EXOSOMES PRODUCED BY MESENCHYMAL STEM CELLS DRIVE DIFFERENTIATION OF MYELOID CELLS INTO IMMUNOSUPPRESSIVE M2-POLARIZED MACROPHAGES IN BREAST CANCER

Abstract

Tumor-associated macrophages are major contributors to malignant progression and resistance to immunotherapy, but the mechanisms governing their differentiation from immature myeloid precursors remain incompletely understood. Here we demonstrate that exosomes secreted by human and mouse tumor-educated mesenchymal stem cells (MSCs) drive accelerated breast cancer progression by inducing differentiation of monocytic myeloid-derived suppressor cells (M-MDSCs) into highly immunosuppressive M2-polarized macrophages at tumor beds. Mechanistically, MSC-derived exosomes, but not exosomes from tumor cells, contain TGF-β, C1q and semaphorins, which promote myeloid tolerogenic activity by driving PD-L1 overexpression in both immature myelo-monocytic precursors and committed CD206⁺ macrophages, and by inducing differentiation of MHC-II⁺ macrophages with enhanced L-Arginase activity and IL-10 secretion at tumor beds. Accordingly, administration of tumor-associated murine MSC-derived exosomes accelerates tumor growth by dampening anti-tumor immunity, and macrophage depletion eliminates exosome-dependent differences in malignant progression. Our results unveil a new role for MSC-derived exosomes in the differentiation of MDSCs into macrophages, which governs malignant growth.

INTRODUCTION

Macrophages with pro-tumoral activity represent the most abundant leukocyte population infiltrating solid tumors (1). It is now generally accepted that tumor-associated macrophages originate primarily from bone marrow-derived blood monocytes and closely related M-MDSC recruited to tumors (2), and only to a lesser extent from tissue-resident proliferating precursors. Thus, pathological expansion and abnormal function of myeloid cells is a hallmark of virtually all solid tumors, and immature myeloid cells with immunosuppressive activity (MDSCs) expand from the bone marrow in response to tumor-derived inflammatory signals, accumulate in the periphery and, in the case of cells of the myelo-monocytic lineage (M-MDSCs) eventually home to tumor beds or pre-metastatic niches (3-5).

The function of immature myeloid cells depends on their environment. Correspondingly, the tumor microenvironment governs the plasticity of intratumoral myeloid cells (1, 6). Hence, after migration to tumor beds, M-MDSC rapidly differentiate to tumor-associated macrophages (4), and converging evidence supports that circulating monocytes/M-MDSCs are essential for macrophage accumulation and activity in solid tumors (7-9). Although the M1/M2 polarization model is too simplistic to explain the heterogeneous spectrum of differentiation and activation that characterizes tumor-associated macrophages(4), it is generally accepted that M2-polarization is associated with tumor-promoting, immunosuppressive activities. Investigation into the mechanisms driving rapid macrophage differentiation from M-MDSCs, and acquisition of M2-like features in the tumor microenvironment has primarily focused on two master regulators: CSF1 (1) and hypoxia (7). However, the mechanistic details of these processes, and the contribution of other factors, remains incompletely understood (2). An attractive candidate for the regulation of phenotypic changes in myeloid cells in cancer is MSCs. Recent studies indicate that, in response to tumor-derived cues, MSCs establish a cross-talk with myeloid cells through the secretion of myeloid chemoattractants and other factors that lead to their functional reprogramming (10). However, tumor-associated MSCs have been associated with controversial effects, which appear to be mediated by their polarization to a proinflammatory or an anti-inflammatory phenotype (11).

In this study, we investigated the association between tumor-educated MSCs and pro-malignant M2-polarized macrophages in breast cancer. Our results indicate that exosomes specifically secreted by tumor MSCs are a major contributing factor to the conversion of M-MDSCs into macrophages, and the acquisition of M2-like, tumor-promoting attributes.

MATERIALS AND METHODS

Animals and tumors

Wild-type, female, 4-6 weeks old C57BL/6 mice were procured from Charles River Laboratories, and maintained by the animal facility of Moffitt. Animal experiments were approved by the Institutional Animal Care and Use Committee at the University of South Florida.

The Brpkp110 primary mammary tumor cell line was generated by culturing a mechanically dissociated C57BL/6 L-Stop-L-Kras^G12Dp53^flx/flxL-Stop-L-Myristoylated p110α−luciferase^flx/+ primary breast tumor mass as previously described (12, 13). Tumors were initiated by injecting 5×10⁵ cells into the axillary flanks. Tumor volume was calculated as: 0.5 × (L × W²), where L is length and W is width. Tumor tissues were dissected mechanically into single cell suspension for flow cytometry, or retained for RNA and protein isolation.

Human samples

Surgically removed fresh primary breast tumor tissues from patients with infiltrating duct carcinoma (IDC) of breast and autologous healthy breast tissues were collected from Saroj Gupta Cancer Centre and Research Institute, Kolkata, India with prior approval from Institutional Ethics Committee (IEC). RNA and proteins were extracted (n= 51) from one part, whereas tumors were dissected and digested (n=15) in RPMI-1640 containing collagenase/hyaluronidase cocktail (Stemcell), filtered through a 70 μ nylon cell strainer (HiMedia) to make single cell suspensions. Pathological information about Stage, RB grade, hormone receptor and Her2 expression status were obtained. A breast tissue microarray (TMA) slide (n=19) was obtained from Christiana Care Health System, Philadelphia, USA with prior IEC approval.

Cell lines and media

Human monocyte cell line THP-1, breast cancer cell lines MDA-MB-231, T47D and normal breast cell line MCF10A were procured from NCCS, India. All cell lines were routinely cultured in R10 media (RPMI-1640, 10% FBS, penicillin (100 I.U./mL), streptomycin (100 μg/mL), L-glutamine (2 mM), sodium pyruvate (0.5 mM)) (Thermo). For MCF10A, medium was additionally supplemented with insulin (10 μg/mL), cholera toxin (100 ng/mL) and EGF (20 ng/mL) (Peprotech).

Murine fibroblast cell line NIH/3T3, procured from ATCC, USA, and breast cancer cell line Brpkp110 were routinely maintained in R10 media, or cultured in 10% exosome-depleted-FBS-containing RPMI.

For conditioned media collection, breast cancer cell lines MDA-MB-231 and T47D were grown to 60-70% confluency before replacing the medium with fresh R10. Both cell lines were then cultured for another 48 hr and media were collected. Further, conditioned media were filtered using 0.45 μ filter (Sartorius) to remove cell debris.

In vitro polarization and differentiation of human M2 macrophages

Human monocyte cell line THP-1 were made to transform into undifferentiated and non-polarized M0 macrophages by 24 hr incubation with phorbol 12-myristate 13-acetate (PMA, LC Laboratories, 150 nM) followed by 24 hr incubation in R10 (14). For a positive control of M2 polarization, PMA-induced M0 THP-1 cells were incubated with IL-4 (20 ng/ml) and IL-13 (20 ng/ml) (Peprotech) for 48 hr. To study the effect of conditioned medium of breast cancer cell lines on differentiation of M0 macrophages, conditioned-R10 medium were collected from 48 hr grown cultures of MDA-MB-231, T47D, and MCF10A, while the M0 THP-1 cells were cultured in these conditioned media according to the combinations.

Sorting and co-culture

Human MSCs (hMSCs) were MACS-sorted from breast tumors (n=3, IDC) using human CD45 and CD271 micro-bead kits (Miltenyi) with manufacturer guidelines. Sorted CD45⁻CD271⁺ MSCs were pooled and cultured in vitro with hMSC proliferation medium (Stemcell). Medium was transitioned to RPMI containing 10% exosome-depleted FBS before utilizing these MSCs in experiments. Purity (CD45⁻CD271⁺ phenotype) was further confirmed by flow cytometry, while experiments using these hMSCs were done in < 5 subsequent passages. MSCs (5 × 10⁴) were placed in the upper chamber of 0.4 μ co-culture inserts placed into a 24 well transwell plate (Thermo), as per the required combinations for indirect co-culture with M0 THP-1 cells. In the lower chamber, PMA-induced M0 THP-1 cells (2 × 10⁵) were placed in breast cancer cell line-conditioned medium or in normal R10. For positive control set of M2 polarization, M0 THP-1 cells were incubated in IL-4 (20 ng/ml) and IL-13 (20 ng/ml)-supplemented R10. Co-cultures were done for 24 or 48 hr.

Mouse MSCs (mMSCs) were FACS–sorted using the following panel: CD45⁻CD11b⁻CD44⁺CD106⁺Sca1⁺ from Brpkp110-tumors (n=3), pooled and cultured in vitro with an mMSC expansion and proliferation medium (Stemcell). Medium was transitioned to RPMI containing 10% exosome-depleted FBS before utilizing these MSCs in experiments. Experiments using these mMSCs were done in < 5 subsequent passages. From dissociated mouse tumors, epithelial tumor cells were sorted using the following panel: CD45⁻EpCAM⁺; Class II MHC negative monocytic cells were sorted using the following panel: CD45⁺CD11b⁺F4/80⁺IA/IE⁻, and class II MHC positive macrophages were sorted using the following panel: CD45⁺CD11b⁺F4/80⁺IA/IE⁺.

Exosome isolation and treatment

For exosome isolation, 5 × 10⁶ cells (pooled human MSCs (n=3) or mouse MSCs (n=3) or breast cancer cells or 3T3 cells) were seeded in T-175 tissue culture flasks and were cultured for 12 hr in RPMI with 10% exosome-depleted serum (Gibco). The cells were washed twice with phosphate buffered saline (PBS) (Himedia) to remove exosome contaminants, and grown in RMPI with 10% exosome-depleted serum (Gibco). Exosomes were isolated using total exosome isolation kit (Invitrogen) according to manufacturer recommendations from conditioned medium of 48 hr grown culture, which provides equivalent purity of exosomes as of the ultracentrifugal method of exosome isolation (15). Exosomes from an entire T-175 flask (~50 μg) were dissolved in 500 μL of PBS (~100 ng/μl); therefore the seeded cell number to reconstituted volume ratio is 10,000 cells: 1 μL. M0 THP-1 cells were treated with exosomes, derived from either breast cancer cell lines or MSCs at a ratio of 1 μL: 50,000 cells. 100 μL of mMSC-derived exosomes or PBS were injected intratumorally or peritumorally after 5 days of Brpkp110 breast tumor-challenge.

Blocking of Exosome biogenesis/secretion in vitro

To prevent biogenesis and secretion of exosomes from MDA-MB-231, T47D and hMSCs, we used a standard chemical inhibitor, GW4869 (Cayman) using a standard protocol (16). GW4869 was dissolved in DMSO (Thermo) and diluted in R10, so that the final DMSO concentration in the medium was 0.005%. Breast cancer cell lines were treated with 20 μM GW4869 in R10, and conditioned media were collected after 48 hr. hMSCs were treated with 20 μM GW4869 12 hr before and during co-culture with M0 THP-1 in the respective combinations.

Quantitative Real-time PCR (Q-PCR)

Total RNA from dissociated tumor chunks and healthy breast tissues were isolated by standardized protocol using TRIzol reagent (Thermo) or RNeasy Plus kit (Qiagen). RNA was reverse transcribed to cDNAs using MMLV-RT (Invitrogen) or SuperScript-IV (Invitrogen), and random hexamers (Invitrogen) or oligo-dT (Invitrogen). Quantification of human CD206, CD271 mRNA and mouse Tgfb mRNA was performed using SYBR green reagent (Applied Biosystems). Expression was normalized by 18S rRNA or GAPDH levels. Fold changes relative to average C_T values in healthy tissues or vehicle group tumors were calculated by the formula 2^−ΔΔC_T.

Western blot (WB)

Cells and mechanically dissociated tumor samples were lysed in RIPA buffer (Thermo) with protease-phosphatase inhibitor cocktail (Sigma) and cleared by centrifugation. Proteins were quantified by Bradford reagent assay or BCA assay. WB was performed using our laboratory optimized protocol (17). Membranes were blotted with anti-CD206 (CST#91992), anti-CD163 (CST#93498), anti-PD-L1 (CST#13684), anti-Snail (CST#3895), anti-Slug (CST#9585), anti-E-cadherin (CST#14472) and anti-β-actin (CST#4970) antibodies. Immunoreactive bands were developed using horse radish peroxidase (HRP)-conjugated secondary antibodies (CST#7074, CST#7076) and Enhanced Chemiluminescence (ECL) substrate (Thermo).

Immunofluorescence

Immunofluorescence for CD206 and CD271 in the breast TMA was performed using our optimized protocol (17). Precisely, the slide was deparaffinized, followed by antigen retrieval, blocking, and incubation overnight at 4°C with anti-CD206 (CST#91992) and anti-CD271 (Invitrogen#MA5-13314) antibodies. The following day, slides were incubated with Alexa Fluor 568 (Invitrogen#A11004) and Alexa Fluor 647 (CST#4414)-conjugated secondary antibodies and mounted using a DAPI-containing mounting reagent (CST). Images were captured in a confocal microscope (LeicaSP8) and quantitative acquisition was performed by an automated slide Scanner (Aperio-Leica). Analyses were done using Definiens Tissue Studio version 4.7 software.

Flow cytometry

Flow cytometry was performed by staining with Zombie Yellow viability dye, blocking with anti-CD16/32, and staining for 30 min at 4°C with the following anti-human antibodies: CD45 (BD:HI30), CD3 (BD:UCHT1), CD11b (Biolegend:M1-70), CD206 (Biolegend:15-2), CD163 (Biolegend:GHI/61), CD271 (Biolegend:ME20.4), PD-L1 (Biolegend:29E.2A3); or anti-mouse antibodies: CD45 (Biolegend:30-F-11), CD11b (Biolegend:M1-70), F4/80 (Biolegend:F4/80), Ly6G (Biolegend:1A8), Ly6C (Biolegend:HK1.4), IA/IE (BD:M5/114), CD206 (Biolegend:C068C2), PD-L1 (Biolegend:10F.9G2), PD-1 (BD:J43), IFN-γ (Biolegend:XMG1.2), CD3 (Tonbo:145-2C11), CD8 (Biolegend:YTS156.7.7), CD4 (BD:GK1.5), CD106 (Biolegend:429-MVCAM.A), EpCAM (Biolegend:G8.8), CD44 (Biolegend:IM7), Sca1 (Biolegend:D7). For IFN-γ, cells were pre-stimulated with PMA (20 ng/mL), Ionomycin (Sigma, 1 μg/mL) and Golgi stop (BD, 0.8 μL/10⁶ cells) for 4 hr. Samples were subsequently run using BD FACS LSRII or sorted using BD FACS ARIA. Data were analyzed using FlowJo.

In vitro polarization of mouse bone marrow-derived monocytic cells, enzyme linked immuno-sorbent assay (ELISA) for IL-10 and Arginase activity assay

Bone marrow cells were collected from wild-type C57BL/6 mice by flushing tibias and femurs. Following red blood cell lysis, 2.5 × 10⁵ cells/mL of R10 media supplemented with recombinant mouse GM-CSF (40 ng/mL) + IL-6 (40 ng/mL) (Peprotech) were cultured for 5 days. These cells were then washed and 0.5 × 10⁶ cells/well were plated in 6 well plates, and treated with purified exosomes from EpCAM⁺ breast tumor cells, MSCs or 3T3 cells as described above, and incubated for 48 hr. Culture-conditioned media were collected and debris was removed.

Murine IL-10 protein concentration in conditioned media was measured by ELISA using a Sandwich ELISA kit (Biolegend) according to the manufacturer-recommended protocol.

L-Arginase activity in these exosome-induced or control monocytic cells were quantified using an Arginase assay kit (Abcam) according to manufacturer-recommended protocol, and Arginase activity fold changes were calculated relative to control cells.

RNA sequencing

Total RNA was isolated from BM-derived myeloid cells using RNA isolation kit (Qiagen) and analyzed for RIN^e. Next gen RNA sequencing was performed by the Moffitt Cancer Center Molecular Genomics facility. Paired-end RNA-seq reads were aligned to the GRCm38 reference genome using STAR (18) (version 2.5.2b) followed by adaptor trimming by cutadapt (version 1.8.1). Uniquely mapped reads were counted by htseq-count (19) (version 0.6.1) using Gencode M21 transcript annotations. Differential expression analysis was performed using DESeq2 (20). Genes with fold-change >2 and false discovery rate (FDR) q-values <0.05 were considered differentially expressed and then subjected for pathway analysis using GeneGO MetaCore (https://portal.genego.com/). Heatmaps were generated with R package ComplexHeatmap (21) using z-score transformed log 2 (1 + normalized count).

Analysis of TCGA data

Molecular data from The Cancer Genome Atlas (TCGA) for Breast Invasive Carcinoma (BRCA) was downloaded from the cBio Cancer Genomics Portal (http://www.cbioportal.org/), Broad Firehose website (https://gdac.broadinstitute.org/), and Genomic Data Commons Data Portal (https://portal.gdc.cancer.gov/). A total of 1,100 patients with matched clinical information and tumor RNA-seq data were used in this study. Clinical data for these patients was obtained from cBio portal. Gene expression values of tumor samples were calculated based on log2 transformed RSEM (RNA-Seq by Expectation Maximization) values (22), which are normalized counts provided in Broad Firehose portal. Multiple correlation analyses were performed using expressions of CD271, CD206, PD-L1, PD-1.

In vivo imaging of mice tumors

Tumor bearing mice were injected intraperitoneally with 2.5 mg/mouse luciferin (Promega), incubated for 10 min and luciferase activity of the Luc⁺ Brpkp110 cells was captured using a Xenogen IVIS200 in vivo imaging system.

Cell proliferation assays

5×10⁴ Brpkp110 cells were plated in 96-well plates and the next morning treated with or without 1 μl mouse MSC-derived exosomes. Cell proliferation was measured after 12 hr or 24 hr by addition of 10 μl MTT (Thiazolyl Blue Tetrazolium Bromide), 2-4 hr incubation at 37°C, dissolving pellet with 100 μl of DMSO and absorbance measurement at 570 nm. Experiment was performed twice with triplicates.

Extracellular matrix invasion assay

Mouse tumor epithelial cells (CD45⁻EpCAM⁺) were sorted and cultured in vitro for < 2 passages in R10 media. For invasion assay, cells were trypsinized and 300 μl of cell suspension containing 1.0 × 10⁶ cells/mL in serum free RPMI were seeded on the upper wells of Cell Invasion Assay chambers (Millipore-ECM550). In the lower wells, 500 μl R10 were filled. Cells were allowed to invade the extracellular matrix (ECM) for 36 hr. Non-migrated cells and the ECM gels were removed by cotton swabs. Invaded cells bound on the lower surface of the membrane were stained, air dried and photographed in an EVOS-Auto microscope. Finally, stained cells were dissolved in 10% acetic acid and absorbance was measured at 560 nm.

ELISpot

BM dendritic cells (BMDCs) were differentiated by culturing healthy C57/BL6 BM cells for 7 days with 20 ng/mL GM-CSF (Peprotech), added on day 0 and 3, and 10 ng/mL GM-CSF added on day 6(22). BMDCs were subsequently primed with irradiated Brpkp110 cells (100 Gy + 30 min UV) at a ratio of 10:1. ELISpot assay was performed by stimulating 1×10⁵ CD8⁺ T cells obtained from reactive lymph nodes from tumor-bearing mice with 1×10⁴ antigen-primed BMDCs in an ELISpot plate (BD) coated with mouse-IFN-γ capture antibody according to manufacturer recommendations, and incubated at 37°C, 5% CO₂ for 72 hr. Positive spots were developed and quantified in an ELISpot reader using Immunospot software (CTL).

Immunosuppression assay

Naïve T cells were harvested from C57/B6 mice spleens by magnetic bead negative selection to remove B220⁺, CD16/CD32⁺, CD11b⁺ and Class II MHC⁺ non-T cells, and labeled with proliferation tracker Cell Trace Violet (CTV) according to manufacturer recommendation. T cell expansion was stimulated by coating plates with anti-CD3 (1 μg/mL, Tonbo) and adding anti-CD28 (100 ng/mL, Tonbo). Mouse tumor sorted M-MDSCs or macrophages, or in vitro polarized BM-myeloid cells subsequently co-cultured with T cells at 1:1, 1:10 or 1:20 ratios, and incubated for 3-4 days prior to flow cytometry analysis.

In vivo PD-L1 blockade

PD-L1 in tumor-bearing mice was neutralized by intraperitoneal (IP) injection of 100 μg of anti-PD-L1 or isotype antibodies (BioXcell) every 3/4 days, starting from 3 days after tumor challenge. All the animals received a total of five doses of neutralization injection.

In vivo myeloid depletion

Myeloid cells in tumor-bearing mice were depleted in vivo by administering 200 μl of Clodronate liposome (Formumax) or Control Liposome (Formumax) IP two days prior to intratumoral or peritumoral administration of exosomes or PBS. Subsequent IP injections were done every 4^th day with 100 μl of respective liposome. All the animals received a total of four doses of neutralization injection.

LC-MS/MS analyses of exosomes

Exosomes were isolated from culture-conditioned media of EpCAM⁺ breast cancer cells, 3T3 and mouse MSCs according to above described protocol. Proteins were extracted from these exosomes, reduced by DTT, digested by trypsin, and subjected to LC-MS/MS analysis by the Moffitt Cancer Center Proteomics Facility. MaxQuant (version 1.5.2.8) was used to analyze the data, identify and quantify the proteins (23).

Statistical analyses

Unless mentioned, all data presented represent mean with SEM. Non-parametric Mann-Whitney-Wilcoxon test has been performed between two groups and Kruskal-Wallis test has been performed for comparisons between more than two groups, unless indicated otherwise. Non-parametric Spearman’s rank correlation has been performed for correlation analysis. Analyses were carried out in Graph Pad Prism 7.0 software. Significance threshold 0.05 for p was used.

RESULTS

Exosomes derived from MSCs contribute to differentiate M-MDSCs into highly immunosuppressive M2-polarized macrophages

M-MDSC differentiate into tumor-associated macrophages (TAMs) after migration to the tumor site (16). This is known to be influenced by factors that include hypoxia, CSF, VEGF and STAT6 activation (24). However, recent studies indicate that reprograming of local tissue MSCs drives a tumor-promoting, immune-modulatory phenotype (25, 26). To understand the association between tumor-educated MSCs and immunosuppressive M2-polarized macrophages at tumor beds, we focused on exosomes secreted by MSCs within the breast TME (Exo^MSC), which are known to contribute to disease progression (27-29). Hence, we first induced the expansion of MDSCs from mouse bone marrow in response to inflammatory cytokines (30), in the presence of exosomes derived from either tumor-derived primary MSCs, tumor cells or control 3T3 cells (all generated in exosome depleted RPMI medium). As shown in Figure 1A, the presence of Exo^MSC from transplantable murine p53/K-Ras/PI3K-driven Brpkp110 breast tumors (12, 13), but not tumor cell-derived exosomes, induced a significant increase in the differentiation of CD11b⁺F4/80⁺Ly6G⁻Ly6C⁺MHC-II^−/low cells into committed MHC-II⁺ macrophages. Moreover, both M-MDSCs and MHC-II⁺ macrophages in these cultures expressed higher CD206 and PD-L1 upon Exo^MSC incubation (Figure 1B-1C). In addition, exosome-educated macrophages showed higher activity of L-Arginase (Figure 1D) and secreted higher levels of IL-10 (Figure 1E), both of which are strong immunosuppressive drivers. Accordingly, RNA-seq and subsequent pathway analysis of BM-myeloid cells treated with Exo^MSC (but not untreated or Exo^EpCAM+ or Exo^3T3 treated cells) confirmed the acquisition of transcriptional patterns associated with immunosuppressive M2 macrophages (Figure 1F & Supplementary Files 1& 2).

Figure 1. MSC-derived exosomes promote conversion of M-MDSCs into immunosuppressive macrophages.

(A) BM-derived GM-CSF + IL-6-induced myeloid cells treated with or without exosomes from EpCAM⁺ tumor cells or 3T3 or MSCs for 48 hr. Bar graphs showing number (percentage) of M-MDSCs (CD11b⁺F4/80⁺Ly6G⁻Ly6C⁺IA/IE⁻) and macrophages (CD11b⁺ F4/80⁺Ly6G⁻Ly6C⁺IA/IE⁺) among the total myeloid population analyzed by flow cytometry. Flow cytometry dot plots of IA/IE in untreated and MSC-derived exosome treated cells gated for viable CD11b⁺F4/80⁺Ly6G⁻Ly6C⁺ cells. Bar graphs showing percentages of IA/IE⁺ cells in different treatment groups. Experiments were performed twice. (B) Dot plots showing CD206 expression in M-MDSCs (upper) and macrophages (lower) in untreated and MSC-derived exosome treated cells. Bar graphs showing percentages of CD206⁺ M-MDSCs and MFI of CD206 in M-MDSCs (upper); and CD206⁺ macrophages and MFI of CD206 in macrophages (lower), in different treatment groups. Experiments were performed twice. (C) Dot plots showing PD-L1 expression in M-MDSCs (upper) and macrophages (lower) in untreated and MSC-derived exosome treated cells. Bar graphs showing percentages of PD-L1⁺ M-MDSCs (upper), and PD-L1⁺ macrophages (lower) in different treatment groups. Experiments were performed twice. (D) Bar graph showing arginase activity of GM-CSF + IL-6-induced BM-derived myeloid cells upon treatment with or without EpCAM⁺ tumor cells or 3T3 or MSCs-derived exosomes for 48 hr. Experiments were performed three times. (E) Bar graphs showing IL-10 concentration in conditioned media from BM-derived myeloid cells upon treatment with or without EpCAM⁺ tumor cells or 3T3 or MSCs-derived exosomes for 48 hr. Experiments were performed three times and ELISA was performed with 4 replicates from each. (F) Heat map showing (left) differences in Z-scores for indicated genes in BM-derived myeloid cells upon treatment with or without EpCAM⁺ tumor cells or 3T3 or MSCs-derived exosomes for 48 hr. Graph (right) showing major upregulated pathways exclusively observed MSC-exosome treated BM-derived myeloid cells, calculated and represented as −log₁₀ values of False Discovery Rate (FDR). Experiments were performed in triplicate. (G) Dilution of Cell Trace Violet in labeled T cells activated with anti-CD3/CD28 antibodies and co-cultured with increasing ratios of BM-derived myeloid cells, treated with or without EpCAM⁺ tumor cells or MSCs-exosomes for 48 hr. Bar graph showing percentage of T cells proliferated after incubation with BM-derived myeloid cells for 3 days. Experiments were performed three times. *, p < 0.05; **, p < 0.01; ***, p < 0.001; NS, not significant.

Accordingly, myeloid cells treated with Exo^MSC exhibited increased immunosuppressive activity, compared to their counterparts incubated with Exo^EpCAM+ or untreated (Figure 1G). Together, these results strongly suggest that exosomes produced by tumor-associated MSCs contribute to the differentiation of M-MDSCs into M2-polarized macrophages at tumor beds, as well as enhancers of their immunosuppressive activity.

Exo^MSC promote accelerated malignant progression through accumulation of immunosuppressive myeloid cells and promoting tumor cell invasiveness

To confirm that exosomes produced by tumor-educated MSCs are sufficient to drive the conversion of M-MDSCs into immunosuppressive macrophages in the breast cancer microenvironment in vivo, we challenged B6 mice with syngeneic Brpkp110 breast tumors, and intratumorally administered Exo^MSC or PBS. As expected, intra-tumor administration of Exo^MSC increased tumor growth, compared to the PBS-injected control mice in multiple experiments (Figure 2A-2B). Similar differences between Exo^MSC and 3T3 cell-derived exosomes (Exo^3T3) were observed (Supplementary Figure 1A). Importantly, accelerated malignant progression was not the result of direct effects on tumor cell proliferation, because Exo^MSC had no effect on in vitro Brpkp110 growth (Supplementary Figure 1B). Rather, Exo^MSC induced a significant increase in the frequency of MHC-II⁻ M-MDSCs (Figure 2C) and, accordingly, MHC-II⁺ macrophages (Figure 2D) among all leukocytes at tumor beds. Notably, we observed a significant increase in the surface CD206 expression among tumor-associated M-MDSCs and macrophages in the Exo^MSC-induced tumors, and also increased proportion of CD206⁺ M-MDSCs and CD206⁺ macrophages number among total leukocytes (Figure 2E&2F).

Figure 2. Mesenchymal stem cell-exosomes drive increased tumor growth, upregulation of CD206 in tumor associated M-MDSCs and macrophages, EMT of cancer cells with superior invasive ability.

(A) In vivo luciferase analysis showing tumor growth in the right axillary flank after 14 days from intratumoral administration of MSC-derived exosomes or PBS into Brpkp110 breast tumors. Mouse MSC-derived exosomes (Exo^MSC) or PBS (Vehicle) was administered after 5 days of tumor challenge. Data are representative of the two independent experiments (n=5/group; one representative experiment of two). (B) Comparison of tumor weight between Exo^MSC and Vehicle group after 14 days of PBS or exosome administration (n=5/group; two experiments). Representative tumors from an individual experiment are depicted. (C) Scattered plot showing number of M-MDSCs (CD45⁺CD11b⁺F4/80⁺Ly6G⁻Ly6C⁺IA/IE⁻) among total tumor-infiltrated leukocytes (viable CD45⁺) in Vehicle and Exo^MSC tumors (n=5/group; two experiments). (D) Scattered plot showing number of macrophages (CD45⁺CD11b⁺F4/80⁺Ly6G⁻Ly6C⁺IA/IE⁺) among total tumor-infiltrated leukocytes (viable CD45⁺) in Vehicle and Exo^MSC tumors (n=5/group; two experiments). (E) Representative flow cytometry plots showing percentage of CD206⁺ cells in gated viable M-MDSCs (CD45⁺CD11b⁺F4/80⁺Ly6G⁻Ly6C⁺IA/IE⁻), and quantification of CD206⁺ M-MDSCs among total tumor-infiltrated leukocytes in Exo^MSC and Vehicle tumors are represented as scattered pots (n=5/group; two experiments). (F) Representative flow cytometry plots showing percentage of CD206⁺ cells in gated viable macrophages (CD45⁺CD11b⁺F4/80⁺Ly6G⁻Ly6C⁺IA/IE⁺), and quantification of CD206⁺ macrophages among total tumor-infiltrated leukocytes in Exo^MSC and Vehicle tumors are represented as scattered pots (n=5/group; two experiments). (G) Western blots of Slug, Snail and E-cadherin in tumors from Exo^MSC and Vehicle group, performed three times from 3 different tumors of each group. β-actin used as loading control. Bar graphs showing western blot intensities of Slug, Snail and E-cadherin, relative to β-actin. (H) Phase contrast microscopy analysis and quantification, by absorbance measurement at 560 nm, of extracellular matrix invasion by CD45⁻EpCAM⁺ cancer cells from Exo^MSC and Vehicle group, performed three times from 3 different tumors of each group. *, p < 0.05; **, p < 0.01.

M-MDSCs and M2 macrophages at tumor beds are known to promote epithelial-mesenchymal-transition (EMT) of tumor cells with increasing ECM-invasive abilities (31). Accordingly, we observed a decrease in epithelial marker E-cadherin and an increase in mesenchymal markers Snail and Slug with stronger ECM invasion ability of tumor-sorted epithelial cells (CD45⁻EpCAM⁺), from Exo^MSC–treated tumors (Figure 2G&2H). Together, these data indicate that Exo^MSC promote the accumulation of immunosuppressive myeloid cells as well as an invasive, pro-metastatic phenotype by breast cancer cells, overall driving accelerated malignant progression.

Exo^MSC thwart protective anti-tumor immunity in tumor-bearing mice by up-regulating PD-L1 expression in myeloid cells and de-repressing PD-1 in T cells

Consistent with the immunosuppressive role of MSCs at tumor beds, Exo^MSC completely abrogated the production of IFN-γ by CD8⁺ T cells from tumor-draining lymph nodes in response to cognate tumor antigens in ELISpot analysis (Figure 3A). Corresponding decreases in IFN-γ production by both intratumoral CD8⁺ and CD4⁺ T cells from Exo^MSC-induced group were observed through FACS analysis of dissociated tumors (Figure 3B). Accordingly, Exo^MSC-induced tumor-derived M-MDSCs and, to an even greater extent, differentiated IA/IE⁺ macrophages, exhibited increased immunosuppressive activity on a per cell basis in multiple independent assays (Figure 3C). Hence, immunosuppressive myeloid cells influenced by Exo^MSC are not only more abundant, but also more active at dampening T-cell-mediated responses.

Figure 3. MSC-derived exosomes dampens anti-tumor T cell responses.

(A) ELISpot analysis of CD8⁺ T cells isolated from draining lymph nodes of Brpkp110-tumor-bearing mice, stimulated with irradiated tumor cell-pulsed BM-derived dendritic cells. (B) Representative flow cytometry analyses of IFN-γ production by CD8⁺ and CD4⁺ T cells from Exo^MSC or Vehicle tumors, pre-induced with PMA (20 ng/mL), Ionomycin (Sigma, 1 μg/mL) and Golgi stop (0.8 μL/10⁶ cells) for 4 hr. An isotype control was utilized to set the gate for intracellular IFN-γ signal. Scattered graphs showing percentages of CD8⁺ and CD4⁺ cells from Exo^MSC or Vehicle tumors expressing IFN-γ; bar graphs showing IFN-γ MFI (n=5/group; two experiments). (C) Dilution of Cell Trace Violet in labeled T cells activated with anti-CD3/CD28 antibodies and co-cultured with increasing ratios of IA/IE⁻ M-MDSCs or IA/IE⁺ macrophages. Experiments were performed twice. (D) Representative flow cytometry analyses showing PD-L1 expression by CD206⁺ M-MDSCs and CD206⁺ M2 macrophages. Scattered graphs showing percentages of CD206⁺ M-MDSCs, and CD206⁺ M2 macrophages from Exo^MSC or Vehicle tumors expressing PD-L1; bar graphs showing PD-L1 MFI (n=5/group; two experiments). (E) Representative quantification of PD-1⁺ T cells (n=5/group; two experiments) and intensity of PD-1 expression on T cells (n=5/group; one representative experiment of two) in tumors from Exo^MSC and Vehicle group. (F) Scattered plot of log2 PD-1 mRNA expression from TCGA dataset (n=1,100) comparing breast tumors with the strongest CD271 (n=250) and weakest CD271 (n=250) expression. (G) Tgfb mRNA expression in tumor-sorted M-MDSCs (left), tumor-associated macrophages (TAMs, middle), and CD45⁻EpCAM⁺ cancer cells (right). Experiment was performed twice. *, p < 0.05; **, p < 0.01; ***, p < 0.001; NS, not significant.

To understand how myeloid cells polarized in response to Exo^MSC acquire enhanced immunosuppressive activity, we focused on our observation that the expression of PD-L1 in both CD206⁺IA/IE⁺ macrophages and their CD206⁺IA/IE⁻ precursors in Exo^MSC tumors was significantly increased by percentage whereas, the level of expression of PD-L1 on a per cell basis is not altered (Figure 3D). In addition, we found that tumor-infiltrating lymphocytes in Exo^MSC tumors express higher levels of PD-1 (Figure 3E), supported by the association between mRNA expressions of CD271 and PD-1 in TCGA human breast tumors (Figure 3F). Accordingly, Exo^MSC–programmed M-MDSCs and macrophages within the TME produced significantly more (~5 fold and ~3 fold, respectively) Tgfb (Figure 3G, left&middle), an immunosuppressive factor that de-represses PD-1 expression (32). In contrast, EpCAM⁺ cancer epithelial cells showed no significant change in Tgfb expression (Figure 3G, right). Again supporting the relevance of these observations, analysis of 1,100 breast tumors in TCGA datasets confirmed a clear difference in PD-1 level between breast tumors with high vs. low CD271 expression levels (p < 1.32e-19; n=250 strongest CD271, and n=250 weakest CD271), suggesting dysregulation of PD-1 expression in breast tumor associated T cells with increased infiltration of CD271⁺ MSCs (Figure 3F). Confirming the requirement of the Exo^MSC–dependent PD-L1:PD-1 axis in the abrogation of anti-tumor immunity, in vivo neutralization of PD-L1 in Brpkp110 tumor-bearing mice eliminated differences in tumor growth elicited by Exo^MSC, while PD-L1 blockade was ineffective in vehicle-treated or Exo^3T3 mice (Figure 4A-4B, Supplementary Figure 1C). Because Brpkp110 tumor cells express very low PD-L1 in vivo (Supplementary Figure 1D), these results indicate that increased expression of PD-L1 elicited by Exo^MSC in M2-polarized myeloid cells is sufficient to explain the abrogation of protective anti-tumor immunity.

Figure 4. MSC-derived exosomes drive breast cancer progression through PD-L1-PD-1 axis, and contain M-MDSC to M2 differentiating factors.

(A) Vehicle and Exo^MSC group mice administered with anti-PD-L1 neutralizing antibodies or anti-isotype control antibodies. Volume comparison at different time points (n=5/group, one representative experiment of two) starting from day of intra-tumor exosome or PBS administration (Day 0). (B) Comparison of tumor weight between anti-PD-L1 or anti-isotype control antibody injected mice groups (n=5/group, two experiments) after resection on day 21 post Brpkp110-challenge. Representative tumors from an individual experiment are depicted. (C) Comparison of intratumoral CD11b⁺ myeloid population in mice administered with Chlodronate liposomes or Control liposomes. (D) Both Vehicle and Exo^MSC groups administered with Chlodronate liposomes or Control liposomes. Volume comparison (n=5/group, one representative experiment of two) at different time points starting from Day 0. (E) Comparison of tumor weight between Chlodronate liposomes or Control liposomes treated mice groups (n=5/group, two experiments) after resection on day 21 after Brpkp110-challenge. Representative tumors from an individual experiment are depicted. (F) List of important molecules contained exclusively or superiorly in MSC-derived exosomes. CD63 is listed as exosome quality control. *, p < 0.05; **, p < 0.01; ***, p < 0.001; NS, not significant.

Depletion of macrophages contain immunosuppressive factors rescues Exo^MSC-driven increased breast tumor growth

To confirm that Exo^MSC-dependent differences in tumor growth are indeed driven by suppressive macrophages we used chlodronate liposomes to deplete macrophages in vivo. As shown in Figure 4C, chlodronate administration resulted in a 4.5-fold reduction in tumor-infiltrating myeloid cells, including M-MDSCs and macrophages. Consequently, this depletion eliminated Exo^MSC–induced differences in malignant progression (Figure 4D-4E).

To understand what elements carried by Exo^MSC are responsible for driving M2-polarization and enhanced immunosuppressive activity in breast cancer myeloid cells, we finally compared the composition of Exo^MSC and exosomes derived from 3T3 and EpCAM⁺ breast tumor cells by tandem mass spectrometry (Supplementary File 3). As shown in Figure 4F, multiple factors known to promote myeloid immunosuppressive activity and macrophage M2-polarization were selectively found in Exo^MSC, but not in exosomes derived from other cell types. They include TGF-β (33) and C1q (34), as well as semaphorins (35). Notably, the purity of exosomes was confirmed with equivalent reads for the exosome marker CD63 from MSCs, 3T3 and EpCAM+ breast tumor cells-derived exosomes (Figure 4F, Supplementary File 3). Therefore, multiple immunosuppressive factors specifically contained in MSC exosomes program myeloid cells at tumor beds to play a necessary role in accelerated malignant progression.

Accumulation of M2-polarized myeloid cells in human breast cancer is associated with increased MSC-infiltration

To support the relevance of our findings in preclinical models, we next quantified mRNA expression of the M2 marker CD206 and the MSC marker NGFR/CD271 in 51 human primary breast tumors of different histological types by Q-PCR. As shown in Figure 5A, we found a significant positive correlation between CD206 and CD271 expression levels, independent of histological type. Accordingly, FACS analysis of 15 freshly dissociated human breast tumors confirmed that tumors with denser infiltrates of CD45⁻CD271⁺ MSCs also accumulate more CD45⁺CD3⁻CD11b⁺CD206⁺CD163⁺ myeloid cells and partially suggests M2 polarization (Figure 5B). In addition, analysis of 1,100 TCGA breast tumors confirmed a positive correlation lies between intratumoral expression of CD206 and CD271 (Figure 5C), while CD206 expression was significantly higher in CD271^high tumors, compared to samples with the weakest CD271 signal (Figure 5D, left). Correspondingly, the highest CD206 expressing tumors also showed significantly higher CD271 levels (Figure 5D, right). More importantly, histological analysis of 19 additional human breast cancer samples of different subtypes confirmed a positive association between infiltration of CD206⁺ myeloid cells and CD271⁺ MSCs (Figure 5E). Together, these data confirm a strong association between MSCs and M2-polarized myeloid cells in human breast cancer, further supporting that the immune-modulatory activity of tumor-educated MSCs could be at least partially driven by the accumulation of suppressive myeloid cells.

Figure 5. Infiltration of MSCs and M2 macrophages in human breast tumors shows a strong positive correlation.

(A) Surgically operated fresh IDC breast tumors were collected and RNA was isolated (n=51). Scattered plat showing correlations between mRNA expressions of CD206 and CD271 with a Spearman’s correlation co-efficient value (r) 0.896. (B) Representative dot plots of flow cytometry with dissociated tumors (n=15), showing tumors with higher percentage of CD45⁻CD271⁺MSCs has a higher percentage of CD45⁺CD3⁻CD11b⁺CD206⁺CD163⁺ M2 macrophages (left). Graph (right) showing correlation (Spearman’s correlation, r=0.83) between intratumoral percentages of MSCs and myeloid cells with features of M2 macrophage. (C) From TCGA dataset of 1,100 primary breast tumors, CD271 and CD206 mRNA expression were analyzed, and log values are represented in Y-axis and X-axis, respectively in the scattered correlation plot, showing a positive correlation (Spearman’s correlation, r=0.26) (D) Graph on left showing log CD206 mRNA expression from TCGA dataset (n=1,100) comparing breast tumors with the strongest CD271 (n=250) and weakest CD206 (n=250) expression. Graph on right showing log CD271 mRNA expression from TCGA dataset (n=1,100) comparing breast tumors with the strongest CD206 (n=250) and weakest CD206 (n=250) expression. (E) Breast TMA (n=19) stained for CD271 (Alexa Fluor 568) and CD206 (Alexa Fluor 647). Nuclei were stained with DAPI. Representative immunofluorescence images showing tumors with both high and both low CD271 and CD206. Scatter graph (upper) showing percentage of CD271⁺ cells in CD206 high (>0.4%; n=9) vs. CD206 low (<0.4%; n=10) tumors. Scattered plot (bottom) showing positive (Spearman’s correlation, r=0.3) correlation between MFI of CD271and MFI of CD206 per unit area of tumor (n=19). *, p < 0.05.

Exosomes derived from MSCs polarize undifferentiated myelo-monocytic cells into CD206⁺PD-L1^high cells

To confirm the effect of human tumor-derived MSCs on macrophage polarization, we induced the acquisition of an undifferentiated M0 phenotype by human monocyte THP-1 cells through PMA treatment, characterized by an increase in CD11b expression (Figure 6A), as reported (14). As shown in Figure 6B, media conditioned by two different breast cancer cell lines elicited some elevation in the expression of the M2 macrophage marker CD206, PD-L1 and, to a lesser extent, PD-L2 (data not shown). Similar elevations in CD206, and PD-L1 in THP-1 cells were elicited under M2-polarizing conditions in the presence of IL-4 and IL-13 (Figure 6B, Supplementary Figure 1E). More importantly, co-culture of M0-polarized THP-1 cells with primary CD45⁻CD271⁺ MSCs isolated from dissociated human breast tumors resulted in further increases in the levels of CD206 and PD-L1, compared to tumor-conditioned media alone or stimulation by IL-4 and IL-13 only (Figure 6B-6C, Supplementary Figure 1E). As expected, incubation with media conditioned by the non-tumor breast cell line MCF10A at the same confluence did not induce any significant effect on the expression of CD206 or PD-L1, compared to normal R10 media (Figure 6B-6C). Flow cytometry analyses further confirmed the up-regulation of cell surface PD-L1 and CD206 upon incubation with tumor-conditioned media (Figures 6D-6E). Again, co-culture with MSCs further increased the proportions of CD206⁺ and PD-L1⁺ cells (Figures 6D-6E). Together, these results indicate that heterotypic interactions with MSCs and breast cancer cells can alter the polarization of myelo-monocytic cells, resulting in a more pronounced suppressive phenotype.

Figure 6. Human mesenchymal stem cells boost CD206^high PD-L1^high M2 macrophage polarization.

(A) Flow cytometry histograms showing expression of CD11b in untreated or PMA-induced THP-1 cells (M0 THP-1). (B) Western blot analysis showing expressions of PD-L1, CD206, CD163 and β-actin in M0 THP-1 cells, with different treatment combinations, representative of two experiments. M0 THP-1 cells treated with IL-4 and IL-13 as a positive control of M2 polarization. In the remaining combinations, M0 THP-1 cells cultured in conditioned media (CM) from either MCF10A or MDA-MB-231 or T47D and co-cultured with or without human MSCs. (C) Densitometry analysis of western blots showing intensities of CD206, PD-L1, CD163 bands relative to respective β-actin bands (D) Flow cytometry dot plots showing percentages of CD206⁺ and PD-L1⁺ and histograms showing MFI of CD206 and PD-L1 in M0 THP-1 cells in different treatment combinations as mentioned, representative of two experiments. (E) Bar graphs showing mean±SEM percentages of CD206⁺, PD-L1⁺ cells. *, p < 0.05; **, p < 0.01; NS, not significant.

Finally, to demonstrate the contribution of human Exo^MSC vs. tumor cell-derived exosomes to the acquisition of tumor-promoting activities by myeloid cells at tumor beds, we cultured primary CD45⁻CD271⁺ MSCs from breast cancer patients, and treated them with GW4869, a chemical inhibitor of exosome biogenesis/release (16), or vehicle. M0 THP-1 cells were then added to the culture, in the presence of conditioned media from GW4869- vs. vehicle-treated breast cancer cells and indirect co-culture with GW4869- vs. vehicle-treated MSCs. As shown in Figure 7A, significant downregulation of PD-L1 expression was only observed when MSCs were inhibited by GW4869, while untreated MSCs drove significant up-regulation of both PD-L1 in M0 THP-1 cells cultured in tumor-conditioned media. In contrast, inhibition of exosome secretion from MDA-MB-231 or T47D tumor cells had no effect on PD-L1 up-regulation (Figure 7A). Correspondingly, M0 THP-1 cells when added in tumor cell-conditioned media and treated with purified exosomes from MSCs, but not with breast cancer cell line exosomes, showed significant upregulation of PD-L1 and CD206 (Figure 7B-7D).

Figure 7. Human mesenchymal stem cells-derived exosomes drive elevated PD-L1 and CD206 expression by macrophages.

(A) Western blot analysis of PD-L1 (left) in M0 THP-1 cells, grown in CM from breast cancer cell lines treated with or without 20 μM GW4869, and co-cultured with mesenchymal stem cells treated with or without GW4869. β-actin used as loading control. Experiments were performed twice. Bar graphs (right) showing mean±SEM intensities of PD-L1 western blot bands relative to β-actin. (B-D) M0 THP-1 cells treated with different combinations (a,b,c and d) of purified exosomes from breast cancer cell lines and mesenchymal stem cells. (B) Western blot analysis of PD-L1, representative of two experiments (left). Bar graphs (right) showing mean±SEM intensities of PD-L1 western blot bands relative to β-actin. (C) Flow cytometry analysis of CD206 and PD-L1, representative of two experiments. (D) Bar graphs showing mean±SEM percentages of CD206⁺, PD-L1⁺ cells. (E) PD-L1 mRNA expression in human breast tumors (n=51) relative to healthy breast tissues, quantified by Q-PCR, and grouped according to CD271 mRNA fold change less or more than 6. (F) Graph showing level of mRNA of PD-L1 in 1,100 breast tumors (showed with color intensity gradient) analyzed from TCGA data where each dot represents individual tumors with X-axis value for log2 CD206 expression, Y-axis value for log2 CD271 expression. Spearman’s Rank correlation co-efficient between CD271 and PD-L1 is 0.06483444; and between CD206 and PD-L1 is 0.497973717. Scattered plot of log2 PD-L1 mRNA expression from TCGA dataset (n=1,100) comparing breast tumors with the strongest CD271-CD206 co-expression (n=250) and weakest CD271-CD206 expressions (n=250). *, p < 0.05; **, p < 0.01; ***, p < 0.001.

Supporting the relevance of these observations, human breast tumors with higher CD271 mRNA in our cohort of 51 patients also showed higher PD-L1 mRNA (Figure 7E). Consistently, TCGA breast tumors with concurrent overexpression of CD271 and CD206 also express maximal PD-L1 (Figure 7F). These data strongly support the relevance of our in vivo experiments demonstrating that Exo^MSC drive M2-polarization and the acquisition of immunosuppressive PD-L1 by tumor-associated myeloid cells at tumor beds.

DISCUSSION

Here we show that exosomes secreted by tumor-educated mesenchymal stem cells accelerate breast cancer progression by driving differentiation of immature M-MDSCs into M2-polarized macrophages with greater immunosuppressive activities, which is sufficient to dampen anti-tumor immunity and promotes the acquisition of EMT features by cancer cells.

Our results provide new insight into how the crosstalk between tumor, mesenchymal and immune cells at tumor beds drives malignant progression by programming immature myeloid cells into macrophages that become necessary and sufficient for accelerated tumor growth. Macrophages with protumoral activity represent the major leukocyte population infiltrating cancers (1). Accumulated evidence indicates a substantial augmentation of immunosuppressive activity of M-MDSCs and rapid differentiation to even more immunosuppressive macrophages in solid tumors (2), but the mechanisms leading to this phenotypic transformation remain unclear. Investigation into the mechanisms driving macrophage programming in the tumor microenvironment have focused on two master regulators: CSF1 (1) and hypoxia (7) that, together, render MDSCs more immune suppressive and accelerate their conversion to macrophages. Our results uncover a novel mechanism diverting MDSC differentiation towards more immunosuppressive macrophages, driven by MSCs at tumor beds. Similar to hypoxia (2), Exo^MSC increase PD-L1 expression in both MDSC and macrophages. Although HIF1α has been implicated in the process of conversion of MDSCs into more inhibitory macrophages, the precise molecular mechanism of this phenomenon remains unclear (2). Interestingly, recent work from Buckanovich and colleagues shows that hypoxia is critical to induce a tumor-promoting phenotype in resident MSCs in different malignancies (25). It is tempting to speculate that exosomes secreted by tumor MSCs re-programmed under conditions of oxygen deprivation are crucial mediators of some of the effects of hypoxia on the conversion of MDSCs into terminally differentiated immunosuppressive macrophages.

Our results also show that Exo^MSC, but not exosomes from other cell types, carry high levels of TGF-β, along with semaphorins and complement factors, all of which are known to enhance immunosuppressive activity and M2-polarization in myeloid cells (33-35). As we reported, TGF-β de-represses PD-1 in tumor-infiltrating lymphocytes through SATB1 downregulation and competition for the same binding site (32). Accordingly, we found higher PD-1 expression in neighboring T cells in Exo^MSC–treated tumors. In addition, tumor epithelial cells attain EMT molecular features and superior invasive ability, another known activity of TGF-β (36). Therefore, the particular re-programming of MSCs at tumor beds results in the production of exosomes that are crucial for rewiring the immune-environment into a permissive milieu for tumor cell invasion and evasion of tumor-reactive T cells.

In summary, our study unveils the role of MSC-derived exosomes in breast cancer progression, and supports that targeting MSCs could be considered as a therapeutic strategy to restore T cell-mediated anti-tumor cytotoxic effects in patients suffering from breast cancer.