5% hydroxypropylmethylcellulose (HPMC) in 1X PBS. Mice underwent survival surgeries on day 34 to remove the primary tumor. Relacorilant or vehicle control treatment resumed one week after surgery. Mice were euthanized on day 62 and tissues were collected.

E0771 analysis of AIB1Δ5 expression

One million E0771 cells were injected subcutaneously in 6 week old female C57BL/6 mice, n=3. Four weeks later, primary tumors were collected via survival surgery for analysis. Mice were euthanized two weeks later and lungs and recurrent tumors were collected for analysis.

E0771 immune cell evaluation

One million E0771 cells were injected bilaterally into the mammary fat pad of 6-8 week old female C57BL/6 mice, n = 5 mice. Survival surgeries to remove the primary tumors were done on days 45 and 46 after cell injections. Primary tumors were FACS sorted for CD45-, CD45⁺ (Alexa Fluor 700 anti-mouse CD45, 103128) and CD3⁺ (Alexa Fluor 647 anti-mouse CD3, 100209) cells. All mice were euthanized on day 60 and spleen, lung, liver and any recurrent tumor were collected and FACS sorted for CD45-, CD45⁺, and CD3⁺ cells. RNA was extracted from all cell populations for evaluation of AIB1Δ5 mRNA levels via RT-qPCR.

E0771-Δ5 overexpression

A total of one million cells were injected bilaterally into the mammary fat pad of 6-8 week old female C57BL/6 mice. The three categories of mice were as follows: E0771 pLX304-empty vector overexpression (E0771-EV), E0771 pLX304-AIB1Δ5 overexpression (E0771-Δ5), and 80% E0771-EV + 20% E0771-Δ5 (E0771-Mix). N=10 mice per group. Tumor take-rate varied between the three groups. Tumors formed in 9/10 E0771-EV group, 5/10 E0771-Δ5 group, and 6/10 E0771-Mix group. Survival surgeries to remove primary tumors were done on days 28 and 31. Any remaining mice were euthanized on day 49 after cell injections. Tumors, lungs and liver were collected for analysis.

67NR cells with relacorilant treatment

A total of 200,000 cells were injected bilaterally into the mammary fat pad of 6-8 week old female BALB/c mice, n=6 mice per treatment group. All mice were injected with 80% 67NR parental + 20% 67NRΔ5 mixed cells. Relacorilant (Corcept Therapeutics), 30 mg/kg, or vehicle control treatment started one week after cells were injected. Drug was administered via oral gavage every Monday, Wednesday and Friday. Relacorilant was prepared in 10% DMSO, 0.1% Tween 80, 0.5% hydroxypropylmethylcellulose (HPMC) in 1X PBS. Mice underwent survival surgeries on day 21 to remove the primary tumor. Relacorilant or vehicle control treatment resumed one week after surgery. Mice were euthanized on day 39 and tissues were collected.

Lentiviral shRNA and siRNA knockdown

For GR knockdown studies, DCIS and DCISΔ4 cells plated in a 6-well plate were treated with pLKO.1 Human NR3C1 shRNA (Millipore Sigma) lentiviruses or non-targeting pLKO.1 shGFP control lentivirus. Cells were infected for 48 hours in the presence of 8 μg/mL polybrene. Selection with 5 μg/mL puromycin (Fisher Scientific, A11138-03) was applied 48 hours after infection.

pLKO.1 Human NR3C1 shRNA: GTGTCACTGTTGGAGGTTATT.

pLKO.1 Human NR3C1 shRNA: TTGGGTGGAGTTTCGTAATTT.

For siRNA knockdown studies, Sigma-Aldrich Validated Mission siRNAs were used to knockdown human MYC. Mission siRNA Universal Negative Control #1 was used as a negative control. Cells were seeded in regular culture media in 6-well dishes. When cells reached approximately 70% confluence, cells were transfected with indicated siRNAs with Lipofectamine RNAiMAX Transfection Reagent (ThermoFisher, 13778100) according to manufacturer’s instructions. siRNA complexes were added to cells to attain a final concentration of 20 nM. 48 hours later, cells were plated for either ECIS cell proliferation or a 3D sphere invasion assay in collagen and matrigel.

Predesigned siRNA for MYC 1: NM_002467; SASI_Hs01_00222676.

Predesigned siRNA for MYC 2: NM_002467, SASI_Hs01_00222680.

Quantitative RT-PCR

Total RNA was extracted using the RNeasy Mini Kit (Qiagen, 74106) according to the manufacturer’s instructions with DNase digestion. RNA sample concentration and quality were measured using a Nanodrop. cDNA was prepared with 500-1000 ng of total RNA using the iScript cDNA Synthesis Kit (Bio-Rad, 1708891) according to manufacturer’s instructions. qPCR was performed with iQ SYBR Green Supermix (Bio-Rad, 170-8882) using a realplex2 eppendorf PCR machine. Primers were obtained from Integrated DNA Technologies (IDT). GAPDH or ACTB were used as reference genes.

Primer sequences (5’ → 3’):

Mouse Gapdh:	TCAACAGCAACTCCCACTCTTCCA
	ACCCTGTTGCTGTAGCCGTATTCA
Mouse Actb:	GGCGCTTTTGACTCAGGATTTAA
	CCTCAGCCACATTTGTAGAACTTT
Mouse AIB1:	ACCGCTTTTACTACAGGCACT
	CCTTCCGGTCTTGCTCATGT
Mouse AIB1Δ5:	AGAGCTCATCTCTGCAAATCTC
	ACCATCCAGTGCCTTGTTCTT
Mouse Fkbp5:	GATGAGGGCACCAGTAACAATG
	CAACATCCCTTTGTAGTGGACAT
Human GAPDH:	GTCTCCTCTGACTTCAACAGCG
	ACCACCCTGTTGCTGTAGCCAA
Human ACTB:	CCTGGCACCCAGCACAAT
	GCCGATCCACACGGAGTACT
Human KLF9:	ATCTGGGTCGAGTCCTTCCC
	GCCGTTCACCTGTATGCACT
Human FN1:	AACACTTACCGAGTGGGTGAC
	TGTAGGACTGACCCCCTTCA
Human NR3C1:	ACCCTGCATGTACGACCAAT
	CTTGGCTCTTCAGACCGTCC
Human MYC:	CCTGGTGCTCCATGAGGAGAC
	CAGACTCTGACCTTTTGCCAGG

Cut&Run

Antibodies used for CUT&RUN were validated by the manufacturer for chromatin applications or previously published studies.

Monoculture

DCIS and DCISΔ4 cells were plated in 100 mm cell culture plates. The following day, cells were washed with 1X PBS and changed to serum- and hydrocortisone- free media. Four hours later, cells were switched to culture media (containing horse serum and hydrocortisone) with the addition of 100 nM dexamethasone (Millipore Sigma, D4902) for one hour. A total of 250,000 cells were collected for each reaction and the CUT&RUN Assay Kit (Cell Signaling Technology, 86652) was performed according to manufacturer’s protocol. In brief, live cells were harvested and washed twice in 1X Wash Buffer. Samples for AIB1 antibody were fixed for 2 minutes with 0.1% formaldehyde methanol-free (Thermo Scientific, 28908) prior to washes. Cells were immobilized on activated Concanavalin A Beads and cell membranes were permeabilized with digitonin. 0.2 - 0.5 μg primary antibody (NCOA3/SRC3: EpiCypher, 13-2013, GR: Cell Signaling Technology, 12041) was added to cell-bead mixture and incubated overnight at 4°C. The next day, samples were washed with Digitonin Buffer followed by the addition of pAG-MNase and incubated at 4°C for 1 hour. pAG-MNase was activated with calcium chloride and samples were incubated at 4°C for 30 minutes. The digestion was stopped with 1X Stop Buffer and samples were incubated at 37°C for 10 minutes to release DNA fragments into the solution. For fixed cells, crosslinking was reversed with 0.1% SDS Solution and proteinase K and incubated at 65°C for 4 hours. DNA was purified using DNA Purification Buffers and Spin Columns (Cell Signaling Technology, 14209) and libraries were prepared using the NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB, E7645). All conditions were done in triplicate. Libraries were analyzed for size distribution with a Bioanalyzer and sent to Novogene Corporation Inc. for paired-end 150 nucleotide sequencing on NovaSeq 6000 S4. The supplied IgG primary antibody was used as the negative control.

Coculture

DCIS-RFP and DCISΔ4-GFP cells were cocultured at a 4:1 ratio (DCIS-RFP:DCISΔ4-GFP) for 48 hours. Cocultures were washed with 1X PBS and then changed to serum- and hydrocortisone- free media. Four hours later, cells were switched to culture media (containing horse serum and hydrocortisone) with the addition of 100 nM dexamethasone (Millipore Sigma, D4902) for one hour. Cell populations were then separated via FACS and the CUT&RUN Assay Kit (Cell Signaling Technology, 86652P) was performed as above.

Data analysis

TrimGalore (v0.6.6) was used to trim paired-end reads to remove adapter sequence and low quality reads with parameters “--paired –q 20”. Reads were then aligned to the human genome (hg38) using Bowtie2 (v2.3.5.1).⁶⁵ MarkDuplicates (Picard, v2.18.14-199 SNAPSHOT) was used to flag duplicated reads which were then removed with Samtools⁶⁶ view (v1.12). Peak calling was done with both SEACR⁶⁷ and MACS3.⁶⁸ For GR peak calling, SEACR was run in stringent mode and MACS3 had a cutoff value of q ≤ 0.05. For AIB1 peak calling, SEACR was run in relaxed mode and MACS3 had a cutoff value of p ≤ 0.005 with the --nomodel option set. SEACR and MACS3 peaks were combined and peaks that were called in at least two biological replicates were considered for downstream analysis. Peaks were annotated and enriched motifs were identified with HOMER (v3.4).⁶⁹ Heatmaps were generated with deepTools (v3.5.1).⁷⁰

Chromatin immunoprecipitation sequencing (ChIP-seq) data analysis

Existing chromatin immunoprecipitation sequencing (ChIP-seq) datasets for H3K27ac and H3K4me3 from DCIS and DCISΔ4 cell lines²⁹ were downloaded from Gene Expression Omnibus under record GSE139367. Peaks were called using MACS3⁶⁸ with a cutoff value of q ≤ 0.05 and annotated with HOMER (v3.4).⁶⁹

ATAC-seq

Library preparation

DCIS and DCISΔ4 cells were plated in 6-well dishes in triplicate. The next day, cells were washed with 1X PBS and then changed to serum- and hydrocortisone- free media. Four hours later, cells were switched to culture media (containing horse serum and hydrocortisone) with the addition of 100 nM dexamethasone (Millipore Sigma, D4902) for one hour. A total of 75,000 cells were collected for each biological triplicate and the ATAC-Seq Kit (Active Motif, 53150) was used. Briefly, cells were washed with ice-cold PBS and then resuspended in 100 μL ATAC Lysis Buffer. Lysed cells were subjected to tagmentation for 30 minutes at 37°C. DNA was purified and PCR amplified for 10 cycles followed by bead clean-up. Libraries were analyzed for size distribution with a Bioanalyzer and sent to Novogene Corporation Inc. for paired-end 150 nucleotide sequencing on NovaSeq X Plus or a NovaSeq 6000 S4. Genomic DNA from each cell line was extracted and fragmented via sonification using a Covaris M220 to 150-200 bp for input control. Input libraries were prepared using the NEBNext Ultra II DNA Library Prep Kit for Illumina.

Data analysis

TrimGalore (v0.6.6) was used to trim paired-end reads to remove adapter sequence and low quality reads with parameters “--paired –q 20”. Reads were then aligned to the human genome (hg38) using Bowtie2 (v2.3.5.1).⁶⁵ MarkDuplicates (Picard, v2.18.14) was used to flag duplicated reads which were then removed with Samtools⁶⁶ view (v1.12). To call peaks, MACS3⁶⁸ with a q value cutoff of 0.01 was used. Consensus peaks were those that were called in at least two of the biological triplicates. Heatmaps were generated with deepTools (v3.5.1).⁷⁰ Differential peak calling was done with the R package DiffBind (v3.18.0). Peak positions were annotated using HOMER (v3.4).⁶⁹

RNA sequencing

Monoculture

DCIS and DCISΔ4 cells were plated in 6-well dishes. The next day, cells were treated in triplicate with either ethanol vehicle control, 5 nM dexamethasone (Millipore Sigma, D4902), or 5 nM dexamethasone and 500 nM Mifepristone/RU486 (Millipore Sigma, 475838) overnight. Total RNA was extracted using RNeasy Mini Kit (Qiagen, 74106) and sent to Novogene Corporation Inc. RNA-seq libraries were prepared using the TruSeq Total RNA library Prep Kit (Illumina) and 150 nucleotide paired-end sequencing was performed on either a NovaSeq X Plus or NovaSeq 6000 S4. Raw FASTQ files were aligned to the human genome (hg38) using HISAT2 (v2.2.1).⁷¹ Gene counts of aligned reads were performed using featureCounts (v2.0.2), followed by differential gene expression analysis using edgeR (v3.42.4).

Coculture

DCIS-RFP and DCISΔ4-GFP cells were cocultured at a 4:1 ratio (DCIS-RFP:DCISΔ4-GFP) for 48 hours. Cocultures were then treated in triplicate with either ethanol vehicle control, 5 nM dexamethasone (Millipore Sigma, D4902), or 5 nM dexamethasone and 500 nM RU486 (EMD Millipore). Following overnight treatment, cell populations were separated via FACS and total RNA was extracted using RNeasy Mini Kit (Qiagen, 74106). RNA was sent to Novogene Corporation Inc. and RNA-seq libraries were prepared using the TruSeq Total RNA library Prep Kit (Illumina). 150 nucleotide paired-end sequencing was performed on either a NovaSeq X Plus or NovaSeq 6000 S4. Raw FASTQ files were aligned to the human genome (hg38) using HISAT2 (v2.2.1).⁷¹ Gene counts of aligned reads were performed using featureCounts (v2.0.2), followed by differential gene expression analysis using edgeR (v3.42.4). Pathway analysis was performed using Ingenuity Pathway Analysis (IPA).

Mouse tumor tissue

Total RNA was extracted from mouse mammary gland tumors using RNeasy Mini Kit (Qiagen, 74106) and MagNA Lyser Green Beads with the MagNA Lyser (Roche). RNA was sent to Novogene Corporation Inc. and RNA-seq libraries were prepared using the TruSeq Total RNA library Prep Kit (Illumina). 150 nucleotide paired-end sequencing was performed on either a NovaSeq X Plus or NovaSeq 6000 S4. Raw FASTQ files were aligned to the mouse genome (mm10) using HISAT2 (v2.2.1).⁷¹ Gene counts of aligned reads were performed using featureCounts (v2.0.2), followed by differential gene expression analysis using edgeR (v3.42.4). Pathway analysis was performed using Ingenuity Pathway Analysis (IPA).

Western blotting

Cells were lysed in 1X NP40 lysis buffer with protease inhibitor cocktail (Sigma-Aldrich) for 30 minutes on ice. Samples were centrifuged at max speed at 4°C for 5 minutes and supernatants were collected. Protein concentration was determined using the Bradford protein assay (Bio-Rad, 5000001). Lysate was mixed with Sample Reducing Agent (Invitrogen, NP0009) and LDS Sample Buffer (Invitrogen, NP0007) and boiled for 10 minutes at 95°C. Whole cell extracts were loaded onto 4-12% Bis-Tris Gels (Invitorgen, NP0321BOX). Proteins were transferred to PVDF membranes and then membranes were blocked with 5% nonfat milk in PBST for 1 hour at room temperature. Membranes were then incubated with indicated primary antibody prepared in 2% milk overnight at 4°C and then washed 3X with PBST. They were then incubation with HRP-conjugated secondary antibody for 1 hour at room temperature followed by 3X washes with PBST. Protein was detected with Immobilon Western Chemiluminescent HRP Substrate (Millipore, WBKLS0500).

Primary antibodies used: SRC-3 (Cell Signaling Technology, 2126), Glucocorticoid Receptor (Cell Signaling Technology, 12041), GAPDH (Cell Signaling Technology, 2118).

Basescope in situ hybridization

Patient derived xenograft (HCI-001) from patients with metastatic disease established by Dr. Alana Welms’ lab were propagated in NOD/SCID mice. Tumors were harvested at week 8, fixed in 10% formalin and embedded in paraffin for sectioning. Unstained slides were processed according to basescope protocol. The AIB1 full length probe was designed to target 282-325 bp of NM_181659.3 aligning to exon 3 and exon 4 splice junction with 1ZZ probe. The AIB1-Δ4 probe was designed to target 281-498 of NM_181659.3 aligning to exon 3 and exon 5 splice junction with 1ZZ probe. All probes did not target mouse AIB1.

Proteomics and mass spectrometry

Immunoprecipitation (IP)

HEK 293T cells were transiently transfected with 3 μg of AIB1Δ5 cDNA. 24 hours later, whole cell lysate was prepared with NP40 lysis buffer and 1 mg of protein was used for immunoprecipitation with AIB1 antibody (Cell Signaling Technology, 2126). Pierce Protein A/G Plus Agarose Beads (Thermo Scientific, 20423) were washed 3X with lysis buffer. 50 uL of washed beads were then added to the lysate to pre-clear the sample. Beads were spun down and supernatant was moved to a new microcentrifuge tube. 35 uL of fresh, washed beads were blocked with 2% BSA for 45 minutes. The AIB1 antibody was then added to the lysate along with the blocked beads and rotated overnight at 4°C. The next day, after extensive washing with lysis buffer, proteins were eluted off beads with 5% SDS.

Identification of the N-terminal amino acid sequence of AIB1Δ5

Proteins were treated with DTT and iodoacetamide and then loaded onto an S-Trap column (ProtiFi, LLC) by following the procedure described previously.⁷³ Proteins were then digested with sequencing-grade Lys-C/trypsin (Promega Corporation) by incubation at 37°C overnight. The resulting peptides were eluted and dried down with a SpeedVac (Fisher Scientific).

Peptides were analyzed with a nanoAcquity UPLC system (Waters) coupled with Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher), with a similar setting as described previously.⁷³ In brief, samples in 0.1% formic acid were loaded onto a C18 Trap column (Waters Acquity UPLC M-Class Trap, Symmetry C18, 100 Å, 5 μm, 180 μm x 20 mm) at 10 μL/min for 4 min. Peptides were then separated with an analytical column (Waters Acquity UPLC M-Class, peptide BEH C18 column, 300 Å, 1.7 μm, 75 μm x 150 mm) with the temperature controlled at 40°C. The flow rate was set as 400 nL/min. A 150-min gradient of buffer A (2% ACN, 0.1% formic acid) and buffer B (0.1% formic acid in ACN) was used for separation: 1% buffer B at 0 min, 5% buffer B at 1 min, 22% buffer B at 90 min, 50% buffer B at 100min, 98% buffer B at 120 min, 98% buffer B at 130 min, 1% buffer B at 130.1 min, and 1% buffer B at 150 min. All data were acquired with an Orbitap Fusion Lumos mass spectrometer using an ion spray voltage of 2.4 kV and an ion transfer temperature of 275°C. Mass spectra were recorded with Xcalibur 4.0. MS parameters were set as below: Detector Type: Orbitrap; Mass range: 375-1500 m/z; Orbitrap Resolution: 120,000; Scan Range: 375-1500 m/z; RF Lens: 30%; AGC Target: Standard; Maximum Injection Time Mode: Auto; Microscans: 1. Charge state: 2-9 s; Exclusion duration: 40 s; Cycle Time: 3 s. MS/MS parameters were set as below: Isolation Mode: Quadrupole; Isolation Window: 1.6 m/z; HCD normalized collision energy: 35%; Detector Type: Orbitrap; Resolution: 30,000; Normalized AGC Target: 200%.Raw files acquired were searched in Proteome Discoverer (Thermo Fisher Scientific, v2.4) with Sequest HT database search engines, by searching against the customized database containing four potential isoforms of mouse protein AIB1. The database-searching parameters were set as below: full tryptic digestion and allowed up to two missed cleavages, the precursor mass tolerance was set at 10 ppm, whereas the fragment-mass tolerance was set at 0.02 Da. Carbamidomethylation of cysteines (+57.0215 Da) was set as a fixed modification, and variable modifications of methionine oxidation (+15.9949 Da), acetyl (Nterminus, +42.011 Da) were allowed. The false-discovery rate (FDR) was determined by using a target-decoy search strategy. The decoy-sequence database contains each sequence in reverse orientations, enabling FDR estimation. On the peptide level, the corresponding FDR was less than 1%. The mass spectra of peptides were then manually checked and annotated.

Bioinformatic and pathway analysis

Pathway analyses were done with Ingenuity Pathway Analysis (IPA) and Enrichr and motif analyses were done with HOMER (v.3.4).⁶⁹

Quantification and statistical analysis

Statistical analyses were performed using GraphPad Prism (v8.3.0) and R (v4.3.1). Data are presented as mean ± SEM unless otherwise stated. Statistical significance was determined using t-test and Fisher’s exact test, with p < 0.05 considered significant. Statistical tests and exact n values for each experiment are reported in the figure legends.

Published: January 22, 2026