BST-2 promotes survival in circulation and pulmonary metastatic seeding of breast cancer cells

Bone marrow stromal antigen 2 (BST-2) mediates various facets of cancer progression and metastasis. Here, we show that BST-2 is linked to poor survival in invasive breast cancer patients as its expression positively correlates with disease severity. However, the mechanisms that drive the pro‐metastatic functions of BST-2 are not fully understood. Correlation of BST-2 expression and tumor aggressiveness was analyzed in human tissue samples. Migration, invasion, and competitive experimental metastasis assays were used to measure the cellular responses after silencing BST-2 expression. Using a mouse model of breast cancer, we show that BST-2 promotes metastasis independent of the primary tumor. Additional experiments show that suppression of BST-2 renders non-adherent cancer cells non-viable by sensitizing cells to anoikis. Embedment of cancer cells in basement membrane matrix reveals that silencing BTS-2 expression inhibits invadopodia formation, extracellular matrix degradation, and subsequent cell invasion. Competitive experimental pulmonary metastasis shows that silencing BST-2 reduces the numbers of viable circulating tumor cells (CTCs) and decreases the efficiency of lung colonization. Our data define a previously unknown function for BST-2 in the i) formation of invadopodia, ii) degradation of extracellular matrix, and iii) protection of CTCs from hemodynamic stress. We believe that physical (tractional forces) and biochemical (ECM type/composition) cues may control BST-2’s role in cell survival and invadopodia formation. Collectively, our findings highlight BST-2 as a key factor that allows cancer cells to invade, survive in circulation, and at the metastatic site.

SCIentIfIC RepoRts | (2018) 8:17608 | DOI: 10.1038/s41598-018-35710-y counteraction of BST-2 activity and has been shown to regulate HIV resistance to interferon (IFN) 17,18 . Thus, in addition to tethering, BST-2 possess antiviral activity as shown by various infection models [19][20][21][22] . BST-2 is a type II transmembrane protein composed of four domains and expressed mainly on the apical side of cells. Expression of BST-2 is regulated by both extrinsic and intrinsic stimuli, including cytokines such as interferons 20,23,24 . In different disease conditions, such as autoimmune diseases 25,26 and different malignancies, BST-2 has been reported to be overexpressed 5,27,28 . BST-2 DNA is hypomethylated in breast cancer cells leading to its overexpression 3 . Increased expression of BST-2 in breast cancer has been shown to mediate various facets of breast cancer progression including cell adhesion, anchorage-independent growth, survival, primary tumor growth, invasion, and metastasis. The effect of BST-2 on both primary tumor growth and metastasis 4,7 suggest that BST-2 may independently regulate both processes as inferred by Mahauad-Fernandez et al., in a correlation analysis 4 . Here we provide evidence that expression of BST-2 in tumors may function as an independent driver of metastasis. Using tumors from an invasive breast cancer (IBC) patient cohort, we found higher BST-2 levels in tumors that progressed beyond a localized state. Furthermore, we found that the cellular mechanism by which BST-2 exerts its pro-metastasis function includes the ability of BST-2 to promote i) formation of invasive structures, ii) survival of cancer cells in circulation, and iii) enhancement of pulmonary seeding of cancer cells and malignant growth of such cells in the metastatic site.

BST-2 levels predict morbidity and mortality in an invasive breast cancer patient cohort.
In previous studies, we utilized publicly available data sets to analyze the levels of BST-2 transcript (TCGA) 3,4 and protein (Human Protein Atlas) 7 to show that high levels of BST-2 positively correlates with features of aggressive breast cancer, such as survival, invasion, migration, and metastasis 3,4,7 . To link BST-2 expression in breast tumors to tumor aggressiveness, we performed BST-2 immunohistochemistry (IHC) using tumor tissues obtained from an unselected, hospital-based cohort of patients (n = 79) with invasive breast cancer (IBC) diagnosed at The University of Iowa between 1986 and 1989 29 . During this period, long-term follow-up data was collected. About 34% (27/79) of patients bear BST-2 + IBC; where BST-2 positivity was defined as tumors with BST-2 staining of 10% of tumor cells or higher. Where positive, membranous and cytoplasmic BST-2 staining of invasive tumor cells was noted with varying levels of intensity (Fig. 1A). In terms of disease severity, BST-2 level is highest in tumors of IBC patients with synchronous IBC followed by those that presented with a 2 nd metachronous IBC (Fig. 1B). In our study, we found that levels of BST-2 in tumors increased with stage of presentation from localized disease, nodal spread, and distant metastasis at the time of diagnosis (Fig. 1C). Kaplan Meier survival plots show that patients with BST-2 + IBC had reduced survival than patients with BST-2 − IBC (Fig. 1D). Although the difference in survival did not reach statistical significance, BST-2 + IBC-bearing patients had a median overall survival (OS) time of 7 years and an AUC of 914.8, while BST-2 − IBC-bearing patients had an OS of 15 years and an AUC of 1,260 (Fig. 1D). These data show that BST-2 + IBCs present at more advanced stages and are more aggressive than IBCs that are BST-2 − .

Suppression of BST-2 expression impairs migratory and invasive capability of breast cancer cells independent of cell proliferation.
Previous studies have shown that endogenous BST-2 promotes non-proteolytic motility (migration) and proteolytic motility (invasion) of aggressive murine breast cancer cell line 4T1 and human MDA-MB-231 cells 28 . Further, exogenous expression of BST-2 in the luminal A human breast cancer cell line MCF-7 confers migratory and invasive abilities to the cells 4,7 . However, whether the effect of BST-2 on cell motility is as a result of increased cell proliferation is unknown. We answered this question by performing 2D and 3D migration experiments in the presence of the proliferation inhibitor mitomycin C using BST-2-expressing and BST-2-suppressed 4T1 4,7 and MDA-MB-231 28 cells. Suppression of BST-2 expression inhibits migration of MDA-MB-231 and 4T1 cells in the presence of mitomycin C ( Fig. 2A,D). To examine the kinetics of BST-2-mediated enhancement of cell migration independent of cell proliferation, we performed time course migration assay in the presence of mitomycin C at 0, 4, 8, and 24 h for MDA-MB-231 cells and 0, 12, and 24 h for 4T1 cells. MDA-MB-231 shBST-2 cells show a reduction in average migration from 8 h when compared with shCTL cells (Fig. 2E). Similarly, 4T1 shBST-2 cells were inhibited for migration as early as 12 h compared to shCTL cells (Fig. 2F,G). Noteworthy is the difference in cell morphology between BST-2-expressing shCTL and BST-2-suppressed shBST-2 cells. Following knockdown of BST-2 expression, cells adopt epithelial-like morphology, as can be observed in Fig. 2F. Furthermore, we used MDA-MB-231 and 4T1 cells to determine if the effect of BST-2 on cell invasion 4,7 is independent of cell proliferation. Suppression of BST-2 expression in both MDA-MB-231 and 4T1 cells inhibits cell invasion in the presence of mitomycin C (Fig. 2H-K). These data confirm that BST-2 has a bonafide effect on cell motility independent of cell proliferation and that suppressing BST-2 expression in breast cancer cells can reverse this invasive phenotype.
Suppression of BST-2 inhibits adhesion-independent survival, and subsequent migration and re-adherence of cells to target structures. To metastasize, cancer cells that acquire anchorage-independent phenotype by resisting the anoikis program have to migrate and colonize target organs. BST-2 has been shown to be critical for anoikis resistance 7 , migration, and invasion of cancer cells 4,7 . To evaluate the role of BST-2 in the migration and re-attachment of anchorage-independent cancer cells, we developed a tandem survival-migration-adhesion assay. BST-2-expressing shCTL and BST-2-suppressed shBST-2 cells were cultured under anchorage-independent conditions on a semi-solid polymer-VitroGel 3D (TheWell biosciences) for 6 days.
As shown in Fig. 3A, BST-2 expression results in adhesion-independent proliferation, since shCTL cells suspended in VitroGel increased in numbers over time. The cells in VitroGel were evaluated for the ability to migrate through the hydrogel and to adhere to the culture plates. Imaging through VitroGel acquired both migrated SCIentIfIC RepoRts | (2018) 8:17608 | DOI:10.1038/s41598-018-35710-y non-adherent and adherent cells. Results show that by day 4 of culture, majority of shCTL cells migrated and subsequently adhered to the culture plates. These adherent shCTL cells display mostly elongated branching morphology (Fig. 3A, upper row, days 4 and 6 zoomed). In contrast, shBST-2 cells were predominantly round reminiscent of non-adherent morphology with few elongated cells (Fig. 3A, lower row, days 4 and 6 zoomed). These results suggest that cancer cells that lost BST-2 expression may also lose the ability to reattach to new targets following transit in circulation.
Since shBST-2 cells that are anchorage-independent (suspended in VitroGel) did not efficiently attach to culture plates, we tested their sensitivity to anoikis when cultured under adherent and anchorage-independent conditions. Thus, after 48 h of Poly-HEMA-mediated suspension of shCTL and shBST-2 cells, the cells were collected for viability analysis using the Trypan blue exclusion and methylthiazole tetrazolium (MTT) assays. Both assays indicate that viability of shBST-2 cells was significantly reduced under anchorage-independent conditions compared to shCTL cells that survived this condition (Fig. 3B), indicating that reduction in BST-2 levels may render cancer cells susceptible to anoikis as previously shown in 4T1 cells 7 . Flow cytometry analysis of the frequency of Annexin V and 7-aminoactinomycin D (7-AAD) stained cells confirmed the induction of cell death in shBST-2 cells cultured under anchorage-independent conditions (Fig. 3C). The apoptotic rates of BST-2-expressing shCTL cells were not significantly changed (12.6% versus 13.75%) after 48 h of suspension culture. In contrast, the apoptotic rates of BST-2-suppressed shBST-2 cells significantly increased (7.3% versus 48.08%) after 48 h of detachment from their extracellular environment (Fig. 3C). Together, our results suggest that reduction in BST-2 expression renders cancer cells anoikis-susceptible. Our results further suggest that loss of BST-2-conferred anoikis susceptibility may render cancer cells less invasive since the suspended cells were unable to re-adhere. It is possible that the elongation and branching observed in BST-2-expressing cells allow cells to form specialized structures for enhanced motility.

Loss of BST-2 expression inhibits the formation of invasive structures in breast cancer cell lines.
Given the importance of BST-2 in mediating anchorage-independent survival and invasion, both of which enhance metastasis, we investigated whether expression of BST-2 plays a role in the formation of invasive structures, also critical for metastasis 30 . shCTL and shBST-2 cells were plated onto Matrigel-coated cover plates as indicated 31 . shCTL cells formed colonies with stellate projection structures invading the surrounding matrix, whereas shBST-2 cells mostly grew as round spheroids without defined stellate projections. Compared to shBST-2 cells, the invasive structures can be observed by high powered microscopy on the periphery of shCTL cells as early as 24 h with complete matrix invasion at 48 h (Fig. 4A, pink arrows). In contrast, shBST-2 cells appeared rounded and by 48 h, some cells present with ruptured membrane morphology (Fig. 4A, blue arrow). By 72 h, majority of shCTL cells have invasive structures (Fig. 4B, cyan arrow heads) and also formed more and larger colonies (Fig. 4B, yellow circles) while shBST-2 cells present with dramatic loss of invasive morphology in 3D Matrigel culture at all times examined. Quantitation of the number of invasive colonies formed over time expressed as a percentage of total number of colonies per dish (invasive and non-invasive) show a significant increase in shCTL invasive structures (Fig. 4C). Additionally, by 72 h, majority of shCTL cells have larger colony diameter (Fig. 4D). Given the observed differences in cell morphology, especially with disrupted cell membrane, we assessed the viability of shCTL and shBST-2 cells suspended in Matrigel. We observed a significant decrease in the viability of shBST-2 cells (Fig. 4E). These results suggest that the effects of BST-2 on the number and length    32 ; and since we have previously shown that BST-2 is an important promoter of breast cancer metastasis, we sought to define the steps at which BST-2 is involved along the metastatic cascade. We performed competitive syngeneic experimental pulmonary metastasis studies (Fig. 6A) by injecting mice via the tail vein with a mixture of 4T1 shCTL: shBST-2: beads in a 2:2:1 ratio (Fig. 6B). Beads were used for normalization of differences in injection efficiencies from mouse to mouse. Blood collected from injected mice was used to determine the relative number of shCTL circulating tumor cells (CTCs) to shBST-2 CTCs over 6 h period. We found that the number of shCTL and shBST-2 4T1 CTCs was similar at time zero and after two hours (Fig. 6C). However, by 4 h, there were more shBST-2 CTCs in blood than shCTL CTCs (Fig. 6C). Furthermore, by 4 h, the percent of live shCTL CTCs was three times higher compared to shBST-2 CTCs (Fig. 6D). At variance, there was about 20% more dead shBST-2 CTCs compared to shCTL CTCs at 4 h (Fig. 6E). These results indicate that BST-2 expression may enhance metastasis through resistance to hemodynamic stress. Because CTCs that resisted hemodynamic shear force and immune clearance are capable of reaching and seeding secondary sites, such as the lungs, we evaluated the ability of shCTL cells to extravasate and form metastatic colonies in lungs. Thus, immediately after injection, there were about three times more shCTL cells in the lungs than shBST-2 cells (Fig. 6F,G) even though there is no difference in cell numbers (Fig. 6C) or survival in the blood (Fig. 6D,E) at this time point. The reason for this difference is unknown but it is plausible that as efficient mediators of cell to cell interaction 7 , shCTL cells may be able to evade stress and interact with other cells compared to shBST-2 cells. Interpretation of this result is convoluted by the fact that overtime, the number of shCTL cells in the lungs decreased, perhaps due to cell death or immune clearance (syngeneic model). Nevertheless, at all times, there were at least twice as many shCTL cells found in the lungs compared to shBST-2 cells (Fig. 6F,G). These data suggest that BST-2 promotes cancer cell extravasation, providing a potential mechanism for BST-2 in the invasive process during intravasation, extravasation, and colonization stages by tumor cells in the metastatic cascade.

Dissemination of cancer cells was significantly inhibited by suppression of BST-2 expression as observed in 4T1
shBST-2 injected mice (Fig. 7A, middle panel) relative to 4T1 shCTL mice (Fig. 7A, top panel). The inhibition of metastasis was reversed by expression of OE BST-2D in shBST-2 cells (Fig. 7A, compare top and bottom panels). Additionally, suppression of BST-2 reduced rate of lung metastasis as indicated by the number of lung nodules observed in mice bearing 4T1 shBST-2 cells compared to 4T1 shCTL and 4T1 OE BST-2D bearing mice (Fig. 7B,C). Further evidence of metastatic disease is shown by analysis of spleen size. Mice bearing shBST-2 cells had normal spleen size whereas shCTL and OE BST-2D bearing mice presented with splenomegaly ( Fig. 7D,E) as previously reported 4 . The morbidity that was caused by the injection of shCTL and OE BST-2D cells resulted in decreased survival of mice bearing shCTL and OE BST-2 cells compared to shBST-2-bearing mice (Fig. 7F). These results therefore demonstrate the requirement of BST-2 in lung metastasis.

Discussion
Cancer cell migration and invasion are highly integrated and dynamic processes that precede metastasis, which is a multi-step process encompassing i) cancer cell infiltration into adjacent tissues, ii) intravasation (trans-endothelial migration) of cancer cells into vessels, iii) survival of such cells in circulation, iv) extravasation (leave the blood stream) of the cells and (v) subsequent attachment and proliferation at secondary sites leading to colonization. During cancer progression, a variety of tumor cells show resistance to detachment-induced cell death (anoikis), as well as alter their plasticity via morphological changes that may include one or a combination of collective to amoeboid transition (CAT) 33 , epithelial to mesenchymal transition (EMT) 34 , and mesenchymal to amoeboid transition (MAT) 35 . Such changes allow cells with metastatic ability to survive harsh conditions while invading incompatible distal sites. Efficient coordination of events in the metastatic cascade is necessary for successful dissemination of cancer cells because alteration in any of the key metastatic processes will eliminate and possibly destroy metastasizing cancer cells. Therefore, it is crucial to identify the factors controlling cancer cell dissemination for development of novel efficacious therapy since most cancer deaths are linked to metastasis. BST-2 is one such factor that have been shown to be important for cancer aggressiveness, including enhanced metastasis 3,4,7-11,36-39 . However, the precise mechanism of how BST-2 functions and its contribution to various stages of metastasis are yet to be fully defined. Prior findings from our group identified BST-2 as an importatnt factor in spontaneous metastasis; which interrogates the full metastatic cascade 4,7 . This current study used pre-clinical and clinical data to provide evidence that BST-2 is directly linked to the aggressive phenotype and metastatic potential of breast cancer cells. Clinical breast tumor specimens from a cohort of invasive breast cancer patients facilitated the correlation of BST-2 levels to disease severity. Although the tumors used in our studies were from patients diagnosed with invasive breast cancer, some of which have metastasized, whether patients' tumors disseminated through lymphatic routes and/or hematogenous routes is unknown. However, comparative analysis of patients presenting with localized disease, lymph node spread, and distant metastases using clinical specimens indicate that disease severity is in line with increasing BST-2 levels. We demonstrate that a key effect of BST-2 expression in cancer cells is the induction of enhanced non-proteolytic and proteolytic cell motility independent of cell proliferation, as well as the ability to overcome anoikis. We previously reported that in breast cancer, BST-2 renders cancer cells resistance to anoikis through the GRB2/ERK/BIM/Cas3 pathway 7 . In addition, in nasopharyngeal carcinoma, high BST-2 expressing cancer cells can overcome anoikis via the induction of NF-kB and expression of anti-apoptotic genes 40 . While our data showing the presence of invadopodia support a role of BST-2 in cancer cell motility, this finding also reveals that BST-2 may actively control invasion events through migration-independent mechanisms that may involve stimulation of degradation of matrix substrates via metalloproteases. It remains to be determined if BST-2 is a component of mature invadopodia and which matrix-degrading factors are present in BST-2-containing invadopodia. Although the BST-2-dependent events that regulate invadopodia formation and the kinetics of elongation as well as invasive protrusion are yet to be determined, it has been shown that various proteins as well as specific phosphoinositide lipids are known to be associated with different stages of invadopodia formation and elongation in cancer cells 41 . For example, the actin cytoskeleton and filopodia-and lamellipodia-associated proteins are involved in the formation of invadopodia, whereas microtubules and vimentin intermediate filament networks play a role in elongation of invadopodia 42 . In our studies, we employed the widely used thin layer of ECM that have been coated directly onto a glass structure 31 . Although the assay clearly depicts the role of BST-2 in invadopodia formation, the glass surface may have blocked to some extent invadopodia elongation. Therefore, the effect of BST-2 on invadopodia formation may have been underestimated and requires a more appropriate physiological environment for accuracy. The fact that BST-2 mediates cell to cell and cell to ECM interactions 4,7 , cell invasion 28 , the formation of invasive structures and degradation of gelatin, suggest the possibility that BST-2 may be part of the cellular processes including CAT, EMT, MAT, ECM degradation, and basement membrane transmigration. It therefore remains to be determined whether BST-2 influences the stability of invadopodia, as well as identify the BST-2-regulated factors that mediate invadopodia formation and function.
The possible link between BST-2 and formation of invasive structures is intriguing for many reasons. Invadopodia formation has been visualized in models of breast cancer 43,44 and these structures have been implicated as key mediators of cancer cell intravasation 45,46 . Further, invadopodia facilitates extravasation of cancer cells at endothelial junctions 32 . One of the most important characteristics for successful metastasis is for cancer cells to withstand hemodynamic (fluid) stress upon intravasation 47 . Expectedly, over 98% of cancer cells that reach blood vessels are cleared within 24 hours 48 in part due to fluid stress. However, cancer cells have developed ways to overcome this hemodynamic stress in circulation 49 . For example, circulating cancer cells (CTCs) can cluster with each other 7,50 , with fibroblasts 51,4,7 , and with platelets 52 . CTCs that form clusters are better able to survive in circulation than CTCs that travel as singlets 50 , in part because CTC clusters travel at slower velocities and are more resistant to anoikis than CTC singlets 53 . Our finding that BST-2 expression confers survival advantage to breast cancer cells in circulation, as demonstrated using competitive syngeneic experimental pulmonary metastasis, may link BST-2 to cancer cell resistance to hemodynamic stress, and support the already known fact that BST-2 mediates interactions between cancer cells and fibroblasts and promotes cancer cell resistance to anoikis. CTCs have their own microenvironment with gradients of signaling molecules that promote cell adhesion and resistance to anoikis 53 . It is possible that in vivo, BST-2 may be induced by IFNs released from cancer-associated macrophages (CAMs) or myeloid cells that are part of these clusters. These gradients of signaling molecules can promote epithelial to mesenchymal transitions of some of the cancer cells found in the clusters explaining the heterocellular characteristic of CTC clusters in vivo 54 . In addition, as an interferon-inducible gene, BST-2 has the capacity to induce the expression of a plethora of inflammatory genes that are involved in platelet activation such as IFNβ 55 .
The efficient pulmonary seeding and metastatic progression of cancer observed with BST-2-expressing cancer cells following intravenous inoculation suggest a primary tumor-independent role for BST-2 in metastasis as well as a potential role for cancer cell autonomous BST-2 in the establishment of metastatic niche. We envision that in a tumor context, BST-2-regulated heterotypic interactions between cancer cells and the tumor microenvironment may promote invadopodia formation, intravasation, extravasation, and metastasis of breast cancer cells. Such BST-2-regulated interactome needs to be identified for development of efficient "stop" signals for metastasis.

Materials and Methods
Ethics approval and consent to participate. The University of Iowa Institutional Review Board (IRB) approved the use of de-identified human tissues and clinical data while the Institutional Animal Care and Use Committee (IACUC) approved the use of mice in this study. All experiments were performed in accordance with the approved University guidelines and regulations. Patient cohort. Dataset and tumor bank 29 from the University of Iowa Department of Pathology containing de-identified 150 females diagnosed with or treated for invasive breast cancer at The University of Iowa hospitals and Clinics between 1986 and 1989 was used in this study. During this time, long-term follow-up data was collected including but not limited to age at time of diagnosis, tumor histological type, tumor grade, stage at diagnosis, lymph node status, cancer recurrence, vital status, date of death, and cause(s) of death. These clinical data were obtained from the Iowa Cancer Registry Database following review of patient charts. Mortality data was obtained from the Iowa Department of Public Health. Out of 150 tissue blocks in this dataset, we used 79 blocks as the other blocks had insufficient tissues for immunohistochemistry analyses.
SCIentIfIC RepoRts | (2018) 8:17608 | DOI:10.1038/s41598-018-35710-y Pathology analysis. All 79 tissue blocks were stained and analyzed by a single pathologist (AB) who was blinded to the clinical data. Histological type and tumor grade data was gathered using Elston-Ellis classification 29 .

BST-2 Immunohistochemistry of human IBC tissue samples. BST2 expression in breast cancer
specimens was assessed using Tissue Microarrays (TMAs) constructed from paraffin-embedded, formalin-fixed breast cancer tissue. Tissue sections were deparaffinized and rehydrated in graded alcohols. Heat mediated antigen retrieval was performed and immunohistochemical staining for BST2 was performed using Dako Envison + Rabbit Peroxidase Detection System (Dako Cytomation) using a rabbit monoclonal anti-BST2 antibody (Abcam, MO, USA). Staining was performed according to manufacturers' protocols. Immunostains were scored according to the intensity of the staining (no staining = 0, weak staining = 1+ , moderate staining = 2+ , strong staining = 3+) and the percentage of cells staining. Staining is considered positive if equal or more than 10% (≥10%) of the tumor cells show staining.
3D Invasion assay. 24-well cell culture inserts (Merck Millipore) were coated with 100 µl of Matrigel at 1.5 mg/ml (Sigma-Aldrich) and incubated at 37 °C for 3 hours to allow solidification. 100,000 MDA-MB-231 shCTL or shBST-2 cells were starved for 4 hours, suspended in 0.1% FBS medium and were plated on top of the Matrigel layer in 100 µl. The basal chamber of the unit was covered with 600 µl of 30% FBS medium and 5 μg/ml fibronectin (Sigma-Aldrich). Cells were allowed to invade for 24 hours at 37 °C at this point inserts were collected and processed for analysis as previously described 4,7 .
3D Migration assay. The migration assay was performed in a similar way as the 3D Invasion assay except that the 24-well cell culture inserts were not coated with Matrigel.
Assessment of cell invasion and growth using VitroGel 3D. 10  Induction of anoikis and assessment of cell viability. Spheroids were formed and collected as described under "Evaluation of cell survival by flow cytometry. " Single cells from the spheroids were gathered using 1x Cell dissociation buffer according to the manufacturer's instructions. Single cells were then used in a MTT assay or in a Trypan blue viability assay as previously described 7 .
Formation of invasive structures. Protocol for colony formation was adopted from 56 . Briefly, a 35 mm cell culture dishes with glass bottom were coated with 100 μl of Matrigel (10 mg/ml). Plates were incubated at 37 °C for 30 minutes for the Matrigel to solidify. 4T1-shCTL and shBST-2 cells were harvested, and re-suspended in 1 ml of serum free RPMI to count the cells. Cells were counted and re-suspended in 1:1 of Matrigel and serum free RPMI in a total volume of 100 µl. Then, 1500 cells (in 100 µl of 1:1 Matrigel and RPMI) were seeded on the pre-coated Matrigel plates. Cells were allowed to embed in Matrigel by incubating the plate at 37 °C for 3 hours. After 3 hours, 2 ml of complete RPMI was added to the dish. Formation of invasive structures and cell colonies were monitored by acquiring images using Olympus ix80 inverted microscope with integrated high precision focus drive or HC PL APO CS2 20x/0.75 Immersion (IMM) objective with Type F immersion oil, utilizing a 6x mechanical zoom. The percentage of stellate colonies relative to total number of colonies formed were assessed at 0 and 72 hours.
Analysis of enzyme kinetics of conditioned medium using fluorogenic DQ ™ -gelatin assay. Gelatinase activity was measured using modified fluorogenic DQ ™ -gelatin assay according to manu- working solution was then removed from the water bath and centrifuged at 10,000 × g for 5 minutes to remove aggregates. Coverslip well bottoms were coated with 90 µl of pre-warmed 1:1 gelatin working solution and allowed to sit at room temperature for 10 minutes. Gelatin was cross-linked with glutaraldehyde (Sigma, catalog #340855), prepared in a 0.1% solution in cold PBS for 20 minutes on ice, then washed with Dulbecco's PBS, and reduced (15 minutes) with 5 mg/ml of sodium borohydride. This was followed by extensive Dulbecco's PBS washing, then sterilization with a 30 minute incubation with 70% Ethanol using aseptic technique in a Class II Biosafety Cabinet (BSC). Plates were stored aseptically in the dark at 4 °C with 10x Penicillin-Streptomycin in PBS until used.

Gelatin Degradation Assay. Gelatin-Coated
Mattek plates were rinsed with Dulbecco's PBS, follow by 1 hour incubation in Phenol-free RPMI media (Gibco, catalogue #11835) in the cell culture incubator for equilibration. Both shCTL and shBST-2 cells were trypsinized, counted, and equivalent numbers (120,000) of cells were plated in each well. The culture plates were placed in the Lionheart FX Live-cell imaging microscope, equipped with full temperature and CO 2 control to maintain 37 °C and 5% CO 2 . Z-stack images were acquired 1, 3, 5, and 21, 23, and 25 hours after plating for 5 fields of view per well at 20x magnification. The total area of gelatin degradation was quantified using Gen5 software (BioTek). The values over time for 5 fields of view for each well were averaged together, which were then averaged again across the 12 wells for shCTL and shBST-2 cells.
Immunofluorescence staining, immunofluorescence microscopy, and image analysis. Cells  shBST-2 or OE BST-2D cell in 100 µl. Prior to luminescence imaging, mice were anesthetized and injected intraperitoneally with D-luciferin (Sigma-Aldrich) as previously described 7 . The Xenogen IVIS three-dimensional optical imaging system (Caliper Life Sciences) was used to image injected mice. Mice were sacrificed when they became moribund or at the end of the experiment. At necropsy, pulmonary nodules were manually quantified.
Competitive experimental pulmonary metastasis. 8-week-old BALB/cAnNCr females were purchased from Harlan and used in competitive experimental metastasis studies. Mice were injected via the tail vein with 2 × 10 6 4T1 shCTL, 2 × 10 6 shBST-2, and 1 × 10 6 beads in 200 µl of PBS. shCTL and shBST-2 cells were previously stained with CellTracker Green CMFDA and CellTracker Red CMTPX respectively following manufacturer's (Thermo Fisher) instructions. Following injection, mice were euthanized at 0, 2, 4, and 6 hours at which point blood was collected in heparin-coated tubes (Microtainer) and lungs were collected, inflated with 1:1 OCT:PBS and flash frozen in OCT blocks.

Assessment of circulating tumor cell (CTC) numbers in the blood.
Blood was collected at 0, 2, 4, and 6 hours post injection from mice injected with shCTL, shBST-2 and beads as described previously. Blood was processed using Histoplaque-1077 (Sigma Aldrich) to collect PBMCs and CTCs according to the manufacturer's instructions. Isolated cells were re-suspended in FACS buffer and stained with either PE-conjugated anti-mouse EPCAM antibody (BioLegend) or appropriate immunoglobulin Gs (IgGs) for 1 hour at 4 °C. Cells were washed once and stained with 7-AAD viability dye for 15 minutes. Using FACS Calibur flow cytometer, 100,000 events were collected per sample. FACS data were analyzed by Flowjo software.
Evaluation of CTC survival by flow cytometry. Following PBMC isolation as described in the previous paragraph, cells were stained with PE-conjugated anti-mouse EPCAM antibody, washed once and re-suspended in Annexin V binding buffer and then cells were stained with Annexin V and 7-AAD using the eGFP Annexin V and PI Apoptosis kit (GeneCopoeia). At least 100,000 events were collected per sample using the FACS Calibur flow cytometer and data were analyzed using FlowJo software. Percent of live cells was determined by gating for EPCAM + , Annexin Vand 7-AAD − cells while dead CTCs were determined by gating for EPCAM + , Annexin + , and/or 7-AAD + cells.
Cryo sectioning, fixation, and mounting of lung slides. Frozen blocks of lungs were cut using a Thermo Microm HM 550 Cryostat (Thermo Scientific). 20 µm-thick slices were collected every 200 µm for a total of 20 slices per lung and they were fixed in 4% PFA for 30 minutes. Following fixation, slides containing lung slices were washed twice with 1x PBS and mounted using DAPI-containing VectaShield (Vector Laboratories Inc.). Lung slices were then covered with coverslips. Lung slides were imaged at 10x using a Nikon Eclipse Ti microscope adjusted with a Nikon digital sight camera. Images were processed, and lung associated shCTL, shBST-2, and beads counted using ImageJ software.
Identification of shCTL and shBST-2 cells in the lungs. To identify and distinguish shCTL, shBST-2 and beads from each other, we used fluorescent and bright field imaging. In fluorescent mode, shCTL cells appear DAPI+ and CMFDA+ (green) and shBST-2 cells appear DAPI+ and CMTPX+ (red). Beads which are not cells are DAPI-with auto-fluorescence in the red channel. In bright field mode, beads appear light blue. Thus, to identify shCTL and shBST-2 cells, they must fulfill the following criteria: 1) be DAPI+ , 2) be green (shCTL) or red (shBST-2), 3) be smaller than beads, and 4) be present within the alveolar tissue. With this criteria in mind, we quantified shCTL and shBST-2 cells that were present in the lung at different time points until we counted a total of 100 beads per time point.
Statistics. The GraphPad Prism software was used to perform statistical analysis of significant differences.
We used unpaired t tests assuming Gaussian distribution with Welch's correction to analyze all data except patient cohort data. For the later, we used unpaired t tests assuming both populations have the same standard deviation. Error bars correspond to standard error of the mean (SEM) for all data except for transcription data in which case we used standard deviation. To analyze Kaplan-Meier survival plots, we used the Gehan-Breslow-Wilcoxon test. A probability (P) value equal to or lower than of 0.05 was considered significant.