Androgen receptor interacts with c-Myc to regulate macrophage-osteoclast axis and drive bone metastasis in triple negative breast cancer

Abstract

Background

Breast cancer distant metastasis is known to exhibit organotropism, with triple negative breast cancer (TNBC) subtypes also displaying organ-specific metastasis. The precise regulatory mechanisms governing this specificity remain unclear.

Methods

Retrospective analysis of metastatic data from public databases was utilized to explore the organotropism of TNBC subtypes. Mouse models combined with single-cell sequencing and immunoprecipitation (CoIP) experiments were utilized to investigate the role and mechanism of androgen receptor (AR) on TNBC bone metastasis. Further analysis of the bone microenvironment combined with CUT&TAG sequencing and osteoclast differentiation experiments was performed to validate the effect of AR and c-Myc interaction on macrophage-osteoclast axis differentiation.

Results

Analysis revealed the luminal androgen receptor (LAR) TNBC subtype had significant bone metastasis propensity. Mouse models showed AR activation promoted LAR TNBC bone metastasis. Using single-cell sequencing, we discovered that c-Myc played a critical role in AR-mediated bone metastasis. Further investigation of the bone microenvironment showed that AR-c-Myc interaction promoted macrophage-osteoclast axis differentiation and macrophage activation via MMP13, ultimately increasing bone resorption.

Conclusions

AR and c-Myc interaction induces macrophage differentiation into osteoclasts and promotes TNBC bone metastasis. These findings elucidate the mechanisms underlying bone metastasis in TNBC subtypes and inform potential interventions for TNBC bone metastasis.

Introduction

The heterogeneity of breast cancer metastasis, i.e., distinct patterns of metastasis to different organs, is associated with different treatment outcomes and prognoses in patients with breast cancer [1,2]. Breast cancer subtypes exhibit differential organotropism for metastasis, with luminal A and luminal B breast cancer patients primarily metastasizing to bone, while HER2-enriched patients preferring liver metastases [3,4]. Among these subtypes, triple-negative breast cancer (TNBC) stands out with its highest rate of distant metastasis [5,6]. TNBC is further classified into six subtypes: 2 basal-like (BL1 and BL2), an immunomodulatory (IM), a mesenchymal (M), a mesenchymal stem-like (MSL), and a luminal androgen receptor (LAR) subtype [7]. The different subtypes of TNBC also exhibit heterogeneity in metastatic patterns, but the precise underlying mechanisms remain elusive [8].

Androgen receptor (AR) is a member of the steroid hormone receptor family, which acts as a transcriptional activator and regulates downstream gene expression in the form of androgen-AR complexes [9,10]. AR is highly expressed in prostate cancer and drives prostate cancer cell proliferation and progression through genetic alterations, dysregulation of AR binding proteins, and androgen synthesis within the tumor [11,12]. With bone metastasis being one of the most important clinical features of prostate cancer, previous studies have identified AR as a key regulator of bone metabolism that interacts with the bone microenvironment [13,14], especially macrophages, to promote the development of prostate cancer bone metastases [15,16]. Estrogen receptor (ER) and AR both belong to the steroid hormone receptor family and have certain similarities in structure and mechanism of action [17,18]. In ER-negative breast cancer, such as in AR-positive and ER-negative TNBC, AR can replace ER and exert similar oncogenic effects as ER [19,20]. Clinical studies have shown a strong tendency for bone metastasis in both AR-positive prostate cancer and ER-positive breast cancer [21,22]. Considering the role and mechanism of AR in prostate cancer bone metastasis, as well as its structural similarity to ER, it is plausible that AR may exert a role similar to that of ER in promoting TNBC bone metastasis, but evidences are lacking.

Building upon these insights, we hypothesized that AR can mediate bone metastasis in TNBC. In this study, we first conducted a retrospective analysis of publicly available databases composed of patients with metastatic TNBC and analyzed the metastatic heterogeneity among different TNBC subtypes. A significant increase was observed in bone metastasis in patients with the LAR subtype, closely associated with AR expression. Using mouse models combined with single-cell sequencing and in vivo and vitro experiments, we found that AR interacted with c-Myc to promote bone metastasis in TNBC. Further analysis of the bone microenvironment revealed that interactions between AR and c-Myc promoted macrophage differentiation to osteoclasts and osteoclast activation through MMP13. In summary, our findings provide important insights into the metastatic heterogeneity of TNBC subtypes and the mechanisms of bone metastasis in LAR subtype, suggesting new avenues and potential therapeutic strategies for TNBC bone metastasis.

Materials and methods

Cell culture

MDA-MB-453, MDA-MB-231, BT549, Hs578t, Py8119, 4T1, RAW264.7 and HEK293T cells were purchased from ATCC. The MDA-MB-453, BT549, Hs578t, and Py8119 cells were cultured in RPMI 1640 (Gibco, California, USA) with 10% fetal bovine serum (Gibco) and 1% penicillin-streptomycin (Gibco) at 37 °C in a 5% CO₂ incubator. Other cell lines were maintained in high-glucose DMEM (Gibco) supplemented with 10% FBS (Gibco) and 1% penicillin-streptomycin (Gibco) at 37 °C in a 5% CO₂ incubator. Mycoplasma contamination was monitored using the Mycoplasma detection kit (Invivogen, California, USA), and short tandem repeat profiling was conducted to ensure the authenticity of all cell lines. All cell lines were used within ten passages.

Construction of bone metastasis model and drug treatment

All animal studies were conducted according to the NIH Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011) and were approved by the Institutional Animal Care and Use Committee of Nanjing Medical University (IACUC-2205019). Female BALB/c nude mice aged 5–8 weeks were purchased from the Animal Center of Nanjing Medical University for the bone metastasis experiment. Blinding was not implemented during surgical procedures or histological analysis due to the necessity of treatment-specific techniques. No samples or animals were excluded. For the construction of the bone metastasis model, the intra-iliac artery injection technique was performed as previously described [23]. In brief, the animals were anesthetized and placed on a warming pad. A 1.2 cm incision was made between the fourth and fifth mammary glands on the right lower abdomen. Approximately 3–4 ×10⁵ MDA-MB-453-LUC and MDA-MB-231-LUC cells were dissolved in 100 µL PBS and injected into the mouse iliac artery using a 31 G insulin syringe (Becton Dickinson, State of New Jersey, USA). Each group consisted of 6–8 mice, randomly assigned for subsequent experimental procedures. After the iliac-artery injection, the post-treatment involved intraperitoneal injections of DHT (10 mg/kg body weight; Selleck, Shanghai, China) every two days, and daily oral administration of the c-Myc inhibitor 10058-F4 (10 mg/kg body weight; Selleck) and the AR inhibitor bicalutamide (10 mg/kg body weight; Selleck).

Quantification of bone metastasis

Mice-bearing bone metastatic lesions were subjected to observation via in vivo bioluminescence imaging, micro-CT, and subsequent 3D reconstruction. Simultaneously, the femurs and tibias of the hind limbs were dissected from the mice for HE staining and TRAP staining. In vivo bioluminescence imaging was performed using the IVIS Spectrum system from PerkinElmer (Massachusetts, USA). In vivo Micro-CT imaging of small animals was performed using the SkyScan 1176 system from Bruker (Massachusetts, USA) with a resolution of 18 μm. The CT images were then subjected to 3D reconstruction and subsequent analysis using CTVox software. For bone tissues, decalcification was performed by incubating the samples in a 14% pH 7.4 EDTA solution for 14 days before embedding. ImageJ was employed for the quantification of the activation status of osteoclasts in the context of resorptive bone metastasis.

Immunofluorescence staining

Immunofluorescence staining of primary tumors was conducted by Servicebio Technology, (Wuhan, China). The following antibodies were incubated overnight: AR (catalog 06-680, Millipore, Massachusetts, USA, for Western blot), c-Myc (catalog 5605, CST, Shanghai, China, for Western blot), and Vimentin antibody (R&D Systems, Minnesota, USA, AF2105). The sections were then incubated with goat anti-mouse IgG antibody (Servicebio, 1:200) at room temperature for 50 min. Sections were washed with TBS in the dark and mounted with DAPI Fluoromount-G (SouthernBiotech, State of Alabama, USA, 0100-20). Images were captured using a Zeiss (Baden-Württemberg, Germany) Axiocam 506 mono microscope. Immunofluorescence images were first exported by ZEN (Zeiss) and subsequently analyzed and quantified using ImageJ.

In vitro osteoclast differentiation assay

Osteoclast precursor cells (OPs) from mice were extracted as described before [24]. In brief, bone marrow cells were obtained from the femur and tibia of 4-week-old BALB/c mice. Bone marrow-derived macrophages (BMMs) were generated by culturing cells in α-MEM medium supplemented with 20 ng/mL macrophage colony-stimulating factor (M-CSF). Osteoclasts were generated by seeding BMMs at a density of 25–150 ×10³ cells/cm² in culture medium with MCSF (50 ng/mL; R&D) and receptor activator of nuclear factor kappa-B ligand (RANKL) (50 ng/mL; R&D) at 37 °C and 5% CO₂. After 4–6 days of culture, cells were fixed and stained using the TRAP staining kit (Sigma-Aldrich, Shanghai, China) for visualization, with TRAP-positive cells containing multiple nuclei being quantified as osteoclasts. The culture medium was refreshed every two days.

Osteoclast precursor cells or RAW264.7 cells migration assay

For the AR-mediated OPs migration assay, MDA-MB-453 and MDA-MB-231 cells were seeded in a 24-well plate at a concentration of 6 ×10⁴ cells per well. For the AR/c-Myc interaction-mediated OPs migration assay, MDA-MB-453 and MDA-MB-231 cells were first transfected with shc-Myc lentiviral plasmids. After successful transfection, 6 ×10⁴ shc-Myc- MDA-MB-453 and MDA-MB-231 cells were seeded in a 24-well plate. After 48 h of stimulation with DHT (10 nM) and T-26c (10 nM), the cells were washed twice with PBS. Subsequently, Transwell inserts with 8 μm pore size (8.0 mm) were placed into the wells, and 1.5 ×10⁵ primary cultured OPs or RAW264.7 cells were added to the upper chamber. The cells were co-cultured for 24 h. At the end of the incubation period, the cells on the bottom side of the insert representing migrated cells were fixed and stained with 0.1% crystal violet. At least four fields of view were captured for each insert, and cell counting was performed using Image-Pro Plus 6.0 software (Media Cybernetics, California, USA).

Co-immunoprecipitation (Co-IP) experiment

For the Co-IP experiment, cells were incubated in the immunoprecipitation buffer (Beyotime, Shanghai, China) containing phenylmethanesulfonyl fluoride (Beyotime) for 30 min on ice. The whole-cell lysates were cleared by centrifugation at 12,000 rpm for 15 min, followed by overnight incubation with the designated primary antibodies in the refrigerator. Subsequently, the lysates were incubated with Protein A/G-PLUS agarose beads (Santa Cruz, Utah, USA) at 4 °C with gentle shaking for 2-4 h. The beads were then washed four times with a lysis buffer (50 mM HEPES [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid]), 125 mM NaCl, and 1% NP-40) at 4 °C using a rotating incubator for 1 h. After washing, the beads were boiled in 40 μL of sample loading buffer (Beyotime) and subjected to immunoblot analysis. The following antibodies were used: anti-HA (11867423001, Roche, Basel, Swiss Confederation), anti-Flag (F1804, Sigma-Aldrich), anti-AR (catalog 06-680, Millipore) and anti-c-Myc (catalog A19032, Abclonal, Wuhan, China).

Construction of AR and c-Myc knockdown cell lines

AR and c-Myc knockdown cell lines were generated by utilizing AR ad c-Myc shRNA lentiviral particles to reduce the expression of mRNA in MDA-MB-453 and MDA-MB-231 cells. The silencing of the gene was confirmed through RT-PCR. For AR knockdown, the annealed sense and antisense shRNA oligonucleotides were cloned into the pLKO.1-puro vector (GenePharma, Shanghai, China) using the following target sequences: CCTGCTAATCAAGTCACACAT. For c-Myc knockdown, the annealed sense and antisense shRNA oligonucleotides were cloned into the pLKO.1-puro vector (GenePharma) using the following target sequences: GGAAACGACGAGAACAGTTGA (shc-Myc#1), GCTCATTTCTGAAGAGGACTT (shc-Myc#2), GAGAATGTCAAGAGGCGAACA (shc-Myc#3), and CCCAGATTATCACCTTTGAAA (shc-Myc#4). With RT-PCR validation, we finally selected the shc-Myc#1 sequence for subsequent experiments.

Western blot analysis

Radioimmunoprecipitation assay (RIPA) buffer (Roche) was used to lyse cells The whole-cell or tissue lysates were separated on 10% SDS-PAGE gel and then transferred onto a polyvinylidene difluoride (PVDF) membrane (Millipore). The membrane was incubated overnight at 4 °C with antibodies against AR (catalog 06-680, Millipore) (1:1000), c-Myc (catalog 5605, CST) (1:1000), Smad2 + Smad3 (catalog ab207447, Abcam, Shanghai, China) (1:1000), Smad2 + Smad3 (phospho T8) (catalog ab272332, Abcam), beta Actin (catalog ab210083, Abcam) and GAPDH (catalog AF0006, SunShine Bio, Hangzhou, China) (1:3000). Subsequently, the membrane was incubated for 1 h at 25 °C with the following secondary antibodies: HRP Goat anti-Mouse IgG (catalog SA00001-1, Proteintech, Wuhan, China) (1:3000) and HRP Goat anti-Rabbit IgG (catalog SA00001-2, Proteintech) (1:3000). Protein bands were visualized using a chemiluminescence kit (Immobilon; Merck Millipore).

RT-PCR

The total RNA from cells was extracted and isolated using TRIzol reagent (Invitrogen). Subsequently, it was reverse transcribed into cDNA using the PrimeScript RT Master Mix (Vazyme, Nanjing, China). RT-PCR was performed on a Thermo ABI QuantStudio 7 Flex Real-Time PCR instrument using AceQ qPCR SYBR Green Master Mix (Low ROX Premixed) (Vazyme). All data were normalized to the reference genes GAPDH or 18S, and quantitative measures were obtained using the comparative CT method. The primer sequences used for RT-PCR experiments are listed in the supplementary table (Supplementary Table 1).

CUT&TAG sequencing and Chip-qPCR

The CUT&Tag assay was performed on MDA-MB-453 cell line either treated or untreated with DHT, using the Hyperactive Universal CUT&Tag Assay Kit for Illumina Pro (Vazyme) following the manufacturer’s protocol. Briefly, cells were harvested and gently resuspended in Wash Buffer (provided in the kit) to maintain >90% viability. Cell suspensions were incubated with ConA.

Beads Pro to immobilize cell membranes and minimize sample loss. c-Myc antibody (Abclonal) incubation was conducted overnight at 4 °C. Secondary antibody (1:100 dilution in Dig-wash Buffer) and hyperactive pA/G-Tn5 transposase (0.04 μM final concentration) were sequentially applied to tag chromatin regions bound to the target protein. DNA libraries were constructed using the TruePrep Index Kit V2 for Illumina (Vazyme) with 12–14 PCR amplification cycles, optimized for 10,000–50,000 input cells. Libraries were purified with VAHTS DNA Clean Beads (Vazyme) and quantified using a Qubit dsDNA HS Assay Kit. Size distribution was validated via Agilent Bioanalyzer. The purified DNA was also analyzed with quantitative RT-PCR. Accordingly, the primer sequences were designed and listed in the Supplementary Table 1.

scRNA-seq

Sample preparation

The mice were injected with tumor cells via the iliac artery and divided into Ctrl/DHT groups, with 6–8 mice per group. After three weeks, the mice were sacrificed, and bone metastatic tumor lesions were isolated and stored in sCelLiveTM Tissue Preservation Solution (Singleron, Nanjing, China). The tumor tissue was cut into small pieces and digested into single-cell suspension diluted in PBS to a concentration of 2 ×10⁵ cells/mL, and was loaded onto a microfluidic chip using the Singleron Matrix® Single-Cell Handling System. Single-cell RNA sequencing libraries were constructed according to the protocol of the GEXSCOPE® Single-Cell RNA Library Kit (Singleron).

Quality control, dimension-reduction, and clustering

Scanpy v1.8.2 was used for quality control, dimensionality reduction, and clustering under Python 3.7 [25]. After filtering, 11,454 (Ctrl)/13,926 (DHT) cells were retained for the downstream analyses, with on average 25,636/25,517 genes and 6762.5/5884 UMIs per cell. Principle Component Analysis (PCA) was performed on the scaled variable gene matrix, and the top 20 principal components were used for clustering and dimensional reduction. Cells were separated into 6 clusters by using Louvain algorithm and setting resolution parameters. Cell clusters were visualized by using Uniform Manifold Approximation and Projection (UMAP) {t-Distributed Stochastic Neighbor Embedding (t-SNE)}.

Differentially expressed genes analysis and pathway enrichment analysis

To identify differentially expressed genes (DEGs), we used the scanpy. tl. rank_genes_groups() function based on Wilcoxon rank sum test with default parameters, and selected the genes expressed in more than 10% of the cells in either of the compared groups of cells and with an average log(Fold Change) value greater than 1 as DEGs. Adjusted p value was calculated by Benjamini–Hochberg correction and the value 0.05 was used as the criterion to evaluate the statistical significance. To investigate the potential functions of tumor cells and macrophages, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis were used with the “clusterProfiler” R package v3.16.1. Pathways with p_adj value less than 0.05 were considered as significantly enriched.

Cell differentiation potential evaluation and pseudotime trajectory analysis

CytoTRACE v0.3.3 (a computational method that predicts the differentiation state of cells from single-cell RNA-sequencing data using gene counts and expression) was used to predict the differentiation potential of cell subpopulations [26]. Cell differentiation trajectory of monocyte subtypes was reconstructed with the Monocle2 v2.10.0 [27]. For constructing the trajectory, top 2000 highly variable genes were selected by Seurat (v3.1.2) FindVairableFeatures, and dimension-reduction were performed by DDRTree. The trajectory was visualized by plot_cell_trajectory function in Monocle2.

Transcription factor regulatory network analysis

Transcription factor network was constructed by pyscenic v0.11.0 using scRNA expression matrix and transcription factors in AnimalTFDB [28]. First, GRNBoost2 predicted a regulatory network based on the co-expression of regulators and targets. CisTarget was then applied to exclude indirect targets and to search transcription factor binding motifs. After that, AUCell was used for regulon activity quantification for every cell. Cluster-specific TF regulons were identified according to Regulon Specificity Scores (RSS) and the activity of these TF regulons were visualized in heatmaps.

Intra-tumoral heterogeneity (ITH) score calculation

The ITH score was calculated by the algorithm described in published work [29]. The ITH score was defined as the average Euclidean distance between the individual cells and all other cells, in terms of the first 20 principal components derived from the normalized expression levels of highly variable genes. The highly variable gene was identified using the FindVariableFeatures function in the Seurat package, with default parameters.

UCell gene set scoring

Gene set scoring was performed using the R package UCell v 1.1.0 [30]. UCell scores are based on the Mann–Whitney U statistic by ranking query genes in order of their expression levels in individual cells.

TNBC subtype metastatic heterogeneity analysis

Breast cancer patients with metastatic information were collected from GSE2603, GSE2034, GSE12276, GSE125989, and GSE110590 datasets. Batch effects were removed using the sva package in R, and TNBC patients were selected based on their expression of ER, PR, and HER2. Subsequently, the TNBCtype tool (http://cbc.mc.vanderbilt.edu/tnbc) was utilized for subtype classification. Six distinct subtypes were identified: BL1, BL2, IM, MSL, M and LAR. For patients with single-site metastasis, the specific site was recorded as the unique metastatic location. In cases where multiple metastatic sites were present (e.g., lung and bone), each site was recorded individually (e.g., lung metastasis and brain metastasis). Differential analyses were performed between LAR and non-LAR patients, as well as between patients with and without different organ metastasis.

Statistics

All measurements were presented as mean ± SEM. For two-group comparisons, unpaired two-tailed Student’s t test was used. For multi-group comparisons, one-way ANOVA and Tukey’s multiple comparisons test was performed. Data analysis was conducted using GraphPad Prism 8.0 software.

Results

AR activation is positively correlated with TNBC osteolytic bone metastasis

To investigate the metastatic heterogeneity of different subtypes of TNBC, we selected breast cancer patients with metastatic information from GEO databases. Finally, five datasets with metastatic breast cancer information were chosen: GSE2603, GSE2034, GSE12276, GSE125989, and GSE110590, with a total of 726 metastatic breast cancer patients. Firstly, based on the expression levels of ER, PR, and HER2, we identified 206 TNBC patients (Fig. 1a). Next, by analyzing gene expression profiles with the TNBCtype tool [31], we classified TNBC into six subtypes: BL1, BL2, M, IM, LAR, and MSL (Supplementary Fig. 1a). We then analyzed the metastatic data of different subtypes, including bone metastasis, lung metastasis, and other sites (Fig. 1b; Supplementary Fig. 1b and Supplementary Table 2). Of note, we observed a significant propensity for bone metastasis in patients with the LAR-subtype of TNBC (Fig. 1b, c). Previous studies have suggested that LAR TNBC is AR-positive and more sensitive to androgen stimulation [8]. To confirm the relationship between AR and TNBC metastasis, we investigated AR expression and pathway activation in metastatic lesions from different organs of TNBC patients. The results revealed that patients with bone metastases exhibited increased AR expression and pathway activation, while there was no significant difference in metastatic lesions from other organs (Fig. 1d and Supplementary Fig. 1c–f). This suggests that AR expression and activation are positively correlated with bone metastasis in the LAR subtype of TNBC.

Fig. 1. AR expression is positively correlated with TNBC osteolytic bone metastasis.

a Flowchart of statistical analysis for organ-specific metastasis in different subtypes of TNBC. Pie charts of the bone metastasis probabilities of different TNBC subtypes (b) and different organ heterogeneity of LAR TNBC metastasis (c). d Androgen pathway activation score in metastatic lesions in patients with or without bone metastasis. e Experimental design of the in vivo bone metastasis mouse model. MDA-MB-453-LUC and MDA-MB-231-LUC cells were injected into the iliac artery of BALB/c nude mice (day 0); then PBS/DHT was injected i.p. into mice every two days for 21 days, and 21 days later, the mice were sacrificed and bone analysis was performed. Representative images of bioluminescent, radiographic, H&E, and TRAP staining in the bones of mice (f) and quantitative analysis of these techniques as well as Ki-67(g–m). Yellow arrows indicate osteolytic lesions; Data indicate the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, as analyzed by one-way ANOVA and Tukey’s multiple comparisons test. Normal Bone, n = 5; Ctrl, n = 6; DHT, n = 6 (g–m); Scale bars: 1 mm (micro-CT), 50 μm (H&E, TRAP). T tumor, B bone. BV/TV Bone volume/Tissue volume, Oc.S/BS osteoclast surface/bone surface. See also Supplementary Figs. S1 and S2.

To further validate the impact of AR on TNBC bone metastasis, we established a TNBC bone metastasis model through intra-iliac artery injection [32]. First, we screened AR expression across TNBC cell lines and selected the representative AR-positive MDA-MB-453 and MDA-MB-231 luciferase-expressing cells for injection into mouse iliac arteries (Supplementary Fig. 1h) [33]. Mice were then divided into three groups: a tumor-free control group, an AR activation group (DHT-treated), and a Ctrl group (PBS-treated), receiving injections every two days (Fig. 1e). In vivo imaging and histological analysis revealed that AR activation elevated AR expression and significantly promoted TNBC bone metastasis (Fig. 1f–i; Supplementary Fig. 1g–l). Given that bone metastasis of breast cancer is characterized by osteolysis [34], we further validated the extent of bone destruction in the three groups. Tartrate-resistant acid phosphatase (TRAP) staining combined with micro-CT 3D reconstruction demonstrated that AR activation significantly reduced bone volume/tissue volume (BV/TV), while increasing osteoclast surface/bone surface (Oc.S/BS) and TRAP positive cells (Fig. 1f, j–l; Supplementary Fig. 1m–o). Beyond promoting bone metastasis, Ki-67 staining showed that AR activation increased tumor proliferation in the bone microenvironment (Fig. 1m; Supplementary Fig. 1p). These data indicate that AR activation exacerbates osteolytic bone metastasis in TNBC.

We next investigated the impact of AR suppression on TNBC bone metastasis by establishing stable AR-knockdown TNBC cell lines and administering the AR pathway inhibitor bicalutamide (Supplementary Fig. 2a, b). In vivo imaging and histological analysis demonstrated that AR knockdown or inhibition significantly suppressed TNBC bone metastatic lesion formation (Supplementary Fig. 2c–g). TRAP staining coupled with micro-CT 3D reconstruction revealed that both genetic AR knockdown and pharmacological AR inhibition attenuated the osteolytic effects of TNBC bone metastasis, manifested by increased bone volume/tissue volume (BV/TV), alongside decreased osteoclast surface/bone surface and TRAP positive cells (Supplementary Fig. 2h–j). Ki-67 staining showed that AR knockdown and inhibition reduced tumor proliferation capacity (Supplementary Fig. 2k).

c-Myc acts as a key factor in promoting AR-mediated bone metastasis

Our previous experiments demonstrated that androgen receptor activation promotes bone metastasis in TNBC. To investigate the underlying molecular mechanisms, we performed 10X single-cell RNA sequencing on bone metastatic tissues from control (Ctrl, PBS-treated) and AR-activated (DHT-treated) mice (Supplementary Fig. 3a). We identified 23,034 cells, predominantly epithelial cells (18,357, 80% of lesions), along with mononuclear phagocytes (3211), fibroblasts (817), neutrophils (452), endothelial cells (140), and adipocytes (15) (Fig. 2a; Supplementary Fig. 3b–d and Fig. 4). Tumor heterogeneity (ITH) scoring and copy number variation (CNV) analysis [29] confirmed all epithelial cells were malignant (Supplementary Fig. 5a–c). DHT treatment increased epithelial cell proportion (82.68% vs 76.09%) and altered microenvironment composition: higher proportion of fibroblasts (20.82% vs 14.54%), neutrophils (11.05% vs 8.45%), and endothelial cells (3.21% vs 2.80%), but fewer mononuclear phagocytes (63.50% vs 73.16%) (Supplementary Fig. 3b–d).

Fig. 2. c-Myc is closely associated with AR-mediated bone metastasis.

UMAP plots of 23,034 all single cells (a) and 18,357 epithelial cells (c) in Ctrl and DHT bone metastatic tumors, colored by cell type (left) and tissue type (top right). The proportions of cell types in Ctrl and DHT tumor tissues are shown on the bottom right (colors as in the left panel). b Volcano map of the DEGs between epithelial cells (tumor cells) of Ctrl and DHT groups. Each point represents a gene; genes marked in blue and red are highly expressed genes in tumor cells of Ctrl and DHT groups, respectively. d Bar graphs of the proportions of major tumor cell subpopulations. e Trajectory of the evolution of eight tumor cell subgroups predicted by monocle. The dashed arrows indicate the direction of evolution. SC1-8 represents epithelial cell1-8, respectively. f GSEA enrichment analysis of the eight tumor cell subgroups, indicating active signature pathways in eight groups. The MYC_TARGETS_V1 and MYC_TARGETS_V2 pathway was highlighted with a red box. The scores were estimated using SCENIC analysis with a Wilcoxon rank-sum test. g GSEA analysis of the enrichment of the MYC_TARGETS_V1 pathway in DHT groups in comparison with Ctrl groups. h Scatter plot showing the specificity of regulons in the DHT group by pySCENIC, with the top5 regulons highlighted. MPs mononuclear phagocytes, ECs endothelial cells, NES normalized enrichment score. See also Supplementary Figs. S3–S7.

Fig. 4. Activation of AR promotes macrophage M2 polarization and exhibits osteoclast differentiation characteristics.

UMAP plots of bone microenvironmental cell populations (a), mononuclear phagocyte populations (b), and macrophage subpopulations (c) in all samples. All cells are colored by cell type. UMAP plots of the M1 gene signature (d) as well as the M2 gene signature (e) in all samples. Expression levels of M1 gene signature (f) and M2 gene signature (g) in each macrophage subpopulation. Volcano map (h) and GSEA enrichment analysis (i) of the DEGs between macrophage Fcna⁺ subpopulations and II1b⁺ subpopulations. RT-PCR analysis of M1 markers (j) and M2 markers (k) in macrophages co-cultured with (i) TNBC cells, (ii) DHT-treated TNBC cells, (iii) c-Myc knockdown (shc-Myc) TNBC cells, and (iv) combination-treated (DHT + shc-Myc) TNBC cells. The results are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, as analyzed by one-way ANOVA and Tukey’s multiple comparisons test. See also Supplementary Figs. S8, S9.

Differential gene expression analysis revealed upregulation of Cp, Cirbp, and c-Myc, and downregulation of Dynll1 and Serpin1 in DHT group (Fig. 2b and Supplementary Table 3). To delineate tumor cell heterogeneity, we classified all tumor cells into eight distinct subclusters. Comparative analysis revealed significantly higher proportions of clusters 1/2/3/5/8 in AR-activated groups, with reduced representation of clusters 4/6/7 (Fig. 2c, d; Supplementary Fig. 3c and Supplementary Table 4). CytoTRACE evaluation [26] identified subclusters 1/3/8 as exhibiting the highest stemness scores, indicative of enhanced self-renewal capacity (Supplementary Fig. 5d). Monocle2 trajectory analysis [27] further established these high-stemness subclusters as the predominant origins of tumor cell differentiation (Fig. 2e; Supplementary Fig. 5e–h). Further pathway enrichment analysis of different subpopulations revealed activation of MYC target-related pathways in subpopulations 1, 3, and 8 (Fig. 2f). By Gene Set Enrichment Analysis (GSEA), we validated the aberrant activation of MYC target-related pathways in the DHT group (Fig. 2g). Building on c-Myc’s established role as a nuclear transcription factor [35], PySCENIC analysis [28] confirmed significant enrichment of c-Myc transcriptional activity in DHT-treated tumor cells (Fig. 2h; Supplementary Fig. 6).

Given that c-Myc is a key downstream target of the TGF-β signaling pathway, we sought to determine whether c-Myc promotes TNBC bone metastasis in a TGF-β–dependent manner. However, immunoblot analysis of bone metastatic lesions from both the Ctrl and DHT groups revealed no significant activation of TGF-β pathway kinases (Supplementary Fig. 3e, f). These results suggest that c-Myc may function independently of TGF-β signaling in this context. To further investigate the specific role of c-Myc in TNBC bone metastasis, we established stable c-Myc knockdown TNBC cell lines and injected them into mice via the iliac arteries (Supplementary Fig. 7a). Mice were then divided into three groups: a tumor-free control group, a c-Myc knockdown group(shc-Myc), and a control (Ctrl) group. In vivo imaging and histological analysis revealed that c-Myc knockdown significantly suppressed TNBC bone metastasis (Supplementary Fig. 7b–d). TRAP staining combined with micro-CT 3D reconstruction demonstrated that c-Myc knockdown significantly increased bone volume/tissue volume, while reducing osteoclast surface/bone surface and TRAP positive cells (Supplementary Fig. 7e–g). Ki-67 staining indicated that c-Myc knockdown inhibited tumor proliferation within the bone microenvironment (Supplementary Fig. 7h). These findings collectively establish c-Myc as the pivotal effector mediating AR-driven TNBC bone metastasis, with c-Myc knockdown suppressing osteolytic bone metastasis.

AR interacts with c-Myc to facilitate TNBC bone metastasis

Through single-cell RNA sequencing, we found that c-Myc plays a critical role in AR-mediated TNBC bone metastasis. To further investigate whether AR interacts with c-Myc to regulate TNBC bone metastasis, we first examined the interaction between AR and c-Myc at both the cellular and tissue levels. Western blot and RT-PCR analysis revealed that c-Myc expression increased in parallel with AR expression, indicating a positive correlation between the two (Fig. 3a, b; Supplementary Fig. 7i, j). Moreover, immunofluorescence staining demonstrated the nuclear co-localization of AR and c-Myc (Fig. 3c; Supplementary Fig. 7n). Given that both AR and c-Myc are nuclear transcription factors [35], we further investigated their molecular interaction. Both endogenous and exogenous co-immunoprecipitation (Co-IP) assays confirmed a physical interaction between AR and c-Myc (Fig. 3d, e; supplementary Fig 7k–m).

Fig. 3. AR interacts with c-Myc to facilitate TNBC bone metastasis.

Western blot analysis (a) and RT-PCR analysis (b) of AR and c-Myc expression in MDA-MB-453 cells treated with DHT (0, 10, 25 nM, 48 h) (n = 3). c Representative images of immunofluorescence staining for AR, c-Myc, and DAPI (4’,6-diamidino-2-phenylindole) in MDA-MB-453 and MDA-MB-231 cells. Scale bars: 10 μm. Representative co-immunoprecipitation (Co-IP) and Western blot assays showing the interaction between AR and c-Myc in HEK293t cells (d) and MDA-MB-453 cells (e). f Experimental design for investigating the impact of AR on bone metastasis after inhibiting c-Myc expression. g–k Representative images of bioluminescent, radiographic, H&E, and TRAP staining in the bones of mice (g) and quantitative analysis of these techniques as well as Ki-67 (h–m). Yellow arrows indicate osteolytic lesions; Data indicate the mean ± SEM. ****P < 0.0001, as analyzed by one-way ANOVA and Tukey’s multiple comparisons test. Ctrl, n = 6; DHT, n = 6; DHT+shc-Myc, n = 6 (h, m); Ctrl, n = 5; DHT, n = 5; DHT+shc-Myc, n = 5 (i–l); Scale bars: 1 mm (micro-CT), 50 μm (H&E, TRAP). T tumor, B bone, BV/TV Bone volume/Tissue volume, Oc.S/BS osteoclast surface/bone surface. See also Supplementary Figs. S7.

We next sought to validate the functional role of AR–c-Myc interaction in TNBC bone metastasis in vivo. Using the same model as described above, mice were injected intra-arterially with control or c-Myc knockdown TNBC cells, followed by intraperitoneal injections of PBS or DHT every two days. This established the control, AR activation, and AR activation + c-Myc knockdown groups, respectively (Fig. 3f). In vivo imaging and HE staining of bone tissues revealed that AR activation promoted the formation of TNBC bone metastases, whereas c-Myc knockdown reversed this pro-metastatic effect (Fig. 3g–i). Furthermore, TRAP staining and micro-CT 3D reconstruction demonstrated that inhibition of c-Myc abrogated the osteolytic bone destruction induced by AR activation (Fig. 3j–l). Ki-67 staining of metastatic lesions showed that AR activation enhanced tumor cell proliferation, while c-Myc knockdown significantly impaired this pro-proliferative effect (Fig. 3m).

Taken together, AR interacted with and depended on c-Myc to promote osteolytic bone metastasis in TNBC.

Activation of AR allows macrophages to exhibit M2 polarization and osteoclast differentiation features

Increasing evidence suggests that bone provides a favorable microenvironment for the homing, proliferation, and colonization of circulating cancer cells [36]. Analysis of bone microenvironmental cells within the bone metastatic lesions of the Ctrl and AR-activated mice revealed macrophages as the predominant cellular component (56.4%) among 4,677 bone microenvironmental cells. Other constituents included mononuclear phagocytes (monocytes, dendritic cells), fibroblasts, neutrophils, vascular endothelial cells, and minor populations of osteoclasts and adipocytes (Fig. 4a, b; Supplementary Fig. 8a, b). Given the documented role of macrophages in promoting tumor progression[37], we investigated their function in TNBC bone metastasis. Macrophages from the DHT group exhibited upregulated M2-polarization genes (Arg1, Chil3) with concomitant downregulation of M1-associated genes and antigen presentation genes (IL1b, H2-Aa, H2-Eb1) (Supplementary Fig. 9a). Single-cell signature scoring confirmed enhanced M2 features and attenuated M1 characteristics (Supplementary Fig. 9b–d).

Differential gene expression classified macrophages into three subsets: Il1b⁺ (Il1b/H2-Eb1/H2-Ab1), Fcna⁺ (Fcna/Ccl8/Saa3), and Top2a⁺ (Top2a/Ccnb2/Pclaf) (Fig. 4c; Supplementary Fig. 8c). UMAP clustering integrated with signature scoring demonstrated M2-feature enrichment in Fcna⁺ subsets and M1-feature enrichment in Il1b⁺ subsets (Fig. 4d–g). M2-type macrophages have been found to play an important pro-carcinogenic role in bone metastasis [38]. To further elucidate the impact of AR on macrophage polarization, and the specific role of M2-type macrophages in TNBC bone metastasis, we performed differential gene and functional enrichment analyses of the Fcna⁺ subpopulation and the II1b⁺ subpopulation, and found that functions related to osteoclast activation, bone resorption and bone remodeling were significantly enriched in the M2-type Fcna⁺ macrophages (Fig. 4h, i).

To validate the AR-mediated regulation of macrophage polarization and osteoclast activation by TNBC cells, co-culture experiments were used and showed that increasing AR activation in the tumor cells progressively attenuated M1 macrophage polarization and enhanced M2 polarization (Fig. 4j, k). Notably, with the increase in M2 polarization, osteoclast differentiation markers also increased (Fig. 4k; Supplementary Fig. 9e). Given that AR interacts with c-Myc to promote metastasis, we further confirmed that c-Myc knockdown in tumor cells reverses AR-driven M2 macrophage polarization and osteoclast differentiation (Fig. 4j, k; Supplementary Fig. 9e). Collectively, AR remodels the TNBC bone metastatic microenvironment by c-Myc-dependent induction of macrophage M2 polarization and osteoclast-activating features.

AR/c-Myc interaction promotes macrophage-to-osteoclast transformation and activation via MMP13

Numerous studies have demonstrated that macrophages serve as a major source of osteoclasts and that alterations in macrophage function significantly influence osteoclast differentiation [39]. Previous studies have shown that AR depended on c-Myc to promote bone resorption during bone metastasis (Fig. 3), accompanied by macrophage M2 polarization and concomitant osteoclast differentiation features (Fig. 4). To investigate the effects of AR and c-Myc interactions on the macrophage-osteoclast axis, we performed osteoclast formation experiments on RAW264.7 cells and mouse OPs (Fig. 5a) [40]. Compared to macrophages cultured alone, macrophages cultured with TNBC cells exhibited enhanced osteoclast differentiation. Furthermore, AR activation in tumor cells significantly increased macrophage-to-osteoclast conversion (Fig. 5b, c; Supplementary Fig. 10a). Osteoclast marker validation confirmed AR-mediated promotion of this transformation (Fig. 5d; Supplementary Fig. 10b). However, c-Myc knockdown in tumor cells reduced osteoclast formation and reversed AR-driven macrophage differentiation into osteoclasts (Fig. 5b–d; Supplementary Fig. 10a, b). Collectively, AR interacts with c-Myc to facilitate macrophage-osteoclast transformation. To assess AR-c-Myc effects on osteoclast recruitment, migration assays were performed after TNBC co-culture (Fig. 5a). Osteoclast precursor recruitment increased proportionally with TNBC cell AR activation. Conversely, c-Myc knockdown abolished AR-enhanced recruitment (Fig. 5e–g; Supplementary Fig. 10c).

Fig. 5. AR interacts with c-Myc to promote macrophage to osteoclast differentiation as well as osteoclast activation.

a Experimental design illustrating the impact of AR/c-Myc interaction on macrophage differentiation and activation. Representative images (b) and quantifications (c) of cell numbers of multinucleated TRAP-positive osteoclasts per square millimeter and percent of total surface area occupied by multinucleated TRAP-positive osteoclasts. d RT-PCR analysis of MMP9, CTSK, and TRAP expression in mouse osteoclast precursors (mOPs) co-cultured with MDA-MB-453 cells or shc-Myc-transfected MDA-MB-453 cells, with or without DHT treatment. Representative images of mOPs subjected to the transwell migration assay with MDA-MB-453 (MDA-MB-231) or c-Myc knockdown MDA-MB-453 (MDA-MB-231) cells treated with DHT (10 nM) (e) and quantification of the number of migrated cells (f, g). These results are presented as mean ± SEM. *P < 0.05, ***P < 0.001, ****P < 0.0001, as analyzed by one-way ANOVA and Tukey’s multiple comparisons test. Scale bars: 50 μm (TRAP, Transwell). n = 3 biological replicates. See also Supplementary Fig. S10.

To further investigate the mechanism by which AR/c-Myc interactions affect macrophage-osteoclast axis differentiation, we performed c-Myc-targeted CUT&Tag sequencing on TNBC cells co-cultured with osteoclasts (Fig. 6a). Integrated Genome Viewer (IGV) analysis revealed that DHT stimulation increased c-Myc binding peaks at promoter and gene body regions (within the first two introns and/or exons) of osteoclast differentiation markers (TRAP, RANKL, CTSK), while decreasing binding at the osteoclastgenesis inhibitor OPG promoter. Additionally, enhanced binding was observed for osteoclast-inducing factors MMP9 and MMP13 (Fig. 6b; Supplementary Fig. 11a, b). Subsequent ChIP-qPCR validation confirmed elevated levels of osteoclast differentiation markers and inducing factors, among which MMP13 showed the most prominent increase, while the osteoclastogenesis inhibitor OPG did not exhibit a significant decrease (Fig. 6c).

Fig. 6. AR/c-Myc interaction promotes osteoclast differentiation and activation through MMP13.

a Experimental design exploring the specific mechanisms of AR/c-Myc interaction in regulating macrophage function. b IGV snapshots showing c-Myc CUT&Tag signals of osteoclast differentiation markers (RANKL, CTSK, MMP13) in control and DHT treated MDA-MB-453 cells. Blue denotes promoter region signals. c ChIP–qPCR analysis of genes related to osteoclast differentiation. Representative images (d) and quantification (e, f) of cell numbers of multinucleated TRAP-positive osteoclasts per square millimeter and percent of total surface area occupied by multinucleated TRAP-positive osteoclasts. Data represented as mean ± SEM (c, e, f). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns not significant, as analyzed by one-way ANOVA and Tukey’s multiple comparisons test. Scale bars: 50 μm (TRAP). n = 3 biological replicates. See also Supplementary Fig. S11.

To determine whether AR/c-Myc interactions function through MMP13, tumor cells were treated with the MMP13-specific inhibitor T-26c. Osteoclast differentiation and migration assays demonstrated that inhibiting MMP13 secretion reversed AR-mediated promotion of both osteoclast differentiation and recruitment (Fig. 6d–f; Supplementary Fig. 11c–e). Collectively, AR/c-Myc interactions promote macrophage-to-osteoclast transformation and enhance osteoclast recruitment via MMP13 regulation, thereby driving osteolytic bone metastasis in TNBC.

Dual-target inhibition of AR and c-Myc is more effective than targeting AR and c-Myc alone in inhibiting TNBC osteolytic bone metastasis

Bone represents one of the most critical target organs for breast cancer metastasis, and currently effective targeted therapies for TNBC bone metastasis patients are lacking [41]. Through intra-iliac artery injection, we found that the inhibition of AR and c-Myc significantly reduced TNBC osteolytic bone metastasis (Supplementary Fig. 2 and Supplementary Fig. 7). Further experiments validated that the interaction between AR and c-Myc promoted TNBC osteolytic bone metastasis (Fig. 3). Therefore, we aimed to investigate whether dual-target inhibition of AR and c-Myc would be more effective than inhibiting AR and c-Myc alone in suppressing TNBC osteolytic bone metastasis. To this end, we used intra-iliac artery injection to inject LUC-expressing TNBC cells into mice. Mice were orally administered with PBS, AR inhibitor bicalutamide, c-Myc inhibitor 10058-F4 and a combination of bicalutamide and 10058-F4, forming the Ctrl group, AR-targeted group, c-Myc-targeted group and AR and c-Myc dual-targeted group, respectively (Fig. 7a). Three to four weeks later, by in vivo imaging and HE staining analysis, consistent with the previous findings, we found that bicalutamide and 10058-F4 reduced osteolytic bone metastases in TNBC. Importantly, the combination of AR and c-Myc inhibitors could more significantly reduce the occurrence of bone metastases compared with AR and c-Myc inhibitors alone (Fig. 7b–d). Combined with TRAP staining and micro-CT 3D reconstruction, we also observed that combined treatment with AR and c-Myc inhibitors could more significantly increase the bone volume/tissue volume, and reduce the osteoclast surface/bone surface and TRAP positive cells, resulting in a better suppression of the extent of TNBC bone resorption (Fig. 7e–g). Furthermore, Ki-67 staining of tumor metastatic lesions showed that the combination treatment with AR and c-Myc inhibitors could more significantly inhibit tumor cell proliferation (Fig. 7h). In summary, dual-target inhibition of AR and c-Myc demonstrates a more significant suppression of both TNBC cell proliferation and the occurrence of osteolytic bone metastasis compared to targeting AR and c-Myc alone.

Fig. 7. In vivo validation of dual-target inhibition of AR and c-Myc versus single-agent inhibition in suppressing TNBC bone metastasis.

a Experimental design of mouse bone metastasis assay. MDA-MB-453-LUC and MDA-MB-231-LUC cells were injected into the iliac artery of mice on day 0, followed by daily oral gavage of PBS, bicalutamide, 10058-F4 or bicalutamide and 10058-F4 from day 1 for 21–28 days. Mice were sacrificed and analyzed for bone metastasis after 28 days. Representative images of bioluminescent, radiographic, H&E, and TRAP staining in the bones of mice (b) and quantitative analysis of these techniques as well as Ki-67 (c–h). Yellow arrows indicate osteolytic lesions; Data indicate the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, as analyzed by one-way ANOVA and Tukey’s multiple comparisons test. Ctrl, n = 5; Bica, n = 5; 10058-F4, n = 5; Bica+10058-F4, n = 5 (c–h); Scale bars: 1 mm (micro-CT), 50 μm (H&E, TRAP). Bica Bicalutamide, T tumor, B bone, BV/TV Bone volume/Tissue volume, Oc.S/BS osteoclast surface/bone surface.

Discussion

Clinically, targeted therapies for TNBC bone metastasis have been lacking, primarily due to a limited understanding of the underlying mechanisms of bone metastasis. Here, we reported the organ-specific role of AR in TNBC bone metastasis. Different subtypes of TNBC exhibited different metastatic patterns, with LAR TNBC being more prone to bone metastasis, which was closely related to the expression and activation of AR. Activation of AR promoted osteolytic bone metastasis in TNBC, whereas inhibition of AR reduced the occurrence of osteolytic bone metastasis. Through single-cell sequencing and in vitro and in vivo experiments, we found that AR interacted with c-Myc to promote TNBC osteolytic bone metastasis. Further analysis of the bone microenvironment revealed that the interaction between AR and c-Myc regulated the macrophage–osteoclast axis. Specifically, AR–c-Myc interaction acted on MMP13 to promote the differentiation of macrophages into osteoclasts and enhanced osteoclast activation, ultimately contributing to bone resorption. Moreover, dual-targeted inhibition of AR and c-Myc inhibited TNBC osteolytic bone metastasis more effectively than single-targeted inhibition of AR and c-Myc. In conclusion, our findings provide new perspectives to understand the intrinsic mechanism of heterogeneity of TNBC metastasis in different subtypes and potential avenues for the treatment of TNBC bone metastasis.

Heterogeneous metastasis of TNBC exists between different subtypes, but the mechanisms underlying this heterogeneity are not yet well understood. Prior studies have demonstrated that LAR TNBC is positively correlated with genes associated with bone recurrence, showing a significant trend towards bone metastasis and lymph node metastasis [42]. Our retrospective analysis of patients with different subtypes of TNBC revealed that the LAR subtype of TNBC had the highest propensity for bone metastasis and was strongly associated with the expression and activation of AR. Activation of AR plays a key role in the development and progression of prostate cancer bone metastases [13,43], making androgen deprivation therapy (ADT) the preferred treatment for patients with recurrent disease [14]. Some studies have also pointed out that AR is closely related to breast cancer bone metastasis [19]. Therefore, to investigate the role of AR in TNBC bone metastasis, we used a bone metastasis model based on iliac artery injection. Our results showed that activation of AR promoted osteolytic bone metastasis in TNBC, whereas knockdown or inhibition of AR attenuated osteolytic bone metastasis. Considering the effectiveness of the AR inhibitor bicalutamide in the treatment of prostate cancer bone metastases, our results further suggest that bicalutamide may be potentially beneficial for the treatment of patients with LAR TNBC bone metastases.

Biologically, the transcription factors AR and c-Myc have been of great interest for researchers, and they are thought to act as key players in a variety of cancer types, but their interactions have been controversial. In prostate cancer, by analyzing gene expression in patients with primary and metastatic CRPC, researchers observed a positive correlation between AR and c-Myc expression [44]. In addition, studies by the Qiu group and other groups have shown that overexpression of MYC antagonized AR target genes without altering AR expression [45]. In breast cancer, especially in TNBC, the interrelationship between AR and c-Myc is also of great interest. Through immunohistochemical staining of tissues from 45 cases of TNBC, researchers found that AR was positively correlated with the expression of c-Myc [46]. Further experiments conducted by Ni et al. confirmed that AR not only directly activated the transcription of c-Myc, but also increased c-Myc activity by destabilizing the c-Myc competitor MAD1 [47]. In addition, the study of Cao et al. went further to reveal that AR promoted c-Myc expression by forming an AR-PDEF-c-Myc axis [35]. To explore the relationship between AR and c-Myc in TNBC bone metastasis, we verified their expression correlation and protein interactions. AR was found to be positively correlated with c-Myc in TNBC bone metastasis, with co-localization in the nucleus and protein interactions, consistent with previous studies. With further validation in a mouse bone metastasis model, we confirmed that c-Myc played a critical role in AR-mediated TNBC bone metastasis, and that the interaction between AR and c-Myc promoted osteolytic bone metastasis in TNBC. Taken together, our study provides new insights on the interaction between AR and c-Myc in TNBC bone metastasis.

Osteoclasts are multinucleated cells responsible for osteolytic bone erosion and are derived from monocyte/macrophage precursors [48]. The macrophage-osteoclast axis in the bone microenvironment is critical for regulating the balance between bone remodeling and resorption [49–51], but the mechanisms influencing macrophage-osteoclast axis differentiation are poorly explored. In this study, by integrating single-cell RNA sequencing and in vitro experiments, we demonstrated that the interaction between AR and c-Myc in tumor cells significantly promoted the differentiation of macrophages into osteoclast-like cells and enhanced osteoclast activation. Further analysis using c-Myc–targeted CUT&Tag sequencing and ChIP-qPCR identified MMP13 as a downstream target of the AR/c-Myc interaction. We further validated that the cooperative interaction between AR and c-Myc regulated the differentiation of the macrophage–osteoclast axis through MMP13, promoted macrophage activation, and ultimately exacerbated bone destruction.

In this study, we employed mouse iliac artery injections to establish bone metastasis models, demonstrating that inhibition of both AR and c-Myc suppresses osteolytic bone metastases in TNBC. Mechanistic investigation revealed the critical involvement of AR/c-Myc interaction in this process. Using the same metastatic model, comparative analysis demonstrated that dual AR/c-Myc inhibition exerted superior efficacy in reducing TNBC osteolytic bone metastases compared to single-target inhibition of either AR or c-Myc alone. These findings expand our understanding of osteolytic bone metastasis mechanisms in TNBC and, more importantly, identify novel therapeutic targets for improving TNBC bone metastasis treatment.

Author contributions

Y. Liu: Conceptualization, data curation, software, formal analysis, investigation, visualization, methodology, writing–original draft. L. Fan: Data curation, software, formal analysis, investigation, visualization, methodology. F. Ye: Investigation, visualization, methodology. Y. Zhao: Investigation, visualization, methodology. Y. Zhu: Data curation, software, formal analysis, investigation. Y. Yang: Data curation, software, formal analysis, supervision. F. Xu: Data curation, supervision. Y. Gu: Data curation, supervision. X. Guan: Conceptualization, resources, data curation, supervision, project administration, writing–review and editing. Y. Liu, L. Fan and F. Ye contributed equally to this work as first authors. All authors read and approved the final paper.

Data availability

The accession number for the scRNA-seq data of Ctrl and DHT bone metastatic tumors and CUT&TAG-seq data generated in this article is GEO(GSE236558) and SRA(PRJNA1268216). Other data analyzed in this study were obtained from GEO with accession numbers GSE2603, GSE2034, GSE12276, GSE125989, and GSE110590. All other raw data are available upon request from the corresponding author. Full, uncut gel images are available in the supplemental materials.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

All animal studies were conducted according to the NIH Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011) and were approved by the Institutional Animal Care and Use Committee of Nanjing Medical University (IACUC-2205019).

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-025-03202-2.