Although most breast cancer metastases in bone cause osteolytic lesions, the osteogenic niche has commonly been described as an initiator of early-stage bone colonization of disseminated cancer cells. Tumor cell–derived extracellular vesicles (EV) have been shown to determine the organotropism of cancer cells by transferring their cargo, such as nucleic acids and proteins, to resident cells at future metastatic sites and preparing a favorable premetastatic niche. Runt-related transcription factor 2 (RUNX2) and its regulated genes have been shown to facilitate the acquisition of osteomimetic features and to enhance the bone metastatic potential of breast cancer cells. In this study, we present in vivo and in vitro evidence to clarify the role of EVs released by breast cancer cells with high RUNX2 expression in the education of osteoblasts to form an osteogenic premetastatic niche. Furthermore, different extracellular vesicular proteins were identified that mediate events subsequent to the specific recognition of tumor-derived EVs by osteoblasts via cadherin 11 (CDH11) and the induction of the osteogenic premetastatic niche by integrin α5 (ITGA5). CDH11high/ITGA5high EVs were demonstrated to be responsible for the formation of a premetastatic niche that facilitates RUNX2 high-expressing breast cancer cell colonization in bone, revealing a potential EV-based premetastatic niche blockage strategy.

Significance:

This study provides mechanistic insights into the generation of an osteogenic premetastatic niche by breast cancer–derived EVs and identifies potential EV-derived diagnostic biomarkers and targets for breast cancer bone metastasis.

Bones are preferentially colonized by disseminated tumor cells in patients with breast cancer (1, 2). The bone tropism of breast cancer is determined by the cross-talk between tumor cells and the bone microenvironment, which has been described by the “seed and soil” theory (3). To prepare a fertile “soil” for metastatic tumor cell (“seed”) colonization, a premetastatic niche can be induced by primary cancer cells before their arrival (4, 5).

In recent years, tumor-derived extracellular vesicles (EV) have been increasingly recognized as contributors to organotropic metastasis (6, 7). EVs are classified on the basis of their size and origin as ectosomes and exosomes (8, 9). Once tumor-derived EVs are internalized by organ-specific resident cells, the nucleic acids and proteins carried by EVs can contribute to the formation of a premetastatic niche (10, 11). Among the cargo proteins within EVs, integrins on the surface of exosomes are key in the determination of organotropic metastasis (12). Hoshino and colleagues (12) found that exosomal integrin α6β4 and α6β1 were associated with lung metastasis, while exosomal integrin αvβ5 was linked to liver metastasis. However, to date, no bone-tropic protein cargo of EVs has been reported.

Runt-related transcription factor 2 (RUNX2) is a critical transcription factor for osteogenic lineage commitment and bone formation. RUNX2 has been found to be ectopically overexpressed in breast cancer with high bone metastatic potential (13). In our previous studies, RUNX2 and its regulated genes were demonstrated to enhance the bone metastatic potential of breast cancer “seeds” by facilitating the acquisition of osteomimetic features (14), promoting chemotactic migration toward osteoblasts, providing a survival advantage in the osteogenic microenvironment and further activating osteoclasts (15, 16). Accumulating evidence indicates that tumor–osteoblast interactions are critical in establishing bone metastasis (17–19). Micrometastases are formed in regions of new bone formation, where differentiating and actively mineralizing osteoblasts are located, supporting the notion that the osteogenic niche promotes cancer cell proliferation (17). Thus, we wondered whether breast cancer cells with high RUNX2 expression could induce the formation of an osteogenic premetastatic niche through their EVs before their arrival.

In this study, we used breast cancer cells with high RUNX2 expression as bone-tropic “seeds” due to their high bone metastatic potential, and we present in vivo and in vitro evidence to clarify the role of EVs released by breast cancer cells with high RUNX2 expression in osteoblast education to prepare a premetastatic niche. Furthermore, we identified extracellular vesicular integrin α5 (ITGA5), a transcriptional target of RUNX2 (16), as a mediator of osteoblast education by RUNX2 high-expressing cancer cells, and extracellular vesicular cadherin 11 (CDH11; ref. 20), an osteoblast cadherin that is ectopically expressed in breast cancer cells, as contributors to the internalization of tumor-derived EVs by osteoblasts. Ultimately, cancer-derived extracellular vesicular CDH11 and ITGA5 were identified to be responsible for the formation of the premetastatic niche that facilitates RUNX2 high-expressing breast cancer cell colonization in bone.

Cells

Human breast cancer MDA-MB-231 (MDA231), BT-549 and MCF7 cells, 4T1 mouse breast cancer cells, and preosteoblast MC3T3-E1 cells were obtained from the ATCC. MDA231, BT-549, and 4T1 cells were maintained in RPMI1640 medium (Thermo Fisher), MCF7 cells were maintained in DMEM (Thermo Fisher), and MC3T3-E1 cells were maintained in minimum essential medium (MEM) alpha (αMEM; Thermo Fisher). DMEM and RPMI1640 media were supplemented with 10% FBS (Thermo Fisher), 100 U/mL penicillin, and 100 μg/mL streptomycin. All cell lines were authenticated by short tandem repeat profiling and tested for Mycoplasma contamination.

Primary mouse osteoblasts (mOB) were isolated from the cranial bone of newborn BALB/c mice (GemPharmatech). Briefly, newborn mouse cranial bone was cut into small pieces, which were then digested with 0.25% trypsin containing 0.02% ethylenediaminetetraacetic acid at 37°C for 20 minutes. The supernatant was discarded, and the precipitate was then digested twice with 2 mg/mL collagenase II at 37°C for 20 minutes. Cell suspensions were collected, resuspended and cultured in αMEM with 10% FBS. Primary mOBs were identified by alkaline phosphatase (Alp) staining and Alizarin S staining after culturing in osteogenic differentiation media (αMEM with 10% FBS, 50 μg/mL L-ascorbic acid, and 10 mmol/L β-glycerophosphate disodium) for 5 days and 10 days, respectively.

Lentivirus construction and infection

Lentiviruses containing fusions of flag-tag with full-length human RUNX2, human CDH11, mouse Runx2, and mouse Cdh11, Tet-on inducible lentivirus with human ITGA5-flag and mouse Itga5-flag and their corresponding control lentiviruses were constructed and produced by GeneChem Co. Ltd. All lentiviruses contained a GFP reporter driven by a separate cytomegalovirus promoter. Cells were infected with lentivirus at a multiplicity of infection (MOI) of 20 for MDA231 cells, an MOI of 50 for BT-549 cells, and an MOI of 200 for MCF7 cells and 4T1 cells. Breast cancer cells with stable overexpression (OE) of the indicated genes and their corresponding control cells (Control) were obtained by puromycin selection. Human ITGA5 and mouse Itga5 were induced OE by 1 μg/mL doxycycline. In addition, stable RUNX2 knockdown (RUNX2-KD) MDA231 cells (shRUNX2) and control cells (shControl) were acquired by infection with lentiviruses expressing short hairpin RNA (shRNA) targeting human RUNX2 (targeting sequence CAAGGACAGAGTCAGATTA; GeneChem) and control shRNA. All indicated proteins were identified by Western blotting.

RNA interference

Small interfering RNAs (siRNA) targeting human ITGA5 (TGGCTCAGACATTCGATCC), human CDH11 (CAACGCAGAGGCCTACATT and CCTCGAAGGACAACCCTAT), and mouse Cdh11(GACTACGACTATCTACAGA) were synthesized by RiboBio Co. (Guangzhou, China). Negative siRNA control was used as a control (siControl). siRNAs were transfected into cells using Lipofectamine 2000 (Thermo Fisher) according to the manufacturer's instructions. Breast cancer cells with KD of the indicated proteins were identified by Western blotting.

EV isolation and characterization

EV-free media were used to collect the conditioned media when the cancer cells were grown to 80% confluence. EVs were isolated from the conditioned culture media using an exoEasy Kit (Qiagen), quantified by measuring the extracellular vesicular protein concentration using Bradford dye reagent (Bio-Rad) and characterized by electron microscopy (Hitachi Ltd.) and immunoblotting analysis for CD9 or TSG101. EV size and particle number were analyzed using a ZetaView nanoparticle tracking analyzer (NTA, Particle Metrix GmbH).

Immunoelectron microscopy

Immunoelectron microscopy was used to identify the location of flag-labeled CDH11 on EVs. Briefly, EV suspensions were mixed 1:1 with 4% paraformaldehyde and then dropped on 200-mesh grids. After blocking with 1% BSA and washing, the grid was incubated with mouse anti-flag primary antibody (1:200, Sigma-Aldrich) for 2 hours at room temperature. Afterward, grids were washed with PBS followed by an additional incubation with the secondary antibody labeled with 6-nm gold particles (1:100, Sigma-Aldrich) for 1 hour at room temperature. After washing, membranes underwent negative staining. After drying, the grids were examined with electron microscopy.

EV uptake analysis

For the in vitro assay, mOB and MC3T3-E1 cells were scraped off (rather than digested by trypsin) and seeded into 24-well plates at an initial density of 1 × 105 cells/well. Twenty-four hours later, EVs were fluorescently labeled with PKH67 fluorescent dye (Sigma-Aldrich). The labeled EVs were washed in PBS and purified using an exoEasy Kit (Qiagen) to remove residual-free PKH67 fluorescent dye. The labeled EVs were then added to the indicated cell culture (1 μg/1 × 105 cells) and incubated for 6 hours or 24 hours in FBS-free αMEM. For 10-day EV uptake analysis, PKH67-labeled EVs were added in mOB and MC3T3-E1 cell culture for 10 days in αMEM with 10% FBS and changed media every other day. Then, the mOB and MC3T3-E1 cells were washed with PBS for microscopic observation or scraped off for flow cytometric analysis.

For the in vivo assay, 20-μg PKH67-labeled EVs were injected into 5-week-old female NOD/SCID mice via the tail vein every other day for 3 times. On day 6, the mice were sacrificed to prepare frozen lung and bone (tibias and femurs) sections. The frozen sections were then used for PKH67 fluorescence observations. In addition, 10-μg flag-labeled EVs were injected into 5-week-old female NOD/SCID mice via the tail vein for 3 weeks (2 doses/week), followed by sacrificing the mice to obtain frozen bone slices. Next, immunofluorescence staining for flag was used to trace the sublocation of EVs. Immunofluorescence staining for osteocalcin (Ocn) was performed to identify mouse osteoblasts, and 4′,6′-diamidina-2-phenylindole (DAPI) was used to stain the nuclei.

Staining for Alp and calcium nodules

Alp staining for adherent osteoblasts and frozen slices was performed using a Gomori Modified Calcium-Cobalt Staining Kit (Solarbio, Beijing, China). Calcium nodules were detected using Alizarin S (Solarbio) staining.

Chemotactic migration assay

For the in vitro assay, the chemotactic migration of breast cancer cells toward a mimic EV-induced osteoblastic microenvironment was assessed using 8-μm pore Transwell inserts (BD Biosciences, Franklin Lakes, New Jersey). Primary mOBs and osteoblast progenitor MC3T3-E1 cells were seeded in the lower chambers, treated with cancer-derived EVs (1 μg/1 × 105 cells, 3 times/week) for 10 days and then used to attract the cancer cells from the upper chambers for 13 hours (MDA231), 9 hours (BT-549), or 18 hours (MCF7). The migrating cells were stained with Giemsa (BBI, Life Sciences) and counted under a microscope. For MDA231 and BT549 cells, the migrating cells in 6 random microscope fields of three independent experiments were counted (cells/field). And for MCF7 cells, the total number of cells migrating through the insert membrane in the chamber of three independent experiments were counted (cells/well).

For the in vivo chemotactic migration assay, 10 μg of EVs was injected into 5-week-old NOD/SCID mice via the tail vein twice a week for 3 weeks. Three days later, 5 × 105 GFP+ cancer cells were inoculated into EV-treated mice by intracardiac injection. After 24 hours, the mice were sacrificed to flush the bone marrow cells from the femurs and tibias or to be used to prepare continuous frozen bone sections. Then, GFP-labeled cancer cells in the bone marrow were counted by flow cytometry. For frozen bone sections, immunofluorescence staining for Ocn was performed to identify osteoblasts, and the location of GFP+ cancer cells relative to osteoblasts in bone was observed by fluorescence microscopy.

3D coculture

MC3T3-E1 cells were treated with cancer-derived EVs (1 μg/1 × 105 cells, 3 times/week) for 10 days and then fluorescently labeled with PKH26 fluorescent dye (Sigma-Aldrich) and cocultured with the corresponding cancer cells in Matrigel. Briefly, 400 μL of Matrigel matrix (BD Biosciences) was precoated into the wells of 24-well plates. A total of 1 × 105 PKH26-labeled MC3T3-E1 cells and GFP+ cancer cells (2.5 × 104 MDA231 cells, 2.5 × 104 BT-549 cells, or 3.5 × 104 MCF7 cells) were mixed and resuspended in 500 μL of 10% Matrigel supplemented with 10% FBS. The cells were continuously cultured for 2 to 3 days, and cell morphology was observed with fluorescence microscopy.

Flow cytometry

All flow cytometric analyses were performed on an LSR Fortessa (BD Biosciences). For in vitro EV uptake analysis, scraped mOBs and MC3T3-E1 cells were washed with PBS, and PKH67+ cells were analyzed. For the in vivo chemotactic migration assay, the bone marrow cells were washed with PBS and fixed with 1% paraformaldehyde after removing the red blood cells (RBC) by using RBC lysis solution (Sigma-Aldrich). Subsequently, GFP+ cancer cells were detected and counted.

To investigate the uptake capability of cancer-derived EVs by MC3T3-E1 cells with different Cdh11 statuses, Cdh11-KD MC3T3-E1 cells were mixed with parental MC3T3-E1 cells in equal proportions and treated with PKH67-labeled EVs for 24 hours. Then, MC3T3-E1 cells were scraped off, washed with PBS, and blocked with 2% mouse serum. PE-conjugated anti-mouse Cdh11 monoclonal antibodies (BioLegend, San Diego, California) were added and incubated with the cells for 1 hour to determine Cdh11 expression statuses. Then, the proportions of PE+/PKH67+ cells were measured by flow cytometry.

Mass spectrometry

The proteins carried by EVs derived from MDA231 and MCF7 cells were analyzed by liquid chromatography-mass spectrometry (LC-MS). LC-MS analyses were performed at GeneChem Co. Ltd. Considering the batch variation within the same specimen, we pooled EVs from three independent isolations in equal proportions of extracellular vesicular proteins into one sample. The abundance of proteins was normalized overall, differentially expressed proteins were identified, and heatmaps were used to visualize the expression levels of differentially expressed proteins. The differentially expressed proteins were further analyzed in the UniProt database, and Gene Ontology analysis was performed to annotate the function and subcellular location of the proteins.

Immunofluorescence

For frozen sections, slices were fixed in cold acetone for 10 minutes, washed in ice-cold PBS, and blocked in 2% BSA in PBS for 1 hour. The slices were incubated overnight at 4°C with rabbit anti-Ocn antibodies (1:500; Abcam) or mouse anti-flag antibodies (1:500; Sigma-Aldrich). The slices were washed with PBS and incubated with Alexa Fluor 594-labeled anti-rabbit secondary antibody (BD Biosciences) or FITC-labeled anti-mouse secondary antibody (BD Biosciences) at room temperature for 1 hour. For adherent mOB or MC3T3-E1 cells labeled with PKH26 fluorescent dye (Sigma-Aldrich), cells were fixed in 4% paraformaldehyde for 10 minutes and then blocked in 2% BSA in PBS for 1 hour. The cells were incubated with monoclonal antibodies against Flag (1:1,000; Sigma-Aldrich) overnight at 4°C and washed three times with PBS, after which, the cells were incubated with FITC-labeled anti-mouse secondary antibodies (BD Biosciences) at room temperature for 1 hour. After nuclear staining with ProLong Gold Antifade Mountant with DAPI (Invitrogen) and slide sealing, the cells were observed, and images were taken using a fluorescence microscope (Carl Zeiss, Oberkochen, Germany).

Quantitative reverse transcription PCR

Total RNA was extracted from EV-treated MC3T3-E1 cells with TRIzol reagent (Invitrogen) and reverse transcription was performed using the cDNA synthesis system (Invitrogen). Quantitative PCR was performed using the SYBR Premix Ex Taq Kit (Takara, Dalian, China) with an ABI 7500 system (Applied Biosystems, Foster City, California). The primers for osteopontin (Opn) were GTGATAGCTTGGCTTATGGAC and GCCCTTTCCGTTGTTGTC, and the primers for bone sialoprotein (Bsp) were CAGAGGAGGCAAGCGTCACT and CCCGTTCTCGTTGTCATAGAC. The mRNA expression levels of Opn and Bsp were normalized to housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (Gapdh). The primers for Gapdh were ATTGTCAGCAATGCATCCTG and ATGGACTGTGGTCATGAGCC. The relative quantification of gene expression was determined by three independent experiments.

Western blotting

Protein lysates from the cells were electrophoresed on sodium dodecyl sulfate–polyacrylamide gels and transferred to polyvinylidene difluoride membranes. Then, the membranes were incubated with primary antibodies (listed in Supplementary Table S1) followed by the corresponding secondary antibodies (Cell Signaling Technology, Danvers, Massachusetts). The bands were visualized with an enhanced chemiluminescence detection system (Promega, Madison, Wisconsin).

Animal experiments

Four-week-old female NOD/SCID mice and BALB/c mice were purchased from GemPharmatech (Nanjing, China) and maintained in a pathogen-free environment. Breast cancer cell MDA231-derived EVs (10 μg in 100 μL of PBS) were injected into NOD/SCID mice by the tail vein for 3 weeks (2 doses/week), followed by intracardiac injection of 5 × 104 luciferase+ MDA231 RUNX2-OE cells or Control cells. Lung metastasis was assessed by in vivo bioluminescence imaging using a Xenogen IVIS 200 Imaging System (Caliper Life Sciences, Hopkinton, Massachusetts) and hematoxylin and eosin (H&E) staining. Bone metastasis was evaluated by in vivo bioluminescence imaging and X-ray. For spontaneous bone metastasis of mouse-derived breast cancer 4T1 cells, 4T1-derived EVs (10 μg in 100 μL of PBS) were injected into BALB/c mice by the tail vein for 2 weeks (2 doses/week), followed by orthotopic injection of 1 × 103 4T1 RUNX2-OE cells into the mammary fat pad. To extend the observation period, tumor resection surgery for BALB/c mice was performed to remove the visible tumor in situ on day 21. The number of bone lesions per mouse and the bone lesion area (erosion surface/bone surface) were determined by X-ray. Metastases in the lung and bone were further confirmed by H&E staining. Tartrate-resistant acid phosphatase (TRAP) staining was used to identify mature osteoclasts in metastatic bone lesions. The activity of osteoclasts was determined by the ratio of osteoclast (TRAP+, ≥3 nuclei) number to the bone surface. All animal experiments complied with the guidelines of animal ethics, and the protocols were approved by the Animal Ethics Committee of Tianjin Medical University Cancer Institute and Hospital.

Gene expression profiling dataset

The published breast cancer gene expression profiling dataset GSE2034 (21), which includes 286 primary breast cancer tissues from the Erasmus Medical Center, was obtained from the Gene Expression Omnibus database. The data were log-transformed and normalized. RUNX2, CDH11, and ITGA5 were identified as high-expression (RUNX2high, CDH11high, and ITGA5high) or low-expression (RUNX2low, CDH11low, and ITGA5low) according to their mRNA expression levels. Average clustering was performed to investigate the relationship between the mRNA expression levels of CDH11 and ITGA5 and bone metastasis in RUNX2high primary breast cancer samples (n = 141).

Statistical analysis

The data are presented as the means ± SD of at least three independent experiments. Statistical analyses were performed using GraphPad Prism 8 software. Kaplan−Meier survival curves and log-rank tests were used to evaluate the overall survival of mice in the animal experiments. χ2 tests or Fisher exact tests were used to estimate the differences in the incidence of lung and bone metastasis in mice educated with different EVs. All other comparisons were performed using repeated measures ANOVA or two-tailed Student t tests.

Data availability statement

The data generated in this study are available within the article and its supplementary data files. Additional data related to this paper are available upon request from the corresponding author.

RUNX2 high-expressing breast cancer–derived EVs facilitate colonization of cancer cells with high RUNX2 expression in bone

The particle sizes of MDA231-derived EVs isolated from RUNX2-OE cells (RUNX2 EVs) were slightly larger than the sizes of EVs isolated from Control cells (Control EVs; Fig. 1A–C). No significant differences were observed in the morphology or production of MDA231 RUNX2 EVs and Control EVs (Fig. 1B–C), although MDA231 RUNX2-OE cells proliferated slightly faster than Control cells (Supplementary Fig. S1A). RUNX2-KD using shRNA did not significantly affect the proliferation capability or EV production of MDA231 cells (Supplementary Fig. S1B–S1D).

Figure 1.

EVs produced by breast cancer cells with high RUNX2 expression facilitate bone colonization by cancer cells with high RUNX2 expression. A, Western blotting of RUNX2 protein levels in MDA231 RUNX2-OE and Control cells. B, Representative electron microscopy imaging of MDA231 RUNX2 EVs and Control EVs. C, Histograms showing the size distribution and concentration of EVs (EV Conc.). D, Diagram of the animal experiment for bone colonization of MDA231-derived cells in SCID mice. E, Kaplan−Meier survival analysis for the overall survival of mice. F, Representative bioluminescence images displaying pulmonary colonization and bone colonization on day 0 (D 0), day 21 (D 21), and day 45 (D 45); representative anatomic images, X-ray images, and H&E staining images showing the lesions of lung and bone; and representative TRAP staining of the activated osteoclasts. G, Histogram showing the incidences of pulmonary colonization and bone colonization. H and I, Histograms of the erosion surface (H) and TRAP+ osteoclast number (I) normalized to the bone surface. *, P < 0.05; **, P < 0.01 compared with the Control EV/Control cell group.

Figure 1.

EVs produced by breast cancer cells with high RUNX2 expression facilitate bone colonization by cancer cells with high RUNX2 expression. A, Western blotting of RUNX2 protein levels in MDA231 RUNX2-OE and Control cells. B, Representative electron microscopy imaging of MDA231 RUNX2 EVs and Control EVs. C, Histograms showing the size distribution and concentration of EVs (EV Conc.). D, Diagram of the animal experiment for bone colonization of MDA231-derived cells in SCID mice. E, Kaplan−Meier survival analysis for the overall survival of mice. F, Representative bioluminescence images displaying pulmonary colonization and bone colonization on day 0 (D 0), day 21 (D 21), and day 45 (D 45); representative anatomic images, X-ray images, and H&E staining images showing the lesions of lung and bone; and representative TRAP staining of the activated osteoclasts. G, Histogram showing the incidences of pulmonary colonization and bone colonization. H and I, Histograms of the erosion surface (H) and TRAP+ osteoclast number (I) normalized to the bone surface. *, P < 0.05; **, P < 0.01 compared with the Control EV/Control cell group.

Close modal

SCID mice were pretreated with MDA231-derived Control EVs or RUNX2 EVs via tail vein injection for 3 weeks, followed by intracardiac injection of MDA231 RUNX2-OE or Control cells (Fig. 1D). As shown in Fig. 1E, all mice (8 of 8) that were pretreated with Control EVs and injected with Control cells (Control EV/Control cell mice) died within 30 days after cancer cell injection. Pretreatment with RUNX2 EVs and RUNX2-OE in cancer cells significantly prolonged the survival of mice. The survival rates of mice that were pretreated with RUNX2 EVs and injected with RUNX2-OE cells (RUNX2 EV/RUNX2 cell mice) at 30 days and 90 days after cancer cell injection were 100% (7 of 7) and 57.1% (4 of 7), respectively.

Pulmonary colonization was observed in 100% (8 of 8) of mice in the Control EV/Control cell group (Fig. 1F and G). RUNX2-OE in cancer cells and pretreatment with RUNX2 EVs decreased the incidence of pulmonary colonization. Only 28.6% (2 of 7) of mice in the RUNX2 EV/RUNX2 cell group developed pulmonary colonization (Fig. 1F and G). Osteolytic lesions were observed in 12.5% (1 of 8) of mice in the Control EV/Control cell group, 28.6% (2 of 7) of mice treated with Control EVs and RUNX2-OE cells (Control EV/RUNX2 cell mice), 33.3% (3 of 9) of mice treated with RUNX2 EVs and Control cells (RUNX2 EV/Control cell mice) and 100% (7 of 7) of mice in the RUNX2 EV/RUNX2 cell group (Fig. 1F and G). Moreover, the erosion surface and number of TRAP+ osteoclasts in the RUNX2 EV/RUNX2 cell group were significantly increased compared with those in the other groups (Fig. 1H and I). Accordingly, mice injected with MDA231 RUNX2-OE cells and treated with EVs isolated from MDA231 shRUNX2 cells (shRUNX2 EVs) had slightly reduced survival, increased pulmonary colonization and reduced bone lesions compared with mice pretreated with the control EVs (shControl EVs), although the difference was not statistically significant (Supplementary Fig. S1E–S1I). These in vivo results demonstrate that EVs released by RUNX2 high-expressing MDA231 cells contribute to bone colonization of cancer cells with high RUNX2 expression.

RUNX2 EVs were preferentially internalized by osteoblasts

To identify the recipient cells of EVs in bone, PKH67-labeled MDA231 RUNX2 EVs and Control EVs were injected into SCID mice via the tail vein, and PKH67 fluorescence was detected in the bones and lungs of these mice (Fig. 2A). As shown in Fig. 2B, PKH67 fluorescence was colocalized with Ocn, a marker of osteoblasts, indicating that tumor cell–derived EVs were taken up by osteoblasts. Furthermore, MDA231 RUNX2 EVs were observed to be taken up more by osteoblasts than Control EVs in vivo (Fig. 2B). Moreover, increased Ocn expression and Alp activity were observed in osteoblasts of mice treated with MDA231 RUNX2 EVs compared with the Control mice (Fig. 2B and C), indicating the contribution of RUNX2 EVs to osteoblast differentiation. However, MDA231 RUNX2 EVs were not observed preferentially recruiting in lung epithelial cells (Fig. 2D). Furthermore, RUNX2 EVs from both MDA231 and BT-549 cells were observed to be preferentially taken up in vitro by primary osteoblast mOB (Supplementary Fig. S2A) and preosteoblast MC3T3-E1 cells compared with their corresponding Control EVs at 6 hours, 24 hours, and 10 days (Fig. 2E–G). Accordingly, the uptake of MDA231 shRUNX2 EVs by mOB and MC3T3-E1 cells was significantly decreased compared with that of shControl EVs (Supplementary Fig. S2B). Thus, these results suggest that RUNX2 EVs are preferentially internalized by osteoblasts to establish an osteogenic niche.

Figure 2.

RUNX2-OE cancer cell–derived EVs are preferentially internalized by osteoblasts. A, Diagram of the experiment used to identify the biodistribution of MDA231-derived EVs in the bone and lung of SCID mice. B, Representative images of PKH67 fluorescence in bone sections. Ocn was used as a marker of osteoblasts. C, Representative images of Alp staining. D, Representative images of PKH67 fluorescence in lung sections. E, Western blotting showing RUNX2 protein levels in BT-549 RUNX2-OE cells and Control cells. F and G, Quantification of MDA231-derived EVs (F) and BT-549-derived EVs (G) internalized by primary mOB or MC3T3-E1 cells at 6 hours (6 H), 24 hours (24 H) and 10 days (10 D). mOB or MC3T3-E1 cells without EV treatment were used as a negative control (NC). The red, green, and blue dashed lines show the peak fluorescence of the Control EV group at 6 hours, 24 hours, and 10 days, respectively. *, P < 0.05; **, P < 0.01; ***, P < 0.001 compared with the Control EV group.

Figure 2.

RUNX2-OE cancer cell–derived EVs are preferentially internalized by osteoblasts. A, Diagram of the experiment used to identify the biodistribution of MDA231-derived EVs in the bone and lung of SCID mice. B, Representative images of PKH67 fluorescence in bone sections. Ocn was used as a marker of osteoblasts. C, Representative images of Alp staining. D, Representative images of PKH67 fluorescence in lung sections. E, Western blotting showing RUNX2 protein levels in BT-549 RUNX2-OE cells and Control cells. F and G, Quantification of MDA231-derived EVs (F) and BT-549-derived EVs (G) internalized by primary mOB or MC3T3-E1 cells at 6 hours (6 H), 24 hours (24 H) and 10 days (10 D). mOB or MC3T3-E1 cells without EV treatment were used as a negative control (NC). The red, green, and blue dashed lines show the peak fluorescence of the Control EV group at 6 hours, 24 hours, and 10 days, respectively. *, P < 0.05; **, P < 0.01; ***, P < 0.001 compared with the Control EV group.

Close modal

RUNX2 EV-educated osteoblasts promoted the recruitment and aggressive spread of cancer cells

To identify the roles of RUNX2 EVs in educating osteoblasts, SCID mice were pretreated with MDA231 RUNX2 EVs and Control EVs, followed by intracardiac injection of GFP-labeled MDA231 RUNX2-OE cells (Fig. 3A). GFP+ cancer cells were more significantly enriched in the bone marrow of mice that were pretreated with RUNX2 EVs than in the bone marrow of Control mice (Fig. 3B). GFP+ cancer cells were observed near osteoblasts whether the mice were pretreated with Control or RUNX2 EVs (Fig. 3C), implying the potential interaction of osteoblasts and cancer cells.

Figure 3.

Osteoblasts educated with EVs derived from RUNX2-OE cancer cells promote the recruitment and aggressive spread of RUNX2-OE cancer cells. A–C, Recruitment and distribution of MDA231 RUNX2-OE cells in the bone marrow of EV-treated SCID mice. A, Diagram displays the experimental process. B, Representative pseudocolor images and chart show the frequencies of GFP+ cells in the bone marrow of mice treated with MDA231 RUNX2 EVs or Control EVs. C, Representative immunofluorescence images show the location of GFP+ cancer cells in the bone marrow. Ocn was used as a marker of osteoblasts. D and E, Representative images and plots showing the chemotactic migration of MDA231 RUNX2-OE (D) or BT-549 RUNX2-OE (E) cells and their Control cells toward mOB or MC3T3-E1 cells pretreated with EVs. F and G, Representative 3D coculture images of EV-treated MC3T3-E1 (PKH26-labeled) cells and GFP-labeled MDA231 RUNX2-OE (F) or BT-549 RUNX2-OE (G) cells or their Control cells. *, P < 0.05; **, P < 0.01; ***, P < 0.001 compared with the Control EV group.

Figure 3.

Osteoblasts educated with EVs derived from RUNX2-OE cancer cells promote the recruitment and aggressive spread of RUNX2-OE cancer cells. A–C, Recruitment and distribution of MDA231 RUNX2-OE cells in the bone marrow of EV-treated SCID mice. A, Diagram displays the experimental process. B, Representative pseudocolor images and chart show the frequencies of GFP+ cells in the bone marrow of mice treated with MDA231 RUNX2 EVs or Control EVs. C, Representative immunofluorescence images show the location of GFP+ cancer cells in the bone marrow. Ocn was used as a marker of osteoblasts. D and E, Representative images and plots showing the chemotactic migration of MDA231 RUNX2-OE (D) or BT-549 RUNX2-OE (E) cells and their Control cells toward mOB or MC3T3-E1 cells pretreated with EVs. F and G, Representative 3D coculture images of EV-treated MC3T3-E1 (PKH26-labeled) cells and GFP-labeled MDA231 RUNX2-OE (F) or BT-549 RUNX2-OE (G) cells or their Control cells. *, P < 0.05; **, P < 0.01; ***, P < 0.001 compared with the Control EV group.

Close modal

In vitro, mOB and MC3T3-E1 cells educated with RUNX2 EVs derived from both MDA231 and BT-549 cells showed remarkable chemotactic capability compared with the corresponding RUNX2-OE cancer cells (Fig. 3D and E). MDA231 shRUNX2 EV-educated mOB and MC3T3-E1 cells, however, exhibited decreased attraction to MDA231 RUNX2-OE cells (Supplementary Fig. S2C). In a 3D coculture system, Control EV-treated MC3T3-E1 cells surrounded RUNX2-OE cancer cells to form large dense cell clusters (Fig. 3F and G), implying that the cancer cells were restricted from spreading by the osteoblasts and could not interact with the surrounding matrix. However, RUNX2-OE cancer cells were not completely surrounded by RUNX2 EV-educated MC3T3-E1 cells, instead forming small scattered cell clusters with exposed tumor cell branches at the edges (Fig. 3F and G), indicating that the RUNX2 EV-educated osteoblasts strengthened the aggressive spread of the cancer cells by increasing the potential connection of cancer cells with the bone matrix and other cells in the bone. Taken together, these results suggest that the osteogenic niche induced by RUNX2 EVs can further recruit cancer cells with high RUNX2 expression and support the spread of cancer cells in bone.

ITGA5 mediates the osteogenic niche induction by RUNX2 EVs

We screened 55 differentially expressed proteins between MDA231 RUNX2 EVs and Control EVs with at least a 3-fold change by MS (Supplementary Fig. S3). We further verified the enrichment of ITGA5, a transcriptional target of RUNX2 identified in our previous study (16), in both MDA231 RUNX2 EVs and BT-549 RUNX2 EVs in comparison with their Control EVs (Fig. 4A; Supplementary Fig. S4A).

Figure 4.

Cancer-derived extracellular vesicular ITGA5 contributes to the education of osteoblasts. A, Western blotting showing ITGA5 enrichment in MDA231 RUNX2 EVs. B, Western blotting showing ITGA5-flag OE in EVs derived from MDA231 ITGA5-OE cells. C and D, Internalization and redistribution of extracellular vesicular ITGA5-flag in MC3T3-E1 and mOB cells at 24 hours (24 H). E–G, Representative images of the morphology (E) and differentiation markers of MC3T3-E1 cells induced by MDA231 ITGA5-OE EVs and Control EVs for 10 days (10 D) determined by RT-qPCR (F) and Western blotting (G). H, Western blotting showing ITGA5-KD in EVs derived from MDA231 RUNX2/ITGA5-KD cells. I and J, The mRNA and protein levels of differentiation markers of osteoblasts were analyzed by RT-qPCR and Western blotting in MC3T3-E1 cells educated by MDA231-derived EVs for 10 days (10 D). K and L, Transwell assays of the chemotactic migration of MDA231 RUNX2-OE cells and Control cells toward MC3T3-E1 cells that were educated with MDA231-derived ITGA5-OE EVs (K) or RUNX2/ITGA5-KD EVs (L) and their corresponding Control EVs. **, P < 0.01; ***, P < 0.001 compared with their corresponding Control EV group. ###, P < 0.001 compared with MDA231 RUNX2 EV group.

Figure 4.

Cancer-derived extracellular vesicular ITGA5 contributes to the education of osteoblasts. A, Western blotting showing ITGA5 enrichment in MDA231 RUNX2 EVs. B, Western blotting showing ITGA5-flag OE in EVs derived from MDA231 ITGA5-OE cells. C and D, Internalization and redistribution of extracellular vesicular ITGA5-flag in MC3T3-E1 and mOB cells at 24 hours (24 H). E–G, Representative images of the morphology (E) and differentiation markers of MC3T3-E1 cells induced by MDA231 ITGA5-OE EVs and Control EVs for 10 days (10 D) determined by RT-qPCR (F) and Western blotting (G). H, Western blotting showing ITGA5-KD in EVs derived from MDA231 RUNX2/ITGA5-KD cells. I and J, The mRNA and protein levels of differentiation markers of osteoblasts were analyzed by RT-qPCR and Western blotting in MC3T3-E1 cells educated by MDA231-derived EVs for 10 days (10 D). K and L, Transwell assays of the chemotactic migration of MDA231 RUNX2-OE cells and Control cells toward MC3T3-E1 cells that were educated with MDA231-derived ITGA5-OE EVs (K) or RUNX2/ITGA5-KD EVs (L) and their corresponding Control EVs. **, P < 0.01; ***, P < 0.001 compared with their corresponding Control EV group. ###, P < 0.001 compared with MDA231 RUNX2 EV group.

Close modal

EVs enriched in ITGA5-flag fusion protein (ITGA5flag) derived from MDA231 and BT-549 cells were prepared via a doxycycline-inducible expression system (Fig. 4B; Supplementary Fig. S4B). The ITGA5flag was used to detect extracellular vesicular ITGA5 in MC3T3-E1 cells treated with ITGA5flag EVs (Fig. 4C; Supplementary Fig. S4C). Immunofluorescence staining showed that ITGA5flag was located in the cytoplasm and on the cell membrane of mOB and MC3T3-E1 cells (Fig. 4D; Supplementary Fig. S4D), indicating the redistribution of ITGA5 on the membrane of MC3T3-E1 cells and its potential activity as a membrane protein after EV internalization. When MDA231 ITGA5-OE EVs and Control EVs were used to induce MC3T3-E1 cells for 10 days, MC3T3-E1 cells educated by ITGA5-OE EVs spread more (Fig. 4E) and displayed higher osteogenic differentiation marker (Opn, Bsp, and Runx2) expression than the Control EV-treated MC3T3-E1 cells (Fig. 4F and G). In contrast, EVs derived from RUNX2-OE MDA231 cells with ITGA5-KD induced using siRNA (RUNX2/ITGA5-KD EVs; Fig. 4H) induced decreased osteogenic differentiation marker expression in MC3T3-E1 cells (Fig. 4I and J). Moreover, ITGA5-OE EV-educated MC3T3-E1 cells induced strong chemotaxis, and RUNX2/ITGA5-KD EVs decreased the chemotaxis of RUNX2-OE cancer cells toward MC3T3-E1 cells (Fig. 4K and L; Supplementary Fig. S4E). Taken together, these results demonstrate that extracellular vesicular ITGA5 mediates osteoblast differentiation, which further increases the recruitment of cancer cells expressing high levels of RUNX2.

Extracellular vesicular ITGA5 did not enhance the uptake of cancer-derived EVs by osteoblasts

Neither ITGA5-OE in MDA231 EVs nor ITGA5-KD in MDA231 RUNX2 EVs affected the uptake of cancer-derived EVs by MC3T3-E1 cells (Supplementary Fig. S4F and S4G). We also obtained unexpected results in MCF7 breast cancer cells, which do not express RUNX2. Although exogenous RUNX2-OE resulted in ITGA5 enrichment in both MCF7 cells and their EVs, MCF7 RUNX2 EVs were not internalized by MC3T3-E1 cells (Supplementary Fig. S5A–S5C). Correspondingly, MCF7 RUNX2 EVs did not induce MC3T3-E1 differentiation and failed to promote the recruitment of MCF7 RUNX2-OE cells toward EV-induced MC3T3-E1 cells (Supplementary Fig. S5D and S5E). When MCF7 RUNX2-OE cells and Control cells were cocultured with EV-educated MC3T3-E1 cells in Matrigel, no significant differences were observed in the cell cluster morphology among the groups (Supplementary Fig. S5F). Moreover, MCF7 ITGA5-OE EVs were not internalized by MC3T3-E1 cells, did not induce MC3T3-E1 cell differentiation and failed to further increase the recruitment and aggressive spread of MCF7 RUNX2-OE cells (Supplementary Fig. S5G–S5K). Thus, we hypothesized that EVs must be internalized by osteoblasts to allow extracellular vesicular ITGA5 to induce osteoblastic niche formation. Otherwise, if EVs cannot be internalized by osteoblasts, ITGA5 cannot play a role in osteogenic niche formation.

CDH11 is essential for the internalization of cancer-derived EVs by osteoblasts

On the basis of the evidence that ITGA5 is effective in educating osteoblasts via MDA231 RUNX2 EVs but not MCF7 RUNX2 EVs, we hypothesized that proteins that are highly expressed in MDA231 RUNX2 EVs but have low expression levels in MCF7 RUNX2 EVs contribute to the specific uptake of EVs by osteoblasts. Among the 45 candidate proteins screened by MS (Supplementary Fig. S6A), CDH11, a classic osteoblast cadherin and adhesion molecule, was notable. CDH11 has been reported to be ectopically expressed in basal-like MDA-MB-231 and BT-549 cells but not in luminal-like MCF7 cells (22). Accordingly, we found that CDH11 was carried by MDA231 EVs and BT-549 EVs, whereas it was not detectable in MCF7 EVs (Supplementary Fig. S6A and S6B). Thus, CDH11 may mediate the internalization of cancer-derived EVs by osteoblasts.

We then stably overexpressed CDH11 in MCF7 cells to obtain MCF7 CDH11-OE EVs (Fig. 5A). As expected, MCF7 CDH11-OE EVs were internalized by mOB and MC3T3-E1 cells (Fig. 5B). We further obtained CDH11-KD EVs from MDA231 RUNX2-OE cells (MDA231 RUNX2/CDH11-KD EVs) using siRNAs (Fig. 5C). CDH11-KD did not significantly affect the size and secretion of MDA231 RUNX2 EVs (Supplementary Fig. S6C); however, MDA231 RUNX2/CDH11-KD EV uptake by MC3T3-E1 cells was significantly reduced (Fig. 5D). Correspondingly, MDA231 CDH11-OE EVs were internalized by MC3T3-E1 cells more than Control cells (Supplementary Fig. S6D). Subsequently, the flag tag was used to trace CDH11-flag (CDH11flag; Fig. 5E). Immunoelectron microscopy demonstrated that CDH11flag was displayed on the surface of MDA231 CDH11flag EVs (Fig. 5F). Once CDH11flag EVs were added to mOB and MC3T3-E1 cell cultures, the flag was detected in the cytoplasm and on the cell membrane of mOB and MC3T3-E1 cells (Fig. 5GI). Thus, these results indicate that extracellular vesicular CDH11 is a key protein in EVs internalized by osteoblasts.

Figure 5.

CDH11 is the key adhesion molecule that mediates the internalization of cancer-derived EVs by osteoblasts. A, Western blotting showing the extracellular vesicular CDH11 levels of MCF7 CDH11-OE and Control cells. B, Quantification of the intake capability of PKH67-labeled EVs by mOB or MC3T3-E1 cells. Histograms and charts show the percentages of PKH67+ mOB or MC3T3-E1 cells. The red and blue dashed lines show the peak fluorescence of the Control group at 6 hours (6 H) and 24 hours (24 H), respectively. C, Western blotting of the extracellular vesicular CDH11 levels of MDA231 RUNX2/CDH11-KD and RUNX2/siControl cells. D, Quantification of the PKH67-labeled EVs intake capability of MC3T3-E1 cells. The red and blue dashed lines show the peak fluorescence of the Control group at 6 hours (6 H) and 24 hours (24 H), respectively. Bar chart showing the percentages of PKH67+ MC3T3-E1 cells. E, Western blotting showing a high abundance of CDH11-flag in EVs derived from CDH11-flag–overexpressing MDA231 cells and MCF7 cells. F, Representative immunoelectron microscopy images displaying the location of CDH11-flag on MDA231 CDH11-OE EVs. G and H, Internalization (G) and redistribution (H) of the extracellular vesicular flag in MC3T3-E1 and mOB cells treated with MDA231-derived CDH11flag EVs and Controlflag EVs. I, Histogram displaying the frequencies of cells with membrane localization. J, Cdh11 levels in MC3T3-E1 Cdh11-KD cells and Control cells were detected by Western blotting. K, Selective EV internalization by MC3T3-E1 cells with high Cdh11 expression. Diagram displaying the experimental procedure and representative pseudocolor image and chart showing the capabilities of MC3T3-E1 cells with different Cdh11 expression statuses to take up CDH11-OE EVs. ***, P < 0.001 compared with the corresponding Control group.

Figure 5.

CDH11 is the key adhesion molecule that mediates the internalization of cancer-derived EVs by osteoblasts. A, Western blotting showing the extracellular vesicular CDH11 levels of MCF7 CDH11-OE and Control cells. B, Quantification of the intake capability of PKH67-labeled EVs by mOB or MC3T3-E1 cells. Histograms and charts show the percentages of PKH67+ mOB or MC3T3-E1 cells. The red and blue dashed lines show the peak fluorescence of the Control group at 6 hours (6 H) and 24 hours (24 H), respectively. C, Western blotting of the extracellular vesicular CDH11 levels of MDA231 RUNX2/CDH11-KD and RUNX2/siControl cells. D, Quantification of the PKH67-labeled EVs intake capability of MC3T3-E1 cells. The red and blue dashed lines show the peak fluorescence of the Control group at 6 hours (6 H) and 24 hours (24 H), respectively. Bar chart showing the percentages of PKH67+ MC3T3-E1 cells. E, Western blotting showing a high abundance of CDH11-flag in EVs derived from CDH11-flag–overexpressing MDA231 cells and MCF7 cells. F, Representative immunoelectron microscopy images displaying the location of CDH11-flag on MDA231 CDH11-OE EVs. G and H, Internalization (G) and redistribution (H) of the extracellular vesicular flag in MC3T3-E1 and mOB cells treated with MDA231-derived CDH11flag EVs and Controlflag EVs. I, Histogram displaying the frequencies of cells with membrane localization. J, Cdh11 levels in MC3T3-E1 Cdh11-KD cells and Control cells were detected by Western blotting. K, Selective EV internalization by MC3T3-E1 cells with high Cdh11 expression. Diagram displaying the experimental procedure and representative pseudocolor image and chart showing the capabilities of MC3T3-E1 cells with different Cdh11 expression statuses to take up CDH11-OE EVs. ***, P < 0.001 compared with the corresponding Control group.

Close modal

Because of the role of CDH11 in cell recognition and adhesion, we expected that osteoblastic CDH11 would be responsible for recognizing CDH11 on EVs. We mixed Cdh11-KD MC3T3-E1 cells (Fig. 5J) with parental MC3T3-E1 cells in one culture system in the same proportion and investigated the uptake of CDH11-OE EVs by MC3T3-E1 cells with different Cdh11 expression levels (Fig. 5K). Cdh11 high-expressing (Cdh11high) MC3T3-E1 cells showed significantly selective uptake of CDH11-OE EVs compared with Cdh11 low-expressing (Cdh11low) MC3T3-E1 cells. Therefore, this evidence demonstrates that the interaction of extracellular vesicular CDH11 with osteoblastic CDH11 is the molecular basis for internalization of cancer-derived EVs by osteoblasts.

Extracellular vesicular CDH11 alone does not determine osteoblast education for the recruitment and aggressive spread of cancer cells

Although CDH11-OE EVs from both MDA231 and MCF7 cells were internalized by MC3T3-E1 cells, these EVs failed to induce a significant increase in osteogenic differentiation markers in MC3T3-E1 cells (Supplementary Fig. S7A). Accordingly, MDA231 CDH11-OE EVs slightly increased the chemotactic capability of MDA231 RUNX2-OE cells, and MCF7 CDH11-OE EVs did not increase the chemotactic capability of MCF7 RUNX2-OE cells toward MC3T3-E1 cells (Supplementary Fig. S7B and S7C). When the cancer cells were cocultured with EV-educated MC3T3-E1 cells, neither MDA231 CDH11-OE EV-educated nor MCF7 CDH11-OE EV-educated MC3T3-E1 cells significantly promoted the aggressive spread of their corresponding RUNX2-OE cells (Supplementary Fig. S7D and S7E).

We further speculated that osteoblast education by EVs may be mediated by ITGA5 loaded on the same EVs after CDH11-mediated internalization. MDA231 RUNX2 EVs from cells with either CDH11-KD or ITGA5-KD in dramatically reduced MC3T3-E1 education and weakened the chemotactic capability and aggressive spread of MDA231 RUNX2-OE cells (Supplementary Fig. S7F and S7G). CDH11-OE in MCF7 RUNX2 EVs, which displayed a high ITGA5 expression phenotype, was observed to contribute to osteoblast education and increase the recruitment and aggressive spread of MCF7 RUNX2-OE cancer cells (Supplementary Fig. S7H and S7I). ITGA5-KD in MCF7 RUNX2/CDH11-OE EVs, however, lost their capability to educate MC3T3-E1 cells for the recruitment and aggressive spread of cancer cells (Supplementary Fig. S7H and S7I).

CDH11-OE/ITGA5-OE EVs determine the premetastatic niche for the bone colonization of breast cancer cells with high RUNX2 expression

In vitro, MC3T3-E1 cells educated with CDH11-OE/ITGA5-OE EVs from both MDA231 and MCF7 cells displayed outstanding abilities to attract their corresponding RUNX2-OE cancer cells (Fig. 6A–C) and supported the aggressive spread of RUNX2-OE cancer cells in the Matrigel (Fig. 6D and E).

Figure 6.

CDH11-OE/ITGA5-OE EVs educate osteoblasts for the recruitment and aggressive spread of RUNX2 high-expressing cancer cells in vitro. A, Western blotting showing the enrichment of CDH11 and ITGA5 in CDH11-OE/ITGA5-OE EVs derived from MDA231 and MCF7 cells. B and C, Transwell assays for the chemotactic migration of MDA231 RUNX2-OE cells (B) and MCF7 RUNX2-OE cells (C) toward EV-educated MC3T3-E1 cells. D and E, Representative 3D coculture images of GFP+ MDA231 RUNX2-OE cells (D) and GFP+ MCF7 RUNX2-OE cells (E) with PKH26-labeled EV-educated MC3T3-E1 cells. ***, P < 0.001 compared with the Control EV group.

Figure 6.

CDH11-OE/ITGA5-OE EVs educate osteoblasts for the recruitment and aggressive spread of RUNX2 high-expressing cancer cells in vitro. A, Western blotting showing the enrichment of CDH11 and ITGA5 in CDH11-OE/ITGA5-OE EVs derived from MDA231 and MCF7 cells. B and C, Transwell assays for the chemotactic migration of MDA231 RUNX2-OE cells (B) and MCF7 RUNX2-OE cells (C) toward EV-educated MC3T3-E1 cells. D and E, Representative 3D coculture images of GFP+ MDA231 RUNX2-OE cells (D) and GFP+ MCF7 RUNX2-OE cells (E) with PKH26-labeled EV-educated MC3T3-E1 cells. ***, P < 0.001 compared with the Control EV group.

Close modal

In vivo, flag-labeled MDA231 CDH11-OE/ITGA5-OE EVs were observed to be internalized more by osteoblasts than Control EVs (Fig. 7A and B). Moreover, increasing Ocn expression and Alp activity were observed in CDH11-OE/ITGA5-OE EV-educated osteoblasts (Fig. 7BC), indicating the role of CDH11-OE/ITGA5-OE EVs in inducing the osteogenic niche. We further investigated the effect of the CDH11-OE/ITGA5-OE EV-educated osteogenic niche on pulmonary and bone colonization of MDA231 RUNX2-OE cells (Fig. 7D). Mice that were treated with CDH11-OE/ITGA5-OE EVs survived, with decreased pulmonary colonization, significantly longer than the Control EV-treated mice (Fig. 7E–G). However, the incidence of bone colonization, erosion surface and the number of TRAP+ osteoclasts were significantly higher in the CDH11-OE/ITGA5-OE EV group than in the Control EV group (Fig. 7FI).

Figure 7.

CDH11-OE/ITGA5-OE EVs facilitate the bone colonization of cancer cells with high RUNX2 expression in vivo. A, Diagram displaying the experimental process used to observe the biodistribution of MDA231-derived EVs in the bone of SCID mice. B, Immunofluorescence staining for flag showing the colocalization of EVs and osteoblasts. Ocn was used as a marker of osteoblasts. C, Representative images of Alp staining. D, Schematic showing the experimental design used to investigate bone colonization of MDA231 RUNX2-OE cells in EV-treated SCID mice. E, Kaplan−Meier survival analysis showing the overall survival of mice. F, Representative bioluminescence imaging on day 0 (D 0) and day 21 (D 21). H&E staining images and X-ray imaging show pulmonary and bone colonization of cancer cells. TRAP staining was used to display the activity of osteoclasts. G, Charts of the incidences of pulmonary and bone colonization in mice. H and I, Histograms show the erosion surface (H) and TRAP+ osteoclast number (I) normalized to the bone surface. *, P < 0.05 compared with the Control EV group.

Figure 7.

CDH11-OE/ITGA5-OE EVs facilitate the bone colonization of cancer cells with high RUNX2 expression in vivo. A, Diagram displaying the experimental process used to observe the biodistribution of MDA231-derived EVs in the bone of SCID mice. B, Immunofluorescence staining for flag showing the colocalization of EVs and osteoblasts. Ocn was used as a marker of osteoblasts. C, Representative images of Alp staining. D, Schematic showing the experimental design used to investigate bone colonization of MDA231 RUNX2-OE cells in EV-treated SCID mice. E, Kaplan−Meier survival analysis showing the overall survival of mice. F, Representative bioluminescence imaging on day 0 (D 0) and day 21 (D 21). H&E staining images and X-ray imaging show pulmonary and bone colonization of cancer cells. TRAP staining was used to display the activity of osteoclasts. G, Charts of the incidences of pulmonary and bone colonization in mice. H and I, Histograms show the erosion surface (H) and TRAP+ osteoclast number (I) normalized to the bone surface. *, P < 0.05 compared with the Control EV group.

Close modal

Subsequently, the effect of Cdh11-OE/Itga5-OE EVs on premetastatic niche formation and spontaneous bone metastases was investigated by implantation of 4T1 Runx2-OE cells into BALB/c mice (Fig. 8A–C). 4T1-derived Cdh11-OE/Itga5-OE EVs did not significantly affect the survival and lung metastasis of mice subcutaneously inoculated with 4T1 Runx2-OE cancer cells (Fig. 8D and E). However, Cdh11-OE/Itga5-OE EV education significantly increased the incidence of bone metastasis, erosion surfaces, the numbers of bone lesions and TRAP+ osteoclasts (Fig. 8F–J). These results suggested that Cdh11-OE/Itga5-OE EV education facilitates bone metastasis of RUNX2 high-expressing cancer cells.

Figure 8.

Extracellular vesicular CDH11 and ITGA5 collaboratively determine the premetastatic niche for bone colonization of cancer cells with high RUNX2 expression. A, Western blotting showing the enrichment of Cdh11 and Itga5 in 4T1 Cdh11-OE/Itga5-OE EVs. B, Western blotting showing Runx2-OE in 4T1 cells. C, Schematic of the 4T1 spontaneous bone metastasis animal experiment. D, Kaplan–Meier survival analysis of the overall survival of mice. E, Representative H&E staining images and a histogram showing the incidence of lung metastasis. F, Representative X-ray, H&E, and TRAP staining images of bone. G, Histogram of the incidence of bone metastasis. H–J, Histograms showing the erosion surface normalized to the bone surface (H), the number of bone lesions (I), and TRAP+ osteoclasts normalized to the bone surface (J). K, Heatmap showing the relationship between tumor CDH11 and ITGA5 mRNA expression levels and bone metastasis (BM) in RUNX2 high-expressing breast cancer tissues based on the published GSE2034 dataset. Red bars, bone metastasis; yellow boxes, proportion with bone metastases (BM%). L, Schematic illustration of the role of cancer-derived CDH11high/ITGA5high EVs in premetastatic niche formation and the bone colonization of cancer cells with high RUNX2 expression. *, P < 0.05 compared with the Control group.

Figure 8.

Extracellular vesicular CDH11 and ITGA5 collaboratively determine the premetastatic niche for bone colonization of cancer cells with high RUNX2 expression. A, Western blotting showing the enrichment of Cdh11 and Itga5 in 4T1 Cdh11-OE/Itga5-OE EVs. B, Western blotting showing Runx2-OE in 4T1 cells. C, Schematic of the 4T1 spontaneous bone metastasis animal experiment. D, Kaplan–Meier survival analysis of the overall survival of mice. E, Representative H&E staining images and a histogram showing the incidence of lung metastasis. F, Representative X-ray, H&E, and TRAP staining images of bone. G, Histogram of the incidence of bone metastasis. H–J, Histograms showing the erosion surface normalized to the bone surface (H), the number of bone lesions (I), and TRAP+ osteoclasts normalized to the bone surface (J). K, Heatmap showing the relationship between tumor CDH11 and ITGA5 mRNA expression levels and bone metastasis (BM) in RUNX2 high-expressing breast cancer tissues based on the published GSE2034 dataset. Red bars, bone metastasis; yellow boxes, proportion with bone metastases (BM%). L, Schematic illustration of the role of cancer-derived CDH11high/ITGA5high EVs in premetastatic niche formation and the bone colonization of cancer cells with high RUNX2 expression. *, P < 0.05 compared with the Control group.

Close modal

We further investigated the relationship between tumor CDH11 and ITGA5 mRNA expression levels and bone metastasis in RUNX2high primary breast cancer tissues based on the published GSE2034 dataset. Patients in the CDH11high/ITGA5high group had a higher risk of bone metastasis than patients in the CDH11low/ITGA5low group, CDH11high/ITGA5low group, or CDH11low/ITGA5high group (Fig. 8K). This clinical evidence suggests that CDH11 and ITGA5 expression in primary cancer determines the bone metastatic capability of RUNX2high breast cancer.

Together, both the in vivo animal experimental results and clinical evidence demonstrate that EVs with high CDH11 and ITGA5 expression (CDH11high/ITGA5high EVs) produced by breast cancer cells contribute to the formation of an osteogenic premetastatic niche in bone, further facilitating RUNX2high cancer cell colonization in and metastasis to bone (Fig. 8L).

Although bone metastatic lesions from breast cancer are osteolytic, the osteogenic niche has been well described as an initiator of the early-stage bone colonization of disseminated cancer cells (18, 23). It has been well elucidated that osteoblasts and tumor cells in metastatic sites induce osteoclast differentiation and osteolytic lesions by initiating a “vicious cycle” (2, 24). As the main cellular components in the osteogenic niche, osteoblasts have been reported to support metastatic cancer cell survival and proliferation during bone metastasis (18, 25). In this study, we identified an extracellular vesicular mechanism that connects primary breast cancer cells with high RUNX2 expression and resident osteoblasts to establish an osteogenic premetastatic niche before the arrival of cancer cells in bone.

In recent decades, the indispensable roles of tumor cell–derived EVs in premetastatic niche formation have been widely revealed (12, 26). Tumor cell–derived EVs transfer their cargos, including nucleic acids and proteins, to resident cells at future metastatic sites after EV internalization by resident cells. Extracellular vesicular miRNAs and proteins educate or reprogram target cells to determine the organotropism of cancer cells by preparing a favorable premetastatic niche that facilitates the recruitment and growth of cancer cells (27, 28). Extracellular vesicular ITGA6 (12) and miR-125b (29) have been reported to induce lung premetastatic niches in breast cancer. miR-181c–containing EVs destroy the blood–brain barrier and prepare a brain premetastatic niche for breast cancer (30). Extracellular vesicular ITGB5 educates Kupffer cells to form a favorable liver premetastatic niche (12). Although exosomal miR-940 has been demonstrated to induce osteogenic premetastatic lesions and facilitate osteoblastic-type bone metastasis of prostate cancer (31), it is still unclear whether breast cancer–derived EV cargo initiates the bone premetastatic niche. Here, we not only discovered the role of EVs derived from breast cancer with high RUNX2 expression in bone metastasis but also identified their protein features (CDH11high/ITGA5high) that establish an osteogenic premetastatic niche. Furthermore, we identified different extracellular vesicular proteins that determine the subsequent specific recognition of tumor-derived EVs by osteoblasts (CDH11) and further induction of the osteogenic premetastatic niche (ITGA5).

CDH11, also known as osteoblast cadherin (OB-cadherin), belongs to the cadherin superfamily, mediates homophilic cell–cell adhesion and is mainly expressed in osteoblasts. During early embryogenesis, CDH11 is expressed predominantly in mesenchymal tissues but not in epithelial tissues (32). CDH11 has been reported to be ectopically overexpressed in basal-like breast cancer cells and has been regarded as a metastasis enhancer (14, 22, 33, 34). In our study, CDH11 was carried by EVs from basal-like MDA-MB-231 and BT-549 cells but not by EVs from luminal-like MCF7 cells. Moreover, RUNX2-OE in basal-like cells upregulated EV-loaded CDH11. EVs not loaded with CDH11 were not taken up by osteoblasts. EVs with high CDH11, however, were robustly internalized by osteoblasts with high CDH11 expression, demonstrating that extracellular vesicular CDH11 determines bone-tropic metastasis through the recognition of tumor cell–derived EVs and resident osteoblasts. However, extracellular vesicular CDH11 alone did not educate osteoblasts for the recruitment and aggressive spread of cancer cells.

Our previous study identified ITGA5 as a transcriptional target of RUNX2 and a mediator of RUNX2-driven bone metastasis of breast cancer (16). Both Pantano F (35) and Croset M (33) also found that ITGA5 silencing in MDA-B02 and Hs-578T cells, which have high bone metastatic potential, reduced bone metastasis and osteolytic lesions in vivo. ITGA5 has been identified as an EV protein cargo (12), and ITGA5 carried by EVs derived from breast cancer cells with high RUNX2 expression was shown to educate osteoblasts to form an osteogenic premetastatic niche in this study. Although Pantano F (12) considered tumor-derived exosomal integrins to dictate exosome adhesion to specific cell types and extracellular matrix molecules in target organs, extracellular vesicular ITGA5 was not shown to contribute to the uptake of EVs by osteoblasts in our study. According to our experimental results, ITGA5 plays a role in educating osteoblasts only when EVs are internalized by osteoblasts because neither extracellular vesicular ITGA5-OE nor ITGA5-KD altered EV internalization by osteoblasts. Thus, both CDH11 and ITGA5 are indispensable extracellular vesicular components that induce the establishment of the osteogenic premetastatic niche.

Tumor-derived EVs are released into the blood from the primary tumor site. Because EV molecular cargoes mirror the cell of origin and are representative of the parental cell (36), circulating tumor-derived CDH11high/ITGA5high EVs may not only be useful in determining the biological characteristics of primary tumors but also become promising liquid biopsy diagnostic biomarkers to predict bone metastasis. Furthermore, CDH11high EVs, because of their specific internalization by osteoblasts, may be a potential delivery system to develop an EV-based premetastatic niche strategy.

No disclosures were reported.

X.-Q. Li: Conceptualization, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing. R. Zhang: Formal analysis, methodology. H. Lu: Formal analysis, methodology. X.-M. Yue: Formal analysis, methodology. Y.-F. Huang: Formal analysis, methodology.

This work was supported by the National Natural Science Foundation of China (Nos. 81672878 and 81201647).

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

1.
Brook
N
,
Brook
E
,
Dharmarajan
A
,
Dass
CR
,
Chan
A
.
Breast cancer bone metastases: pathogenesis and therapeutic targets
.
Int J Biochem Cell Biol
2018
;
96
:
63
78
.
2.
Mandal
CC
.
Osteolytic metastasis in breast cancer: effective prevention strategies
.
Expert Rev Anticancer Ther
2020
;
20
:
797
811
.
3.
Paget
S
.
The distribution of secondary growths in cancer of the breast. 1889
.
Cancer Metastasis Rev
1989
;
8
:
98
101
.
4.
Peinado
H
,
Zhang
H
,
Matei
IR
,
Costa-Silva
B
,
Hoshino
A
,
Rodrigues
G
, et al
.
Premetastatic niches: organ-specific homes for metastases
.
Nat Rev Cancer
2017
;
17
:
302
17
.
5.
Liu
Y
,
Cao
X
.
Characteristics and significance of the premetastatic niche
.
Cancer Cell
2016
;
30
:
668
81
.
6.
Mo
Z
,
Cheong
JYA
,
Xiang
L
,
Le
MTN
,
Grimson
A
,
Zhang
DX
.
Extracellular vesicle–associated organotropic metastasis
.
Cell Prolif
2021
;
54
:
e12948
.
7.
Li
FX
,
Liu
JJ
,
Xu
F
,
Lin
X
,
Zhong
JY
,
Wu
F
, et al
.
Role of tumor-derived exosomes in bone metastasis
.
Oncol Lett
2019
;
18
:
3935
45
.
8.
Kalluri
R
,
LeBleu
VS
.
The biology, function, and biomedical applications of exosomes
.
Science
2020
;
367
:
367
.
9.
He
C
,
Li
L
,
Wang
L
,
Meng
W
,
Hao
Y
,
Zhu
G
.
Exosome-mediated cellular cross-talk within the tumor microenvironment upon irradiation
.
Cancer Biol Med
2021
;
18
:
21
33
.
10.
Guo
Y
,
Ji
X
,
Liu
J
,
Fan
D
,
Zhou
Q
,
Chen
C
, et al
.
Effects of exosomes on premetastatic niche formation in tumors
.
Mol Cancer
2019
;
18
:
39
.
11.
Becker
A
,
Thakur
BK
,
Weiss
JM
,
Kim
HS
,
Peinado
H
,
Lyden
D
.
Extracellular vesicles in cancer: cell-to-cell mediators of metastasis
.
Cancer Cell
2016
;
30
:
836
48
.
12.
Hoshino
A
,
Costa-Silva
B
,
Shen
TL
,
Rodrigues
G
,
Hashimoto
A
,
Tesic Mark
M
, et al
.
Tumor exosome integrins determine organotropic metastasis
.
Nature
2015
;
527
:
329
35
.
13.
Vishal
M
,
Swetha
R
,
Thejaswini
G
,
Arumugam
B
,
Selvamurugan
N
.
Role of Runx2 in breast cancer–mediated bone metastasis
.
Int J Biol Macromol
2017
;
99
:
608
14
.
14.
Tan
CC
,
Li
GX
,
Tan
LD
,
Du
X
,
Li
XQ
,
He
R
, et al
.
Breast cancer cells obtain an osteomimetic feature via epithelial–mesenchymal transition that have undergone BMP2/RUNX2 signaling pathway induction
.
Oncotarget
2016
;
7
:
79688
705
.
15.
Li
XQ
,
Du
X
,
Li
DM
,
Kong
PZ
,
Sun
Y
,
Liu
PF
, et al
.
ITGBL1 is a Runx2 transcriptional target and promotes breast cancer bone metastasis by activating the TGFbeta signaling pathway
.
Cancer Res
2015
;
75
:
3302
13
.
16.
Li
XQ
,
Lu
JT
,
Tan
CC
,
Wang
QS
,
Feng
YM
.
RUNX2 promotes breast cancer bone metastasis by increasing integrin alpha5-mediated colonization
.
Cancer Lett
2016
;
380
:
78
86
.
17.
Riquelme
MA
,
Cardenas
ER
,
Jiang
JX
.
Osteocytes and bone metastasis
.
Front Endocrinol
2020
;
11
:
567844
.
18.
Wang
H
,
Yu
C
,
Gao
X
,
Welte
T
,
Muscarella
AM
,
Tian
L
, et al
.
The osteogenic niche promotes early-stage bone colonization of disseminated breast cancer cells
.
Cancer Cell
2015
;
27
:
193
210
.
19.
Wang
S
,
Li
GX
,
Tan
CC
,
He
R
,
Kang
LJ
,
Lu
JT
, et al
.
FOXF2 reprograms breast cancer cells into bone metastasis seeds
.
Nat Commun
2019
;
10
:
2707
.
20.
Chen
JH
,
Huang
WC
,
Bamodu
OA
,
Chang
PM
,
Chao
TY
,
Huang
TH
.
Monospecific antibody targeting of CDH11 inhibits epithelial-to-mesenchymal transition and represses cancer stem cell-like phenotype by upregulating miR-335 in metastatic breast cancer, in vitro and in vivo
.
BMC Cancer
2019
;
19
:
634
.
21.
Wang
Y
,
Klijn
JG
,
Zhang
Y
,
Sieuwerts
AM
,
Look
MP
,
Yang
F
, et al
.
Gene expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer
.
Lancet
2005
;
365
:
671
9
.
22.
Assefnia
S
,
Dakshanamurthy
S
,
Guidry Auvil
JM
,
Hampel
C
,
Anastasiadis
PZ
,
Kallakury
B
, et al
.
Cadherin 11 in poor prognosis malignancies and rheumatoid arthritis: common target, common therapies
.
Oncotarget
2014
;
5
:
1458
74
.
23.
Wang
H
,
Yu
C
,
Zhang
XH
.
Devil's wake: early-stage bone colonization by breast cancer
.
Mol Cell Oncol
2016
;
3
:
e1026526
.
24.
Ren
G
,
Esposito
M
,
Kang
Y
.
Bone metastasis and the metastatic niche
.
J Mol Med
2015
;
93
:
1203
12
.
25.
Kolb
AD
,
Shupp
AB
,
Mukhopadhyay
D
,
Marini
FC
,
Bussard
KM
.
Osteoblasts are “educated” by cross-talk with metastatic breast cancer cells in the bone tumor microenvironment
.
Breast Cancer Res
2019
;
21
:
31
.
26.
Xie
X
,
Lian
S
,
Zhou
Y
,
Li
B
,
Lu
Y
,
Yeung
I
, et al
.
Tumor-derived exosomes can specifically prevent cancer metastatic organotropism
.
J Control Release
2021
;
331
:
404
15
.
27.
Yuan
X
,
Qian
N
,
Ling
S
,
Li
Y
,
Sun
W
,
Li
J
, et al
.
Breast cancer exosomes contribute to premetastatic niche formation and promote bone metastasis of tumor cells
.
Theranostics
2021
;
11
:
1429
45
.
28.
Costa-Silva
B
,
Aiello
NM
,
Ocean
AJ
,
Singh
S
,
Zhang
H
,
Thakur
BK
, et al
.
Pancreatic cancer exosomes initiate premetastatic niche formation in the liver
.
Nat Cell Biol
2015
;
17
:
816
26
.
29.
Vu
LT
,
Peng
B
,
Zhang
DX
,
Ma
V
,
Mathey-Andrews
CA
,
Lam
CK
, et al
.
Tumor-secreted extracellular vesicles promote the activation of cancer-associated fibroblasts via the transfer of microRNA-125b
.
J Extracell Vesicles
2019
;
8
:
1599680
.
30.
Tominaga
N
,
Kosaka
N
,
Ono
M
,
Katsuda
T
,
Yoshioka
Y
,
Tamura
K
, et al
.
Brain metastatic cancer cells release microRNA-181c-containing extracellular vesicles capable of destructing blood–brain barrier
.
Nat Commun
2015
;
6
:
6716
.
31.
Hashimoto
K
,
Ochi
H
,
Sunamura
S
,
Kosaka
N
,
Mabuchi
Y
,
Fukuda
T
, et al
.
Cancer-secreted hsa-miR-940 induces an osteoblastic phenotype in the bone metastatic microenvironment via targeting ARHGAP1 and FAM134A
.
Proc Natl Acad Sci U S A
2018
;
115
:
2204
9
.
32.
Simonneau
L
,
Kitagawa
M
,
Suzuki
S
,
Thiery
JP
.
Cadherin 11 expression marks the mesenchymal phenotype: towards new functions for cadherins?
Cell Adhes Commun
1995
;
3
:
115
30
.
33.
Croset
M
,
Pantano
F
,
Kan
CWS
,
Bonnelye
E
,
Descotes
F
,
Alix-Panabieres
C
, et al
.
miRNA-30 family members inhibit breast cancer invasion, osteomimicry, and bone destruction by directly targeting multiple bone metastasis–associated genes
.
Cancer Res
2018
;
78
:
5259
73
.
34.
Satriyo
PB
,
Bamodu
OA
,
Chen
JH
,
Aryandono
T
,
Haryana
SM
,
Yeh
CT
, et al
.
Cadherin 11 inhibition downregulates beta-catenin, deactivates the canonical WNT signaling pathway and suppresses the cancer stem cell-like phenotype of triple negative breast cancer
.
J Clin Med
2019
;
8
:
148
.
35.
Pantano
F
,
Croset
M
,
Driouch
K
,
Bednarz-Knoll
N
,
Iuliani
M
,
Ribelli
G
, et al
.
Integrin alpha5 in human breast cancer is a mediator of bone metastasis and a therapeutic target for the treatment of osteolytic lesions
.
Oncogene
2021
;
40
:
1284
99
.
36.
Taverna
S
,
Giusti
I
,
D'Ascenzo
S
,
Pizzorno
L
,
Dolo
V
.
Breast cancer derived extracellular vesicles in bone metastasis induction and their clinical implications as biomarkers
.
Int J Mol Sci
2020
;
21
:
3573
.

Supplementary data