Preexisting Commensal Dysbiosis Is a Host-Intrinsic Regulator of Tissue Inflammation and Tumor Cell Dissemination in Hormone Receptor–Positive Breast Cancer

Breast cancer is a major global health issue, with 1 in 8 women expected to be diagnosed with breast cancer in their lifetime. The majority of breast cancers (∼65%) are diagnosed as hormone receptor–positive (HR⁺; estrogen and progesterone receptor⁺, HER2^neg/low; ref. 1). Because of the effectiveness of adjuvant hormonal therapies, HR⁺ breast cancer has a more favorable short-term prognosis compared with more aggressive subtypes such as triple-negative breast cancer (TNBC; ref. 2). However, patients diagnosed with HR⁺ breast cancer have heterogeneity in tumor aggressiveness and overall survival, with a subset of patients remaining at significant risk for developing recurrent disease after 5 years of standard endocrine therapy (3). This heterogeneity in outcome results in a small but sustained risk for developing recurrent metastatic disease 5 years or more after initial remission (4).

Metastatic dissemination to distal lymph nodes and the lungs occurs early in HR⁺ breast cancer progression (5). Although it remains unknown whether host-intrinsic differences in immune function contribute to increased early dissemination of tumor cells, lymph node involvement is one of the most reliable predictors of recurrent disease (6, 7) and identifies patients that would benefit from extended adjuvant hormonal therapy. This therapy reduces late recurrence for some high-risk patients but also commonly results in undesirable and potentially serious side effects (8). Identification of early or preexisting host-associated factors that influence the evolution of disseminated breast cancer could therefore aid in the development of alternative interventions for patients that are at highest risk for recurrent metastatic disease.

Factors such as obesity, genetic polymorphisms, race, diet, and chronic use of antibiotics have been associated with adverse outcomes in breast cancer (9–12). All of these factors are also linked to a disruption in the homeostasis of the commensal microbiome. The commensal microbiota is a host-intrinsic regulator of systemic innate and adaptive immune function (13–15). Altered diversity in the microbiota, known as dysbiosis, can influence the systemic immune environment (16) and lead to increased production of inflammatory mediators that associate with a poor outcome for multiple diseases, including cancer (17–19). Indeed, infection of APC^Min/+ mice with H. hepaticus led to increased inflammation and significant incidence of breast cancer in female mice (20), supporting the idea that perturbation of commensal organisms can influence breast cancer risk (21). However, it is currently unknown whether preexisting commensal dysbiosis leads to changes within the tissue and/or tumor microenvironment that promote tumor aggressiveness and metastatic dissemination.

We set out to determine whether commensal microbes influence breast cancer progression: specifically, if preexisting dysbiosis in the commensal microbiome associates with poor outcomes in HR⁺ breast cancer. Here, we demonstrate that a disruption in commensal homeostasis promotes tumor cell dissemination to distal sites, enhances fibrosis, and collagen deposition both systemically and locally within the tissue and tumor microenvironment, and results in significant early inflammation and myeloid infiltration into the mammary tissue and tumor. These effects are initiated by signals from the gut, as commensal dysbiosis induced both by treatment with nonabsorbable antibiotics and by fecal microbiota transplantation (FMT) of dysbiotic cecal contents recapitulates these outcomes. This study identifies commensal dysbiosis as a host-intrinsic regulator of tissue inflammation, myeloid recruitment, fibrosis, and dissemination of tumor cells–all of which contribute to reduced survival in HR⁺ breast cancer.

Five- to 8-week-old female C57BL/6 mice were purchased from Charles River Laboratories. L-Stop-L-KRas^G12Dp53^flx/flxL-Stop-L-Myristoylated p110α−GFP⁺ mice on a C57BL/6 background (22) were bred and maintained in house. All animals were maintained in pathogen-free barrier facilities at the University of Virginia (Charlottesville, VA). All experiments in this study were approved by the University of Virginia Institutional Animal Care and Use Committee.

The poorly metastatic HR⁺ mouse mammary cancer cell line BRPKp110 has been described previously (23). 5E5 BRPKp110 cells were injected orthotopically into the abdominal mammary fat pad. The highly metastatic mouse mammary cancer cell line PyMT, which expresses luciferase, was cloned from a metastatic tumor derived from the HR⁺ MMTV-PyMT model as described previously (24). In experiments using this cell line, 1E5 PyMT cells were injected orthotopically into the abdominal mammary fat pad. Cell lines were authenticated by maintaining at less than four passages, monitoring of morphology, and testing for Mycoplasma. Autochthonous mammary tumors were induced in L-Stop-L-KRas^G12Dp53^flx/flxL-Stop-L-Myristoylated p110α−GFP⁺ mice on a C57BL/6 background as described previously (22). Briefly, tumors were initiated via intraductal injection of 2.5E7 PFU of adenovirus-Cre through the nipple in the fourth mammary fat pad.

For most experiments, mice were orally gavaged for 14 days with antibiotic cocktail or water as a vehicle control as described previously (25). After 14 days of gavage, mice were left untreated for four days before tumor initiation to allow for the establishment of commensal dysbiosis. Mice treated with an absorbable (systemic) antibiotic cocktail received 100 μL of a cocktail containing vancomycin (0.5 mg/mL), ampicillin (1 mg/mL), metronidazole (1 mg/mL), neomycin (1 mg/mL), and gentamicin (1 mg/mL). Vehicle-treated mice received 100 μL of water as a control. Mice treated with nonabsorbable (nonsystemic) antibiotics received 200 μL of a cocktail containing neomycin (2 mg/mL) and bacitracin (2 mg/mL). Vehicle-treated mice received 200 μL of water as a control. All antibiotics were purchased from Sigma-Aldrich with the exception of vancomycin (Gold Biotechnology). For FMT experiments, dysbiosis was induced in donor mice as described above. Cecal contents from dysbiotic and nondysbiotic animals were collected, homogenized, and frozen at −80°C in sterile 1:1 glycerol/PBS. Recipient mice were orally gavaged with absorbable antibiotics for 7 days, followed by 3 consecutive days of oral gavage with either dysbiotic or nondysbiotic cecal contents. After the final day of gavage, recipient mice were rested for 7 days before tumor initiation to allow for microbial engraftment.

Tumors, mammary glands, and lungs were removed, weighed, homogenized (gentleMACS Dissociator, Miltenyi Biotec), and digested with collagenase D (Sigma-Aldrich) for 30 minutes at 37°C prior to 70-μm filtration. Tumor-draining lymph nodes were homogenized using glass slides and processed as the other tissues. Flow cytometry was performed by staining single-cell suspensions with Zombie Aqua viability dye (BioLegend), blocking Fc receptors with anti-CD16/32 (93, purified), and staining with anti-mouse antibodies at the manufacturer's recommended dilution. Myeloid cell infiltration was quantitated in tumors and normal-adjacent mammary glands by surface staining with anti-mouse CD45 (30-F11, Pacific Blue), CD11b (M1/70, PE/Cy7), F4/80 (BM8, PerCP/Cy5.5), CD86 (GL-1, BV650), CD206 (C068C2, PE/Dazzle-594), Ly6C (HK1.4, APC/Cy7), and Ly6G (1A8, FITC) and intracellular staining using anti-mouse arginase-1 (IC5868P, PE) and IL6 (MP5-20F3, APC). Tumor cell dissemination was quantitated using single-cell suspensions from lungs and tumor-draining lymph nodes through surface staining with anti-mouse CD45 (30-F11, PE) and intracellular staining with anti-GFP (FM264G, APC). Circulating tumor cells from the blood were quantitated as described previously (26). 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 detached and stained as described for quantitation of disseminated tumor cells. All antibodies were purchased from BioLegend with the exception of anti-arginase-1 (R&D Systems). Counting beads (AccuCount, Spherotech) were added to samples at the manufacturer's recommended concentration, and the samples were subsequently run on a Beckman Coulter CytoFLEX cytometer and analyzed using FlowJo.

Cytokines and chemokines were analyzed in serum and mammary glands using a custom multiplex U-PLEX assay from Meso Scale Diagnostics. Blood was collected by cardiac puncture and allowed to clot at room temperature for 30 minutes before centrifugation and collection of serum. Mammary glands were flash frozen in liquid nitrogen, ground in a mortar and pestle with lysis buffer (150 mmol/L NaCl, 20 mmol/L Tris, 1 mmol/L EDTA, 1 mmol/L EGTA, 1% Triton X-100, pH 7.5), sonicated, and centrifuged. The supernatants were then collected, protein concentrations were quantified by BCA assay (Pierce, Thermo Fisher Scientific), and supernatants were used for U-PLEX analysis. Mammary gland readouts were normalized to total protein input.

To quantitate fibrosis, tissues were fixed in neutral-buffered formalin, paraffin-embedded, and cut in 5-μm sections (Research Histology Core, University of Virginia, Charlottesville, VA). Slides were stained using PicroSirius Red (0.1% Direct Red 80 in saturated aqueous picric acid, Sigma-Aldrich). Fibrosis was quantified using ImageJ software by calculating the area of tissue and the intensity of red staining. To examine mammary gland pathology, hematoxylin and eosin–stained slides were scored by a pathologist in a blinded fashion. To determine estrus cycling, vaginal lavages were collected daily after the 14-day antibiotic gavage period and examined for vaginal cytology. Mice were considered to be cycling when they progressed from proestrus to estrus.

Mice were injected intravenously with 3 mg of d-luciferin potassium salt (Gold Biotechnology) and allowed to saturate systemically for 2 minutes. Mice were then euthanized, and lungs were removed for ex vivo imaging to detect bioluminescent light emission using a Caliper IVIS Spectrum bioluminescence instrument (PerkinElmer; Molecular Imaging Core, University of Virginia, Charlottesville, VA). Bioluminescence from ex vivo lungs was calculated as photons per second within a standardized region of interest and normalized to final tumor burden.

Sequencing was performed by the University of Maryland Institute for Genome Science. Sequences were demultiplexed using the mapping file split_libraries_fastq.py, a QIIME-dependent script. Fastq files were split by using seqtk (https://github.com/lh3/seqtk), primer sequences removed using TagCleaner (0.16), followed by downstream processing using DADA2 Workflow for Big Data and dada2 (v. 1.5.2; https://benjjneb.github.io/dada2/bigdata.html). Forward and reverse reads were trimmed using lengths of 255 and 225 bp, respectively, to contain no ambiguous bases, have a minimum quality score of 2, and contain less than two expected errors based on quality score. Reads were assembled and chimeras removed 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 species-level assignments. Read counts for ASVs assigned to the same taxonomy were summed for each sample.

Alpha diversity of microbiome samples was measured using Shannon alpha diversity measure and a Kruskal–Wallis test to measure significance. Beta diversity of microbiome sequences was assessed using Bray–Curtis dissimilarity measures based upon relative abundance data and a Permanova test was performed to measure statistical significance.

Metabolomic analysis was performed by Metabolon, Inc. Principal component analysis was performed using the “prcomp()” R function (https://www.R-project.org/). Samples were colored according to tumor status, dysbiosis, or both (sample annotations). The association between the sample annotations and the first two principal components was tested by one-way ANOVA. A significant association suggests that the variability explained by a principal component is driven by a corresponding sample annotation. Calculations and visualizations were performed in the R/Bioconductor environment v.3.5.1.

P values were generated using Student t tests (unpaired, two-tailed, at 95% confidence interval) or ANOVA when appropriate. Statistical significance is designated when P < 0.05.

To evaluate the impact of commensal dysbiosis on metastatic HR⁺ breast cancer, we used the poorly metastatic GFP⁺ BRPKp110 syngeneic model (23) of HR⁺ mammary cancer. Our goal was to determine whether preexisting commensal dysbiosis affects mammary tumor metastasis and progression. To that end, mice were orally gavaged with a broad-spectrum cocktail of antibiotics (25) for 14 days, a sufficient duration to enable outgrowth of antibiotic-resistant commensal species (27). Mice were then rested for four days following cessation of antibiotics to allow for dysbiotic reacquisition of commensal species (28), and mammary tumors were initiated by injecting tumor cells into the fourth abdominal mammary fat pad (Fig. 1A).

Bacterial composition was evaluated in the feces of mice on the day of tumor initiation using 16S rDNA sequencing. Compared with vehicle-treated mice, those receiving antibiotics showed shifts in bacterial communities at both the phylum and genus levels (Supplementary Fig. S1A). Using a Bray–Curtis dissimilarity measure, beta diversity between the two treatment groups differed significantly (Supplementary Fig. S1B). Antibiotic-treated mice had a significant reduction in community richness within the gut, as indicated by calculation of Shannon alpha diversity index (Supplementary Fig. S1C), as well as enlarged ceca (Supplementary Fig. S1D), demonstrating that antibiotic-treated mice had commensal dysbiosis prior to tumor initiation.

To evaluate the impact of antibiotic-driven gut dysbiosis in the breast tissue, we first performed histologic examination of the mammary gland at the time of tumor implantation. Blinded histopathologic scoring of the mammary tissues indicated that the antibiotic treatment resulted in minimal changes to tissue morphology (Supplementary Fig. S2A and S2B).

Metabolic output often reflects the status of the commensal ecosystem. To indirectly evaluate whether functional changes to the commensal ecosystem within the mammary tissue microenvironment occur in response to antibiotic treatment or tumor status, we compared metabolites in normal-adjacent mammary tissue from antibiotic- or nonantibiotic-treated mice, with or without a tumor 6 days after tumor cell inoculation. As a positive control, we collected feces from treated and nontreated mice. Similar to changes observed with 16S rDNA sequencing at day 0, global fecal metabolites significantly differed when compared by dysbiotic status, but were independent of tumor status (Supplementary Fig. S2C). On the contrary, metabolites within the mammary tissue remained unchanged despite dysbiosis or tumor status (Supplementary Fig. S2D). Thus, at the time of tumor implantation, histologic and metabolic changes within the mammary tissue were similar between treatment groups. Together, these data suggest that the mammary tissue is minimally affected by antibiotic treatment.

Using the poorly metastatic syngeneic BRPKp110 model, we evaluated tumor cell dissemination to lungs and axillary lymph nodes by flow cytometric quantitation of GFP⁺CD45⁻ mammary tumor cells (Supplementary Fig. S3A). Significantly more disseminated tumor cells were detected within the lungs (Fig. 1B and C), peripheral blood (Fig. 1D; Supplementary Fig. S3B), and tumor-draining axillary lymph nodes (Fig. 1E; Supplementary Fig. S3C) of dysbiotic mice. Importantly, although commensal bacteria have been shown to modulate estrogen levels, which could impact the estrus cycle and tumor growth kinetics of HR⁺ mammary tumors, we did not observe differences in primary tumor growth kinetics in mice with or without commensal dysbiosis (Fig. 1F). Estrus cycle monitoring between the last day of antibiotic treatment until the day of tumor implantation using vaginal lavage confirmed no substantial differences in cycle patterns (Supplementary Fig. S4A) or in the proportions of animals that were cycling, as defined by progressing from proestrus to estrus (Supplementary Fig. S4B). These data suggest that commensal dysbiosis influences tumor cell seeding to distal sites, independent of changes to the tissue microenvironment or estrus cycle.

We next evaluated whether preestablished commensal dysbiosis would impact tumor growth or metastatic dissemination in a more aggressive and metastatic tumor model. For these experiments, we used a tumor cell line that was derived from an advanced primary mammary tumor arising in the autochthonous MMTV-PyMT mouse model (24). Similar to what we observed using the BRPKp110 model, we found significantly greater dissemination of tumors into the lungs, independent of tumor volume (Fig. 1G and H), and tumors progressed with equal kinetics regardless of the dysbiotic status of the animals (Fig. 1I). Together, these results suggest that commensal dysbiosis has a sustained impact on the dissemination of HR⁺ mammary cancer, and that increased dissemination in mice with commensal dysbiosis is occurring independently of tumor growth kinetics.

We also tested the effects of commensal dysbiosis using a GFP⁺-inducible autochthonous model (22) that is established by targeting latent mutations with intraductal delivery of adenovirus-Cre. Using a similar experimental approach as depicted in Fig. 1A, despite variable tumor kinetics, no significant differences in primary tumor growth between dysbiotic and nondysbiotic groups were observed (Supplementary Fig. S5A). However, this variability makes it challenging to evaluate effects on tumor cell dissemination. To confirm whether dysbiosis-driven, host-intrinsic differences induce dissemination in the autochthonous model, we performed paired tumor transfers into dysbiotic or nondysbiotic recipient mice. Autochthonous mammary tumors were induced in wild-type mice without commensal dysbiosis, sterilely excised when they reached approximately 1 cm in diameter, and equal-weight fragments from the same tumor section were surgically implanted in the mammary tissue of recipient animals pretreated with antibiotics or vehicle gavage as described in Fig. 1A. After the implanted tumors grew to an advanced size, lungs and axillary lymph nodes from the recipient mice were examined for the presence of disseminated tumor cells. Similar to what we observed using the syngeneic models, lungs from dysbiotic mice showed significantly greater frequencies of disseminated tumor cells (Supplementary Fig. S5B and S5C), whereas there was a nonsignificant increase in GFP⁺ tumor cells in distal lymph nodes (Supplementary Fig. S5D). Together, these results confirm that commensal dysbiosis enhances tumor cell dissemination independently of primary tumor growth, and suggest that host-associated changes in response to commensal dysbiosis, not tumor-intrinsic differences in invasive capacity, enhance tumor dissemination.

To identify possible microenvironmental factors associated with increased dissemination in tumor-bearing mice with commensal dysbiosis, we first defined the immunologic changes occurring within the primary tumor microenvironment, specifically within the mammary tissue. Previous studies have demonstrated that inflammation within the mammary tissue associates with an increased risk for breast cancer (29). However, it is not well understood how this inflammation arises. The commensal microbiome is important for the maintenance of systemic and local immune homeostasis, giving rise to the possibility that commensal dysbiosis may result in a disruption of mammary tissue homeostasis and increased inflammation within the tissue. To examine the effects of commensal dysbiosis on the mammary gland during tumor progression, mammary tissues were harvested during early (day 12) and advanced (day 27) stages of tumor progression (Fig. 2A). Mammary glands from nontumor-bearing mice with or without dysbiosis were also evaluated in parallel. To normalize the systemic effects of hormones on mammary gland homeostasis and immune function, estrus cycles were synchronized in all animals prior to analysis using a modified Whitten effect (30).

Macrophages within the mammary gland associate with an increased risk of developing metastatic breast cancer (31). We therefore determined whether commensal dysbiosis influenced the frequencies and numbers of macrophages during early or advanced stages of mammary tumor progression. During early stages of tumor progression, the presence of a tumor was sufficient to significantly enhance myeloid accumulation into the normal-adjacent mammary gland (Fig. 2B). This effect persisted into advanced stages of tumor progression, resulting in a substantial increase in myeloid frequencies within the normal-adjacent mammary gland (Fig. 2C). However, commensal dysbiosis lead to a dramatic and significant increase in total accumulated myeloid cells within the normal-adjacent mammary glands at both early (Fig. 2B) and advanced (Fig. 2C) stages of tumor progression. Because tumors progressed with similar kinetics, these results suggest that commensal dysbiosis synergizes with a developing tumor to enhance myeloid cell infiltration within mammary tissue. Importantly, these data indicate that the effects of preestablished commensal dysbiosis on myeloid-driven inflammation within the mammary gland are sustained throughout tumor progression. Similar to published studies in humans (31), myeloid cells infiltrating into normal-adjacent mammary tissue in dysbiotic mice expressed high levels of the inflammatory mediators arginase-1 and IL6, specifically in the context of an early developing mammary tumor (Fig. 2D). However, during advanced stages of tumor progression, myeloid production of arginase-1 and IL6 was unchanged (Fig. 2E). These results demonstrate that commensal dysbiosis promotes the early recruitment of inflammatory myeloid cells into the mammary tissue microenvironment.

We next wanted to define signals that arise in response to commensal dysbiosis and could enhance the recruitment of inflammatory myeloid cells into the breast tissue. Therefore, mammary tissues were harvested prior to tumor initiation (pretumor; day 0) and during early (early tumor; day 12) stages of tumor progression. Similar to the experiment depicted in Fig. 2A, estrus cycles were synchronized prior to each experimental endpoint (Fig. 2F). We focused our analysis on three myeloid chemoattractants that also associate with adverse outcomes in breast cancer: CXCL10 (32, 33), CCL2 (34–36), and CXLC2 (37).

Compared with nondysbiotic mice, commensal dysbiosis resulted in a significant upregulation of both CXCL10 and CCL2 within mammary tissue prior to tumor initiation (pretumor mammary tissue; day 0), while CXCL2 was elevated but not significantly different (Fig. 2G). These data demonstrate that the systemic and/or local response to commensal dysbiosis culminates, at the very least, in low-level nonpathologic inflammation within the mammary gland. Although the mammary gland inflammation eventually subsided in nontumor-bearing dysbiotic mice, CXCL10, CCL2, and CXCL2 remained significantly elevated within normal-adjacent mammary glands in tumor-bearing mice with preestablished dysbiosis when compared with all other experimental groups (Fig. 2G). Together, these data demonstrate that preexisting commensal dysbiosis results in inflammation within the mammary gland that persists in the presence of a tumor during early stages of mammary tumor progression. In addition, these data suggest that cross-talk occurring between an inflamed tissue and a developing tumor amplifies the accumulation of myeloid cells into the microenvironment, and that the effects of this cross-talk are likely sustained throughout early and advanced stages of tumor progression.

In addition to promoting local inflammation within breast tissue, commensal dysbiosis also enhanced systemic inflammation. To examine the systemic impact of commensal dysbiosis, serum was collected prior to tumor initiation (pretumor serum; day 0) and at early (early tumor serum; day 12) and advanced (advanced tumor serum; day 27) stages of tumor progression. Serum was analyzed for the presence of GM-CSF, CCL2, and CXLC2, chemokines involved in myeloid cell generation and migration; and IL22 and IL23, cytokines that mediate inflammatory responses and that are associated with gut barrier function. We focused upon these two cytokines as expression of both IL22 and IL23 are enhanced in the serum of patients with breast cancer and have been associated with tumor progression and decreased overall survival (38, 39).

Interestingly, in response to commensal dysbiosis, all cytokines and chemokines tested were significantly increased in the serum of mice with established commensal dysbiosis prior to tumor initiation. While no significant differences were observed in serum from dysbiotic or nondysbiotic mice with or without early-stage mammary tumors, dysbiotic mice bearing advanced mammary tumors showed significantly increased serum CCL2 and a low but significant increase of IL23, while levels of GM-CSF, CXCL2, and IL22 were elevated but not significantly different (Fig. 3A). Importantly, because tumors progress at similar rates regardless of dysbiotic status, enhanced serum cytokines were not due to differences in primary tumor burden.

Macrophages are one of the most abundant cell types within the breast tumor microenvironment (40) and are a significant prognostic indicator of reduced survival for patients diagnosed with HR⁺ breast cancer (41). Confirming this, the majority of myeloid infiltrates within the mammary tumor microenvironment at early and advanced stages of tumor progression were M2-like macrophages, based upon CD206 expression (Fig. 3B). Similar to what we observed in normal-adjacent mammary glands, preestablished commensal dysbiosis resulted in enhanced total myeloid cell accumulation within both early and advanced mammary tumors. Importantly, advanced tumors from dysbiotic mice had significantly increased numbers of infiltrating tumor-promoting M2-like macrophages as compared with nondysbiotic controls with equal tumor burden. Together, these data demonstrate that the systemic expression of inflammatory mediators is increased in dysbiotic tumor–bearing mice, enhancing myeloid recruitment into mammary tumors.

Enhanced stromal density, or dense breasts, is a well-established risk factor for developing metastatic breast cancer (42) and tumor-promoting inflammation in the breast (43). To determine whether commensal dysbiosis enhanced fibrosis within the tissue or tumor microenvironment or at systemic sites of metastatic dissemination, we stained tissues from dysbiotic and nondysbiotic mice with PicroSirius Red to visualize collagen deposition. We found that preestablished commensal dysbiosis resulted in significant enhancement of collagen deposition within the normal-adjacent mammary gland (Fig. 4A) and tumors (Fig. 4B) of advanced tumor-bearing mice. We also observed a slight, but significant, increase in collagen accumulation within the lungs of advanced tumor-bearing dysbiotic mice (Fig. 4C). Together, these data demonstrate that enhanced fibrosis at both local and distal sites are a long-term consequence of commensal dysbiosis during mammary cancer.

The mammary gland is considered a mucosal surface with mammary-specific commensal microorganisms. Although significant metabolic changes within the mammary tissue of dysbiotic or nondysbiotic mice were not observed (Supplementary Fig. S2C), suggestive of low to minimal impact on the commensal population residing within the mammary gland, the systemic effects of the antibiotic cocktail could theoretically contribute to enhanced tumor dissemination. Metronidazole is highly absorbed from the gut, whereas ampicillin has moderate systemic bioavailability after oral administration. To determine whether tumor dissemination and tissue inflammation arise due to a direct effect of antibiotics upon the tissue microflora or due to distal changes within the gut, commensal dysbiosis was induced using antibiotics that have no to minimal systemic bioavailability (44, 45). This nonabsorbable antibiotic cocktail consisting of bacitracin and neomycin is well-established to specifically deplete bacteria within the gut (46).

Commensal dysbiosis initiated using nonabsorbable antibiotics recapitulated the changes within the tissue and tumor microenvironments similar to what was observed using absorbable antibiotics. Myeloid cells infiltrating into normal-adjacent mammary glands of dysbiotic mice produced IL6 or arginase-1 when commensal dysbiosis was established using absorbable or nonabsorbable antibiotics (Fig. 5A and B). Early myeloid cell recruitment into both mammary glands (Fig. 5C) and mammary tumors (Fig. 5D) was also enhanced in dysbiotic mice using both absorbable and nonabsorbable antibiotics. During advanced tumor progression, fibrosis was enhanced both in mammary tumors (Fig. 6A and B) and normal-adjacent mammary glands (Fig. 6C and D) from dysbiotic mice using both absorbable and nonabsorbable antibiotics.

Importantly, preestablished dysbiosis using nonabsorbable antibiotics resulted in significant dissemination of tumor cells both into the peripheral blood (Fig. 6E; Supplementary Fig. S6A) and to the lungs (Fig. 6F; Supplementary Fig. S6B), similar to what was observed in animals treated with absorbable antibiotics. Although it cannot be ruled out that nonabsorbable antibiotics indirectly affect the tissue commensal microorganisms or enhance bacterial translocation from the gut to distal sites such as the mammary tissue, these data demonstrate that targeted disruption of the gut microbiome is sufficient to initiate enhanced metastatic dissemination during breast cancer.

To determine whether a dysbiotic microbiome is sufficient to enhance mammary tumor cell dissemination, an FMT approach was used. Cecal contents from dysbiotic and nondysbiotic donor mice were collected after 14 days of antibiotic gavage followed by a 4-day rest, to recapitulate a dysbiotic microbiome at the time of tumor initiation. Cecal contents were transferred over 3 consecutive days into antibiotic-sterilized recipient SPF mice. BRPKp110 mammary tumors were initiated 7 days following FMT gavage (Fig. 7A).

FMT of a dysbiotic or nondysbiotic microbiome did not impact tumor kinetics (Supplementary Fig. S7A). Mice receiving FMT of dysbiotic cecal contents mirrored the phenotype of dysbiotic mice that had received extended gavage of absorbable or nonabsorbable antibiotics. Specifically, mice receiving dysbiotic FMT showed enhanced infiltration of inflammatory myeloid cells into early mammary tissue (Fig. 7B–D) and increased myeloid cell accumulation into tumors (Fig. 7E). Similar effects were observed during advanced stages of tumor progression, both within the mammary tissue (Fig. 7F) and advanced tumors (Fig. 7G). Mammary tissue (Fig. 7H) and tumors (Fig. 7I) from mice receiving dysbiotic FMT also showed enhanced tissue fibrosis. Importantly, mice receiving dysbiotic FMT, but not “normal” FMT, had significantly enhanced tumor cell dissemination into the peripheral blood (Fig. 7J; Supplementary Fig. S7B), within the lungs (Fig. 7K; Supplementary Fig. S7C), and to the distal axillary lymph nodes (Fig. 7L; Supplementary Fig. S7D). These data demonstrate that a dysbiotic microbiome, but not a normal microbiome, is sufficient to enhance metastatic dissemination during mammary cancer.

Our study demonstrates that commensal dysbiosis occurring prior to mammary cancer initiation is a host-intrinsic factor that contributes to increased tumor cell dissemination and metastatic seeding, leading to adverse outcomes. Our data support the idea that inflammation within the normal tissue microenvironment, in this case driven by a systemic response to commensal dysbiosis, precedes the development of aggressive breast cancer. In our model of HR⁺ mammary cancer, although primary tumor growth remained unchanged, mice with preestablished commensal dysbiosis showed significantly enhanced tumor cell dissemination to the peripheral blood, lungs, and distal lymph nodes. These seemingly paradoxical data suggest that host factors associated with the growth of HR⁺ tumors are distinct from those that contribute to metastatic dissemination. More investigation into myeloid function and tumor immune surveillance are necessary to understand these differences.

To understand the role of dysbiosis in driving metastatic dissemination, several key factors known to be involved in promoting metastasis and adverse outcome in HR⁺ breast cancer were evaluated. The syngeneic model of commensal dysbiosis used in this study recapitulates features of breast cancer that have been independently associated with adverse outcome: enhanced inflammation both systemically and within the tissue environment; increased macrophage infiltration within tumors and normal-adjacent mammary tissue; and enhanced fibrosis within mammary tissue, tumors, and lungs. Importantly, enhanced dissemination as a result of commensal dysbiosis was observed regardless of the metastatic potential of the HR⁺ mammary tumor models used in this study. Therefore, our data demonstrate that a disruption in commensal homeostasis results in the establishment of a tissue and/or tumor microenvironment that favors enhanced metastatic dissemination.

During commensal dysbiosis, chemokines involved in myeloid cell chemotaxis and recruitment were enhanced both systemically in serum and within the mammary tissue microenvironment of both tumor- and nontumor-bearing dysbiotic mice. Whereas inflammation eventually resolved in dysbiotic nontumor-bearing animals, the presence of a tumor in dysbiotic mice resulted in a persistent elevation of myeloid chemoattractants within the tissue microenvironment. Thus, it is likely that cross-talk occurs between the tumor and tissue microenvironments, leading to sustained inflammation and possible myeloid recruitment into these tissues in mice with commensal dysbiosis. Indeed, significantly more myeloid cells–particularly M2-like macrophages–infiltrated into mammary tumors and normal-adjacent mammary glands during early and advanced stages of tumor progression. Macrophages are one of the most abundant cell types within the breast tumor microenvironment. In patients, increased frequencies of macrophages within stroma and tumors associate with significantly reduced survival in HR⁺ breast cancer (47). Macrophages increase angiogenesis, evasion of antitumor immunity, and promote metastatic dissemination and seeding of cells into distal sites. Thus, the accumulation of M2 macrophages likely plays a direct role in dysbiosis-dependent enhancement of tumor cell dissemination.

Inflammation within the mammary gland also increases susceptibility to developing breast cancer, as women with benign breast disease who later develop breast cancer have significantly greater proportions of inflammatory macrophages present within the tissue compared with women who do not go on to develop cancer (21, 31). Similar to these human studies, in our model of HR⁺ mammary cancer, significantly higher numbers of inflammatory macrophages producing IL6 or arginase-1 were observed within normal-adjacent mammary tissue of mice with commensal dysbiosis during early stages of tumor development. Nonabsorbable antibiotics that specifically target gut commensals and FMT of dysbiotic cecal contents also increased macrophage infiltration and inflammation within the mammary gland. Together, our studies demonstrate that commensal dysbiosis leads to a disruption in immune homeostasis and increased inflammation within both tumor- and nontumor-associated tissue.

Supporting the idea that commensal dysbiosis contributes to the evolution of a more aggressive and high-grade disease, the induction of commensal dysbiosis using either absorbable or nonabsorbable antibiotics or FMT of dysbiotic cecal contents resulted in significant fibrosis within both tumors and the tissue microenvironment. Collagen accumulation and stiffening within the extracellular matrix of tumors and within the tumor microenvironment is associated with increased dissemination of tumor cells into distal lymph nodes and lungs (48). These results suggest that a disruption of the gut microflora, not mammary tissue flora, prior to tumor initiation is sufficient to facilitate collagen accumulation within the tissue and tumor microenvironments. Furthermore, these data suggest that in HR⁺ breast cancer, commensal dysbiosis leads to systemic changes that correspond with more invasive disease.

Factors that influence the development of dysbiosis are increasing worldwide, including rates of obesity and antibiotic use, highlighting the importance of elucidating the mechanisms by which commensal dysbiosis may promote cancer development and progression. While other studies have examined the impact of commensal microorganisms and/or dysbiosis in response to cancer treatments (49, 50), this study is the first, to our knowledge, to directly assess whether commensal dysbiosis that is present at the time of tumor initiation impacts tumor aggressiveness and dissemination. Future studies will investigate the relationship between commensal microbe composition and fibrosis and myeloid cell invasion within premalignant mammary tissue. In addition, it will be important to evaluate whether commensal dysbiosis during triple-negative or HER2⁺ breast cancer results in similar outcomes. Overall, our study identifies commensal dysbiosis as a preexisting, host-intrinsic factor that promotes adverse outcomes in HR⁺ breast cancer. Based upon this study, we speculate that commensal dysbiosis could thus serve as a potential biomarker or as a therapeutic target to reduce tumor-promoting inflammation within the tissue microenvironment.

No potential conflicts of interest were disclosed.

Conception and design: C. Buchta Rosean, M.R. Rutkowski

Development of methodology: C. Buchta Rosean, R.R. Bostic, J.C.M. Ferey, M.R. Rutkowski

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Buchta Rosean, R.R. Bostic, J.C.M. Ferey, T.-Y. Feng, F.N. Azar, P.D. Bos, M.R. Rutkowski

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Buchta Rosean, R.R. Bostic, T.-Y. Feng, F.N. Azar, K.S. Tung, M.G. Dozmorov, E. Smirnova, P.D. Bos, M.R. Rutkowski

Writing, review, and/or revision of the manuscript: C. Buchta Rosean, R.R. Bostic, J.C.M. Ferey, T.-Y. Feng, K.S. Tung, M.G. Dozmorov, E. Smirnova, P.D. Bos, M.R. Rutkowski

Study supervision: M.R. Rutkowski

This study was supported by a Susan G. Komen Career Catalyst award CCR17483602 (to M.R. Rutkowski), IRG-17-097-31 (ACS; to M.R. Rutkowski), the University of Virginia Cancer Center and support from NCI Cancer Center Support grant P30CA44570 as startup funds (to M.R. Rutkowski), and training grant 5T32AI007496-24 (to C. Buchta Rosean). The authors would like to acknowledge the University of Virginia Research Histology, Advanced Microscopy, Biorepository and Tissue Research Facility, Molecular Imaging, and Biomolecular Analysis Core Facilities, as well as the Carter Immunology Center. In addition, we would like to thank Jose R. Conejo-Garcia for the BRPKp110 HR⁺ breast tumor cell line and Steven N. Fiering for the L-Stop-L-KRas^G12Dp53^flx/flxL-Stop-L-Myristoylated p110α−GFP⁺ mice.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.