In the bone marrow (BM), breast cancer cells (BCC) can survive in dormancy for decades as cancer stem cells (CSC), resurging as tertiary metastasis. The endosteal region where BCCs exist as CSCs poses a challenge to target them, mostly due to the coexistence of endogenous hematopoietic stem cells. This study addresses the early period of dormancy when BCCs enter BM at the perivascular region to begin the transition into CSCs, which we propose as the final step in dormancy. A two-step process comprises the Wnt-β-catenin pathway mediating BCC dedifferentiation into CSCs at the BM perivascular niche. At this site, BCCs responded to two types of mesenchymal stem cell (MSC)–released extracellular vesicles (EV) that may include exosomes. Early released EVs began the transition into cycling quiescence, DNA repair, and reorganization into distinct BCC subsets. After contact with breast cancer, the content of EVs changed (primed) to complete dedifferentiation into a more homogeneous population with CSC properties. BCC progenitors (Oct4alo), which are distant from CSCs in a hierarchical stratification, were sensitive to MSC EVs. Despite CSC function, Oct4alo BCCs expressed multipotent pathways similar to CSCs. Oct4alo BCCs dedifferentiated and colocalized with MSCs (murine and human BM) in vivo. Overall, these findings elucidate a mechanism of early dormancy at the BM perivascular region and provide evidence of epigenome reorganization as a potential new therapy for breast cancer.

Significance:

These findings describe how the initial process of dormancy and dedifferentiation of breast cancer cells at the bone marrow perivascular niche requires mesenchymal stem cell–derived exosomes, indicating a potential target for therapeutic intervention.

Despite aggressive methods to diagnose breast cancer early and, the exhaustive efforts to develop new anticancer drug treatments, breast cancer remains a clinical problem (1). The ability of breast cancer cells (BCC) to remain dormant for decades for later resurgence underscores the clinical challenges (2, 3). In addition, BCCs can survive as dormant cells at distant tissue sites, long before clinical diagnosis (4–6). These dormant BCCs, including those in bone marrow (BM), can be the source of metastatic cancer (4, 7–9). Thus, it is important to understand how the tissue microenvironment facilitates BCCs to transition into cancer stem cells (CSC), providing an advantage to become dormant cancer cells.

BCCs show preference for the BM, leading to poor prognosis (9–11). Dormant BCCs exhibit functional similarities with stem cells, such as expression of stem cell genes, self-renewal, and tumor initiation (12–14). Other studies identified dormant BCCs as CSCs, which are linked to drug resistance and immune evasion (9, 12, 15, 16). Targeting dormant BCCs in the BM remains a challenge. This is mostly due to the similarity between dormant BCCs/CSCs and resident hematopoietic stem cells (HSC), along with their shared location in the BM, for example, endosteum (17, 18). Thus, the challenge is to specifically target dormant BCCs without untoward effect on HSCs. Particularly, HSCs are the source of immune and blood cells throughout life and must be protected (19).

Resident BM niche cells facilitate the transition of BCCs into a dormant state (7, 17). The BM niche, which includes fibroblasts, macrophages, and mesenchymal stem cells (MSC), creates a microenvironment to support breast cancer dormancy (17, 20). The mechanisms of dormancy include direct intercellular interaction through gap junction between CSCs and BM niche cells and indirect interaction via cytokine and exosomal secretome (7, 15, 21, 22). MSCs can also protect dormant BCCs in the BM by skewing the immune response to an immunosuppressive environment, increased regulatory T cells and decreased natural killer–cell activity (10, 23).

To target dormant BCCs, it is necessary to understand how BCCs survive as dormant cells and what drives them out of quiescence to establish metastasis. Dormancy can be established at any time during the disease, including years prior to clinical detection (24). Thus, the development of safe and efficacious treatment will allow for quick elimination of existing dormant BCCs, as well as prevent those entering dormancy. Thus, a suitable solution to target dormant BCCs in BM must consider two phases: methods to reverse the already established dormant cells for effective treatment and establish strategies to prevent further transition of BCCs entering dormancy. Indeed, the scientific community has acknowledged the need to reverse dormancy (20, 25). However, the process toward dormancy is highly complex and includes multiple factors and the specific BM niche. This study addresses the mechanisms by which BCCs begin the transition into dormancy at the interface of the peripheral vascular system. Thus, we capture the changes occurring in BCCs immediately and after entry into the BM cavity.

We performed a systematic approach to address the steps toward dormancy. Specifically, we examined how BCCs instruct MSCs to release exosomes with distinct RNA cargo to impart their transition into dormancy. We selected MSCs because they are among the first set of cells encountered by BCCs when they enter the perivascular niche of BM (26, 27). In addition, miRNA cargo in MSC-released exosomes can induce cycling quiescence in BCCs (17). Senescent MSCs also facilitate BCC invasion and tumor condition media can prime MSCs into carcinoma-associated fibroblasts (28, 29).

Exosomes are extracellular vesicles (EV), with size averaging 100 nm in diameter, and are derived from the endosomes. The EV cargo includes lipids, proteins, and nucleic acid, such as mRNA and noncoding RNAs, such as miRNAs (30, 31). EVs modulate several biological processes, including cancer cell behavior, such as promoting angiogenesis and drug resistance (17, 32). Here, we report the specific effects of MSC-derived EVs on the behavior of BCCs. We discuss the involvement of canonical and noncanonical Wnt/β-catenin pathway in dedifferentiation of BCCs into fully dormant cells. In addition, we discuss the role of DNA repair and epigenetic restructuring during dedifferentiation of BCCs.

Cell lines

BCCs, triple negative (MDA-MB-231 and MDA-MB-468) and triple-positive T47D, were purchased from the ATCC and cultured as per their instructions. BCCs, stably transfected with pEGFP1-Oct3/4 or pEdsRED1-Oct3/4 have been described previously (21). The transfectants were maintained in media with 500 μg/mL G418. We authenticated all cell lines with ATCC STR database (www.atcc.org/STR_database.aspx) and also performed weekly assessment for Mycoplasma. The RNA sequencing (RNA-seq) studies used MDA-MB-231. All other studies used both cell lines, hence the use of “BCCs” to represent both cell lines.

Human subjects

Rutgers Institutional Review Board (Newark, NJ) approved the use of paraffin sections from BM biopsy of patients with breast cancer and BM aspirates from healthy donors (18–35 years). All subjects provided and signed the informed written consent.

Preparation of MSC-derived EVs

Naïve and primed preparation

The method to isolate naïve and primed microvesicles (MV) from MSCs was described previously (17, 20, 33). Primed MVs were prepared as follows: we added equal amounts (5 × 104) of BCCs and MSCs in DMEM containing 10% FCS to the inner and outer wells of 6-well transwell plates (0.4 μm membranes), respectively. Naïve exosomes contained media in the inner wells. After 24 hours, we removed the inner wells and aspirated the media from the outer wells with MSCs. The MSCs were washed twice with PBS, and then reincubated with DMEM containing 2% exosome-free FCS. After 48 hours, we collected the media for exosome/MV isolation by gradient centrifugation or a kit with <5 mL media.

Differential centrifugation

Cells were cleared from media (>5 mL) by centrifugation at 2,000 × g for 20 minutes. We transferred the cell-free supernatants to sterile 1.5 mL tubes to clear for remaining cell debris and large vesicles by centrifugation at 10,000 × g for 30 minutes. We transferred the supernatants to 5 or 30 ml Ultracentrifuge Tubes (Thermo Fisher Scientific), and then centrifuged at 100,000 × g for 80 minutes. The supernatants were aspirated and the pellets, which contained the exosomes, were then washed with PBS. The exosomes were subjected to a final spin at 130,000 × g for 80 minutes and the exosome pellet was resuspended in 200 μL sterile PBS.

Isolation using a kit

In cases of experimental repeats when the resulting media were <5 mL, we isolated the exosomes with the Total Exosome Isolation Kit (Invitrogen) as per the manufacturer's guidelines. We observed similar function, size distribution, and phenotype with particles isolated with the kit and gradient centrifugation.

Quantification

The size and numbers of exosomes were determined with the nanoparticle tracking analysis (NTA). The NTA uses light scattering and Brownian motion, a syringe pump, and script control system. The NanoSight LM10 system, which is equipped with a 405-nm laser, recorded videos. The NTA software (NanoSight version 2.3) analyzed the data.

Nanoimaging

Purified exosomes using CD63 Dynabeads were stained for CD9-Alexa 647 or the lipophilic dye, FM-143. We incubated the particles with anti-CD9 at 1/200 for 20 minutes at room temperature, washed with PBS, and imaged with the Nanoimager S Mark II from ONI (Oxford Nanoimaging) using laser, 640 nm/1W. We acquired images of CD9 clusters with NimOS 1.3.7511 imaging software, 1000X and 1.4 NA objective.

EV-treated BCCs

We have described previously the method to treat BCCs with MSC-released EVs (17). BCCs (105) were seeded in 100-mm tissue culture plates containing 6 mL of DMEM and 2% exosome-depleted FCS. After overnight adherence at 37°C (day 0), the BCCs were treated with naïve or primed MSC exosomes (105–109) at days 1, 3, and 5. Fractionated addition of EVs mimics the in vivo condition where BCCs would be continually exposed to MSC-derived EVs. Thus, there was no media change during the culture period. At different times, we collected the cells for functional studies and molecular analyses. We selected 108 EVs for the studies, based on the optimal decreases in cyclin D1 and CDK4 (Supplementary Fig. S1A and S1B; ref. 17). In addition to confocal microscopy to verify EV entry into cells, we now show real-time live images of the EVs bouncing on the cell surface and their movement within the cell (Supplementary Videos; ref. 17).

Data analyses

DropSeq (single-cell RNA-seq)

Data visualization was performed by hierarchical clustering, principal component analysis (PCA) using Partek Flow (2018), and Morpheus (https://software.broadinstitute.org/morpheus), with Pearson correlation.

RNA-seq

Partek Flow software was used to align the reads to Homo sapiens assembly hg38. Transcript abundance of the aligned reads used Partek (version 7) optimization of the expectation-maximization algorithm (Partek Inc.). PCA detected data trends and outliers. The RNA-seq data were normalized with log2 counts per million with an offset of 1. We removed batch effect between two sequencing rounds of the same biological samples. Differentially expressed genes between the treatment groups were selected for further analyses with genes, −1.5- to +1.5-fold change and P > 0.05. This resulted in 1,500 genes for the development of a heatmap and pathway analyses. The latter used gene set enrichment analysis with an FDR of 0.05 as cutoff to select significantly up- and downregulated pathways. Pathways linked to hematologic functions were selected using ingenuity pathway analyses (IPA; Qiagen Inc., https://digitalinsights.qiagen.com/products-overview/discovery-insights-portfolio/analysis-and-visualization/qiagen-ipa/?cmpid=QDI_GA_IPA&gclid=Cj0KCQiAs5eCBhCBARIsAEhk4r6SqKb29DgkGETdcO1BWzHY7vx-tIaK-2xPFYkdShSqVgiqbgdwWOEaAiTdEALw_wcB). We ranked the pathways by significance score, defined as −log10B-H P value times activation z-score. We retained the pathways associated with hematologic functions that passed a significance score threshold of >2.

Interpopulation variance analyses of Oct4alo

Previous Affymetrix data with Oct4alo MDA-MB-231 (n = 4) were analyzed in R. The log2 mean intensity probes versus log2 interpopulation variance and a least squares regression line were used to create a linear fit. The top approximately 1% (300 probes) probes showing positive deviation from the least squares regression line were retained and queried for gene identity. We collapsed the list to remove the redundant genes. The remaining genes were determined to be heterogeneous within the Oct4alo population.

VarElect analysis

We used VarElect (http://varelect.genecards.org/), a comprehensive tool that used an algorithm to identify genes common to diseases in all next-generation sequencing (NGS) data. We selected the genes with scores >30 and then submitted to STRING (https://string-db.org/) for protein–protein interaction and also to Panther (http://pantherdb.org/), a classification system for pathways.

Statistical analyses

Statistical analyses applied ANOVA and Tukey–Kramer multiple comparisons test. Single-cell RNA-seq and NGS expression analyses were performed, average linkage was used for clustering, and Pearson correlation analysis was used for distance measurement to generate heatmaps and hierarchically cluster genes. P < 0.05 was considered significant.

Data availability

We deposited all raw and processed sequencing data to NCBI Genome Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) under the following accession numbers: GSE86861 for the Affymetrix data on BCC subsets (34), GSE150152 for the series of dataset for single-cell sequencing of BCCs, and GSE138435 for MSC-derived exosomes.

Naïve versus primed EVs

MSC-derived EVs can induce functions linked to breast cancer dormancy, for example, cell-cycle quiescence and chemoresistance (35, 36). Because MSCs are located at the peripheral-BM cavity interface, we sought the mechanisms by which MSC-released exosomes start the process of BCCs entering dormancy. We used a coculture method to isolate naïve (MSCs never contacted BCCs) and primed (MSCs exposed to BCCs) exosomes, which we added to fresh BCCs (Fig. 1A). We demonstrated exosomal entry into BCCs by confocal microscopy (17), as well as dSTORM imaging to track the MV within BCCs, first demonstrating the EV bouncing on and off BCCs (Supplementary Video SA) and then its movement into BCCs (Supplementary Video SB).

Figure 1.

Exosome characterization and cycling phase of BCC subsets. A, Illustration of the method used to purify naïve and primed exosomes from MSCs. Also shown is the addition of the purified exosomes to fresh BCCs. B, FACS analysis with naïve and primed exosomes for CD63, CD9, and CD90. The exosomes were isolated from the two triple-negative BCC lines and from T47D, and then coupled to anti-CD63 coupled beads. C, Western blot analysis for TSG101 and ALIX with MV extracts from MDA-MB-231. D, Western blot analyses for Alix and GrP78 with extracts from MV and whole-cell extract. E, Western blot analysis for CD63 with extracts of particles from the pellets of 2,000 and 10,000 × g centrifugation during MV isolation. F, Representative transmission electron microscopy of MVs that were selected by gradient centrifugation or using the Exosome Isolation Kit (Invitrogen). G, Histogram of naïve and primed exosomes/mL versus size using Nanosight tracking analysis. The histogram represents exosomes from both cell lines. H, Total RNA was isolated from 108 exosomes from both BCC lines using the miRCURY RNA Isolation Kit (Exiqon), and then quantitated and presented as mean total RNA (ng/μL) ± SD, n = 5. I, Super resolution image for CD9 on naïve and primed exosomes from MDA-MB-231. The graph represents the distribution of CD9/exosome (bottom). J, BCCs (both triple-negative lines), treated with 108 naïve or primed exosomes for 48 hours, were labeled with 7-AAD and pyronin Y. The cells were analyzed by flow cytometry for G0-phase. ±SD, n = 7; each experiment was performed in triplicate. K, Adaptation of known BCC hierarchy to show how specific BCC subset responds to naïve and/or primed exosomes (black arrow) with respect to G0-phase transition of the cell cycle. Extended information in Supplementary Fig. S1. NS, nonsignificant.

Figure 1.

Exosome characterization and cycling phase of BCC subsets. A, Illustration of the method used to purify naïve and primed exosomes from MSCs. Also shown is the addition of the purified exosomes to fresh BCCs. B, FACS analysis with naïve and primed exosomes for CD63, CD9, and CD90. The exosomes were isolated from the two triple-negative BCC lines and from T47D, and then coupled to anti-CD63 coupled beads. C, Western blot analysis for TSG101 and ALIX with MV extracts from MDA-MB-231. D, Western blot analyses for Alix and GrP78 with extracts from MV and whole-cell extract. E, Western blot analysis for CD63 with extracts of particles from the pellets of 2,000 and 10,000 × g centrifugation during MV isolation. F, Representative transmission electron microscopy of MVs that were selected by gradient centrifugation or using the Exosome Isolation Kit (Invitrogen). G, Histogram of naïve and primed exosomes/mL versus size using Nanosight tracking analysis. The histogram represents exosomes from both cell lines. H, Total RNA was isolated from 108 exosomes from both BCC lines using the miRCURY RNA Isolation Kit (Exiqon), and then quantitated and presented as mean total RNA (ng/μL) ± SD, n = 5. I, Super resolution image for CD9 on naïve and primed exosomes from MDA-MB-231. The graph represents the distribution of CD9/exosome (bottom). J, BCCs (both triple-negative lines), treated with 108 naïve or primed exosomes for 48 hours, were labeled with 7-AAD and pyronin Y. The cells were analyzed by flow cytometry for G0-phase. ±SD, n = 7; each experiment was performed in triplicate. K, Adaptation of known BCC hierarchy to show how specific BCC subset responds to naïve and/or primed exosomes (black arrow) with respect to G0-phase transition of the cell cycle. Extended information in Supplementary Fig. S1. NS, nonsignificant.

Close modal

We validated the particles for endosomal tetraspanin proteins by flow cytometry and Western blotting (Fig. 1B and C). The particles did not blot of GrP78 endosomal protein (Fig. 1D). The specificity of endosomal positivity was determined by analyzing the pellets for CD63 from 10,000 × g centrifugation (Fig. 1E). We noted smeared bands in the Western blot with extracts from the 2,000 × g centrifugation pellets, suggesting whole-cell fragments (Fig. 1E). Analyses of particles, isolated by gradient centrifugation, by transmission electron microscopy and NTA indicated <120 nm (Fig. 1F and G). We noted a wider distribution among the naïve exosomes as compared with primed exosomes, which contained four times more total RNA (Fig. 1H).

The EV, <50 nm, could be exomeres, which are CD9 (37). Super resolution microscopy with anti-CD9-Alex 647 showed fluorescence clusters (Fig. 1I, red spots), indicating lack of exomeres. The histogram below the images represents CD9 distribution within one vesicle. On the basis of the size and expression of endosomal proteins on the isolated EVs, we deduced that our EVs comprised of exosomes.

Cycling quiescence of EV-treated BCC subsets

Dose–response studies using cyclin D1 (Western blotting and reporter gene activity) and Oct4a levels in BCCs as readouts determined that 108 MSC particles were the optimal dose (Supplementary Figs. S1A–S1C and S2A). We studied the effects of EVs in cycling of specific BCC subsets by treating BCC-pOct4a-GFP with naïve and primed EVs for 48 hours. We labeled the BCCs with pyronin Y and 7-AAD, and then gated different subsets on the basis of GFP intensity (Fig. 1J, top) to determine the percentage of cells in G0-phase (Fig. 1J, bottom). G0-phase was significantly (P < 0.05) increased when Oct4alo BCCs were exposed to prime or naïve exosomes, relative to untreated cells (right bars). Positive control Oct4ahi BCCs, which are mostly CSCs, showed similar results, regardless of MV treatment (left bars; ref. 38). The Oct4amed subsets (Med-1–3) showed heterogeneity with respect to G0-phase, the response depended on the EV type. Each of the three Oct4amed subsets responded similarly to naïve exosomes with respect to G0-phase of the cell cycle. Oct4a-Med-2 responded equally to both naïve and primed exosomes. In summary, BCCs, which were catalogued previously into hierarchy using Oct4a expression as low, medium, or high, indicated that Oct4alo BCCs were most responsive with respect to robust entry to G0-phase with naïve or primed exosomes, relative to untreated corresponding subset (Fig. 1K).

MVs induce stem cell-like properties in BCCs

Because Oct4alo BCCs showed the highest percentage of cells in cycling quiescent after 48-hour exposure to EVs (Fig. 1), we asked whether prolonged exposure to EVs would provide BCCs with stem cell properties. We treated MDA-MB-231-pOct4a-dsRed with naïve or primed exosomes at 2-day intervals. At 1 week, we analyzed the cells for dsRed intensity, which served as a surrogate for the expression of the stem cell gene, Oct4a (21). The untreated BCCs (baseline) at days 0 and 7 showed weak dsRed intensities (Fig. 2A). However, exosome treatment caused a right shift, indicating Oct4a increase (21). This led us to ask whether exosomes induced the expression of other stem cell–associated genes.

Figure 2.

Stem cell transition in exosome-treated BCCs. A, MDA-MB-231 with stable pEdsRED1-Oct3/4 were treated with naïve and primed exosomes or untreated as for Fig. 1A. Day 0 (baseline) and day 7–treated BCCs were analyzed by FACScan for dsRED. The figure represents five different experiments. B, The studies in A were repeated and total RNA was analyzed by real-time PCR for the genes shown in the graph. The normalized values are presented as mean 2ΔΔCt ± SD, n = 4. Each experiment used exosomes from a different MSC donor. *, P < 0.05 versus primed exosomes. C, The studies in B were repeated, except for Western blot analyses with whole-cell extracts. The membranes were developed with specific antibodies for the proteins shown in the blot. D, Tumorsphere assay used unsorted BCCs, untreated or treated with naïve or primed exosomes. The data are presented as the mean tumorsphere/100 cells ± SD, n = 5. E, Representative images of tumorspheres from serial-passaged BCCs that were untreated or treated with naïve or primed exosomes. F, BCCs were untreated or treated with naïve or primed exosomes. At day 7, the cells were treated with 2 μmol/L doxorubicin. At different times, cell viability was assessed with 7-AAD and the results are shown as percentage mean dead cells ± SD, n = 4. **, P < 0.01 vs. similar time point with exosome treatment. G, Dye efflux analyses were performed with BCCs, untreated or treated with naïve or primed exosomes for 7 days. The results are presented as the mean percentage dye retention ± SD, n = 4. *, P < 0.05 vs. exosome treatment; **, P < 0.05 vs. naive exosome. H, Real-time PCR for MDR1 mRNA with RNA from unsorted BCCs, untreated or treated with naïve or primed exosomes for 7 days. The Ct values for untreated cells were assigned a value of 1 and the experimental value is presented as mean fold change ± SD, n = 4. *, P < 0.05 versus untreated cell. I, Experiments in H were repeated, except for Western blots with whole-cell extracts for P-gp1. J, Autophagic vacuole detection in unsorted BCCs, untreated or treated with naïve or primed exosomes for 7 days. Positive control cells were treated with 500 μmol/L rapamycin. Images were taken with EVOS FL2 at ×100 magnification. K, The experiment in J was repeated for Western blot for autophagy-associated ATG5. Scale bar, 175 μm. Extended information in Supplementary Fig. S2.

Figure 2.

Stem cell transition in exosome-treated BCCs. A, MDA-MB-231 with stable pEdsRED1-Oct3/4 were treated with naïve and primed exosomes or untreated as for Fig. 1A. Day 0 (baseline) and day 7–treated BCCs were analyzed by FACScan for dsRED. The figure represents five different experiments. B, The studies in A were repeated and total RNA was analyzed by real-time PCR for the genes shown in the graph. The normalized values are presented as mean 2ΔΔCt ± SD, n = 4. Each experiment used exosomes from a different MSC donor. *, P < 0.05 versus primed exosomes. C, The studies in B were repeated, except for Western blot analyses with whole-cell extracts. The membranes were developed with specific antibodies for the proteins shown in the blot. D, Tumorsphere assay used unsorted BCCs, untreated or treated with naïve or primed exosomes. The data are presented as the mean tumorsphere/100 cells ± SD, n = 5. E, Representative images of tumorspheres from serial-passaged BCCs that were untreated or treated with naïve or primed exosomes. F, BCCs were untreated or treated with naïve or primed exosomes. At day 7, the cells were treated with 2 μmol/L doxorubicin. At different times, cell viability was assessed with 7-AAD and the results are shown as percentage mean dead cells ± SD, n = 4. **, P < 0.01 vs. similar time point with exosome treatment. G, Dye efflux analyses were performed with BCCs, untreated or treated with naïve or primed exosomes for 7 days. The results are presented as the mean percentage dye retention ± SD, n = 4. *, P < 0.05 vs. exosome treatment; **, P < 0.05 vs. naive exosome. H, Real-time PCR for MDR1 mRNA with RNA from unsorted BCCs, untreated or treated with naïve or primed exosomes for 7 days. The Ct values for untreated cells were assigned a value of 1 and the experimental value is presented as mean fold change ± SD, n = 4. *, P < 0.05 versus untreated cell. I, Experiments in H were repeated, except for Western blots with whole-cell extracts for P-gp1. J, Autophagic vacuole detection in unsorted BCCs, untreated or treated with naïve or primed exosomes for 7 days. Positive control cells were treated with 500 μmol/L rapamycin. Images were taken with EVOS FL2 at ×100 magnification. K, The experiment in J was repeated for Western blot for autophagy-associated ATG5. Scale bar, 175 μm. Extended information in Supplementary Fig. S2.

Close modal

Real-time PCR showed significant (P < 0.05) increases of Oct4a, Nanog, KLF4, Hes1, and ID2 mRNA in primed exosome–treated cells, relative to untreated cells (Fig. 2B). The levels of Nanog, KLF4, Hes1, and ID2 were significantly (P < 0.05) higher in primed exosomes as compared with naïve exosomes (Fig. 2B). Western blots confirmed the increases in Oct4a and Nanog (Fig. 2C; Supplementary Fig. S2B). Active NF-κB, which is linked to stemness (39), showed increased band density for phosphorylated p65 (NF-κB subunit) with primed exosomes (Fig. 2C; Supplementary Fig. S2B; ref. 40). Because Notch (mRNA and protein) was unchanged, we deduced that this gene might not have a major role in the exosome-mediated changes in BCCs.

The increases in stem cell–associated genes led us to analyze 7-day exosome-treated BCCs for tumorsphere formation at one cell per low attachment well (21). The results indicated significantly (P < 0.05) more wells forming clusters of cells resembling tumorsphere when the BCCs were exposed to naïve or primed exosomes (Fig. 2D). Positive control with CSCs (Oct4ahi) showed serial passaging (Fig. 2E, top; ref. 21). Spheres from untreated BCCs failed to undergo serial passaging, whereas naïve or primed exosome–treated cells could be serially passaged (Fig. 2E). However, the passaged spheres from the exosome-treated cells were relatively smaller as compared with the positive control, suggesting that the transition to stem cells was an ongoing process.

Because we noted chemoresistance by CSCs, and observed BCCs transitioning into cells with stem cell properties after exposure to EVs, we asked whether EV treatment caused the noted chemoresistance. We subjected the EV-treated BCCs to doxorubicin (Adriamycin, 2 μmol/L) and observed significant (P < 0.01) protection by EVs (Fig. 2F). This correlated with significant (P < 0.05) dye efflux in the EV-treated BCCs (Fig. 2G; Supplementary Fig. S2C). The primed EVs were more efficient at dye efflux than naïve EVs. Prime EV–treated BCCs showed a bright band for P-gp in Western blot as compared with naïve EVs and showed increased levels of the corresponding mRNA (Fig. 2H and I; Supplementary Fig. S2D).

We analyzed EV-treated BCCs for autophagy because of their involvement with stem cell survival (41). EVs, particularly primed, showed autophagic vacuoles in cells with high GFP intensity, as compared with untreated cultures (Fig. 2J, yellow cells). This strongly suggested an autophagic response in BCCs with stem cell properties. Atg5 protein, which is required for autophagy, was increased with naïve and primed EV exposure (Fig. 2K; ref. 42). In summary, EV treatment caused BCCs to adapt properties consistent with stemness, indicating transition into CSCs.

Metabolic changes in exosome-treated BCCs

A balanced metabolic activity in stem cells is essential to prevent their exhaustion and to maintain self-renewal (43). Thus, we determined whether exosome treatment affected the metabolic activity of BCCs. Their proliferation was significantly (P < 0.05) reduced with EV treatment (Fig. 3A). Reactive oxygen species (ROS), which correlate with metabolic activity (44), were significantly (P < 0.05) reduced with EV treatment, relative to untreated BCCs (Fig. 3B). Mitochondrial distribution using MitoTracker (MitoRed) indicated a significant (P < 0.05) decrease of MitoSOX in BCCs treated with naïve and primed exosomes, as compared with untreated cells (Fig. 3C). Next, we analyzed Oct4alo BCCs for MitoSOX, before and after dedifferentiation with EVs (Fig. 3D). Immunofluorescence imaging indicated a significant (P < 0.05) decrease in MitoSOX when the Oct4alo BCCs dedifferentiated to GFPhi BCCs (Fig. 3E). Thus, there was a negative correlation between dedifferentiation and mitochondrial activity. Superoxide, a mitochondrial byproduct, which can lead to cell damage, can be detoxified by manganese superoxide dismutase (MnSOD) to prevent cellular detriment (45). Thus, reduced MitoSOX should lead to reduced MnSOD, which we confirmed by Western blotting using protein extracts from EV-treated BCCs (Fig. 3F). Together, these results show a decrease in the metabolic activity of BCCs exposed to MSC-driven EVs as a plausible protective mechanism against cellular damage.

Figure 3.

Metabolic activity of BCCs treated with MSC-released MVs. A, The proliferative response of BCCs was assessed with CyQUANT cell proliferation assay kit. The cells were treated with naïve or primed MVs, or untreated. The data are presented as mean number of viable cells ± SD, n = 4. *, P < 0.05 versus similar time points with MV treatment. B, ROS activity in BCCs that were untreated (none) or treated with naïve or primed MVs for 7 days. Positive control cells were treated with 55 μmol/L tert-butyl hydroperoxide (TBHP). The results are presented as mean ROS activity ± SD, n = 4. *, P < 0.05 versus MV treatment. C, Left, MitoRed mitochondrial staining was done with BCCs, untreated or treated with naïve or primed MVs for 7 days. Right, the images were quantitated with ImageJ, and then presented as the mean fluorescence intensity (MFI) ± SD, n = 4. D, The studies in C were repeated, except for labeling for MitoSOX. White arrows depict stained cells within the GFP bright BCCs. Shown are quantitated images of four experiments as for C. E, The fluorescence intensity of MitoSOX within the GFP bright cells in D was quantified and is presented as mean ± SD, n = 4. *, P < 0.05 versus MV treatment. F, The experiment in E was repeated, except for isolation of whole-cell extracts, which was used in Western blot analysis for MnSOD. The data were normalized to β-actin. Below are the mean normalized band densities for three independent experiments, ±SD. *, P < 0.05 versus MV treatment.

Figure 3.

Metabolic activity of BCCs treated with MSC-released MVs. A, The proliferative response of BCCs was assessed with CyQUANT cell proliferation assay kit. The cells were treated with naïve or primed MVs, or untreated. The data are presented as mean number of viable cells ± SD, n = 4. *, P < 0.05 versus similar time points with MV treatment. B, ROS activity in BCCs that were untreated (none) or treated with naïve or primed MVs for 7 days. Positive control cells were treated with 55 μmol/L tert-butyl hydroperoxide (TBHP). The results are presented as mean ROS activity ± SD, n = 4. *, P < 0.05 versus MV treatment. C, Left, MitoRed mitochondrial staining was done with BCCs, untreated or treated with naïve or primed MVs for 7 days. Right, the images were quantitated with ImageJ, and then presented as the mean fluorescence intensity (MFI) ± SD, n = 4. D, The studies in C were repeated, except for labeling for MitoSOX. White arrows depict stained cells within the GFP bright BCCs. Shown are quantitated images of four experiments as for C. E, The fluorescence intensity of MitoSOX within the GFP bright cells in D was quantified and is presented as mean ± SD, n = 4. *, P < 0.05 versus MV treatment. F, The experiment in E was repeated, except for isolation of whole-cell extracts, which was used in Western blot analysis for MnSOD. The data were normalized to β-actin. Below are the mean normalized band densities for three independent experiments, ±SD. *, P < 0.05 versus MV treatment.

Close modal

Stem cell–associated pathways in late progenitors (Oct4alo)

Oct4alo BCCs, which are not functionally CSCs, responded to MSC-derived EVs by dedifferentiating into stem-like cells (Figs. 1–3; ref. 21). To gain insights into the reason for this observation, we revisited a previously deposited Affymetrix database (GSE86861) for Oct4alo and Oct4ahi (CSCs) BCCs (34). Using an unbiased approach, we screened the genes in the technical replicates for Oct4alo BCCs. The data, plotted as log2 expression variance versus log2 mean expression intensity, showed a linear correlation (Fig. 4A). The top 300 genes (1%) that deviated from the linear relationship (red) indicated those with the highest variance. The replicates showing the varied gene expression suggested that these genes comprised the heterogeneity within Oct4alo BCCs.

Figure 4.

Bioinformatics with data from Oct4alo BCCs; in vivo dedifferentiation of Oct4alo BCCs. A, Affymetrix microarray analysis of log variance versus log means expression intensity with the variant genes within Oct4alo cells (red) using the data from accession number GSE86861. B, Panther gene ontology analysis of the variant genes showing the top five signaling pathways in Oct4ahi and Oct4alo BCCs. The data for Oct4ahi BCCs also used the data from GSE86861. C, Western blot analysis for β-catenin and p53 with extracts from BCCs, untreated or treated with exosomes. D, The gating scheme used to isolate BCC subsets (left), fluorescence images of sorted BCCs, based on GFP intensity (middle), and protocol used to inject 106 sorted Oct4alo or Oct4amed BCCs intravenously in female 8-week-old NSG mice (right). After 3 days, the femurs were analyzed and the results are shown in E–G. E, Real-time PCR for GFP with RNA isolated from nucleated cells of mice femurs. The normalized Ct values for sorted Oct4alo BCCs was assigned 1 and this was used to determine the fold changes of the experimental points. The data are presented as mean fold change ± SD, n = 5. F, IHC of BM sections for Oct4a and cytokeratin. The images represent staining of femurs from five different mice. G, Tissues at the endosteal regions of femurs were scraped and then analyzed on the EVOS FL2 for GFP. The images represent five analyses, each from femur of a different mouse. Extended information in Supplementary Figs. S3 and S4.

Figure 4.

Bioinformatics with data from Oct4alo BCCs; in vivo dedifferentiation of Oct4alo BCCs. A, Affymetrix microarray analysis of log variance versus log means expression intensity with the variant genes within Oct4alo cells (red) using the data from accession number GSE86861. B, Panther gene ontology analysis of the variant genes showing the top five signaling pathways in Oct4ahi and Oct4alo BCCs. The data for Oct4ahi BCCs also used the data from GSE86861. C, Western blot analysis for β-catenin and p53 with extracts from BCCs, untreated or treated with exosomes. D, The gating scheme used to isolate BCC subsets (left), fluorescence images of sorted BCCs, based on GFP intensity (middle), and protocol used to inject 106 sorted Oct4alo or Oct4amed BCCs intravenously in female 8-week-old NSG mice (right). After 3 days, the femurs were analyzed and the results are shown in E–G. E, Real-time PCR for GFP with RNA isolated from nucleated cells of mice femurs. The normalized Ct values for sorted Oct4alo BCCs was assigned 1 and this was used to determine the fold changes of the experimental points. The data are presented as mean fold change ± SD, n = 5. F, IHC of BM sections for Oct4a and cytokeratin. The images represent staining of femurs from five different mice. G, Tissues at the endosteal regions of femurs were scraped and then analyzed on the EVOS FL2 for GFP. The images represent five analyses, each from femur of a different mouse. Extended information in Supplementary Figs. S3 and S4.

Close modal

Using Panther gene ontology database, we analyzed the 300 genes for top signaling pathways and the output indicated those linked to stem cell maintenance (Fig. 4B). We subjected these genes in the top pathways to STRING for protein interactions and the output supported links to multipotent genes (Supplementary Fig. S3A). These findings were similar to similar analyses using the genes deposited for Oct4ahi BCCs on the same series of data (Fig. 4B; Supplementary Fig. S3B; refs. 21, 34). Notably, in both pathway analyses, the Wnt-β-catenin pathway was upregulated with twice the number of genes as compared with Oct4ahi/CSCs. This correlated with the Western blot analysis showing increased band for β-catenin (Fig. 4C) using extracts from MDA-MB-231. Interestingly, we observed an increase in p53 with primed exosomes.

We subjected the variant genes to The Cancer Genome Atlas database and the output indicated a worse outcome (Supplementary Fig. S4). These findings indicated that baseline Oct4alo BCCs express genes similar to CSCs and might explain the ease by which this subset dedifferentiated into CSC-like cells following EV exposure.

Dedifferentiation of Oct4alo BCCs in vivo

Functional and bioinformatics evidence indicated dedifferentiation of Oct4alo BCCs, but did not provide evidence of similar dedifferentiation in vivo. Thus, we studied murine BM because this organ contains cells that can support BCC dormancy (17, 21). We injected NSG mice intravenously with sorted Oct4alo and Oct4amed BCCs (106), which were GFPlo/– (Fig. 4D). After 72 hours, real-time PCR evaluated the nucleated femur cells for GFP. The normalized Ct for baseline (sorted) Oct4alo BCCs was assigned 1. After injection, the Oct4alo BCCs showed 2-fold more GFP, relative to freshly sorted Oct4alo BCCs (Fig. 4E). Because the Oct4amed were GFP+, we used this group of mice as positive control for the qPCR (Fig. 4E). We confirmed these findings at the level of protein with IHC analyses of femur sections. We labeled the sections for pan-cytokeratin (blue)/BCCs, and then analyzed the blue cells for GFP (GFP/green + blue = purple). Indeed, we observed purple cells, confirming dedifferentiation of Oct4alo BCCs (Fig. 4F). These findings correlated with bright GFP cells in scraped endosteal tissues (Fig. 4G). In summary, these studies showed in vivo dedifferentiation of Oct4alo/med BCCs.

Single-cell sequencing of BCCs exposed to MSC EVs

The source of EVs (naïve vs. primed) influenced the efficiency of Oct4alo BCC dedifferentiation (Figs. 14). We also confirmed that dedifferentiation of Oct4alo BCCs occurred in vivo in murine femur (Fig. 4). To gain an understanding of these changes, we performed single-cell RNA-seq with BCCs, untreated or treated with naïve or primed EVs. The t-NSE plots and dot plots showed EV-induced distinct changes with respect to the resulting BCC populations (Fig. 5A; Supplementary Fig. S5A). A heatmap of genes (SD ± 1.5) highlighted the distinct gene clusters among the different treatment groups (Fig. 5B; Supplementary Fig. S5B and S5C). We selected the unique genes (Fig. 5B, boxed regions A–C, hereafter referred as unique genes) for further analyses.

Figure 5.

Wnt/β-catenin pathway genes in MSC exosome–treated BCCs. A, t-SNE plot of MDA-MB-231, untreated (red) or treated with naïve (blue) or primed exosomes (yellow). B, Heatmap of genes shown in A. C, Molecular function of genes unique to untreated cells within the boxed regions (A) of the heatmap. D–F, Signaling pathways of the unique/increased genes (boxed regions in heatmap shown in B) in BCCs treated with naïve (D) or primed (E and F) exosomes. G, t-SNE plot of distinct BCC subsets. H–J, Gene ontology using the top five pathways within the major breast cancer subsets in the treatment groups shown in G. K, Flow cytometry for GFP in Oct4alo BCCs, treated with Wnt inhibitor (8 μmol/L JW74) and/or naïve or primed exosomes for 7 days. The data are presented as mean fluorescence intensity (MFI) ± SD, n = 4. L, The experiments in K were repeated and cell extracts were analyzed by Western blotting (left) for the shown proteins. Representative blot with the mean ± SD normalized band densities for three independent experiments at right. *, P < 0.05 versus vehicle. Extended information in Supplementary Fig. S5.

Figure 5.

Wnt/β-catenin pathway genes in MSC exosome–treated BCCs. A, t-SNE plot of MDA-MB-231, untreated (red) or treated with naïve (blue) or primed exosomes (yellow). B, Heatmap of genes shown in A. C, Molecular function of genes unique to untreated cells within the boxed regions (A) of the heatmap. D–F, Signaling pathways of the unique/increased genes (boxed regions in heatmap shown in B) in BCCs treated with naïve (D) or primed (E and F) exosomes. G, t-SNE plot of distinct BCC subsets. H–J, Gene ontology using the top five pathways within the major breast cancer subsets in the treatment groups shown in G. K, Flow cytometry for GFP in Oct4alo BCCs, treated with Wnt inhibitor (8 μmol/L JW74) and/or naïve or primed exosomes for 7 days. The data are presented as mean fluorescence intensity (MFI) ± SD, n = 4. L, The experiments in K were repeated and cell extracts were analyzed by Western blotting (left) for the shown proteins. Representative blot with the mean ± SD normalized band densities for three independent experiments at right. *, P < 0.05 versus vehicle. Extended information in Supplementary Fig. S5.

Close modal

Top pathways within the unique genes in the untreated group (box A) were linked to G-protein coupled receptors, including chemokine receptors (Fig. 5C). The unique genes (box B) in the naïve EV–treated group identified pathways involved in repair of DNA double-strand break, mismatch sequences, and nucleotide excision repair (Fig. 5D). Similar analyses for BCCs treated with primed EVs (box C) showed canonical and noncanonical Wnt pathways—Wnt/Ca+, Wnt/β-catenin, integrin, and cell-cycle regulation (Fig. 5E; ref. 46). Interestingly, primed EVs, which are relevant to transition BCCs into stemness, decreased immune response pathways, immunoglobulin and complement (Fig. 5F).

Further analyses of the untreated and EV-treated groups showed distinct clusters (Fig. 5G). The clusters derived from primed EVs showed relatively more homogeneity as compared with naïve EV treatment. This observation was consistent with the clustered EV size released from primed MSCs (Fig. 1G). Using Panther ontology, we stratified the clusters in Fig. 5G, on the basis of similarities. We aligned the genes within each cluster with the data deposited for Oct4ahi BCCs, which contained mostly CSCs (GSE86861; ref. 33). We identified clusters 11, 13, and 4 in the untreated, naïve, and primed treatment groups, respectively, as those with the most expressed stem-linked genes (Fig. 5G, clusters in center). IPAs of each cluster (Fig. 5G) for top five pathways (Fig. 5H–J) indicated similar pathways between untreated and naïve exosome, clusters 5 and 2 and 14 and 7, respectively. Wnt pathway, which is activated in triple-negative BCCs, was changed within the distinct subsets after specific EV treatment (47).

Wnt in exosome-mediated dedifferentiation

The source of EV influenced Wnt signaling pathways in specific BCC clusters. The top Wnt pathways resulting with primed EV treatment were both canonical and noncanonical, as evidenced by Wnt/Ca+ and Wnt/β-catenin, whereas naïve EVs showed only canonical Wnt pathway (Fig. 5E and H–J; ref. 47). We, therefore, determined whether Wnt signaling has a role in EV-mediated dedifferentiation by focusing on the canonical pathway. We treated late breast cancer progenitors (Oct4alo) with naïve or primed EVs, in the presence of inhibitor of Wnt destruction complex (JW74) or vehicle. JW74 caused a significant (P < 0.05) decrease in GFP intensity relative to vehicle (Fig. 5K), indicating a role for the canonical Wnt-β-catenin pathway in Oct4alo dedifferentiation into stem-like cells. Western blot analysis showed double bands for β-catenin, suggesting phosphorylation, which is required for β-catenin degradation (Fig. 5L). JW74 added to cells without EV showed a bright band for β-catenin, indicating an accumulation of phospho-catenin in the stabilized degradation complex. JW74 resulted in a double band in naïve EV–treated cells, which was consistent with a blunted activation of β-catenin. There was no marked difference with JW74-treated cells containing primed EVs, suggesting the possible involvement of the noncanonical (Wnt-Ca2+) pathway with primed EV treatment (Fig. 5E).

Because dedifferentiated/stem-like cells would be in cycling quiescence, we studied cyclin D1 and CDK4 levels in EV-treated (with or without JW74) BCCs by Western blotting. Both proteins were decrease with JW74 treatment (Fig. 5L). We noted double bands for β-catenin with JW74 even without EV. This indicated that baseline BCCs contained phosphorylated β-catenin, making this protein susceptible for degradation. In summary, MSC-derived EVs (primed and naïve) induced dedifferentiation, and control cell cycle of Oct4alo BCCs via canonical Wnt-β-catenin.

Validation of Wnt-catenin in genes and pathways commonly associated with breast cancer

The experimental results point to a role for Wnt-catenin in BCC dedifferentiation. Thus, we performed a robust search using VarElect for common genes linked to breast cancer. Figure 6A lists the genes with scores >30. More importantly, the top gene was catenin. We subjected the genes in Fig. 6A to STRING for protein interaction (Fig. 6B). The results showed catenin with links to genes associated with drug resistance and stemness, further pointing to Wnt-β-catenin as key in the dedifferentiation of BCCs. Next, we subjected the genes in Supplementary Table S6A to Panther for top pathways and identified Wnt, cadherin, and presenilin (Fig. 6C). These findings indicated that the pathways identified in the single-cell RNA-seq database were relevant to the known genes and pathways associated with breast cancer.

Figure 6.

MSC-derived exosome cargo. A, VerElect algorithm was applied to identify common genes that are linked to breast cancer. Shown are the genes scoring >30. B, STRING analyses for protein interaction using the genes in A. C, Panther gene ontology analysis of signaling pathways of the genes in A. D, MSC exosomes (naïve and primed) were subjected to RNA-seq. Shown is the PCA plot of the data. E, Heatmap of the RNA-seq genes in naïve and primed exosomes. F, Canonical pathways identified with IPA of the genes in primed versus naïve exosomes. G, IPA overlapped the information on Wnt pathway with the dataset from primed versus naïve exosome treatment shown in E. Extended information in Supplementary Figs. S6 and S7.

Figure 6.

MSC-derived exosome cargo. A, VerElect algorithm was applied to identify common genes that are linked to breast cancer. Shown are the genes scoring >30. B, STRING analyses for protein interaction using the genes in A. C, Panther gene ontology analysis of signaling pathways of the genes in A. D, MSC exosomes (naïve and primed) were subjected to RNA-seq. Shown is the PCA plot of the data. E, Heatmap of the RNA-seq genes in naïve and primed exosomes. F, Canonical pathways identified with IPA of the genes in primed versus naïve exosomes. G, IPA overlapped the information on Wnt pathway with the dataset from primed versus naïve exosome treatment shown in E. Extended information in Supplementary Figs. S6 and S7.

Close modal

RNA-seq—exosomal cargo

To advance our understanding on the link between Wnt pathways and breast cancer dedifferentiation, we sought the contents of the EV cargo (naïve and prime) by RNA-seq. The PCA plot from three biological samples indicated clustering of the genes from each group, although one donor in the primed group was distant from the others within the same set (Fig. 6D). The data when presented in a heatmap indicated similarities among the biological replicates, but distinct genes between naïve and primed EVs (Fig. 6E; Supplementary Fig. S6A).

We selected the uniquely expressed genes between primed versus naïve EVs and subjected them to IPA for top canonical pathways. The output indicated that primed EVs contained genes within the Wnt-β-catenin pathway, drug resistance, metabolism, development, and multipotency (Fig. 6F; Supplementary Fig. S6B). The identified Wnt-catenin genes within the EV cargo correlated with the functional studies, as well as the top genes in the database of specific diseases, in this case breast cancer (Figs. 5L and 6A–C). We, therefore, overlapped the data from the RNA-seq data (primed vs. naïve) with the dataset for canonical Wnt in IPA. The output showed Wnt target genes linked to reprograming genes, such as Myc, KLF4, and other stem cells genes, for example, ID2 and Oct4 (Fig. 6G). This further supported a role for Wnt pathways in dedifferentiation, and partly linked partial reprogramming Myc and KLF4. This might explain the ease by which Oct4alo BCCs dedifferentiate into cells with stem cell phenotype. In total, the MSC-secreted EV cargo contains RNAs that have the potential to induce dedifferentiation through activation of the Wnt-catenin pathway.

In vivo interaction between MSCs and Oct4alo BCCs

Although we showed in vitro and in vivo roles for MSC-derived EVs in the dedifferentiation of Oct4alo BCCs, the question is the extent of in vivo MSC–BCC interaction especially because MSCs are located throughout the marrow and at the perivascular niche (26). Thus, upon entry into the BM, BCCs quickly encounter MSCs and interact, partly via membrane-bound CXCR4 and CXCL12 (22). We surmise that MSC–BCC interaction serves as an advantage to BCCs. Specifically, such close encounter allows the BCCs to incorporate MSC-released exosomes (naïve), resulting in rapid transition into cycling quiescence (Fig. 1H). In addition, MSCs can confer immune protection to the incoming BCCs (23).

Figure 7A summarizes our data in the context of the literature on dormant CSCs in the endosteal niche. This figure depicts the first step when naïve MSC EVs facilitate Oct4lo BCCs to enter cycling quiescence. The second step entails exposure to the delayed release of EVs from primed MSCs. This leads to the Oct4lo BCCs transitioning into cells with stem properties to prepare them for complete dormancy close to the endosteal niche (10, 20, 21).

Figure 7.

In vivo interaction between MSCs and Oct4alo BCCs. A, Summary of the findings supporting Oct4alo BCCs transitioning to CSCs, in the context of known information. Shown are the early response to naïve exosomes followed by primed exosome, resulting in a phenotype consistent with stemness. B, IHC for Numa (Alexa 405/blue) and CD146-PE/red pericytes/MSCs with sections from mice femurs. Oct4alo BCCs were injected into mice, as for Fig. 4D. Representative image represents five experiments, each with sections from a different mouse. Inset with white arrows shows the GFP+/green cells distant from the blood vessels but closer to the endosteum. C, IHC for cytokeratin-Alexa 405/blue (cancer cells) and CD29-PE/red (MSCs) in sections of BM biopsies from two patients with breast cancer. Zoomed region showing the purple labeling of each image (right). D, BCCs (untreated and naïve exosome treatment) were assessed for migration across endothelial cells (EC) and MSCs using the method shown in the diagram at right. The data are presented as mean (± SD) fluorescence, n = 3. *, P < 0.05 versus naïve BCCs + MSCs + endothelial cells. E, Network of unique genes taken from Fig. 5B within the boxed regions: naïve versus no exosome (left) and primed versus no exosome (right). F, Immunofluorescence for PARP, H2AX, and Oct4A. Inset shows an enlarged nuclei depicting the concentration of H2AX. G, Dot plot of epigenetic restructuring proteins from the RNA-seq data. H, The diagram summarizes the data in this figure and incorporates how the stepwise process occurs.

Figure 7.

In vivo interaction between MSCs and Oct4alo BCCs. A, Summary of the findings supporting Oct4alo BCCs transitioning to CSCs, in the context of known information. Shown are the early response to naïve exosomes followed by primed exosome, resulting in a phenotype consistent with stemness. B, IHC for Numa (Alexa 405/blue) and CD146-PE/red pericytes/MSCs with sections from mice femurs. Oct4alo BCCs were injected into mice, as for Fig. 4D. Representative image represents five experiments, each with sections from a different mouse. Inset with white arrows shows the GFP+/green cells distant from the blood vessels but closer to the endosteum. C, IHC for cytokeratin-Alexa 405/blue (cancer cells) and CD29-PE/red (MSCs) in sections of BM biopsies from two patients with breast cancer. Zoomed region showing the purple labeling of each image (right). D, BCCs (untreated and naïve exosome treatment) were assessed for migration across endothelial cells (EC) and MSCs using the method shown in the diagram at right. The data are presented as mean (± SD) fluorescence, n = 3. *, P < 0.05 versus naïve BCCs + MSCs + endothelial cells. E, Network of unique genes taken from Fig. 5B within the boxed regions: naïve versus no exosome (left) and primed versus no exosome (right). F, Immunofluorescence for PARP, H2AX, and Oct4A. Inset shows an enlarged nuclei depicting the concentration of H2AX. G, Dot plot of epigenetic restructuring proteins from the RNA-seq data. H, The diagram summarizes the data in this figure and incorporates how the stepwise process occurs.

Close modal

On the basis of the above rationale, we asked whether MSCs interact with dedifferentiated BCCs in vivo, and whether this interaction expands throughout the marrow cavity. Oct4alo BCCs (GFP–/lo) were injected into NSG mice as for Fig. 4D. After 3 days, the injected GFP-/Oct4alo BCCs became cells with bright GFP fluorescence, indicating that the Oct4alo BCCs have undergone in vivo dedifferentiation (Fig. 4D). We analyzed the mouse femurs for MSC-BCC colocalization by labeling with anti-CD146-PE for MSCs/pericytes and Numa-Alexa 405 for human cells (Fig. 7B, left, Supplementary Fig. S6C and S6D). We noted colocalized MSCs and BCCs (purple cells) around the blood vessels, CD146/red (MSCs) and Numa/blue (BCCs; Fig. 7B). CD146, which is a marker of pericytes, allowed us to determine whether pericytes/MSCs continue to interact with BCCs as the latter moved away toward the endosteum. Indeed, we detected colocalized MSCs and BCCs (purple stain) away from the blood vessel (Fig. 7B). However, as the cells moved further away, we noted yellow cells, which indicated bright GFP+ BCCs (green) and red pericytes/MSCs (Fig. 7B). This indicated that the BCCs have transitioned into intensely green (increased Oct4a-GFP expression) cells, therefore, dominating the blue fluorescence of Numa labeling.

We next performed a pilot study with BM biopsies from 2 patients with breast cancer, before and after chemotherapeutic treatment. We observed colocalization of MSCs (anti-CD29-PE) and BCCs (anti-cytokeratin-Alexa 405-blue; Fig. 7C), similar to mouse femurs (Fig. 7B). There was distinct purple staining in patient A, who completed her treatment. In contrast, there was diffused purple staining throughout the section in patient B (stage IV breast cancer), who did not begin treatment. Patient B showed clusters of red MSCs and blue cytokeratin BCCs. Together, these results showed in vivo interaction between Oct4alo BCCs and pericytes/MSCs.

Endothelial cell-MSC influence on BCC migration

We have reported previously the ease of BCC migration across endothelial cell-MSC (48). This study asked whether BCCs that enter BM could return to the periphery after exposure to naïve EVs. We used a transwell system (Fig. 7D, right diagram) to study BCC (untreated or treated with naïve EVs) migration across endothelial cell and then through MSCs.

Untreated BCCs required endothelial cell and MSCs for migration (Fig. 7D, left group), based on the significant (P < 0.05) migration, relative to BCC alone and BCC+ endothelial cell (Fig. 7D, right group). Naïve EV–treated BCCs showed significantly (P < 0.05) less migration with MSCs −/+ endothelial cells as compared with endothelial cell alone (Fig. 7D, right group). The latter observation is consistent with the differentiation of EV-exposed BCCs beginning to acquire dormancy, placing them in quiescence with less migratory ability (Figs. 16). Together, this section showed a requirement for endothelial cells for initial BCC migration, but after exposure to naïve MSC exosomes, the cells showed less migratory ability.

Molecular changes in Oct4alo BCCs in response to exosomes

Gap junctional communication between MSCs and BCCs could lead to breast cancer dormancy (10, 23). The single-cell RNA-seq data identified pathways involving tight junction and DNA repair pathways with naïve exosome–treated BCCs (Fig. 5D). We, therefore, analyzed the single-cell RNA-seq data for genes linked to stemness. Thus, we subjected naïve versus no exosome treatment genes to STRING. The output indicated links between DNA repair genes and those associated with dormancy, such as the gap junction (GJA1) and stem cell (Sox2; red asterisks; Fig. 7E, left; Supplementary Fig. S6A and S6B). Similar analysis of primed versus no exosome treatment showed genes linked to multipotency, including ID2 (Fig. 7E, right). The bioinformatics analyses indicated that naïve and primed exosomes might exert two sequential steps in BCCs: DNA repair to prepare the cells for response to primed exosomes when the BCCs dedifferentiate to CSCs for dormancy. These DNA remodeling methods could include the basic excision repair (BER) genes. Indeed, active DNA demethylation requires the oxidation of methylated cytosines by the ten-eleven translocation (TET) family of enzymes, TET1, TET2, and TET3 (49). TET enzymes catalyze the successive conversion of 5-methylcytosine into 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC), where both 5fC and 5caC can be excised by thymine-DNA glycosylase (TDG) for repair by BER factors (49). Thus, to understand the involvement of BER in BCCs exposed to naïve exosomes, we examined the exosomal dataset for its mRNA cargo and found a significant increase of mRNAs encoding for TET2 and TDG (Fig. 7F).

In addition to epigenetic remodeling, BCCs would also need to repair their DNA. Indeed, immunofluorescence for DNA damage (H2AX), DNA repair (PARP), and stemness (Oct4A) indicated loss of γH2AX with naïve exosome and concomitant presence of PARP (Fig. 7G). Because primed exosomes induced Oct4A, the BCCs exposed to these exosomes would be bright GFP. Thus, the data indicated that DNA repair occurs in the dedifferentiated BCCs. In summary, this section showed evidence of early DNA repair upon exposure to naïve exosomes, as the BCCs prepare for transitioning into cells with stem cell properties. The results also showed evidence for transcriptional remodeling, potentially due to epigenetic regulators, such as the TET enzymes.

This study reports on the early mechanisms of BCCs entering dormancy at the BM interface. MSCs, which comprised the first set of BM niche cells encountered by BCCs, change the behavior of BCCs, directly and indirectly by EV secretome (17, 22). Despite our minimal characterization for exosomes, it appears that our EVs contained exosomes and these particles might contribute to the functional data described in this report. However, there is no definitive unbiased method to isolate one subset of EVs, which led us to conclude that our data could be the sum of different EVs (31, 50, 51).

MSC-released EVs induced cycling quiescence and chemoresistance in a small subset of BCCs (17). This study identified the BCC subset that responded to MSC EVs and described the stepwise process by which the responding BCCs dedifferentiate into cells with CSC properties. The data in this study used two triple-negative BCCs for the functional studies. We validated key RNA-seq data from MDA-MB-231 with other cell lines, referred as “BCCs.” The results with T47D cells were similar to the data with MDA-MB-231. The composition of naïve and primed MSC EVs depended on exposure to BCCs. Baseline (naïve) MVs are released, regardless of communication with the BCCs. The cargos are distinct after communication with MSCs (primed), resulting in different behavior of BCCs.

Although this study focused on the early event occurring in BCCs entering the BM, other ongoing studies are examining late-stage dormancy close to the endosteum. In addition to the early events shown for MSC with respect to dormancy, MSCs could be active in sustained dormancy at the later stage due to their location throughout the marrow. We showed evidence of cellular chaperone between MSCs and BCCs toward the endosteum (Fig. 7B). Such interaction could offer BCCs with immune protection (10, 23). As Oct4alo BCCs moved toward to endosteum, GFP intensity was increased (Fig. 7B), suggesting continued dedifferentiation along BM cavity. Going forward, we will determine whether this process involves MV and/or soluble secretome. Gap junction between MSCs and CSCs (10) might lend to sharing of molecules through the connexin channels. MSCs can maintain dormancy close to the endosteum by supporting gap junction between M2 macrophage and CSCs (20). Overall, the findings, when combined with other reports, indicate a key role for MSCs during early- and late-stage dormancy.

We were surprised that the least mature BCCs (Oct4alo) expressed transcripts similar to CSCs, despite lacking stem cell–related function. This might partly explain why the Oct4alo BCCs were most sensitive to MSC-derived exosomes with respect to dedifferentiation into CSCs. These exosome-mediated changes were stepwise, beginning with the naïve exosomes facilitating DNA repair and cycling quiescence, followed by dedifferentiation when exposed to primed exosomes (Fig. 7G). Single-cell RNA-seq indicated distinct BCC subsets after exposure to naïve or primed EVs. Notably, we showed how dedifferentiated BCCs interacted with MSCs in vivo, suggesting cellular chaperoning of BCCs to offer a survival advantage.

We can dissect the model in mice because of functional similarity between murine and human MSCs. Thus, the in vivo findings recapitulate the in vitro studies with human MSCs. In addition, we observed in vivo dedifferentiation of Oct4alo BCCs in murine femurs similar to human BM biopsy from patients with breast cancer (Figs. 4 and 7). The dedifferentiated CSCs would be able to establish gap junction with other BM niche cells to ensure long-term survival (10, 20, 21). Exposure to BCCs did not affect the function of MSCs (Supplementary Fig. S7). The in vivo studies suggested that MSCs protect BCCs in femurs, as well as ensure full dedifferentiation into long-term dormant BCCs.

We have reported previously on MSC-derived EVs inducing cycling quiescence in triple-positive and -negative BCCs (17). This study focused on triple-negative BCCs because this type of breast cancer has no targeted therapy and results in poor prognosis. Because of the lack of targeted treatment, this type of BCCs have a high probability to survive in BM, resulting in rapid resurgence. The data in this study are relevant to breast tumors with different hormone receptor because they have similar integration process to acquire dormancy within the BM stromal compartment (20, 21).

In our previous studies, we stratified BCCs, on the basis of Oct4a expression, in a hierarchical distribution (21). Interestingly, Oct4alo BCCs, which are developmentally distant from the CSCs, express low level of Oct4a. These cells quickly responded to MSC exosomes to become cycling quiescent cells (Fig. 1). Such transition occurs within 48 hours, but prolonged exposure (7 days) to EVs, including primed, led to the expression of stem cell genes and the acquisition of other properties linked to stemness, including drug resistance and tumorsphere formation (Fig. 2). To recapitulate the in vivo framework, we exposed BCCs to exosomes using a fractionated and prolonged method. Specifically, upon entry of BCCs in BM, they are immediately exposed to naïve EVs with later exposure to primed EVs.

We gained insights into the ease by which Oct4alo BCCs responded to the EVs by identifying gene expression signatures similar to those in CSCs, including genes within the Wnt-β-catenin pathway (Fig. 4B). We were initially dismissive of the Wnt pathway due to its activation in triple-negative BCCs (47). However, our studies elucidated a new role for the Wnt-β-catenin pathway during dedifferentiation of BCCs into CSCs (Figs. 2 and 5K). We confirmed the relevance of Wnt-β-catenin by single-cell RNA-seq of BCCs exposed to EVs (Fig. 5H–J). We gained further information when we identified the Wnt–β-catenin signaling genes in the EV cargo (Fig. 6F). The data from the exosome RNA-seq, when overlapped with the Wnt pathway in IPA, indicated a gene network with a composite function linked to stemness (Fig. 6G). More importantly, VerElect algorithm scored the catenin pathway high for known breast cancer genes (Fig. 6A).

Single-cell RNA-seq indicated a distinct pattern of BCC subsets, depending on the type of EV exposure (naïve and primed vs. no exposure; Fig. 5). Primed EVs resulted in a relatively more homogenous BCC population, whereas naïve EVs and untreated controls showed distinct subsets (Fig. 5G). The release of primed EVs is subsequent to naïve EV release because their release requires BCC–MSC communication. It is anticipated that the EV cargo of primed MSCs would be relatively homogenous as compared with those released from baseline and naïve MSCs. Such homogeneity would be due to the BCCs preparing MSCs for specific cargo to assist them to survive. Indeed, we noted a more homogenous size and increased RNA in the primed EVs, as compared with naïve EVs, suggesting similar cargo (Fig. 1).

A key question that arose from this study is why Oct4alo BCCs express genes similar to CSCs (Fig. 4A and B), despite showing no function of stem cells. This led us to examine the database of all NSG sequence information for genes common to breast cancer. Consistent with our RNA-seq data for single cells and EVs, the top gene was Wnt-Catenin, as well as other genes linked to stemness, such as EpCAM and ALDH1 (Fig. 6A). More importantly, when these genes were studied for their interaction, it was clear that Wnt-β-catenin is connected the dormancy-related network (Fig. 6B). We have reported previously on roles for miRNAs within MSC EVs in BCC quiescence and chemoresistance (17). Thus, going forward, we will need to determine whether miRNAs and other noncoding RNAs, such as circular miRNAs, could be involved in controlling the fate of Oct4alo BCCs to survive as CSCs. Moreover, it will be important to determine how interrogation of EV cargo would counteract the complex RNA network that leads to dedifferentiation of BCCs into CSCs.

It is likely that there is a link between cell fate changes in BCCs and reorganization of the epigenome. Indeed, we noted changes in the expression of TET enzymes and TDG in BCCs, depending on the type of EV exposure (Fig. 7F). TDG excision of TET-mediated DNA oxidations (5fC and 5caC) allows BER to impart DNA demethylation. Thus, upregulated TET2 and TDG were significant to the dedifferentiation process, which is consistent with TET2-mediated DNA demethylation in somatic cellular reprograming, stem cell maintenance, and cancer cell quiescence (52, 53). Thus, because TET2 mRNA is enriched in naïve EVs, which are involved the initial steps in BCC dedifferentiation, we propose a link between TET2-mediated DNA demethylation and dedifferentiation for the acquisition of stem cell signatures. These exciting findings will form the basis for future studies on in-depth genome-wide changes as BCCs respond to the complex BM microenvironment to undergo dormancy.

It will be important to map the global epigenetic changes involved in the stepwise process of dedifferentiation (Fig. 7H). Nevertheless, the data analyses led us to propose the following: an epigenetic mechanism that reverts DNA methylation to reactivate the expression of genes to promote cellular stemness in BCCs when exposed to naïve EVs. More specifically, our single-cell RNA-seq analysis indicates that amplification of TET2 and TDG could facilitate BER-mediated DNA demethylation in BCCs exposed to naïve EVs, as the initial steps toward the acquisition of stemness. Furthermore, we propose a link between TET2-TDG-BER and Wnt signaling, through β-catenin, as a mechanism involved in the response of BCCs toward naïve EVs. Our ongoing studies will identify the targeted genes whose expressions are reactivated by TET2-TDG-BER, in combination with Wnt signaling, during conversion of EV-mediated BCCs into CSCs. The findings of pathways leading to BER genes indicated that the double-stranded break repair shown in Fig. 7G requires more robust studies, including the incorporation of epigenetic studies.

Although preliminary, it appears that MSCs also interact with BCCs in patients' BM (Fig. 7C). There are several advantages of MSCs being colocalized with dedifferentiated BCCs, including immune protection (10). Interestingly, primed EVs decreased immune response pathways, immunoglobulin and complement, in BCCs (Fig. 5F). Together, these findings indicated that in addition to MSCs mediating dedifferentiation of BCCs, their role might be more complex, including immune protection (10). Also of interest is the identification of cadherin and presenilin (Fig. 6C), which are relevant to maintain gap junction–mediated dormancy of CSCs close to the endosteal region. This suggested that MSC-secreted EVs could be preparing BCCs for gap junctional intercellular communication with other BM cells, such as macrophages and stroma As the BCCs arrive close to the endosteum, it becomes a challenge to treat due to the presence of the endogenous HSCs (20, 21). Overall, we unraveled mechanistic pathways involving EVs released by MSCs, which are capable of instructing BCCs to become dormant CSCs with chemoresistance capacity, which holds potential for the development of innovative intervention.

No disclosures were reported.

O.A. Sandiford: Conceptualization, methodology, writing–review and editing. R.J. Donnelly: Data curation, software, writing–original draft, writing–review and editing. M.H. El-Far: Data curation, software, writing–review and editing. L.M. Burgmeyer: Methodology, writing–review and editing. G. Sinha: Methodology, writing–review and editing. S.H. Pamarthi: Methodology, writing–review and editing. L.S. Sherman: Data curation, software, methodology, writing–review and editing. A.I. Ferrer: Methodology, writing–review and editing. D.E. DeVore: Methodology, writing–review and editing. S.A. Patel: Conceptualization, writing–review and editing. Y. Naaldijk: Methodology, writing–review and editing. S. Alonso: Methodology, writing–review and editing. P. Barak: Conceptualization, data curation, formal analysis, methodology, writing–review and editing. M. Bryan: Resources, writing–review and editing. N.M. Ponzio: Conceptualization, writing–review and editing. R. Narayanan: Data curation, software, formal analysis, writing–review and editing. J.-P. Etchegaray: Validation, methodology, writing–review and editing. R. Kumar: Conceptualization, writing–review and editing. P. Rameshwar: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, methodology, writing–review and editing.

This work was supported, in part, by grants awarded by The New Jersey Commission of Health (DCSH20PPC041 to A.I. Ferrer) and METAvivor Foundation (to P. Rameshwar). We thank Nathanael Joseph for assisting us with the single-cell RNA-seq, Dr. Nataki Douglas, Dr. Chandra Namas, Dr. RamaRao Venkata Kakulavarapu, and Mr. Luke Fritzky for assistance with the tissue sectioning, and Dr. Jie-Gen Jiang for providing the human bone marrow sectioning.

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.
Henley
SJ
,
Thomas
CC
,
Lewis
DR
,
Ward
EM
,
Islami
F
,
Wu
M
, et al
Annual report to the nation on the status of cancer, part II: progress toward healthy people 2020 objectives for 4 common cancers
.
Cancer
2020
;
126
:
2250
66
.
2.
Dittmer
J
. 
Mechanisms governing metastatic dormancy in breast cancer
.
Semin Cancer Biol
2017
;
44
:
72
82
.
3.
Braun
S
,
Kentenich
C
,
Janni
W
,
Hepp
F
,
de Waal
J
,
Willgeroth
F
, et al
Lack of effect of adjuvant chemotherapy on the elimination of single dormant tumor cells in bone marrow of high-risk breast cancer patients
.
J Clin Oncol
2000
;
18
:
80
6
.
4.
Hosseini
H
,
Obradović
MMS
,
Hoffmann
M
,
Harper
KL
,
Sosa
MS
,
Werner-Klein
M
, et al
Early dissemination seeds metastasis in breast cancer
.
Nature
2016
;
540
:
552
58
.
5.
Husemann
Y
,
Geigl
JB
,
Schubert
F
,
Musiani
P
,
Meyer
M
,
Burghart
E
, et al
Systemic spread is an early step in breast cancer
.
Cancer Cell
2008
;
13
:
58
68
.
6.
Harper
KL
,
Sosa
MS
,
Entenberg
D
,
Hosseini
H
,
Cheung
JF
,
Nobre
R
, et al
Mechanism of early dissemination and metastasis in Her2(+) mammary cancer
.
Nature
2016
;
540
:
588
92
.
7.
Ono
M
,
Kosaka
N
,
Tominaga
N
,
Yoshioka
Y
,
Takeshita
F
,
Takahashi
RU
, et al
Exosomes from bone marrow mesenchymal stem cells contain a microRNA that promotes dormancy in metastatic breast cancer cells
.
Sci Signal
2014
;
7
:
ra63
.
8.
Weigelt
B
,
Peterse
JL
,
van 't Veer
LJ
. 
Breast cancer metastasis: markers and models
.
Nat Rev Cancer
2005
;
5
:
591
602
.
9.
Braun
S
,
Vogl
FD
,
Naume
B
,
Janni
W
,
Osborne
MP
,
Coombes
RC
, et al
A pooled analysis of bone marrow micrometastasis in breast cancer
.
N Engl J Med
2005
;
353
:
793
802
.
10.
Patel
SA
,
Dave
MA
,
Bliss
SA
,
Giec-Ujda
AB
,
Bryan
M
,
Pliner
LF
, et al
Treg/Th17 polarization by distinct subsets of breast cancer cells is dictated by the interaction with mesenchymal stem cells
.
J Cancer Stem Cell Res
2014
;
2014
:
e1003
.
11.
Tjensvoll
K
,
Nordgard
O
,
Skjaeveland
M
,
Oltedal
S
,
Janssen
EAM
,
Gilje
B
. 
Detection of disseminated tumor cells in bone marrow predict late recurrences in operable breast cancer patients
.
BMC Cancer
2019
;
19
:
1131
.
12.
De Angelis
ML
,
Francescangeli
F
,
Zeuner
A
. 
Breast cancer stem cells as drivers of tumor chemoresistance, dormancy and relapse: new challenges and therapeutic opportunities
.
Cancers
2019
;
11
:
1569
.
13.
Carcereri de Prati
A
,
Butturini
E
,
Rigo
A
,
Oppici
E
,
Rossin
M
,
Boriero
D
, et al
Metastatic breast cancer cells enter into dormant state and express cancer stem cells phenotype under chronic hypoxia
.
J Cell Biochem
2017
;
118
:
3237
48
.
14.
Chen
C
,
Okita
Y
,
Watanabe
Y
,
Abe
F
,
Fikry
MA
,
Ichikawa
Y
, et al
Glycoprotein nmb is exposed on the surface of dormant breast cancer cells and induces stem cell-like properties
.
Cancer Res
2018
;
78
:
6424
35
.
15.
Tivari
S
,
Lu
H
,
Dasgupta
T
,
De Lorenzo
MS
,
Wieder
R
. 
Reawakening of dormant estrogen-dependent human breast cancer cells by bone marrow stroma secretory senescence
.
Cell Commun Signal
2018
;
16
:
48
.
16.
Massague
J
,
Obenauf
AC
. 
Metastatic colonization by circulating tumour cells
.
Nature
2016
;
529
:
298
306
.
17.
Bliss
SA
,
Sinha
G
,
Sandiford
OA
,
Williams
LM
,
Engelberth
DJ
,
Guiro
K
, et al
Mesenchymal stem cell-derived exosomes stimulate cycling quiescence and early breast cancer dormancy in bone marrow
.
Cancer Res
2016
;
76
:
5832
44
.
18.
Phadke
PA
,
Mercer
RR
,
Harms
JF
,
Jia
Y
,
Frost
AR
,
Jewell
JL
, et al
Kinetics of metastatic breast cancer cell trafficking in bone
.
Clin Cancer Res
2006
;
12
:
1431
40
.
19.
Dzierzak
E
,
Bigas
A
. 
Blood development: hematopoietic stem cell dependence and independence
.
Cell Stem Cell
2018
;
22
:
639
51
.
20.
Walker
ND
,
Elias
M
,
Guiro
K
,
Bhatia
R
,
Greco
SJ
,
Bryan
M
, et al
Exosomes from differentially activated macrophages influence dormancy or resurgence of breast cancer cells within bone marrow stroma
.
Cell Death Dis
2019
;
10
:
59
.
21.
Patel
SA
,
Ramkissoon
SH
,
Bryan
M
,
Pliner
LF
,
Dontu
G
,
Patel
PS
, et al
Delineation of breast cancer cell hierarchy identifies the subset responsible for dormancy
.
Sci Rep
2012
;
2
:
906
.
22.
Greco
SJ
,
Patel
SA
,
Bryan
M
,
Pliner
LF
,
Banerjee
D
,
Rameshwar
P
. 
AMD3100-mediated production of interleukin-1 from mesenchymal stem cells is key to chemosensitivity of breast cancer cells
.
Am J Cancer Res
2011
;
1
:
701
15
.
23.
Patel
SA
,
Meyer
JR
,
Greco
SJ
,
Corcoran
KE
,
Bryan
M
,
Rameshwar
P
. 
Mesenchymal stem cells protect breast cancer cells through regulatory T cells: role of mesenchymal stem cell-derived TGF-beta
.
J Immunol
2010
;
184
:
5885
94
.
24.
Talmadge
JE
. 
Clonal selection of metastasis within the life history of a tumor
.
Cancer Res
2007
;
67
:
11471
5
.
25.
Shimizu
H
,
Takeishi
S
,
Nakatsumi
H
,
Nakayama
KI
. 
Prevention of cancer dormancy by Fbxw7 ablation eradicates disseminated tumor cells
.
JCI Insight
2019
;
4
:
e125138
.
26.
Bianco
P
,
Riminucci
M
,
Gronthos
S
,
Robey
PG
. 
Bone marrow stromal stem cells: nature, biology, and potential applications
.
Stem Cells
2001
;
19
:
180
92
.
27.
Crisan
M
,
Yap
S
,
Casteilla
L
,
Chen
CW
,
Corselli
M
,
Park
TS
, et al
A perivascular origin for mesenchymal stem cells in multiple human organs
.
Cell Stem Cell
2008
;
3
:
301
13
.
28.
Ghosh
D
,
Mejia Pena
C
,
Quach
N
,
Xuan
B
,
Lee
AH
,
Dawson
MR
. 
Senescent mesenchymal stem cells remodel extracellular matrix driving breast cancer cells to a more-invasive phenotype
.
J Cell Sci
2020
;
133
:
jcs232470
.
29.
Mishra
PJ
,
Mishra
PJ
,
Humeniuk
R
,
Medina
DJ
,
Alexe
G
,
Mesirov
JP
, et al
Carcinoma-associated fibroblast-like differentiation of human mesenchymal stem cells
.
Cancer Res
2008
;
68
:
4331
9
.
30.
Dassler-Plenker
J
,
Kuttner
V
,
Egeblad
M
. 
Communication in tiny packages: exosomes as means of tumor-stroma communication
.
Biochim Biophys Acta Rev Cancer
2020
;
1873
:
188340
.
31.
Kalluri
R
,
LeBleu
VS
. 
The biology, function, and biomedical applications of exosomes
.
Science
2020
;
367
:
eaau6977
.
32.
Kalluri
R
. 
The biology and function of exosomes in cancer
.
J Clin Invest
2016
;
126
:
1208
15
.
33.
Munoz
JL
,
Bliss
SA
,
Greco
SJ
,
Ramkissoon
SH
,
Ligon
KL
,
Rameshwar
P
. 
Delivery of functional anti-miR-9 by mesenchymal stem cell-derived exosomes to glioblastoma multiforme cells conferred chemosensitivity
.
Mol Ther Nucleic Acids
2013
;
2
:
e126
.
34.
Bliss
SA
,
Paul
S
,
Pobiarzyn
PW
,
Ayer
S
,
Sinha
G
,
Pant
S
, et al
Evaluation of a developmental hierarchy for breast cancer cells to assess risk-based patient selection for targeted treatment
.
Sci Rep
2018
;
8
:
367
.
35.
Braun
S
,
Pantel
K
,
Muller
P
,
Janni
W
,
Hepp
F
,
Kentenich
CR
, et al
Cytokeratin-positive cells in the bone marrow and survival of patients with stage I, II, or III breast cancer
.
N Engl J Med
2000
;
342
:
525
33
.
36.
Otvos
B
,
Silver
DJ
,
Mulkearns-Hubert
EE
,
Alvarado
AG
,
Turaga
SM
,
Sorensen
MD
, et al
Cancer stem cell-secreted macrophage migration inhibitory factor stimulates myeloid derived suppressor cell function and facilitates glioblastoma immune evasion
.
Stem Cells
2016
;
34
:
2026
39
.
37.
Zhang
H
,
Freitas
D
,
Kim
HS
,
Fabijanic
K
,
Li
Z
,
Chen
H
, et al
Identification of distinct nanoparticles and subsets of extracellular vesicles by asymmetric flow field-flow fractionation
.
Nature Cell Biol
2018
;
20
:
332
43
.
38.
Mao
CP
,
Wu
T
,
Song
KH
,
Kim
TW
. 
Immune-mediated tumor evolution: Nanog links the emergence of a stem like cancer cell state and immune evasion
.
Oncoimmunology
2014
;
3
:
e947871
.
39.
Gomes
I
,
de Almeida
BP
,
Dâmaso
S
,
Mansinho
A
,
Correia
I
,
Henriques
S
, et al
Expression of receptor activator of NFkB (RANK) drives stemness and resistance to therapy in ER+HER2- breast cancer
.
Oncotarget
2020
;
11
:
1714
28
.
40.
Ghasemi
F
,
Sarabi
PZ
,
Athari
SS
,
Esmaeilzadeh
A
. 
Therapeutics strategies against cancer stem cell in breast cancer
.
Int J Biochem Cell Biol
2019
;
109
:
76
81
.
41.
Kim
H
,
Kim
Y
,
Jeoung
D
. 
DDX53 promotes cancer stem cell-like properties and autophagy
.
Mol Cells
2017
;
40
:
54
65
.
42.
Ye
X
,
Zhou
X-J
,
Zhang
H
. 
Exploring the role of autophagy-related gene 5 (ATG5) yields important insights into autophagy in autoimmune/autoinflammatory diseases
.
Front Immunol
2018
;
9
:
2334
34
.
43.
Ito
K
,
Ito
K
. 
Metabolism and the control of cell fate decisions and stem cell renewal
.
Annu Rev Cell Dev Biol
2016
;
32
:
399
409
.
44.
Panieri
E
,
Santoro
MM
. 
ROS homeostasis and metabolism: a dangerous liason in cancer cells
.
Cell Death Dis
2016
;
7
:
e2253
.
45.
Miriyala
S
,
Holley
AK
,
St Clair
DK
. 
Mitochondrial superoxide dismutase–signals of distinction
.
Anticancer Agents Med Chem
2011
;
11
:
181
90
.
46.
Flanagan
DJ
,
Vincan
E
,
Phesse
TJ
. 
Wnt signaling in cancer: not a binary ON:OFF switch
.
Cancer Res
2019
;
79
:
5901
06
.
47.
Pohl
SG
,
Brook
N
,
Agostino
M
,
Arfuso
F
,
Kumar
AP
,
Dharmarajan
A
. 
Wnt signaling in triple-negative breast cancer
.
Oncogenesis
2017
;
6
:
e310
.
48.
Corcoran
KE
,
Trzaska
KA
,
Fernandes
H
,
Bryan
M
,
Taborga
M
,
Srinivas
V
, et al
Mesenchymal stem cells in early entry of breast cancer into bone marrow
.
PLoS One
2008
;
3
:
e2563
.
49.
Rasmussen
KD
,
Helin
K
. 
Role of TET enzymes in DNA methylation, development, and cancer
.
Genes Dev
2016
;
30
:
733
50
.
50.
Mathieu
M
,
Martin-Jaular
L
,
Lavieu
G
,
Théry
C
. 
Specificities of secretion and uptake of exosomes and other extracellular vesicles for cell-to-cell communication
.
Nat Cell Biol
2019
;
21
:
9
17
.
51.
Théry
C
,
Witwer
KW
,
Aikawa
E
,
Alcaraz
MJ
,
Anderson
JD
,
Andriantsitohaina
R
, et al
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): a position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines
.
J Extracell Vesicles
2018
;
7
:
1535750
.
52.
Li
C
,
Lan
Y
,
Schwartz-Orbach
L
,
Korol
E
,
Tahiliani
M
,
Evans
T
, et al
Overlapping requirements for Tet2 and Tet3 in normal development and hematopoietic stem cell emergence
.
Cell Rep
2015
;
12
:
1133
43
.
53.
Cimmino
L
,
Dolgalev
I
,
Wang
Y
,
Yoshimi
A
,
Martin
GH
,
Wang
J
, et al
Restoration of TET2 function blocks aberrant self-renewal and leukemia progression
.
Cell
2017
;
170
:
1079
95
.