Reciprocal Interactions Between the Gut Microbiome and Mammary Tissue Mast Cells Promote Metastatic Dissemination of HR⁺ Breast Tumors

Despite successful first-line treatments, a substantial population of hormone receptor–positive (HR⁺: estrogen and/or progesterone receptor⁺, HER2^–) patients with breast cancer develop metastasis and succumb to the disease. According to the Surveillance, Epidemiology and End Results database, HR⁺ breast cancer is the most common molecular subtype (73.1%) of the metastatic breast cancer population in the US (1). Well-established treatment strategies have significantly increased long-term survival of women diagnosed with HR⁺ breast cancer, but accurate prediction of patient recurrence remains a significant barrier to reducing the mortality associated with disease.

Using a novel murine model of HR⁺ breast cancer (2–4), we previously demonstrated that gut commensal dysbiosis (an inflammatory gut microbiome with low biodiversity) enhances metastatic dissemination of HR⁺ tumors, despite primary tumor growth remaining unchanged (5), demonstrating that the gut microbiome acts as a host-intrinsic mediator of breast tumor metastasis. Sequencing of the gut microbiome from treatment-naïve patients newly diagnosed with breast cancer similarly demonstrated that differences in the gut microbiome affected tumor aggressiveness, prognosis, and therapy response (6), supporting the notion that the gut microbiome can systemically and negatively impact breast cancer outcomes.

The mechanism by which changes in the gut microbiome impact breast tumor metastasis remains unknown. We previously found that dysbiosis elevated levels of CCL2 in mammary tissue before tumor initiation. When increased CCL2 precedes breast cancer, there is an increased risk of developing breast cancer (7) or metastatic disease after diagnosis (8). Considering the heterogenous outcomes for HR⁺ breast cancer and in light of our previous findings, we hypothesized that commensal dysbiosis changes the immunological composition of the normal mammary tissue, and that the cellular and molecular changes that arise in response to dysbiosis enhance the risk of metastatic spread when a tumor is present. Here, we sought to address this hypothesis by investigating how commensal dysbiosis-induced mammary tissue inflammation influenced dissemination of HR⁺ tumor cells.

Mast cells are evolutionarily conserved innate immune cells that exhibit a diverse range of effector functions in multiple physiological and disease-associated settings. Mast cells mediate stromal rearrangement needed for the development of mammary ducts (9), suggesting a role for mast cells during mammary tissue homeostasis. Despite this, the role of mast cells during breast tumor progression remains controversial (10, 11). Studies demonstrate both positive and negative associations between mast cells and breast cancer outcomes (12). Mechanistically, Majorini and colleagues (13) demonstrate that tumor-associated mast cells promote a luminal tumor phenotype by stimulating estrogen receptor expression. McKee and colleagues (14) have demonstrated that antibiotic treatment enhances mast cell accumulation into mammary tumors, increasing tumor fibrosis and growth kinetics, suggesting that modulation of the gut microbiome might dictate mast cell-mediated differences in breast cancer outcomes. Despite this, it remains unknown whether mammary tissue-associated mast cells impact breast cancer metastasis.

In this study, we demonstrate that commensal dysbiosis enhances accumulation of mast cells in the mammary tissue of non–tumor-bearing mice, and that this occurred in response to elevated mammary tissue CCL2. In the presence of a tumor, commensal dysbiosis enhanced infiltration of mammary tissues with profibrogenic mast cells, culminating in increased tissue collagen levels and the promotion of early metastatic dissemination of HR⁺ tumors. We confirmed that increased mast cell abundance correlated with high collagen levels in breast tissues of women diagnosed with HR⁺ breast cancer and risk for recurrence. To our knowledge, this is the first study revealing that commensal microbiome-mediated recruitment and phenotypic modulation of mammary tissue-associated mast cells before tumor initiation influence the metastatic spread of HR⁺ tumors.

5- to 8-week-old female C57BL/6 mice were purchased from Charles River Laboratories. B6.Cg-Kit^W-sh/HNihrJaeBsmJ (Kit^w-sh/w-sh; Sash; refs. 15–17) mice were purchased from The Jackson Laboratory and maintained at the University of Virginia. All animals were maintained in specific pathogen-free barrier conditions. All experiments in this study were approved by the University of Virginia Institutional Animal Care and Use Committee.

The poorly metastatic syngeneic luminal A, C57BL/6-derived GFP⁺ breast tumor cell line, BRPKp110 (2, 3), was cloned from an inducible autochthonous model of breast cancer after activation of latent KrasG12D and myristoylated-GFP–tagged-p110α (myr-GFP-p110α) in addition to deletion of p53 (4)—all relevant mutations in HR⁺ breast adenocarcinoma (4). BRPKp110 was a kind gift from Dr. Jose Conejo-Garcia, Moffitt Cancer Center and Research Institute. For confirmatory experiments, the highly metastatic mouse mammary cancer cell line, PyMT, which expressed luciferase and was cloned from a metastatic tumor derived from the HR⁺ MMTV-PyMT model as previously described (18), was a kind gift from Dr. Paula Bos, Virginia Commonwealth University. The NIH-3T3 murine fibroblast cell line was isolated from an NIH/Swiss embryo (19). These cells were a gift from Dr. Kimberly Kelly, the University of Virginia. Cell lines were authenticated using MAP testing and bi-monthly Mycoplasma testing to ensure they were free of viral and other pathogens. Cells were monitored for changes in morphology before inoculation into mice. To limit the opportunity for genetic drift, cells were maintained at less than five passage numbers and maintained as frozen stocks at −180°C and expanded only for inoculation into mice. Tumor cell lines were cultured in RPMI (11875093, Gibco), 10% FBS (Sigma), 2 mmol/L of l-glutamine (25030081, Gibco), 1 mmol/L of sodium pyruvate (11360070, Gibco), 50 μmol/L of β-mercaptoethanol (M6250, Sigma), and 100 U/mL of Penicillin/Streptomycin (15140122, Gibco). NIH-3T3 fibroblasts were cultured in DMEM (10313021, Gibco) with 10% FBS (Sigma), 2 mmol/L of l-glutamine (25030081, Gibco), and 100 U/mL of Penicillin/Streptomycin (15140122, Gibco).

Commensal dysbiosis was initiated in mice by oral gavage daily for 14 days with 100 μL of an antibiotic cocktail containing vancomycin (0.5 mg/mL, Gold Biotechnology), ampicillin (1 mg/mL), metronidazole (1 mg/mL), neomycin (1 mg/mL), and gentamicin (1 mg/mL; all from Sigma-Aldrich), as previously reported (5). Vehicle-treated mice received 100 μL of water. Mice were then left untreated for 4 days to allow the establishment of commensal dysbiosis before tumor implantation with BRPKp110 or PyMT, as described below.

Fecal microbiota transplantation (FMT) experiments were performed as previously described (5). Dysbiosis was induced in donor mice as described above. Cecal contents from dysbiotic and nondysbiotic vehicle-treated animals were collected and manually homogenized by passage through a 1-mL syringe to break apart solids. This was followed by vortexing and filtration of solid materials through a 40-μm filter. Cecal solutions were frozen in aliquots at −80°C in sterile 1:1 glycerol/PBS mixture and thawed each day of transplant. Recipient mice were then orally gavaged with the antibiotic cocktail used for inducing dysbiosis for 7 days, followed by 3 or 4 consecutive days of oral gavage with 200 μL of either dysbiotic or nondysbiotic cecal contents. After the final day of gavage, recipient mice were rested for 7 days before tumor initiation with BRPKP110 or PyMT, as indicated below, to allow for microbial engraftment.

The poorly metastatic HR⁺ mouse mammary cancer cell line BRPKp110 has been described previously (2). 5×10⁵ BRPKp110 cells were injected orthotopically into the abdominal mammary fat pad of C57BL/6 mice. For experiments using the highly metastatic mouse luciferase-expressing mammary cancer cell line PyMT (18), 1×10⁵ PyMT cells were injected orthotopically into the abdominal mammary fat pad of C57BL/6 mice. As indicated, tumors were allowed to grow until day 12 for analysis of the tumor and mammary tissue environment, blood for circulating tumor cells, and distal lymph nodes; or day 27 for analysis of metastasized tumor cells in the lungs. To normalize the systemic effects of hormones on mammary gland homeostasis and immune function, 4 days before each temporal point, mice were synchronized into estrus, using a modified Whitten effect (20). All tumor experiments were performed in accordance with UVA IACUC standards. For tumor growth assays, tumors become advanced after 25 to 27 days after injection, a time point when tumor cells are quantifiable in the lungs. Because tumors progress as solid masses, tumors were monitored every 3 days beginning on day 5 of tumor progression using calipers. For experiments extending beyond 12 days, beginning on day 15, or when tumors reach 10 mm total dimension, mice were monitored daily. Mice were euthanized once tumors reach no more than 15 mm total diameter, if tumors were ulcerated, or if tumor growth impaired the animals’ ability to ambulate or eat and drink.

Solutions with mast cell stabilizers ketotifen fumarate (Sigma-Aldrich) and cromolyn (Alfa Aesar) were freshly prepared before each treatment by diluting in sterile water, followed by passage through a 0.22-μmol/L syringe filter. Mice were orally gavaged with 10 mg/kg of ketotifen or an equal volume of water daily, beginning 1 day after tumor implantation, and continuing until day 5 after tumor. For cromolyn, mice were intraperitoneally injected with 10 mg/kg cromolyn or equal volume PBS once per day, beginning 3 days before 12 days after tumor initiation with BRPKp110 or PyMT tumors.

For CCL2 inhibition, mice were intraperitoneally injected with 200 μg of monoclonal anti-mouse CCL2 (Bio X Cell) or 200 μg of the isotype matched control IgG (Bio X Cell) once per day at 8, 6, 4, and 2 days before the implantation of BRPKp110 or PyMT tumors.

Isolated tissues were weighted and then placed in sterile 6-well plate with 3 mL of RMPI (11875093, Gibco) with 5% FBS (Sigma) on ice. Tissues were minced into 2-mm pieces, followed by digestion with collagenase type 1 (17100017, Gibco). Briefly, 2 mg of collagenase type I was added to each well and incubated on a shaker at 130 rpm at 37°C for 30 minutes. To make single-cell suspensions before staining with antibodies, the digested tissues were then passed through 70-μm cell strainers (352350, Corning) using mechanical force with the rubber end of a 5-mL syringe. For in vitro coculture experiments and sorting of mammary tissue mast cells, all tissues were processed in sterile conditions.

CD45⁺cKit⁺FcεRIα⁺ mast cells were sorted from dissociated mammary tissues from nondysbiotic or dysbiotic C57BL/6 mice using a BD FACS Aria Fusion flow sorter (Flow Cytometry Core Facility, University of Virginia). Sorted mast cells were then co-cultured with NIH-3T3 fibroblasts (mast cell: NIH-3T3 fibroblast = 1:20) for 24 hours in serum-free DMEM (10313021, Gibco) at 37°C, 5% CO₂. To test the role of PDGF (platelet-derived growth factor) signaling, 1×10⁴ fibroblasts were pretreated with the tyrosine kinase inhibitor imatinib (0.1 μg/mL) for 20 minutes, followed by incubation with mast cells (mast cell: NIH-3T3 fibroblast = 1:20) in the presence of imatinib (0.1 μg/mL) in serum-free DMEM at 37°C, 5% CO₂ for 24 hours. At the end of incubations, the phenotype of NIH-3T3 fibroblasts was analyzed by surface staining with anti-mouse CD45-PE, PDGFRα-BV605, and intracellular staining with anti-SMA-AF647 and anti-collagen I-AF488, as described below.

Mast cells (CD45⁺cKit⁺FcεRIα⁺ cells) were sorted from mammary tissues of nondysbiotic or dysbiotic C57BL/6 mice as described above. Equal numbers of mammary tissue mast cells were reconstituted in 100 μL of PBS and orthotopically transferred to the fourth abdominal mammary fat pad (approximately ∼300 mast cells/mammary tissue were injected for each experiment) of Kit^w-sh/w-sh (Sash) mice. Eighteen hours after the transfer of sorted mast cells, 5×10⁵ BRPKp110 tumor cells were injected into the same mammary fat pad, as described above.

Total RNA was extracted from murine mast cells sorted from adjacent mammary tissues or day 6 tumors of using an RNeasy Micro Kit (Qiagen), followed by reverse transcription using High-Capacity cDNA Reverse Transcription Kits (Thermo Fisher Scientific). For amplicon detection, the Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) was used as described previously by the manufacturer, using 10 ng of cDNA. Duplicate technical replicates were analyzed for each of 5 biological replicates for each experimental group. Real-time PCR was performed using a QuantStudio 6 Real-Time PCR Systems (Thermo Fisher Scientific) as follows: Initial denaturation at 95°C for 10 minutes; amplification for 35 cycles of denaturation (95°C for 1 minutes), annealing (55°C for 2 minutes), and extension (72°C for 3 minutes). Specificity of the amplicon was determined by melting curve. The relative expression of mRNA was determined by the comparative C_t method and normalized by housekeeping genes Gapdh and β-actin. The following primers were used:

Pdgfa: forward primer 5′-GTGCGACCTCCAACCTGA-3′ and reverse primer 5′-GGCTCATCTCACCTCACATCT

Pdgfb: forward primer 5′-CGGCCTGTGACTAGAAGTCC-3′ and reverse primer 5′-GAGCTTGAGGCGTCTTGG-3′

Tgfb1 forward primer 5′-TGGAGCAACATGTGGAACTC-3′ and reverse primer 5′-CAGCAGCCAATTACCAAG-3′

Cpa3 forward primer 5′-ATCGCAGGCACGCACAGTTAT-3′ and reverse primer 5′-AACCCAGTCTAAGGAAGAGCC-3′

Mcpt4 forward primer 5′-ACCACTGAGAGAGGGTTCACAGC-3′ and reverse primer 5′-GAAGACTCTGATGCACGCAGG-3′

Mcpt5 forward primer 5′-CTGAGAACTACCTGTCGGCCTGC-3′ and reverse primer 5′-TCCAGTTCCAGATTTCCTCACGG-3′

Mcpt2 forward primer 5′-CCACTAAGAACGGTTCGAAGGAG-3′ and revere primer 5′-GCTGGGATGAACTCAGAGGTACC-3′

β-actin forward primer 5′-GTGGGCCGCTCTAGGCACCAA-3′ and reverse primer 5′-CTCTTTGATGTCACGCACGATTTC-3′

Gapdh forward primer 5′-CATCACTGCCACCCAGAAGACTG-3′ and reverse primer 5′-ATGCCAGTGAGCTTCCCGTTCAG-3′

16S rDNA Sequencing was performed by the University of Maryland Institute for genome science, as previously reported (5). Briefly, DNA was extracted from fecal pellets using the Qiagen DNeasy Powerlyzer Power Soil Kit (#12855–100) Per Kit instructions. Amplification of the V3-V4 regions of the 16S rRNA gene was performed using a two step-PCR in which the sample-specific barcode is added during the second PCR, to maximize target amplification. Amplicons were visualized on a 2% agarose gel, quantified by qPCR, and pooled in equimolar concentration and SPRI purified before loading on an Illumina Hiseq 2500 modified to generate 300 Bp paired-end sequence reads. All samples were processed in 96-well plates, along with two extraction and two PCR negative controls.

The MSL has developed an in-house pipeline for sequence quality controls. Briefly, the sequences were demultiplexed using the dual-index strategy, the mapping file generated on the robotic platform and split_libraries_fastq.py, a QIIME-dependent script (21). The resulting forward and reverse fastq files were split by sample using seqtk (https://github.com/lh3/seqtk), and primer sequences removed using TagCleaner (0.16; ref. 22). Further processing followed the DADA2 Workflow for big data and dada2 (v. 1.5.2; https://benjjneb.github.io/dada2/bigdata.html; ref. 23). Forward and reverse reads were each trimmed using lengths of 255 and 225 bp, respectively, and filtered to contain no ambiguous bases, minimum quality score of 2, and required to contain less than two expected errors based on their quality score. Reads were assembled and chimeras removed as per dada2 protocol.

Taxonomy was assigned to each amplicon sequence variant (ASV) generated by dada2 using a combination of the SILVA v128 database and the RDP naïve Bayesian classifier as implemented in the dada2 R package (23, 24) species-level assignments. Read counts for ASVs assigned to the same taxonomy were summed for each sample. A table, including total sequence count for each taxon and for each sample was generated and used for downstream statistical analyses.

Alpha diversity of microbiome samples was measured using Shannon alpha diversity measure. Beta diversity of microbiome sequences was assessed using Bray–Curtis dissimilarity measures based upon relative abundance data.

Mammary tissues, tumors, and tumor-draining lymph nodes were collected, weighed, and processed as described above. Samples were stained with Zombie Aqua (BioLegend), per manufactures instructions in serum-free media. Cells were blocked using anti-CD16/32 (BioLegend). Surface staining was performed on single-cell suspensions with a master mix of antibodies at 4°C for 20 minutes. After surface staining, cells were washed with ice-cold FACS buffer, PBS (10010023, Gibco) containing 3% of FBS (Sigma), and 2 mmol/L of EDTA (AAJ15694AE, Thermo Fisher Scientific). For intracellular staining, cells were resuspended in 100 μL of fixation/permeabilization solution (BD554714, BD) for 10 minutes at room temperature in the dark. Cells were washed with the perm/wash buffer (BD554714, BD Biosciences), followed by staining with the master mix of antibodies for 60 minutes at room temperature in the dark. After staining, cells were washed two times with the Perm/Wash buffer, followed by resuspending in FACS buffer.

Fibroblasts were identified by surface staining with anti-mouse CD45-PE-Cy7 (30–F11), CD31-PerCP-Cy5.5 (390), gp38-APC-Cy7 (8.1.1), and PDGFRα-BV605 (APA5) and intracellular staining with anti-mouse smooth muscle actin (SMA)-AF647 (SPM32) and type I collagen-AF488 (1310–30). Mast cells and basophils were identified by surface staining with anti-mouse CD45-APC-Fire (30–F11), Lineage cocktail-Biotin (145–2C11; RB6–8C5; RA3–6B2; Ter-119; M1/70), CD49b-FITC (HMa2), cKit-BV650 (ACK2), FcεRIα-Biotin (MAR_1), CCR2-AF647 (SA203G11), CXCR4-PE/Dazzle (L276F12), CXCR3-PerCP-Cy5.5 (S18001A), Streptavidin-PerCP-Cy5.5, Streptavidin–PE-Cy7, and intracellular staining with anti-mouse PDGFB-AF488 (F-3). Myeloid cell infiltration was identified by surface staining with anti-mouse CD45-APC (30–F11), CD11b-PE-Cy7 (M1/70), Ly6C-APC-Cy7 (HK1.4), Ly6G-FITC (1A8), F4/80-PerCP-Cy5.5 (BM8), CD206-PE/Dazzle (C068C2), CD86-BV650 (GL-1). T cells, NK cells, and NKT cells were detected by surface staining with anti-mouse CD45-PE (30–F11), CD3-FITC (17A2), CD4-PE-Cy7 (GK1.5), CD8-APC-Cy7 (53–6.7), and NK1.1 (PK136). All antibodies were purchased from BioLegend, except for anti-SMA (Novus Biologicals), anti-type I collagen (SouthernBiotech), anti-GFP (Santa Cruz Biotechnology), and anti-luciferase (Santa Cruz Biotechnology).

Counting beads (AccuCount, Spherotech) were added into samples before acquisition on the flow cytometer. All antibodies and counting beads were added into samples at the manufacturer's recommended concentration. Samples were run on an Attune NxT flow cytometer (Thermo Fisher Scientific) and analyzed with FlowJo software. Absolute numbers were quantitated per the manufacturer's instructions.

Tumor cell dissemination was quantitated using single-cell suspensions from lungs and tumor-draining lymph nodes, prepared as described above, followed by surface labeling with anti-mouse CD45 (30-F11, APC) and intracellular staining with anti-GFP (B-2, PE) or anti-luciferase (C-12, PE). Circulating tumor cells from the blood were isolated and quantitated as previously described (5, 25). Briefly, equal volumes of EDTA-treated blood underwent red blood cell lysis, were plated into 6-well culture dishes, and incubated for 7 days. Cells were then detached and stained as described for quantitation of disseminated tumor cells.

Mammary tissues and tumors were snap-frozen, and protein was extracted by using a mortar and pestle in liquid nitrogen. Proteins were extracted using RIPA lysis buffer, substituted with proteinase inhibitor cocktail (Thermo Fisher Scientific). Equal amounts of protein (5–10 μg/lane) were loaded into precast gels (Thermo Fisher Scientific) and transferred onto polyvinylidene difluoride membranes (Thermo Fisher Scientific). Membranes were blocked using 5% powdered milk solution or 5% BSA solution in PBS with 0.1% Tween 20 (Thermo Fisher Scientific) for one hour and probed with primary antibodies overnight at 4°C, followed by incubation with horseradish peroxidase (HRP)–linked secondary antibodies for one hour. Membranes were imaged using ChemiDoc Imaging System (Bio-Rad) after adding to the washed membrane in a 1:1 solution of Clarity Western Peroxide Reagent:Clarity Western Luminol/Enhancer reagent, per manufacturer's instructions (1705060, Bio-Rad). Antibodies used were: anti-mast cell tryptase antibody (EPR9522, Abcam; 1:1,000 dilution), anti-collagen I antibody (ab34710, Abcam; 1:1,000), goat-anti rabbit IgG H&L (HRP; 1:10,000; Abcam), anti-PDGF-B (F-3, Santa Cruz Biotechnology; 1:1,000), anti-GAPDH (6C5, Santa Cruz Biotechnology; 1:1,000), and mouse-IgGκ-HRP (Santa Cruz Biotechnology; 1:10,000). Immunoblots were quantified using ImageJ.

Archival, deidentified breast tumor tissues (N = 14, Fig. 6) from the University of Virginia Biorepository and Tissue Research Facility tissue bank were fixed in neutral-buffered formalin, paraffin-embedded, and cutoff in 5-μm sections (Research Histology Core, University of Virginia). A second cohort of tissue samples was obtained from the Barts Cancer Institute Breast Cancer Now Tissue Bank (Tissue Request TR241, Supplementary Table S1). These samples were stratified into two groups: No recurrence (n = 18) and Recurrence (to distant sites - bone, lung, liver; n = 23). For detecting mast cells, sections from both cohorts were stained using 0.1% toluidine blue solution (pH 2.0–2.5). To investigate collagen deposition, sections from both cohorts were also stained using PicroSirius Red (0.1% Direct Red 80 in saturated aqueous picric acid, Sigma-Aldrich). Collagen deposition was quantified using ImageJ by calculating the stained area per section. Total numbers of mammary tissue mast cells were normalized to the area of mammary tissues. The sections were imaged with a Slide Scanner (Leica) or a NanoZoomer S210 slide scanner (Hamamatsu).

Statistical analysis was performed using Prism software (GraphPad). Statistical differences between datasets were evaluated using the two-tailed Mann–Whitney U test or unpaired t test, as indicated in the figure legends. Correlation analysis was performed using linear regression. P values < 0.05 were considered significant.

All histological, gene expression analysis, and flow cytometric data presented within this article are available upon request to the corresponding author. Raw 16S rDNA sequencing analysis was performed by the University of Maryland Institute for Genome Science and are available upon request to the primary investigator and University of Maryland. Derived data and ASV data are available from the corresponding author upon request.

We have previously shown that the gut microbiome is a distal host-intrinsic mediator of breast tumor metastasis. When commensal dysbiosis is established before breast tumor initiation, metastatic dissemination of HR⁺ tumors is significantly increased (5). Importantly, dissemination of HR⁺ tumor cells does not occur due to differences in tumor growth, as tumor volumes were unchanged (5). Using the antibiotic treatment scheme in Fig. 1A to initiate commensal dysbiosis, metagenomic 16S rDNA sequencing from feces of mice at the day of tumor initiation (day 0) revealed a substantial difference in beta diversity (inter-individual differences) and a reduction in alpha diversity (intra-individual commensal richness) for mice treated with antibiotics compared with vehicle-treated animals (Supplementary Fig. S1A and S1B). We confirmed our previously reported findings that commensal dysbiosis was established before tumor initiation (5). Twelve days after the initial analysis, reduced differences in both beta and alpha diversity were observed, regardless of tumor status (Supplementary Fig. S1A and S1B). The significant decline in microbiome composition at day 0 relative to diminished differences at day 12 suggests that microenvironmental changes due to dysbiosis, before tumor initiation, prime the mammary tissue to enhance metastatic dissemination.

Gut commensal dysbiosis also elevates chemokines CCL2 and CXCL10 in normal mammary tissues, both of which remain elevated in adjacent mammary tissues during HR⁺ tumor progression (5). To determine whether cellular changes were arising in the mammary tissue in response to CCL2 and CXCL10, the immune cell composition was investigated in normal and tumor-adjacent tissues from mice with or without commensal dysbiosis. Commensal dysbiosis was established, and mice were synchronized into estrus to normalize cellular changes occurring in response to the estrus cycle (Fig. 1A), as previously described (5). Macrophages and myeloid cells in the mammary gland increase risk of metastatic breast cancer (26, 27) and can be recruited in response to CCL2 (28) and CXCL10 (29). Commensal dysbiosis alone did not increase myeloid populations in the mammary gland of non–tumor-bearing mice (Supplementary Fig. S2A), despite our previous report showing enhanced the accumulation of myeloid cells in tumor-adjacent mammary tissue and tumors of dysbiotic HR⁺ tumor-bearing mice (5). Although CD4⁺ T cells were slightly elevated in mammary tissues of non–tumor-bearing dysbiotic mice, the differences were not statistically significant and did not persist with a tumor (Supplementary Fig. S2B and S2D). Similarly, no significant changes in the number of T cells, NK cells, and NKT cells were observed (Supplementary Fig. S2B–S2D).

Mast cells are highly responsive to the microenvironment in which they reside, and in other non-cancer disease models have been shown to induce fibrosis (30) and inflammation (31), both of which are relevant to the changes that occur in the mammary tissue of tumor-bearing mice with preexisting commensal dysbiosis (5). We used flow cytometry to evaluate mast cells (gating scheme in Supplementary Fig. S2E). Mast cell numbers were significantly increased in mammary tissues of non–tumor-bearing mice with commensal dysbiosis (Fig. 1B), and changes in mast cell numbers mirrored the changes in mammary tissue CCL2 and CXCL10, as demonstrated in our previous publication (5). Specifically, mast cell numbers remained elevated in tumor-adjacent mammary tissues of dysbiotic mice 12 days after tumor implantation (Fig. 1C and D). We confirmed these results by quantitating mast cells in tumor-adjacent mammary tissues of dysbiotic or nondysbiotic mice using histochemistry and toluidine blue staining (Fig. 1E and F). Tryptase, a serine protease predominantly produced by mast cells, serves as a specific mast cell marker in both animal and human studies (32). As expected, significantly enhanced protein expression of tryptase were detected in tissue lysates of tumor-adjacent mammary tissues from dysbiotic mice (Supplementary Fig. S2F–S2G). Dysbiotic mice bearing the metastatic PyMT tumors (18) also had increased mast cell abundance in tumor-adjacent mammary tissues (Fig. 1G and H), suggesting that the observed changes in mast cell numbers are not model-specific. Despite observing numeric changes of mast cells in adjacent mammary tissues of dysbiotic tumor-bearing mice, no differences in the numbers of tumor-associated mast cells were observed (Supplementary Fig. S2H). However, the abundance of mast cells in the lungs was increased both in response to dysbiosis and tumor challenge (Supplementary Fig. S2I), highlighting the systemic effect of commensal dysbiosis on the abundance of mast cells in lungs and mammary tissues, but not in the tumors, of mice with commensal dysbiosis.

We have previously shown that transferring a dysbiotic microbiome via FMT enhances metastatic dissemination of HR⁺ breast tumors (5). As an additional readout of dysbiosis-induced mast cell expansion, we tested whether an FMT with a dysbiotic microbiome is sufficient to enhance numbers of mast cells in the mammary tissues of tumor-bearing mice. To do this, cecal contents from dysbiotic and nondysbiotic donor mice were engrafted into antibiotic-sterilized recipient mice (Supplementary Fig. S2J). Flow cytometry was then used to analyze normal non–tumor-bearing mammary tissues. Similar to animals with dysbiosis, mice receiving an FMT of a dysbiotic microbiome had significantly greater numbers of mast cells in mammary tissues compared with those that received an FMT of a nondysbiotic microbiome (Supplementary Fig. S2K). In tumor-bearing mice receiving an FMT, tumor-adjacent mammary tissues were also analyzed 12 days after tumor-initiation (Supplementary Fig. S2L). These results indicate that commensal dysbiosis is sufficient to enhance mast cells in mammary tissues during HR⁺ tumor progression.

We next asked whether mast cell activation influenced the dissemination of HR⁺ breast tumor cells. To prevent mast cell activation, mice were orally gavaged with the mast cell stabilizer ketotifen fumarate (Fig. 2A). Ketotifen treatment significantly reduced dysbiosis-mediated tumor dissemination into the lungs (Fig. 2B; Supplementary Fig. S3A). However, ketotifen treatment did not affect the growth of primary BRPKp110 tumors (Supplementary Fig. S3B). Ketotifen treatment also significantly reduced the lung dissemination of PyMT tumors (Fig. 2C; Supplementary Fig. S3C). Because the treatment with ketotifen also reduced primary PyMT tumor burden (Supplementary Fig. S3D), the numbers of PyMT tumor cells in the lungs were normalized to final tumor burden (Fig. 2C). Ketotifen treatment also reduced tumor dissemination in nondysbiotic mice bearing PyMT tumors (Fig. 2C). Treatment of mice with cromolyn, another mast cell stabilizer (Supplementary Fig. S3E), also significantly reduced dissemination of BRPKp110 tumor cells into the lungs of dysbiotic mice (Supplementary Fig. S3F and S3G), suggesting that the observed differences in lung tumor burden were not due to off-target effects of either inhibitor. Altogether, these data suggest that mast cell activation contributes to dysbiosis-induced metastatic dissemination of HR⁺ breast tumors.

Next, we sorted mast cells from mammary tissues of non–tumor-bearing dysbiotic and nondysbiotic mice. Equal numbers of sorted mast cells were orthotopically transferred into the left inguinal mammary fat pad of mast cell-deficient Kit^w-sh/w-sh mice (Sash mice; ref. 33). The following day, BRPKp110 tumor cells were injected into the same mammary fat pad that received mast cells (Fig. 2D). Using flow cytometry, we confirmed that equivalent numbers of mast cells remained in the mammary tissues of Sash mice 12 days following tumor initiation (Fig. 2E), indicating that both groups of sorted mast cells were equally capable of engrafting into the mammary tissue of Sash mice. Sash mice that received mast cells sorted from mammary tissues of dysbiotic mice had significantly increased dissemination of tumor cells into the blood and to the axillary lymph nodes compared with that observed in Sash mice receiving mast cells from mammary tissues of nondysbiotic mice (Fig. 2F–H).

To test whether dysbiosis alone could affect tumor dissemination in the absence of mast cells, Sash mice or wild-type C57BL/6 mice were given a fecal transplant with a dysbiotic microbiome (Supplementary Fig. S3H). Although dysbiotic Sash mice showed comparable tumor growth compared with wild-type mice (Supplementary Fig. S3I), dissemination of tumor cells into the blood was significantly reduced compared with wild-type mice transplanted with the same dysbiotic microbiome (Supplementary Fig. S3J and S3K). Collectively, these data indicate that mast cells are required for dysbiosis-induced metastatic dissemination of HR⁺ tumors. Furthermore, these results suggest that mammary tissue mast cells are functionally changed into a pro-metastatic phenotype in response to commensal dysbiosis and before tumor initiation.

Changes in mast cell numbers mirrored changes in chemokine levels of CCL2 and CXCL10 (5), leading us to hypothesize that mast cells were accumulating in response to CCL2 and/or CXCL10. To test this hypothesis, we first measured CCR2 and CXCR3 levels on mast cells in the mammary tissue of dysbiotic and nondysbiotic mice via flow cytometry, with the assumption that CCR2 or CXCR3 expression corresponded with accumulation in response to each respective ligand, CCL2 or CXCL10. Mast cells from mammary tissues of non–tumor-bearing dysbiotic mice showed approximately 60% of mast cells expressed CCR2 compared with 30% in mammary tissues of nondysbiotic mice, and no differences were observed for CXCR3-expressing mast cells between the two groups (Fig. 3A). These data suggest that mast cells are accumulating in response to CCL2.

To test whether mast cells accumulated in the mammary tissue in response to CCL2 signaling, CCL2 was neutralized using a blocking antibody. Dysbiosis was established, and monoclonal anti-CCL2 or an IgG isotype control was administered to mice before tumor initiation (Fig. 3B). Neutralization of CCL2 significantly reduced the abundance of total (Fig. 3C; Supplementary Fig. S4A) and CCR2⁺ (Fig. 3D; Supplementary Fig. S4B) mast cells infiltrating into adjacent mammary tissues of tumor-bearing dysbiotic mice. Reduced mast cell accumulation was not due to changes in tumor growth, as overall tumor burden was similar across groups (Supplementary Fig. S4C). Concomitant with the reduction of mast cells during CCL2 blockade, tumor dissemination into the peripheral blood (Fig. 3E and F) and tumor-draining lymph nodes (Fig. 3G and H) was significantly reduced in dysbiotic mice. Together, these results indicate that CCL2 is required for commensal dysbiosis-induced accumulation of mammary tissue mast cells and subsequent dissemination of HR⁺ tumors.

Mast cells contribute to fibroblast activation and tissue fibrosis in various organs of both humans and animals (30). Activated fibroblasts synthesize and remodel the extracellular matrix, creating a conduit for tumor cells to egress out of the primary tissue into peripheral organs (34). To determine whether mast cells affected fibroblast activation and matrix remodeling, activation [collagen I⁺ and smooth muscle actin (SMA)⁺] and expansion of mammary tissue fibroblasts were analyzed before and during early tumor progression (Fig. 4A) using flow cytometry (Supplementary Fig. S5A). Commensal dysbiosis enhanced fibroblast expansion and activation in normal mammary tissues of non–tumor-bearing mice, which was evident before tumor initiation (Fig. 4B). Although the numbers of activated fibroblasts diminished in the absence of tumors (Fig. 4C), tumor-bearing mice with preestablished commensal dysbiosis had a significant expansion of activated fibroblasts in adjacent mammary tissues (Fig. 4D). Collagen I protein expressions in adjacent mammary tissues of dysbiotic mice bearing BRPKp110 (Fig. 4E; Supplementary Fig. S5B) or PyMT (Fig. 4F; Supplementary Fig. S5C) tumors were correspondingly increased compared with nondysbiotic mice with equivalent tumor burden.

Using FMT, transfer of a dysbiotic microflora, but not a nondysbiotic microflora, significantly induced collagen I protein expression in adjacent mammary tissues of BRPKp110 tumor-bearing mice (Fig. 4G; Supplementary Fig. S5D). Despite not observing significant changes in tumor-associated mast cells (Supplementary Fig. S2G), collagen I expression in tumors from dysbiotic mice was significantly increased (Supplementary Fig. S5E), and collagen I expression in adjacent mammary tissue significantly and positively correlated with tumor collagen I expression (Supplementary Fig. S5F). This association was not observed in tumors from nondysbiotic mice (Supplementary Fig. S5G). Fibroblasts in mammary tumors are recruited from the adjacent mammary tissue environment (35), suggesting that dysbiosis-induced changes in tissue fibroblast activation correspondingly increased collagen I density in mammary tissue and tumors.

We next tested whether inhibition of mast cell activation would reduce activation of tissue fibroblasts and collagen I expression. Dysbiosis was initiated, and mice were treated with the mast cell stabilizer ketotifen, as depicted in Fig. 2A. Ketotifen significantly reduced the accumulation and activation of fibroblasts in adjacent mammary tissues of tumor-bearing mice with commensal dysbiosis (Supplementary Fig. S5H), and treatment also significantly reduced protein expression of collagen I in adjacent mammary tissues of dysbiotic mice (Fig. 4H and I). Next, we investigated whether fibroblasts were activated in the absence of mast cells. Compared with C57BL/6 mice, dysbiotic Sash mice had reduced numbers of collagen I⁺SMA⁺ fibroblasts in adjacent mammary tissues (Fig. 4J and K). Similar observations were made after reducing mammary tissue mast cell accumulation via neutralization of CCL2 (Fig. 4L and M). Together, these results suggest that commensal dysbiosis-induced mammary tissue mast cells initiate fibroblast activation and shift the tissue milieu toward a profibrogenic microenvironment.

Transfer of mast cells into Sash mice (Fig. 2D–H) suggested that mast cells from dysbiotic mice were phenotypically distinct from those in mammary tissues of nondysbiotic mice. We evaluated expression of genes related to profibrogenic pathways from mast cells sorted from adjacent mammary tissues and tumors of dysbiotic and nondysbiotic mice 12 days after tumor initiation. Pdgfb expression was increased in mammary tissue mast cells from tumor-bearing dysbiotic mice, whereas other profibrogenic mediators, including Mcpt4, Mcpt5, Mcpt2, Pdgfa, and Tgfb1, remained unchanged (Supplementary Fig. S6A). The mast cell marker Cpa3 also remained comparable between the two groups (Supplementary Fig. S6A). Tumor-associated mast cells from dysbiotic mice showed phenotypic differences, with increased Mcpt2, Pdgfb, and Cpa3 expression compared with tumor-associated mast cells from nondysbiotic mice (Supplementary Fig. S6B). These data suggest that the phenotype of mast cells is influenced by dysbiosis-induced changes to the local tissue environment.

Profibrogenic mediator, platelet-derived growth factor B (PDGF-B), is known to be involved in mast cell-mediated activation of fibroblasts (36). Indeed, before tumor initiation, numbers of PDGF-B–expressing mast cells in mammary tissues of dysbiotic mice were significantly increased compared with mast cells in mammary tissues of nondysbiotic mice (Fig. 5A). PDGF-B⁺ mast cell numbers remained significantly elevated in adjacent mammary tissues during early tumor progression (Fig. 5B), and changes in PDGF-B-expressing mast cell numbers corresponded with a significant increase in the expression of PDGFR alpha (PDGFRα) on mammary tissue fibroblasts before (Fig. 5C) and during tumor progression (Fig. 5D). Global protein expression of PDGF-B in mammary tissues was also measured using immunoblot. Similar to that observed in mast cells, dysbiotic tumor-bearing mice had enhanced PDGF-B protein expression in adjacent mammary tissues compared with nondysbiotic mice (Supplementary Fig. S7A and S7B). A dysbiotic microflora was also found to be sufficient to induce PDGF-B protein expression in adjacent mammary tissues (Supplementary Fig. S7C–S7E). Tumor-bearing Sash mice had significantly lower PDGF-B in response to FMT-induced dysbiosis compared with wild-type mice with dysbiosis (Supplementary Fig. S7F and S7G). As expected, a mast cell deficiency also reduced expression of tryptase in the mammary tissue (Supplementary Fig. S7F and S7H). TGFβ1 is another potent profibrogenic and prometastatic mediator in multiple disease models, including breast cancer (37). However, total protein expressions of TGFβ1 in the adjacent mammary tissues and tumors of dysbiotic and nondysbiotic mice were comparable (Supplementary Fig. S7I and S7J). Together, these data implicate mast cell-mediated PDGF signaling in the activation of mammary tissue fibroblasts during commensal dysbiosis.

To test whether mammary tissue mast cells from dysbiotic mice promoted fibrosis via PDGF signaling, mast cells sorted from mammary tissues of dysbiotic or nondysbiotic mice were cocultured with NIH-3T3 fibroblasts for 24 hours (Fig. 5E). The following day, activation of PDGFRα⁺ and PDGFRα-negative fibroblasts was evaluated by measuring SMA and collagen I expression. Mast cells sorted from mammary tissues of dysbiotic mice induced significantly increased SMA and collagen I in PDGFRα⁺ fibroblasts (Fig. 5F and G). Although collagen expression between PDGFRα⁺ fibroblasts cultured with dysbiotic versus nondysbiotic mast cells were not statistically different, collagen I expression was significantly increased when comparing NIH-3T3 fibroblasts cultured alone versus those cultured with mast cells from dysbiotic mice (Fig. 5F and I). In contrast, mast cells from dysbiotic mice induced low, but similar, expression of SMA and collagen I in PDGFRα-negative fibroblasts (Fig. 5F, H, and J), suggesting that PDGF signaling may be involved in mast cell-mediated fibroblast activation. When sorted mast cells from mammary tissues of dysbiotic tumor-bearing mice were co-incubated with fibroblasts and the tyrosine kinase inhibitor imatinib (to inhibit PDGF signaling), fibroblast activation was also significantly reduced (Fig. 5K–M). These data indicate that mast cell-derived PDGF signaling mediates fibroblast activation and collagen I deposition.

To determine the clinical relevance of mast cell-mediated fibroblast activation with breast cancer, we evaluated formalin-fixed, paraffin-embedded tissue sections, containing tumor-adjacent mammary tissue, from a small archival cohort (N = 14) of patients with HR⁺ breast cancer. Collagen deposition in tumor-adjacent tissue positively and significantly correlated with mast cell numbers (Fig. 6A and B), suggesting that mast cell abundance coincides with collagen in patients with HR⁺ breast cancer.

We next asked whether a correlation between mast cell numbers and collagen I expression associated with increased risk of metastatic disease. We first analyzed collagen I and tryptase protein levels via immunoblot, a specific mast cell marker in both animal and human studies (32), from adjacent mammary tissues of tumor-bearing mice with or without commensal dysbiosis. There was a significant and positive association between collagen I and tryptase protein expression in adjacent mammary tissues from mice with dysbiosis, but not in mice without commensal dysbiosis (Fig. 7A and B). To investigate whether this association existed in adjacent mammary tissues of patients that eventually developed metastatic disease, we evaluated a second cohort of archival tissues obtained from the Barts Cancer Institute Breast Cancer Now Tissue Bank [n = 18 patients diagnosed with HR⁺ breast cancer who did not experience metastatic recurrence; n = 23 patients diagnosed with HR⁺ breast cancer who developed metastatic disease (Supplementary Table S1)]. There was a positive, although not significant, linear correlation between the numbers of mast cells and collagen expression in adjacent mammary tissues for patients that developed recurrence (Fig. 7C and D). In contrast, no association was observed in patients with no recurrence. The difference in the linear correlation between mast cells and collagen of each group was significant, indicating that the data follow two separate linear trajectories between each group of patients (Fig. 7D). The mast cell density was enriched in some patients without recurrence, suggesting that qualitative, rather than quantitative, changes in the mammary tissue mast cell populations might be more representative for increased risk of metastatic disease.

Here, we demonstrated that mammary tissue mast cells enhanced metastatic dissemination of HR⁺ tumors in response to gut commensal dysbiosis. The role of mast cells in breast cancer has been controversial (11), likely due to the focus of associating quantitative changes in tumor- or lymph node–associated mast cells with outcomes, instead of evaluating mast cells qualitatively or temporally. In doing so, some studies have identified a positive association with outcome (38, 39, 40), whereas others have associated mast cells with a poor prognosis (40–43). The work from Majorini and colleagues (13) was one of the first studies to mechanistically define how mast cells influence the progression of breast tumors. In this seminal study, they demonstrated mast cells directly modulate estrogen receptor expression on breast tumor cells via activation of cMET, enhancing breast tumor growth. Somasundaram and colleagues (44) further went on to demonstrate that mast cells enhance immune suppression and failure of immune therapy with PD-1 blockade. By evaluating the tumor microenvironment before and after neoadjuvant chemotherapy, Reddy and colleagues (45) demonstrated that having a high density of mast cells is a reliable indicator that patients will not respond to therapy, underscoring the clinical relevance of mast cells for breast cancer.

Mast cells are responsive to the microenvironment in which they reside (46), and in other non-cancer disease models, they induce endothelial cell tube formation (47), fibrosis (30), and inflammation (31). Thus, it remains unknown how and when mast cell function is modulated to favor tumor growth, immune evasion, and metastasis. We observed that tumor-associated mast cells from dysbiotic mice expressed increased expression of Cpa3, Mcpt2, and Pdgfb, whereas mammary tissue-associated mast cells expressed higher Pdgfb. Our results demonstrate that mast cells residing in distinct tissue microenvironments are heterogenous and implicate the gut microbiome as a major contributor to mast cell functional heterogeneity. McKee and colleagues (14) have demonstrated that antibiotic-induced perturbations in the gut microbiome significantly increases the density of mast cells in stromal-dense regions of luminal A and B tumors, resulting in accelerated tumor growth. Although we did not observe differences in tumor growth with our model of dysbiosis, this study supports the idea that changes to the gut microbiome can distally affect mast cells to favor poor breast cancer outcomes. We also demonstrate that microbiome-induced changes to mammary-tissue mast cells are established before tumor initiation.

Our findings also suggest that mast cells enhance the reservoir of tumor-associated fibroblasts through activation of tissue-associated fibroblasts (35) via PDGF signaling. PDGF-B⁺ mast cells and PDGFRα⁺ fibroblasts are increased during commensal dysbiosis. Mast cells are an important cellular source of PDGF, a well-known profibrogenic signaling molecule. Indeed, mast cells sorted from mammary tissues of dysbiotic mice induced significantly higher expression of SMA and collagen I in PDGFRα⁺ 3T3 fibroblasts after in vitro co-culture compared with mast cells from nondysbiotic mice, and activation of fibroblasts by mast cells sorted from mammary tissues of dysbiotic mice was inhibited in vitro with imatinib. Although PDGF signaling associates with higher rates of breast cancer recurrence (48), treatment with the PDGF receptor inhibitor imatinib has limited efficacy in patients with metastatic HR⁺ breast cancer (49–51). Given the contribution of mast cells in the upregulation of PDGF signaling, imatinib therapy may benefit HR⁺ patients with a high abundance of PDGF⁺ mast cells in the adjacent mammary tissue or in patients with commensal dysbiosis. Alternatively, mast cell stabilizers could potentially enhance the efficacy of imatinib for these patients.

Clinically, high levels of fibrosis associate with increased invasiveness and poor outcomes for patients with breast cancer (52). Indeed, for BRPKp110 and PyMT tumors, orthotopic models of luminal A (5, 53) and luminal B (54, 55) breast cancer, respectively, collagen levels in adjacent tissues and tumors are increased (5). Chemical inhibition of mast cells was sufficient to reduce collagen, suggesting that for luminal A and B breast cancer mast cells may be involved in increasing tissue stiffness and matrix reorganization, at least in the context of changes to the microbiome. Future studies will need to evaluate whether pre-existing dysbiosis affects the stroma and metastasis of triple-negative or Her2⁺ breast cancer subtypes.

Mechanistically, blockade of CCL2 before tumor initiation reduced mast cell accumulation and early dissemination of mammary tumor cells. These data uncover a tumor-independent role for CCL2 in promoting metastatic spread. Although CCL2 is a well-known driver of breast tumor growth (56) and metastasis (28), tumor-independent triggers of CCL2 have not been investigated. Supporting the idea that CCL2 contributes to tissue fibrosis in the absence of breast tumors, Sun and colleagues (7) demonstrated that overexpression of CCL2 in the pre-malignant mammary epithelium not only increased mammary tissue density, but also susceptibility to carcinogen-induced breast cancer. Although the cell populations producing CCL2 in the mammary tissue in response to dysbiosis are undefined in our model, it is tempting to speculate that mast cells may be important for the fibrotic phenotype in the Sun and colleagues (7) model. It would also be interesting to determine whether the CCL2–CCR2 axis is involved in programming of mast cells toward a more profibrogenic phenotype.

Breast cancer risk factors, such as obesity, use of antibiotics, age, diet, and race, all associate with altered composition of the commensal microbiome (57). This highlights the necessity to define how an unbalanced commensal microbiome facilitates dissemination of breast tumor cells. There is a growing body of evidence, indicating that the commensal microbiome affects multiple cancer outcomes and therapy responses (58, 59). This study identifies a novel mechanism, whereby an unbalanced and inflammatory microbiome enhances metastatic breast cancer through the promotion of cellular and molecular changes in normal mammary tissue. Considering the finding that specific gut microbiome differences in treatment-naïve patients with breast cancer significantly associate with increased tumor aggressiveness, poor prognosis, and diminished therapy response (6), it is critical to address whether changes in the microbiome that favor metastatic breast cancer are present before diagnosis. Because it is not yet known how changes to the gut microbiome modulate mast cell phenotype and function, future studies will also be focused on more extensively characterizing mast cell phenotypic changes occurring in response to dysbiosis and defining how mast cells directly affect tumor cell stemness or cells in the immune environment. Importantly, to interrogate how changes in the tissue environment affect breast cancer outcomes, phenotypic evaluation of mast cells in the adjacent mammary tissues and possibly tumors of patient samples should also be performed. On the basis of this study, targeting mast cells in patients with early-stage HR⁺ breast cancer could serve as an adjuvant to existing therapies aimed at reducing recurrence with metastatic disease.

T.Y. Feng reports grants from NCI and Susan G. Komen during the conduct of the study. M.T. McGinty reports grants from NCI and Susan G. Komen during the conduct of the study. M.R. Rutkowski reports grants from NCI/NIH 1R01CA253285, Susan G. Komen CCR17483602, and other support from UVA Comprehensive Cancer Center, grants from American Cancer Society, other support from NIH T32AI007496-21 and NIH T32CA009109-46 during the conduct of the study. No disclosures were reported by the other authors.

T.Y. Feng: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft. F.N. Azar: Investigation, writing–review and editing. S.A. Dreger: Data curation, formal analysis, investigation, writing–review and editing. C.B. Rosean: Investigation, writing–review and editing. M.T. McGinty: Investigation, writing–review and editing. A.M. Putelo: Investigation, writing–review and editing. S.H. Kolli: Investigation, writing–review and editing. M.A. Carey: Formal analysis, visualization, methodology. S. Greenfield: Data curation, visualization. W.J. Fowler: Formal analysis, investigation, writing–review and editing. S.D. Robinson: Data curation, formal analysis, supervision, investigation, writing–review and editing. M.R. Rutkowski: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration.

This study was supported by a CCR grant from Susan G. Komen CCR17483602 and 1R01CA253285 from the NCI, with aid from Grant #IRG 81-001-26 from the American Cancer Society. T.Y. Feng was supported by a trainee fellowship from the University of Virginia Comprehensive Cancer Center, CBR was supported by T32AI007496-21, MTM was supported by T32CA009109-46. S.D. Robinson and S.A. Dreger were supported by the BBSRC Institute Strategic Program Gut Microbes and Health BB/R012490/1 and its constituent project (BBS/E/F/000PR10355). S.D. Robinson and W.J. Fowler were also supported by Cancer Research UK (grant number C18281/A29019). The authors also wish to acknowledge the roles of the Breast Cancer Now Tissue Bank in collecting and making available the samples and/or data, and the patients who have generously donated their tissues and shared their data to be used in the generation of this publication. The authors are also grateful for the extraordinary support from the University of Virginia Comprehensive Cancer Center, the Center for Research Histology, Biorepository and Tissue Research Facility, Flow Cytometry Core Facility, Center for Comparative Medicine, Trans University Microbiome Initiative and the Carter Immunology Center. In addition, we would like to thank Dr. Jose R. Conejo-Garcia for the BRPKp110 HR⁺ breast tumor cell line, Dr. Paula D. Bos for the PyMT cell line, Dr. Kimberly D. Kelly for the NIH-3T3 fibroblast cell line, and Dr. Sanja Arandjelovic for critical reading of the article.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.