A Procarcinogenic Colon Microbe Promotes Breast Tumorigenesis and Metastatic Progression and Concomitantly Activates Notch and β-Catenin Axes

The abundance of commensal microorganisms colonizing the human body can be appreciated by the fact that the number of microbial cells living within and on the human body is roughly equal to the total number of human cells (1). Though accounting only for ∼2% to 7% of biomass owing to the minuscule size of microbes, the human microbiome encodes for 100-fold more genes than the human genome, indicating an important role in human health (2). Microbiota and host maintain a dynamic equilibrium referred to as eubiosis that actively influences many physiologic processes and is generally beneficial to the host. However, a state of disequilibrium or dysbiosis may evolve, contributing to various disease states. A major advance in the microbiome field was achieved in the last decade with the completion of the Human Microbiome Project (HMP), which identified, characterized, and elucidated the role of microbes in five major sites, including the nasal passages, oral cavity, skin, gastrointestinal tract, and urogenital tract (3–5). More recent developments show the existence of microbiota in other body sites initially considered “sterile,” such as bladder, lung, endometrium, prostate, and breast (6–11).

Breast cancer is a heterogeneous disease with multiple subtypes, and, interestingly, microbial signatures may differ between the subtypes. Triple-negative breast cancer (TNBC) and triple-positive breast cancer have distinct signatures that differ from estrogen receptor–positive and HER2-positive breast cancer, which share similar microbial profiles (12). Two distinct microbial signatures are proposed from a screening of ∼100 TNBC samples, suggesting a possibility of further segregating TNBC subtype based on associated microorganisms (13). Comparison of breast tumor tissue and paired normal adjacent tissue highlights distinct alterations in bacterial species enriched in breast tumor tissues, demonstrating that the total bacterial DNA load is lower in tumor tissue in contrast to normal breast (14). Bacterial species having the ability to induce DNA double-strand breaks are more abundant in patients with breast cancer compared with the breast tissue of healthy subjects, suggesting a possibility of DNA damage leading to chromosomal aberrations (15). Not only is the breast microbiota different between normal healthy tissue and tumor tissue, an enrichment of low abundant taxa, including genera Hydrogenophaga, Atopobium, Fusobacterium, and Gluconacetobacter, is observed in malignant disease in comparison with benign disease, showing how malignancy associates with unique microbial signature (16). Furthermore, the presence of microbiota is also reported in nipple aspirate fluid (NAF), showing that unique microbes inhabit the ductal system of the human breast, and, interestingly, community composition differs significantly between the NAF of patients with breast cancer and healthy women (17). Diversity in composition of the breast microbiota (18, 19) indicates that microbial dysbiosis may play an important role in breast cancer growth as well as metastatic progression.

In addition to breast microbiota, some studies have shown that gut microbiota may also influence breast cancer. Analysis of fecal microbiota shows that postmenopausal women with breast cancer harbor compositionally different gut microbiota than healthy volunteers (20) and exhibit enrichment of several bacterial species (21). Bacteroides fragilis (B. fragilis) is a commonly found colon colonizer (22), and individuals can be asymptomatically colonized by enterotoxigenic Bacteroides fragilis (ETBF; ref. 23) whose virulence is attributed to a 20-kDa zinc metalloprotease toxin termed the B. fragilis toxin (BFT; ref. 24). B. fragilis forms a small fraction of total gut bacteria, estimated at about ∼0.1% to 0.5% (22), but is regarded as an important symbiont that can function as a potent pathogen and a determinant of the structure of microbial communities based on its secretory products (25). These rogue symbionts, in addition to causing diarrhea and inflammatory bowel disease, are capable of inducing oncogenic transformation in the gut mucosa, leading to formation of spontaneous tumors (22, 23, 25, 26). Owing to its unique virulence traits, ETBF has been proposed as an “alpha bug” capable of direct pro-oncogenic actions, remodeling the bacterial community to enhance its own induction as well as selective “crowding out” of protective microbial species (22).

We identified the presence of B. fragilis in breast tumor tissue using meta-analyses of breast cancer microbiome studies, forming the rationale of our work that the growth and progression of breast cancer may be affected by the pro-oncogenic actions of ETBF. Consequently, we aim to decipher (i) whether gut or intraductal colonization with ETBF affects normal breast tissue, (ii) the effect of BFT exposure on breast epithelial cells and the underlying molecular networks, and (iii) whether ETBF colonization of gut or breast ducts can aid breast cancer growth and metastasis. Our results demonstrate both the distant (via gut colonization) and local (via intraductal colonization) effects of ETBF on breast and involvement of the β-catenin and Notch1 pathways in mediating the oncogenic effects of BFT, thereby proposing ETBF as a potential pathogen in breast carcinogenesis.

We started this investigation by meta-analysis of clinical data examining the microbiota of the breast, specifically selecting studies that compared differences in microbial community composition between benign breast tumors and malignant breast cancer (PRJNA335375, EBI-ENA) and NAFs of breast cancer survivors and healthy volunteers (SRP071608, NCBI-SRA). 16S rRNA gene sequencing data were retrieved from the sequence read archives and analyzed using One Codex. We observed that B. fragilis is consistently detected in all the breast tissue samples from benign and malignant breast cancer as well as nipple aspirate fluids (Fig. 1A). Since the 1970s, B. fragilis has been known to be the most invasive of the colon anaerobes. In other words, B. fragilis is the anaerobe most likely to enter the bloodstream. Since that time, the pathogenicity of B. fragilis has expanded even further by the discovery of toxin-producing strains of B. fragilis (ETBF). ETBF is known to trigger a colon mucosal cascade resulting in colitis and colon neoplasia in mouse models and has also been demonstrated to be common in human populations. In these murine models, ETBF has been shown to modulate systemic immune responses including increasing IL17, a known contributor to multiple types of cancer. Collectively, identification of B. fragilis in the breast microbiota by our metagenomic analyses combined with our understanding of the role B. fragilis and ETBF play in gut and systemic immune responses led us to test the hypothesis that B. fragilis, and specifically ETBF, contributes to breast oncogenesis. To examine the impact of ductal dysbiosis on normal mammary tissue, we initially used an intraductal approach to colonize mammary ducts of mice with ETBF or its isogenic mutant with an in-frame deletion of the chromosomal Bft gene (086Mut; Fig. 1B). Intraductal injection of mouse teats with 10⁸ colony-forming units (CFU) of ETBF or 086Mut resulted in mammary gland colonization (Fig. 1C). Mammary glands of mice harboring ETBF infection showed the presence of BFT, whereas no BFT was detected in 086Mut-infected mice (Fig. 1D and E). Marked differences in mammary tissue architecture were observed in the ETBF group exhibiting widespread local inflammation and tissue fibrosis with increased epithelial cell proliferation as evident by Ki-67 and proliferating cell nuclear antigen (PCNA) staining, higher T-cell infiltration indicated by CD3 staining, and significantly altered expression of pan-keratin in comparison with 086Mut and sham controls (Fig. 1F and G).

ETBF is a gut commensal in some individuals and colon disease–associated (e.g., diarrhea, colitis, tumorigenesis) in others (22, 23, 25, 26), but extra-colonic disease links are unknown. Hence, we queried whether gut infection with ETBF is capable of inducing distant effects on mammary gland epithelium. C57BL/6 mice orally infected with ETBF developed brief diarrhea by 2 to 3 days that resolved by 4 to 5 days after colonization with high-level persistent gut colonization (≥ × 10⁹ CFU). We observed the presence of ETBF as well as BFT in mammary glands of mice harboring gut ETBF infection, whereas 086Mut and sham-control mice exhibited none (Fig. 2A and B). ETBF-gut-colonized mice also showed circulating serum levels of BFT, which peaked at week 1 followed by a slight decline at week 3, whereas no BFT was detected in sham-control mice (Fig. 2C). Intriguingly, mammary glands in ETBF gut-colonized mice showed significantly enlarged terminal end buds, with more prominent bifurcations indicating a proliferative and inflammatory response (Fig. 2D). Blinded scoring of focal hyperplasia (score range, 0–3) showed that all of the mice in the ETBF group scored 2 or 3 at week 3 after infection, whereas none of the sham-control mice scored 2 or 3 (Fig. 2D). Only ETBF gut-colonized mice showed a marked increase in thickening of the breast duct lining and hyperproliferation of breast epithelium. Blinded scoring based on the thickness of the duct lining and cellularity of the inner margins of the ducts (score range, 0–3) classified most of the mice harboring gut ETBF infection in the score 3 group, and all of the 086Mut and sham-control mice showed normal ducts (score 0–1; Fig. 2E). Histopathology of mammary glands from ETBF gut-colonized mice confirmed increases in stromal infiltration, collagen deposition, hyperplasia, and T-cell infiltration as evident by trichrome, Ki-67, pan-keratin, and CD3 staining compared with 086Mut and sham-control mice (Fig. 2F). Overall, these data indicate the presence of B. fragilis in cancerous breast and show that ETBF is capable of exerting pathogenic effects on mammary glands locally as well as remotely plausibly via BFT.

Virulence of ETBF is ascribed to a 20-kDa zinc metalloprotease termed the BFT (24). Thus, we examined the effect of BFT exposure on normal breast epithelial cells and breast cancer cells. MCF10A and MCF7 cells were treated with varying concentrations of BFT, ranging from 25 to 150 ng/mL (∼1–7 nmol/L), and structural changes were examined. Cells exhibited membrane blebbing and increased intracellular spacing in response to 100 ng/mL (5 nmol/L) BFT, which was deemed the optimum concentration for exerting a morphologic effect on breast cells (Supplementary Fig. S1A). Indeed, treatment with 100 ng/mL (5 nmol/L) BFT exhibited a temporal decrease of the tight junction protein E-cadherin, consistent with structural changes in the cells (Supplementary Fig. S1B). Immunofluorescence analysis with an antibody specific for the intracellular domain of E-cadherin showed a loss of membrane-bound E-cadherin upon BFT treatment (Supplementary Fig. S1C). Rhodamine–phalloidin staining of F-actin showed that BFT-treated cells underwent a gradual but prominent cytoskeletal organization, increased intracellular separation, membrane blebbing, and an increase in F-actin stress fibers, whereas the control cells showed uniform cuboidal, smooth-edged structures (Supplementary Fig. S2A). Next, we examined the effect of BFT treatment on cell viability and clonogenicity and observed that MCF7 and MCF10A cells treated with 5 nmol/L BFT did not show any significant alterations in cell viability or colony formation (Supplementary Fig. S2B and S2C). A closer examination of colonies formed from the BFT-treated cells showed spindle-shaped cells emanating from the colonies, indicating a migratory phenotype, a distinctive feature not observed in control colonies (Fig. 3A). To further elucidate these phenomena, MCF7 cells were exposed to 5 nmol/L BFT for 48 hours and subjected to RNA-sequencing (RNA-seq) analysis. A differential expression analysis was conducted to characterize global differences in RNA transcript levels induced upon BFT exposure. Genes associated with cytoskeletal remodeling, cell movement, and migration are highlighted in a volcano plot (in green; Fig. 3B), and genes specifically associated with cell movement of breast cancer cell lines are presented in a functional annotation graph (Fig. 3C; Supplementary Fig. S2D). These data point to a possibility that BFT exposure might impart a migratory and invasive phenotype to breast cells.

To query if BFT exposure truly leads to a migratory/invasive phenotype, single-cell migration was investigated using a microfluidic device where cells were allowed to migrate across microfluidic channels of different widths ranging from 3 to 50 μm toward a gradient of EGF used as a chemoattractant (27, 28). Following a brief 12-hour exposure to 5 nmol/L BFT, MCF7 cells migrated with significantly higher velocity and persistence through the microchannels compared with the control cells (Fig. 3D–F; Supplementary Fig. S3A). MCF10A cells, owing to their larger size and tendency to move in clusters, did not migrate far along in the microchannels as single cells, and their velocity and persistence could not be calculated, but visually, BFT-exposed cells seemed to migrate more than control MCF10A cells (Supplementary Fig. S3A). Rapid wound closure within 48 hours was observed for both cell lines in a scratch-migration assay. Control MCF7 cells migrated at a speed of 0.212 μm/h, whereas BFT-treated cells migrated at a speed of 0.657 μm/h. In the case of MCF10A, the migration rate increased from 0.709 to 1.34 μm/h in response to BFT treatment (Supplementary Fig. S3B and S3C). BFT-exposed MCF7 and MCF10A cells showed increased invasion potential (Fig. 3G). Cell-adhesion assay suggested a reduced adhesion to collagen I, as significantly fewer BFT-exposed MCF10A and MCF7 cells attached to collagen I–coated plates (Fig. 3H and I). We further validated our observations using a 3-D spheroid migration assay. When spheroids of MCF7 and MCF10A cells were treated with 5 nmol/L BFT, increased migration of cells from the spheroids was observed (Supplementary Fig. S3D). Cells migrated an average distance of 19.55 ± 3.051 μm in control MCF7 spheroids (n = 3), which increased to 24.81 ± 3.009 μm in BFT-treated MCF7 spheroids (n = 3) within 48 hours. In the case of MCF10A spheroids, the distance traversed increased from 24.39 μm to 44.39 ± 1.932 μm in BFT-treated spheroids (Supplementary Fig. S3E). Multiple human breast cancer cell lines, upon exposure to BFT, showed increased invasion and migration potential (Supplementary Fig. S4A–S4D). BFT treatment also decreased adhesion and increased migration and invasion of 4T1 mouse mammary cancer cells (Supplementary Fig. S5A–S5D). These results suggest that BFT treatment induces cytoskeletal reorganization in the cells; consequently, they exhibit increased migration and invasion potential.

In addition to an enrichment of genes associated with a migratory phenotype, RNA-seq analysis revealed an upregulation of an embryonic stem cell pluripotency pathway in BFT-exposed MCF7 cells (Supplementary Fig. S6A). Higher expression of a set of 38 embryonic stemness-associated genes was observed in BFT-exposed MCF7 cells in comparison with control cells (Supplementary Fig. S6B). Encouraged by the RNA-seq results implicating an enhancement of stemness in BFT-exposed cells, we examined the self-renewal potential of BFT-treated MCF7 and MCF10A cells using mammosphere formation assay. BFT-treated cells formed a higher number of mammospheres in both solid and liquid media (Supplementary Fig. S6C and S6D). Additionally, the liquid mammospheres from BFT-exposed cells continued to form higher numbers of secondary and tertiary mammospheres in two subsequent generations (Supplementary Fig. S6C). Collectively, these results show that breast cells undergo distinct morphologic and molecular changes, supporting an invasive and migratory phenotype with enhanced stemness potential upon BFT exposure.

Encouraged by our in vitro findings, we investigated whether BFT exposure affects tumorigenicity of breast cancer cells. MCF7 and MCF10A cells were pretreated with 5 nmol/L BFT for 72 hours followed by mammary gland implantation in NOD/SCID mice, and tumor incidence and progression was monitored for 7 weeks. More rapid tumor progression was observed in the BFT-pretreated group compared with the control group (Fig. 4A). To our surprise, starting at week 3, tumors in the BFT-pretreated group extended locally, forming multifocal tumors resembling local metastases (Fig. 4A). Tumors were dissociated into single cells and evaluated for their functional potential pertaining to migration, invasion, adhesion, and mammosphere formation. As expected, tumor cells from the BFT-pretreated group exhibited higher invasion and migration potential with reduced adhesion (Fig. 4B). Mammosphere formation potential was also elevated in tumor cells from the BFT-pretreated group, a characteristic sustained for three generations forming a higher number of primary, secondary, and tertiary mammospheres (Supplementary Fig. S7A and S7B). Primary and tertiary mammospheres from the BFT-pretreated group showed elevated expression of Oct4, NANOG, and Sox2 compared with control mammospheres (Supplementary Fig. S7C). Histology of tumor sections surprisingly revealed widespread fibrosis and stromal infiltration in tumors formed by BFT-pretreated MCF7 cells. Masson trichrome staining showed regions rich in connective tissue. IHC staining showed c-Myc-positive nuclei and denser Ki-67 nuclear staining in BFT-pretreated tumors, indicating higher proliferation. Also, these tumors were more richly vascularized as confirmed by enhanced CD31 staining (Fig. 4C; Supplementary Fig. S7D). BFT-pretreated MCF10A cells did not form tumors because they are nontransformed cells. However, BFT-pretreated MCF10A-Kras cells formed more aggressive tumors compared with control cells (Supplementary Fig. S8A–S8C). Dissociated tumor cells from the BFT-pretreated MCF10A-KRAS group exhibited significantly higher invasion and migration, and lower adhesion potential compared with control group, qualities similar to tumor cells from the BFT-pretreated MCF7 group (Supplementary Fig. S8D–S8F). Mammosphere formation potential was also elevated in tumor cells from the BFT-pretreated MCF10A-KRAS group, as exhibited by the formation of a higher number of primary, secondary, and tertiary mammospheres and elevated expression of OCT4, NANOG, and SOX2 compared with mammospheres generated from the control group (Supplementary Fig. S8G–S8I).

The formation of multifocal tumors by BFT-pretreated MCF7 cells and the fact that tumor-dissociated cells continue to form higher numbers of mammospheres through three generations prompted us to investigate if these tumors possessed increased tumor-initiating cells. Tumors formed by BFT-pretreated MCF7 cells were excised and dissociated to single cells followed by secondary transplants in limiting dilution (Fig. 4D). Monitoring of tumor incidence and progression showed that the BFT-pretreated MCF7 tumor group formed larger secondary tumors with a shorter tumor-free survival in comparison with the control MCF7 tumor group (Fig. 4E and F). The BFT-pretreated MCF7 tumor group exhibited significantly increased tumor-initiating frequencies when transplanted into secondary hosts at limiting dilutions. The frequency of breast tumor-initiating cells in the BFT-pretreated MCF7 tumor group was determined to be 1 in 38,481 compared with 1 in 151,950 in the control group at week 4 and 1 in 11,644 compared with 1 in 35,573 at week 6 (Fig. 4G). Tumor cells dissociated from the secondary tumors from the BFT-pretreated group retained higher matrigel invasion, spheroid migration, and mammosphere formation along with reduced adhesion characteristics (Supplementary Fig. S9A–S9C). Elevated expression of Oct4 and Nanog was also observed in primary and tertiary mammospheres formed with the tumor cells dissociated from the secondary tumors from the BFT-pretreated MCF7 group (Supplementary Fig. S10A–S10C). To uncover the underlying molecular changes, the secondary tumors formed in the in vivo limiting dilution assay were subjected to RNA-seq analysis. Interestingly, an enrichment of stemness markers was observed in the BFT-pretreated group (Supplementary Fig. S10D). In addition, the BFT-pretreated group showed higher expression of genes associated with cell invasion potential, cell movement of tumor cell lines, and cell movement (Supplementary Figs. S11–S13). Together, these data provide the intriguing notion that a 72-hour-long exposure of breast cancer cells to BFT is sufficient to induce morphologic and molecular changes that impart a highly migratory, invasive, and stemness-rich phenotype to MCF7 cells that is maintained through in vitro and in vivo generations lasting up to 13 weeks.

We next asked which signaling pathways underlie the biological effects of BFT. To address this query, we further analyzed the RNA-seq data obtained from the secondary tumors from the BFT-pretreated MCF7 group and the control group. Differential expression analysis of the global differences in RNA transcript levels induced in the BFT-pretreated versus control group revealed a significant upregulation of the β-catenin and Notch1 pathways. Genes associated with β-catenin (marked in blue) and Notch1 (marked in pink) pathways are highlighted in the volcano plot (Fig. 5A), and genes specifically associated with the Wnt–β-catenin pathway are presented in a functional annotation graph (Supplementary Fig. S14).

Further exploration of the direct involvement of β-catenin in BFT function showed increased expression of β-catenin in breast cancer cells treated with 5 nmol/L BFT (Fig. 5B and C). Dephosphorylation of β-catenin prevents its ubiquitination and subsequent proteasomal degradation, leading to its cytoplasmic accumulation and increased nuclear localization enabling transcriptional activity (29). BFT exposure resulted in higher expression of total β-catenin protein and a reduction in phospho-β-catenin levels (Fig. 5C). Increased levels of nuclear β-catenin and a decrease in cytoplasmic levels of β-catenin were observed in BFT-treated MCF7 cells, whereas MCF10A cells exhibited an increase in nuclear as well as cytoplasmic β-catenin (Fig. 5D). Immunofluorescence analysis confirmed nuclear accumulation of β-catenin upon BFT treatment in MCF7 cells (Fig. 5E). Increased expression of β-catenin–responsive genes, including SLUG, c-MYC, Jagged, and TWIST, was noted in BFT-treated MCF7 and MCF10A cells (Fig. 5F and G). Further, IHC analysis of tumors from the BFT-pretreated MCF7 group compared with the control group showed a higher expression of β-catenin in the BFT group (Fig. 5H; Supplementary Fig. S15A). These results explicitly show that BFT activates the β-catenin pathway in breast cancer and breast epithelial cells (Fig. 5I). Because we observed an enrichment of Notch1 pathway genes in the RNA-seq of secondary tumors from the BFT-pretreated MCF7 group, we next examined the direct effect of BFT on Notch1 in breast cells. Notch receptors are membrane-bound receptors with a cytoplasmic region or Notch intracellular domain (NICD). NICD translocates to the nucleus upon activation and mediates transcriptional activity (30). We found that BFT treatment indeed increased NICD expression in a temporal manner in MCF7 and MCF10A cells and increased its nuclear localization (Fig. 6A and B). IHC analysis of tumors from the BFT-pretreated MCF7 group and the control group showed a higher expression of NICD in the BFT group (Supplementary Fig. S15B). BFT increased the expression of NICD and β-catenin and reduced the level of phospho-β-catenin in 4T1 cells, whereas no significant effect was observed with biologically inactive mutant BFT-H352Y (Supplementary Fig. S15C). BFT-treated cells also showed elevated expression of NICD-responsive genes, including p21/Waf and Hes1 (Fig. 6C). We then evaluated the interdependence of NICD and β-catenin in BFT-treated cells. As evident in the immunofluorescence analysis, BFT-treated cells showed increased nuclear accumulation of β-catenin and NICD (Supplementary Fig. S16A and S16B). Cotreatment of cells with BFT and the Notch inhibitor DAPT (a γ-secretase inhibitor) resulted in lower nuclear accumulation of β-catenin, whereas the DAPT-only group showed cytoplasmic localization of β-catenin (Supplementary Fig. S16A). Breast cancer cells cotreated with BFT and the β-catenin inhibitor ICG001 showed reduced nuclear accumulation of NICD, and the ICG001-alone group showed cytoplasmic localization of NICD (Supplementary Fig. S16B). Further evaluation of the functional importance of these pathways in the context of BFT treatment showed that a single treatment with DAPT and ICG001 could significantly reduce BFT-mediated elevated invasion and migration of cells, whereas cells treated with a combination of DAPT and ICG001 further improved abrogation of BFT effect (Supplementary Fig. S16C–S16E).

Examining the in vivo relevance of our in vitro findings, MCF7 cells were exposed to BFT and treated with DAPT and/or ICG001 in vitro followed by mammary gland implantation in mice. BFT-exposed MCF7 cells formed significantly larger tumors (4-fold) than control cells. Reduced tumor growth was observed in BFT-exposed cells treated with DAPT or ICG001 alone whereas cotreatment with DAPT and ICG001 led to even more inhibition. DAPT and ICG001-treated control cells also showed tumor inhibition (Fig. 6D). Tumor-dissociated cells, examined for the presence of stemness marker CD49f, showed higher expression (60.3%) in the BFT-alone group in comparison with control (23.5%), whereas BFT-exposed cells with DAPT and/or ICG001 treatment showed lower CD49f expression (19.6%, 12.6%, and 21%; ref. Fig. 6E). In vitro treatment with DAPT and/or ICG001 could inhibit BFT's tumorigenic and stemness effect on breast cancer cells. Next, we examined whether in vivo treatment with DAPT and/or ICG001 could inhibit tumors formed by BFT-exposed breast cancer cells. Tumors formed with BFT-exposed MCF7 cells were significantly larger (2–4-fold) than control tumors. Mice treated with ICG001 achieved a 58.5% regression whereas DAPT reduced the tumor load by 72.2% in the BFT-exposed group. Interestingly, combined treatment with DAPT and ICG001 further regressed the tumors by 86.4% in the BFT group. The DAPT and/or ICG001-treated control group (no BFT exposure) also showed tumor inhibition (Fig. 6F). Histopathology of tumors showed increased Ki-67 (Fig. 6G) in the BFT-pretreated group, whereas the DAPT- and ICG001-treated groups showed reduced expression. Importantly, BFT-induced “memory effect” was inhibited with in vivo DAPT/ICG001 treatment as evident from abrogated mammosphere formation (Fig. 6H), transwell migration, matrigel invasion, and elevated adhesion potential of dissociated tumor cells (Supplementary Fig. S17A and S17B). Increased CD31 (Supplementary Fig. S17C) was observed in the BFT-pretreated group compared with the DAPT- and ICG001-treated groups. Taken together, these results show that β-catenin and NICD present an important functional node in mediating the biological effects of BFT in breast cells.

To evaluate the impact of gut or ductal colonization with ETBF on breast cancer progression, we used a mammary intraductal (MIND) model in syngeneic mice. Our goal was to determine whether preexisting gut or breast duct infection with ETBF affects mammary tumorigenesis and metastatic progression. For gut colonization, BALB/c mice were given an antibiotic cocktail for 1 week followed by oral infection with 10⁸ CFU of ETBF or 086Mut; high-level gut colonization persisted for the duration of the experiment (≥ × 10⁹ CFU). Mammary gland ducts of mice harboring gut colonization with ETBF or 086Mut were then injected (via teats) with 20,000 4T1-Luc2 cells to initiate mammary tumors in the ductal tree (Supplementary Fig. S18A). We monitored the tumor progression by bioluminescence imaging. Mice harboring gut colonization with ETBF exhibited increased tumor progression compared with 086Mut and sham-control mice (Fig. 7A). RNA-seq data obtained from the secondary tumors from the BFT-pretreated MCF7 group and the control group showed higher expression of genes associated with migration, homing, and metastasis (Supplementary Fig. S18B), indicating that exposure to BFT owing to gut colonization with ETBF might enhance metastatic progression. Intriguingly, significantly higher levels of lung as well as liver metastases were observed in ETBF-colonized mice compared with either the sham-control group or 086Mut-colonized mice (Fig. 7B and C; Supplementary Fig. S18C). Analysis of tumor sections showed sheets of tumor cells with mesenchymal-like phenotype in the ETBF group, whereas no such phenotype was observed in any tumors in sham-control or 086Mut groups (Supplementary Fig. S18D). Liver and lung sections from ETBF-colonized mice exhibited a higher number and increased area of metastatic lesions in comparison with sham-control and 086Mut-colonized mice (Fig. 7D–F; Supplementary Fig. S18E). Next, we queried the impact of mammary duct colonization with ETBF on breast cancer progression. To this end, 10⁸ CFU of ETBF or 086Mut was introduced in mammary ducts via teats (Supplementary Fig. S18A); high-level ductal colonization persisted for the duration of the experiment (Supplementary Fig. S19A). The presence of BFT was observed in mammary glands of mice harboring ductal ETBF infection, whereas no BFT was detected in 086Mut-infected mice (Supplementary Fig. S19B). 4T1-Luc2 cells injected (via teats) in mammary gland ducts showed increased tumor progression and formed ∼3.9-fold larger tumors in the ETBF-infected group in comparison with the 086Mut or sham-control group (Fig. 7G and H). Increased lung and liver metastases were observed in mice harboring ductal colonization of ETBF in comparison with either the sham-control group or 086Mut-colonized mice (Fig. 7I–L; Supplementary Fig. S19C). Tumors formed in the ETBF group were more proliferative as evident from Ki-67 staining, showed a mesenchymal phenotype, and exhibited markedly higher stromal infiltration evident from trichrome staining compared with the sham-control or 086Mut group (Supplementary Fig. S19D). Next, we queried the impact of gut colonization with ETBF on MCF7 tumor formation. Indeed, mice harboring gut ETBF infection exhibited significantly increased (3.3-fold larger tumors) tumor progression than the sham-control and 086Mut groups (Fig. 7M). MCF7 tumors formed in ETBF-infected mice were more proliferative and richly vascularized, as evident from Ki-76– and CD31-specific IHC (Fig. 7N). These results using multiple preclinical murine models unequivocally show that gut or ductal colonization with ETBF markedly enhances breast tumorigenesis and metastatic progression, pointing to a novel role of ETBF.

To our knowledge, the results presented here are the first to demonstrate a direct role for ETBF infection and its secreted toxin, BFT, in breast carcinogenesis. We found the presence of B. fragilis in human breast tissues and nipple aspirates using meta-analysis of clinical data and then developed in vivo approaches to examine its involvement in breast tumorigenesis. This involved using a MIND model to introduce ETBF directly into the breast ducts or gut colonization with ETBF, both leading to the development of ductal hyperplasia as well as metastatic progression. BFT is the only identified virulence factor of ETBF, and, notably, in the course of these studies we show that even a short exposure to BFT elicits a long-term oncogenic memory in breast cells. Particularly significant is the attainment of a highly migratory, invasive, stemness-rich phenotype that is supported by the drastic change in molecular machinery as expression of the genes supporting cell movement, embryonic pluripotency pathway, and metastasis is induced in BFT-exposed cells. Mechanistic evaluations identify the involvement of the β-catenin and Notch1 pathways, whose inhibition leads to the abrogation of BFT-mediated cell migration and invasion, underscoring their biological importance. Few recent studies report the presence of breast microbiota and have identified some bacterial species as selective inhabitants of breast tumors (10, 14, 16–19, 31). Methylobacterium radiotolerans is abundantly present in tumor tissues, whereas Sphingomonas yanoikuyae is enriched in normal breast (14). The presence of Lactobacillus and Bifidobacterium, species known for health-promoting effects, in addition to taxa generally associated with pathogenesis such as Enterobacteriaceae, Pseudomonas, and Streptococcus agalactiae in breast tissues has been noted (10). A higher abundance of Escherichia coli and Staphylococcus epidermidis has been observed in breast tumors in comparison with healthy breast tissue (10, 15). Although these studies establish the existence of breast microbiota and some even identify the differential existence of distinct microbial species in normal versus tumor samples, their biological impact on breast cancer initiation and progression has not been investigated.

Our study implicates an oncogenic role of gut colonization with ETBF in breast cancer initiation and metastatic progression. Direct mammary ductal colonization or indirect gut colonization with ETBF is sufficient to initiate mammary gland hyperplasia within three weeks of ETBF infection, and, unexpectedly, gut colonization with ETBF potentiates mammary tumor growth and metastatic progression. We observed the presence of ETBF in mammary ducts of mice harboring gut ETBF infection, but it is unclear whether ETBF traveled internally from gut to breast or gut-infected mice acquired mammary gland infection through environment. A large body of work establishing the pathogenic role of this gut commensal in colonic inflammation and colon cancer (23, 32–34) has helped form the “Alpha-bug hypothesis,” proposing that a single pathogenic microbe, such as ETBF, possessing unique virulence traits can direct oncogenesis acting alone or via modulating the bacterial community in the organ and selectively “crowding out” microbes with protective roles (22, 23). Interestingly, BFT is a zinc-dependent metalloprotease toxin that is structurally similar to other important bacterial toxins like botulinum toxin or tetanus toxin (32). Although BFT has been shown to induce proliferation of human colonic epithelial cells (35), its impact on the invasion, migration, and stemness potential of cancer cells has never been examined. Cytoskeletal remodeling is a salient feature of epithelial-to-mesenchymal transition (36), and it also regulates cell migration and invasion in response to extracellular stimuli and aids in metastasis (37). Upon BFT exposure, breast cells undergo distinctive morphologic changes, acquiring mesenchymal-like phenotype, and acquire a highly migratory and invasive phenotype. Tumors may harbor cancer stem cells that characteristically possess the capability of self-renewal and differentiation and are key for tumor initiation, metastatic progression, and therapy resistance as well as relapse (38, 39). In addition to the enrichment of gene expression associated with cell movement and invasion potential, BFT-exposed breast cells also show an enhanced embryonic stem cell pluripotency pathway. Importantly, increased stemness potential is observed in BFT-treated breast cells as reflected in the formation of mammospheres and demonstrated as well by the in vivo–limiting dilution assay. Of note, the nontoxigenic mutant Bacteroides fragilis (086Mut) does not secrete BFT and exhibits no oncogenic effect on breast tumor growth and metastatic progression, suggesting that BFT is central to ETBF's actions. Our data show that ETBF acts as an “Alpha bug” triggering an oncogenic cascade resulting in breast cancer progression via its toxin.

Our results advance the understanding of the molecular mechanisms underlying ETBF/BFT and breast cancer progression. RNA-seq analyses of secondary tumors developed with BFT-pretreated cells and breast cancer cells treated with BFT show enrichment of the β-catenin pathway. Several canonical and noncanonical β-catenin–responsive genes exhibit increased expression in response to BFT. Wnt–β-catenin signaling is important for human breast development as well as breast cancer progression and is elevated in multiple breast cancer subtypes. Higher expression of β-catenin is associated with higher tumor grade and poor prognosis (29). These results are consistent with previous studies presenting BFT-induced activation of β-catenin via cleavage of E-cadherin in colonic epithelial cells (35). We have also discovered that BFT triggers the activation of the Notch1 pathway, and the expression of several canonical and noncanonical Notch-responsive genes is enriched in breast cancer cells. In addition to growth and progression of breast cancer, Notch1 is also implicated in the maintenance of mammary stem cells (40). It is interesting to note that the Wnt–β-catenin and Notch signaling pathways cross-talk via GSK3β and Jagged1 (41). Inhibiting Notch and β-catenin results in abrogation of BFT-mediated migration and invasion of breast cells. Several inhibitors of the Wnt–β-catenin pathway, including the porcupine inhibitors LGK974 and ETC159, the pathway antibodies vantictumab and ipafricept, and the β-catenin/TCF inhibitor PRI-724, are under investigation (42). Preclinical and clinical studies are also examining γ-secretase inhibitors, peptide inhibitors, and anti-Notch monoclonal antibodies for efficient inhibition of the Notch pathway (43). Our results indicate a possibility of overcoming the molecular impact of ETBF infection via inhibiting the actionable key molecular nodes.

An intriguing relationship has emerged between infectious diseases and cancer pathogenesis where a chronic infection with a virus such as human papilloma virus is associated with cervical cancer, hepatitis B and C viruses are associated with liver cancer, and Helicobacter pylori is associated with gastric cancer. B. fragilis is a common colonizer of the gut (22), and 5% to 35% of studied populations are asymptomatically colonized by ETBF (23). Although it forms a small fraction of the total gut biome (22), it is regarded as a potent pathogen, and its pathogenic role is well established in colitis and colon cancer (32–34). Despite multiple established risk factors, a large number of breast cancers arise in women harboring none of the established risk factors, indicating the need to look beyond those factors. We conclude that microbial perturbations may associate with breast cancer development. Our study is a first step to show the involvement of a common, often commensal, colon microbiome member, ETBF, in breast carcinogenesis; additional studies are needed to clarify whether ETBF can be the sole driver to directly trigger the transformation of breast cells in humans and/or if other microbiota members also display procarcinogenic activity for the breast. Drugging or targeting a specific microbiota member is very challenging yet a very exciting goal for microbiome–breast cancer research. Combining microbiome analysis and molecular subtyping of breast tumors may help identify new contributors to the pathogenesis of breast cancer as well as patients most likely to benefit from prevention strategies.

Breast cancer cell lines MCF7, HCC1806, HCC1569, B474, and HCC1937 and normal breast epithelial cell line MCF10A were procured from the ATCC and maintained at 37°C in 5% CO₂ and 95% humidity. MCF10A-Kras cells were a gift from Dr. Ben Park (Vanderbilt University Medical Center). 4T1-luc cells were a gift from Dr. Saraswati Sukumar (Johns Hopkins School of Medicine). Cells were used for experiments within 10 to 20 passages from thawing. All cells were authenticated via short tandem repeat testing. Mycoplasma detection was routinely performed using the MycoAlert Detection Kit (Lonza, LT07-218). Cultures of enterotoxigenic Bacteroides fragilis (ETBF) strain 86-5443-2-2 (BFT2-secreting strain) and its isogenic nontoxigenic mutant that does not secrete BFT due to an in-frame deletion of the chromosomal Bft gene were maintained anaerobically at 37°C. Bacterial pellets were washed and resuspended with 1 × Dulbecco PBS (1 × PBS free of calcium chloride and magnesium chloride) for mouse inoculums.

BFT was HPLC purified from culture supernatants of ETBF as previously described (44) and stored at −80°C. Briefly, BFT-producing wild-type B. fragilis strain or biologically inactive mutant were cultured anaerobically in BHC medium [37 g of brain heart infusion base (Difco Laboratories)/liter, 5 g of yeast extract (Difco Laboratories)/liter, and 1 μg of vitamin K/mL, 5 μg of hemin/mL, and 0.5 g of l-cysteine/mL] for 24 hours. After removing the bacterial cells by centrifugation at 4°C and sterilization by filtering through 0.22-μm pore size filter, the supernatant containing BFT2 protein or biologically inactive mutant BFT-H352Y was concentrated 5- to 6-fold at 4°C by ultrafiltration using a membrane with a molecular mass exclusion of 10 kDa (Millipore Corporation). The ultrafiltrate was chromatographically (FPLC) purified under the denaturing condition in urea. Purified BFT and BFT-H352Y were stored at −80°C. For in vitro studies, cells were treated with different concentrations of BFT. The optimum concentration of BFT was 5 nmol/L (100 ng/mL; ref. 45). Rabbit monoclonal BFT2 antibody was a generous gift from Dr. Saraspadee Mootien, L2 Diagnostics LLC (46). For Western blot and IHC, rabbit monoclonal anti–E-cadherin, anti–E-cadherin, anti-GSK3β, anti–p-GSK3β, anti–p-β-catenin, anti-Slug, anti-Snail, anti-Vimentin, anti-c-Myc, anti-Notch1, anti-NICD, anti-RBPSUH, anti-MAML1, anti-Hes1, anti-cyclin D3, anti-p21/Waf1, anti-Oct4, anti-Sox2, anti-Nanog, anti-KLF4, and anti–Ki-67 antibodies were purchased from Cell Signaling Technology. Anti-β-catenin, anti-CD31, anti-Jagged1, anti-Twist, anti-Snail, anti-Occludin, and anti-Lamin B were purchased from Santa Cruz Biotechnology. IHC-specific rabbit monoclonal c-Myc was procured from Abcam and mouse monoclonal β-actin was procured from Sigma-Aldrich. Horseradish peroxidase–conjugated goat anti-rabbit IgG, goat anti-mouse IgG and donkey anti-goat IgG were purchased from Sigma-Aldrich. DAPT and ICG001 were procured from Sigma-Aldrich. Chemiluminescent peroxidase substrate and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) were procured from Sigma-Aldrich. Rhodamine phallaoidin was procured from Invitrogen Corporation.

Raw data from the studies evaluating the breast microbiome in patients with breast cancer were accessed from NCBI-SRA, EMBL-EBI, and NCBI-GEO. Data sets PRJEB4755, SRP071608, and PRJNA335375 examining local microbiota of normal breast tissue, cancerous breast, benign breast cancer, malignant breast cancer, and nipple fluid aspirate from healthy women and breast cancer survivors were analyzed using One Codex. Abundant and rare species were identified from each cohort, and further strain-level classification was performed. Comparisons were made across studies as well as between cohorts. Potential pathogenic organisms were identified for further experimental studies.

All animal studies were in accordance with the guidelines of Johns Hopkins University Animal Care and Use Committee. For mammary gland colonization, twice parous BALB/c mice (obtained from Charles River and maintained in house) were given antibiotic cocktail (clindamycin 0.1 g/L and streptomycin 5 g/L) in water bottles (Hospira and Amresco) for 7 days and discontinued. Mice were also injected with antibiotic cocktail intraductally to clear ductal microbiome at the same time. Mice in ETBF or 086Mut group were then injected with ∼10⁸ CFU of ETBF or 086Mut in 1× PBS via intraductal administration. For sham control, mice were intraductally injected with 1× PBS. For gut colonization, female, 3-week-old, C57BL/6 mice (obtained from The Jackson Laboratories and maintained in house) were given antibiotic cocktail (clindamycin 0.1 g/L and streptomycin 5 g/L) in water bottles (Hospira and Amresco) for 7 days and discontinued. Mice in ETBF or 086Mut group were given ∼10⁸ CFU of ETBF in 1× PBS via oral gavage. For sham control, mice were gavaged with 1× PBS. We quantified fecal bacterial colonization as CFU per gram stool (34). Mice were euthanized, and serum was collected 1 week and 3 weeks after infection as indicated. Whole mammary glands were excised and immediately spread on a glass slide and fixed in Carnoys's fixative (ethanol, chloroform, and glacial acetic acid in the ratio 6:3:1) for 48 hours. After fixation, the whole mammary glands were incubated in 70% ethanol for 1 hour and then stained with carmine alum stain for 2 days. Stained mammary glands were then dehydrated in a series of graded alcohol, 1 hour each and immersed in xylene for 4 days to get rid of the immense fat deposits of the breast. Once clear of most of the fat deposits, tissue was mounted using high-viscosity Cytoseal 260. Ductal structure of the mammary gland was examined microscopically.

Mammary gland tissue and tumor tissue excised from MCF7 xenografts, MCF10A-Kras xenografts, and 4T1 intraductal tumors were fixed in 10% formalin, paraffin-embedded and sectioned, and IHC analyses were performed using anti–Ki-67, anti–pan-keratin, anti-CD3, anti-PCNA, anti-CD31, anti-β-catenin, anti–E-cadherin, anti–N-cadherin, and anti-c-Myc antibodies. Images were captured using Leica microscope at 20× magnification. The presence of BFT in the serum of ETBF-colonized mice was determined using ELISA by the antigen capture method. For tissue ELISA, whole breast was excised from 3-week-old mice bearing ETBF or 086Mut infection and corresponding shams. Tissue was lysed in Sigma CellLytic mammalian tissue lysis buffer as per the manufacturer's protocol. Plates (96-well) were coated with lysate containing 100 μg protein made up to 30 μL in PBS. BFT was quantified by antigen capture ELISA. Standards were used for quantitation.

Cellular morphology of BFT-treated cells was monitored using phase-contrast microscopy. For rhodamine/phalloidin staining, cells were washed with PBS and stained with rhodamine-conjugated phalloidin to stain the F-actin filaments for 30 minutes followed by counterstaining with DAPI nuclear stain for 5 minutes. Cytoskeletal changes were examined and imaged under Leica E800 fluorescent microscope. Cell viability was measured by MTT dye reduction assay. The percentages of live cells were calculated and plotted using Prism software (GraphPad Prism). For clonogenicity assay, colonies containing >50 normal-appearing cells were counted, and pictures were taken using a digital camera.

Microfluidic devices containing an array of parallel microchannels of 3 to 50 μm dimensions were fabricated by standard lithography and coated with 20 μg/mL collagen type I (BD Biosciences). Cells were allowed to migrate toward a gradient of EGF used as a chemoattractant. Phase-contrast time-lapse images were captured at 30-minute intervals for up to 10 hours on an inverted Nikon microscope (10× objective) at multiple stage positions via stage automation (Nikon Elements). Cell speed, velocity, and persistence were computed using a custom-written Matlab program (The MathWorks).

Serum-starved cells were treated with 5 nmol/L BFT for 48 hours prior to the assay. For adhesion assay, wells of a microtiter plate were coated with 30 μL collagen I (40 μg/mL in PBS) for 16 hours at 4°C followed by the introduction of BFT-treated cells. Adhered cells were fixed with 4% paraformaldehyde, stained with 0.5% crystal violet, and imaged using a brightfield microscope. For spheroid migration assay, migration of single cells from the tumor spheroids over time was monitored using phase-contrast microscopy. Distance migrated was measured using Lieca image scope software and plotted as a measure of cell movement using GraphPad Prism 5 software.

To assess the migratory and invasive potential of BFT-treated MCF7 and MCF10A cells, we utilized the conventional Matrigel invasion, scratch migration, and transwell migration assays. For Matrigel invasion assay, BFT (5 nmol/L)-treated MCF7 and MCF10A cells (20,000) were seeded in a Matrigel invasion chamber from BD Biocoat Cellware. For transwell migration assay, BFT (5 nmol/L)-treated MCF7 and MCF10A cells (20,000) were seeded in the top chamber with an 8-μm pore size, and cells were allowed to invade or migrate through the Matrigel or filter for 24 to 48 hours. The number of invaded/migrated cells on representative sections of each membrane was counted under light microscope. For scratch migration assay, monolayers of MCF7 and MCF10A were allowed to form; a 1-mm wide scratch was made across the cell layer using a sterile pipette tip, and media were replaced with fresh serum-free media containing 5 nmol/L BFT or vehicle control. Plates were photographed immediately after scratching, and migration of cells was followed for various time intervals. Wound closure was quantified from distance between edges using Leica ImageScope software. Speed of migration was calculated, and wound closure was plotted using GraphPad Prism 5 software.

For liquid mammosphere assay, 5,000 MCF7 or MCF10A cells were seeded in 2 mL of liquid mammosphere media in 30 mm ultra-low attachment plates. Cells were treated with vehicle or 5 nmol/L BFT and allowed to grow for 7 days. For solid mammosphere assay, 5,000 MCF7 or MCF10A cells were suspended in 2 mL mammosphere medium containing methylcellulose, plated on ultra-low attachment plates, and incubated for 7 days. Cultures were observed under microscope and spheres (>50 μm) were counted.

Whole-cell lysates were prepared using modified RIPA buffer. For protein lysates of tumor samples, tumor tissues were homogenized using a tissue homogenizer in mammalian tissue lysis buffer on ice. For subcellular fractionation, nuclei were separated from the cytosolic fraction resuspended in nuclear protein extraction buffer.

For RT-PCR, cells/tissue/mammosphere samples were lysed in TRIzol, RNA was isolated by chloroform-isopropanol method, and cDNA was synthesized using an iScript cDNA Synthesis Kit. RT-PCR was performed and imaged using the Gel Doc image system (Bio-Rad). For immunofluorescence, fixed cells were permeabilized using 0.1% Triton-X-100 followed by overnight incubation with primary antibody at 1:100 dilution in 3% BSA. Cells were then incubated with FITC/TRITC-tagged secondary antibody. Cells were examined under Lieca E800 fluorescent microscope. Images were captured at 60× magnification using oil immersion objective with Lieca Elements software.

The expression profile of CD24 (anti-human #555427, BD Biosciences) and CD49f (anti-human/mouse #313616, BioLegend) in tumor-derived cells was analyzed by flow cytometry. Briefly, 10⁶ cells were stained with respective antibodies following the antibody manufacturer's specific protocol. Labeled cells were acquired by BD FACS LSR II and analyzed using FACSDiva 6.0 software.

NOD/SCID mice (female, 6–8 weeks old) were acquired from Sidney Kimmel Comprehensive Cancer Center (SKCCC) animal facility and maintained in house. Exponentially growing MCF7 or MCF10A-Kras cells, treated with 5 nmol/L BFT for 48 hours (5 × 10⁷ cells in 100 μL matrigel), were implanted in the fourth mammary fat pad on either side. Tumor volumes were monitored. Tumors were excised and processed for further analysis. For limiting-dilution orthotopic xenograft model (47), cells dissociated from primary tumors formed by BFT-treated MCF7 cells were injected at limiting dilutions (5 × 10⁶–5 × 10³) into the mammary fat pads of immunocompromised NOD/SCID mice. Tumor incidence was regularly monitored, and stem cell frequency was calculated based on tumor incidence.

For in vitro drug treatments, MCF7 cells were pretreated with 5 nmol/L BFT alone or in combination with β-catenin inhibitor ICG001 or γSecretase inhibitor DAPT or both for 48 hours. Pretreated cells (5 × 10⁶) were then implanted into the mammary fat pads of 4- to 6-week-old NOD/SCID mice previously implanted with E2 pellets subcutaneously. Tumor progression was monitored for 7 weeks. For in vivo drug treatments, MCF7 cells were pretreated with 5 nmol/L BFT for 72 hours. MCF7 or BFT pretreated MCF7 cells (5 × 10⁶) were then implanted into the mammary fat pads of 4- to 6-week-old NOD/SCID mice previously implanted with E2 pellets subcutaneously. Tumors were allowed to reach a minimum volume (200 mm³ in BFT pretreatment group), and mice were randomized into eight experimental groups. The animals were then treated with ICG001 (20 mg/kg body weight, daily in 2% DMSO + 50% PEG300 + 5% Tween 80 + ddH2O) or DAPT (20 mg/kg, alternate days in corn oil) or a combination of both drugs through the course of the experiment. Tumor progression was monitored regularly. Tumors were excised at the end of the experiment and processed for further analyses.

Twice parous BALB/c mice were obtained from Charles River and maintained in house. Mice were given antibiotic cocktail (clindamycin 0.1 g/L and streptomycin 5 g/L) in water bottles (Hospira and Amresco) for 7 days and discontinued. Mice in ETBF or 086Mut group were then oral gavaged with ∼10⁸ CFU of ETBF or 086Mut strains of B. fragilis in 1× PBS and allowed to colonize for 3 days. Intraductal infection groups received ∼10⁸ CFU of ETBF or 086Mut strains of B. fragilis in 1× PBS via teat injection. For sham control, mice were oral gavaged or injected in teats with 1× PBS. To establish intraductal mammary tumors, 20,000 4T1-Luc2 cells were injected directly into the mammary ducts of mice on one side. Tumor progression was regularly monitored by bioluminescence imaging using IVIS spectrum at SKCCC animal resources. For bioluminescence imaging, animals were injected with 10 μL/g d-luciferin (15 mg/mL in PBS) intraperitoneally, and images were captured 8 to 10 minutes after injection. At the end of the experiment, ex vivo bioluminescence images of major organs, lungs, and liver were captured to investigate metastatic progression. Briefly, animals were given an intraperitoneal injection of D-luciferin and were euthanized after 10 minutes. Lungs and liver were excised, and images were captured using the IVIS system. Tumors were excised, weighed, and preserved for subsequent studies. For the MCF7 tumor model, 4- to 6-week-old NOD/SCID mice were given a cocktail of antibiotics as described above in drinking water for 1 week. Mice were then infected with 10⁸ CFUs of ETBF or 086Mut via oral gavage and allowed gut colonization for 1 week followed by implantation of 5 × 10⁶ MCF7 cells in mammary fat pads. Tumor progression was monitored for 7 weeks. Tumors were excised, weighed, and preserved for subsequent studies.

Lungs and liver excised from 4T1 Luc2 tumor–bearing BALB/c mice were harvested in DMEM F12 medium, minced into pieces, and transferred into 15-mL tubes containing 2.5 mL of respective digestion cocktail; RPMI containing 10 mg/mL collagenase A for liver and RPMI containing 10 mg/mL of collagenase A⁺ 10 mg/mL of hyaluronidase for lungs. The organs were then placed in shaking water bath at 37°C for 30 minutes to allow complete dissociation and filtered through separate 70-μm nylon cell strainer to remove large chunks of undigested tissue. Samples were collected in 50 mL tubes, centrifuged for 5 minutes/1,500 rpm/room temperature and supernatants were discarded. Samples were washed twice by centrifugation in PBS. Pellets were resuspended in culture media containing 10 μL of 60 mmol/L 6-thioguanine and plated onto 6-well culture plates. Plates were incubated in 37°C/5% CO₂ to allow growth of colonies for 3 to 7 days (48).

RNA extraction, sequencing, and sequence analysis were performed by the Johns Hopkins Transcriptomics and Deep Sequencing Core. Detailed method is provided in the supplementary section.

All experiments were performed thrice in triplicates. RNA-sequence expression analysis was done using Partek software, and functional pathway analyses were done using Spotfire and Ingenuity software. Measurements of micrographs and IHC quantitation were done using Leica Aperio ImageScope, Leica Biosystems. Western blot and RT-PCR quantification were performed using GelQuant. Statistical analyses were done using GraphPad Prism 5. Results were considered to be statistically significant if P < 0.05. Results were expressed as mean ± SE between triplicate experiments performed thrice. For comparison between multiple groups, statistical significance was determined by one-way ANOVA and Bonferroni analysis. Comparison between two groups was done using the Student t test.

K. Konstantopoulos reports grants from NIH/NCI during the conduct of the study. C.L. Sears reports grants from Bloomberg Philanthropies during the conduct of the study; grants from Bristol Myers Squibb and Janssen, and personal fees from Biomx outside the submitted work. No other disclosures were reported.

S. Parida: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. C.C. Talbot: Investigation. K. Konstantopoulos: Investigation. K.L. Gabrielson: Investigation. C.L. Sears: Investigation. D. Sharma: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, project administration, writing–review and editing. S. Wu: Investigation, methodology. S. Siddharth: Investigation, methodology. G. Wang: Investigation, Methodology. N. Muniraj: Investigation, methodology. A. Nagalingam: Investigation, methodology. C. Hum: Investigation. P. Mistriotis: Investigation. H. Hao: Investigation.

This work was supported by NCI NIH R01CA204555 (to D. Sharma), Breast Cancer Research Foundation (BCRF) 90047965 (to D. Sharma), NCI NIH CA183804 (to K. Konstantopoulos), and Bloomberg Philanthropies (to C.L. Sears). We acknowledge Dr. Xinqun Wu for her technical help.