Regulation of mesenchymal stem cell function by TGFβ-1 on mast cell extracellular vesicles – role of endosomal retention

Extracellular vesicles (EVs) convey biological messages between cells, either by surface-to-surface interaction, or by shuttling of bioactive molecules to a recipient cell cytoplasm. Here we show that EVs released by human primary mast cells or transformed human mast cells (HMC1), carry TGFβ-1 on their surface. EV-associated TGFβ-1 enhance the migratory activity of human mesenchymal stem cells (MSCs) compared to free TGFβ-1, as both knockdown of TGFβ, or a TGFβ-antibody, attenuate the effect. The MSCs respond by increasing matrix metalloproteinase-2 and −9 (MMP) activity. Further, EVs given to MSCs are retained in the endosomal compartments at a time of biological function, prolonging EV-associated TGFβ-1 signaling vs free TGFβ-1. When exposed to EVs, MSCs home more toward allergen-exposed lung in a mouse allergen model, resulting in attenuated allergic inflammation. Our results show that mast cell-EVs are decorated with TGFb-1, are retained in endosomes, which influences both MSC phenotype and function.


INTRODUCTION
Inflammatory responses involve intercellular communication through secreted soluble mediators such as chemokines and cytokines (1-4). These molecules induce different signaling pathways in cells, and modulate gene expression. Beyond soluble mediators, cells also release extracellular vesicles (EVs) that can influence recipient cell phenotype. EVs are nanometer-sized, lipid bilayer-enclosed membrane structures that carry an array of bioactive molecules, including proteins, nucleic acids, and lipids (5)(6)(7)(8)(9)(10)(11)(12). EVs can rapidly be taken up by recipient cells, and the contents of the EVs can induce a vast array of phenotypic changes in a recipient cells, which has been implied in both health and disease (13)(14)(15)(16).We and others have shown that EVs carry bioactive molecules on their surface, including c-KIT, Wnt and IL-1β (6,17).
It has been shown that mesenchymal stem cells (MSCs) can regulate immune responses (18)(19)(20)(21)(22), which has made them attractive therapeutic candidates in several inflammatory diseases, including multiple sclerosis, rheumatoid diseases, tissue transplant rejection and inflammatory respiratory disorders (23)(24)(25). It is known that cytokines can alter the MSC phenotype in numerous ways (26)(27)(28)(29), but very little is known whether or how EVs can influence MSC functions, and exactly by which mechanism this may be induced.
In theory, MSC biology in vivo may be regulated by cells in their vicinity, and thus the EVs released by other cells and taken up by the MSCs within the body have the capacity to regulate MSCs. Further, it could also be proposed that EVs may be used to manipulate MSCs in vitro to improve their therapeutic potential, putatively increasing their immunomodulatory effects as cell therapies in disease. We here hypothesize that mast cell-derived EVs influence the phenotype of human MSCs, and thereby their therapeutic potential in a model of respiratory inflammation. To verify this, we have utilized EVs from a both primary human mast cells as well as a human mast cell line, and tested the ability of EVs to influence the phenotype of primary bone marrow derived mesenchymal stem cells, including identifying intracellular signaling pathways that may be involved in an induced response. Further, we could identify a molecular component of EVs that can influence recipient cell biology. Lastly, the effects of EV-treated MSCs on a lung allergic inflammation model are tested.

Mast cell-derived EVs enhance the migration of MSCs in-vitro.
EVs were isolated from the human mast cell line, HMC-1.2 by using differential ultracentrifugation. The mast cell-derived EVs had a size of 125.4 ± 3.9 nm according to nanoparticle tracking analysis (NTA, Supplementary Fig. 1a), and were positive for EVenriched proteins such as CD63, CD81, and CD9 as detected by flow cytometry ( Supplementary Fig. 1b). EVs from primary human derived mast cells, (cultured from a human peripheral blood stem cell population positive for CD133 + ) have similar size and structure on EM (40-120 nm) ( Supplementary Fig. 1c), and express CD63 as detected by flow cytometry (Supplementary Fig. 1d). HMC-1 EVs were efficiently taken up by primary human MSCs in a time and temperature-dependent manner, which suggests that the EV uptake is a biologically active process ( Supplementary Fig. 2a-c). Adding EVs to the cell cultures induced a n elongated morphology of the MSCs (Fig. 1a), but did not alter the ability of the MSCs to differentiate into adipocytes or osteocytes ( Supplementary Fig. 4). In vitro analysis of an MSC scratch assay, suggest that the EVs increase wound healing activity (Fig. 1b).
Enhanced migratory activity of MSCs upon EV stimulation was also associated with increase in transcripts of matrix metalloproteinases (MMP-2 and MMP-9) mRNA in cells, and with increased secretion of MMP-2 and MMP-9 proteins into the cell culture supernatant, associated with increased gelatinolytic activity (Fig. 1c-d). It has previously been shown that multiple growth factors can influence MSC migration behavior by regulating signaling cascades (30). We therefore exploited fetal bovine serum (FBS) as a pan-chemoattractant in the next experiment, and could show that EV-treated MSCs have a higher migratory tendency towards the FBS compared to non-treated MSCs, in a dose-depended manner (Fig. 1e).
Collectively, these results show unequivocally that mast cell-derived EVs enhance the migration of MSCs in vitro.

TGFβ-1 is present on the surface of mast cell-derived EVs
To identify surface components of mast cell-derived EVs that could potentially regulate the observed MSC migration, we first determined the membrane proteome of the EVs. This proteomics approach reduces extra and intra-vesicular proteins, but enriches membrane proteins and therefore identifies various low-abundance bioactive proteins that are not detected in standard EV proteomics assays (31). In total we identified in total 1743 proteins (Supplementary Table 1), of which 504 were membrane-associated proteins, including several receptors such as TGFβ and insulin receptors. Interestingly, we also detected cytokines such as TGFβ-1 that are rarely identified with mass spectrometry due to their low abundance.
Multiple approaches were used to validate the presence of TGFβ-1 on the EVs. Firstly, we developed a classical sandwich ELISA to detected the relative abundance of TGFβ-1 in TGFβ-1+ as well as CD63+ EVs in various fractions of HMC-1 EVs after iodixanol density gradient separation. Confirming our proteomics finding we determined that the majority of TGFβ-1floated in fractions no. 1 and 2, with density in the range of 1-1.08 g/ml ( Supplementary Fig 3 and Fig. 2a. This was coincident with the EV markers TSG101 and CD81. In addition, the association of TGFβ-1 with EVs was observed by fluorescence correlation spectroscopy (FCS). Specifically, we found that the signals for TGFβ-1 and the EV marker CD63 were co-localized in the same time lapse, and had the same diffusion time in the FCS analysis (Fig. 2b), which strongly suggests that they are co-localized on the same individual EV. With this method, we determined that approximately 17 % of CD63 positive EVs were also positive for surface expression of TGFβ-1, but some EVs carried either marker alone. Finally, we performed a sandwich ELISA to quantify the amount of TGFβ-1 associated with vesicles bound to anti-CD63-coated beads. We could detect approximately 12 pg of total TGFβ-1 on 30 µg CD63-positive EVs from HMC-1 cells, out of which 7.1pg was of the active form of TGFβ-1 (Fig. 2c). Taken together these experiments conclusively show that TGFβ-1 is associated with EVs on their surface.

TGFβ-1 on the surface of EVs induces enhanced migration of MSCs
As it has been shown before that free, soluble TGFβ-1 can stimulate MSC migration (27) and since we in the proteomics observed TGFβ-1is present in EV membrane isolates, we set out to determine whether this molecule is responsible for the observed EV-mediated enhanced migration of the cells. Importantly, SMADs are a family of proteins that are involved in signal transduction and transcriptional regulation (32), and SMAD2 is activated in response to TGFβ signaling ( Fig. 3a) and could be involved in MSC migration (33). Indeed, we observed that the ratio between nuclear and cytoplasmic SMAD2 in MSCs increases following EV exposure, indicating that EVs induce nuclear translocation of this transcription factor (Fig.   3c). Consistent with activation of SMAD2, we also observed an increase in Tgfb1 mRNA transcript following stimulation with EVs, in a time-and dose-dependent manner. This suggests that EVs enhance Tgfb1 expression in MSC, most likely via SMAD2 signaling (Fig.   3d). This is consistent a previous study documenting that soluble TGFβ-1 regulates its own expression in an autocrine manner (34).
To further determine whether the EV-associated TGFβ-1 is responsible for the observed enhanced migration and TGFβ-1signaling in MSCs we first isolated EVs from TGFB1 knock-down (35) HMC-1 cells, using a doxycycline-inducible CRISPR/Cas-9 system (Supplementary Fig. 6a-b). Indeed, EVs derived from TGFβ-1 KD mast cells resulted in reduced phosphorylation of SMAD2 in recipient MSCs compared to controls (Fig. 3e), as well as reduced migration towards the pan-chemoattractant FBS (Fig. 3f). To further confirm this, we pre-incubated EVs with a TGFβ-neutralizing antibody that blocks the TGFβ-1 interaction with its receptor. Our results show that TGFβ-neutralized EVs evoke less phosphorylation of SMAD2 ( Supplementary Fig.6) and significant reduction in the migration of the MSCs towards the pan-chemoattractant FBS (Fig. 3g). In addition, we found a drastic drop in TGFβ-1 secretion into the growth media by the MSCs when the TGFβ-1 receptor was blocked by a specific small molecule inhibitor (LY215799) (Fig. 3h). Collectively, these results suggest that TGFβ-1 on the EV surface induces an enhanced migration phenotype and TGFβ-1 response in MSCs.

Sustained TGFβ-1 signaling by endosomal retention of EVs and lysosome evasion
It is well known that TGFβ-1 rapidly activates signaling by phosphorylation of SMADs, which then quickly lessens (36,37). Having observed similar signaling activation and functional outcome for EV-bound TGFβ-1, we asked whether free TGFβ-1 induces similar temporal signaling activation in MSCs in our system. To test this, we used dose-matched concentrations of TGFβ-1 in its free form and in the EV-bound form, and performed Western blotting analysis for phosphorylated SMAD2 (pSMAD2) in the recipient MSCs. First, we determined the amount of TGFβ-1 associated with our total EV-isolates after floating them on a density gradient. We found approximately 18.5 pg of total TGFβ-1 in 30 µg total EVs, out of which approximately 40% was in the active form of TGFβ-1 ( Supplementary Fig.7).
Hence, 0.3 pg of active TGFβ-1 was considered to be present per microgram of total EVs, which was utilized for further experiments ( Supplementary Fig. 7). Interestingly, TGFβ-1 displayed altered activation kinetics when bound to EVs, producing a lower amplitude (Supplementary Fig. 8 (30 mins)), but more sustained (Fig. 4a (60 mins)) activation of pSMAD2. We further found that EVs stimulate MSC migration more efficiently than free TGFβ-1 at dose-matched concentrations (≥30 pg/ml) (Fig. 4b). These data suggest that the sustained, low amplitude signaling elicited by EV-bound TFGβ-1 has more functional effect than the transient, high amplitude signal evoked by free TFGβ-1.
To elucidate the difference in the signaling efficiency of free versus EV-bound TGFβ-1, we assessed the fate of EVs in intracellular compartments, primarily comparing lysosomes and early endosomes. It has previously been reported that free TGFβ-1 is rapidly processed via lysosomes (38), and we hypothesized that EV-bound TFGβ-1 taken up by recipient cells evades or is delayed in trafficking to lysosomes, producing a prolonged TGFβ intracellular signal. As it is not feasible to perform these experiments in large scale using MSCs, we performed these specific organelle localization experiments on HEK-293T cells. Briefly, we performed high-resolution fractionation of cell organelles, to physically separate lysosomes and early endosomes from HEK-293T cells that were exposed to biotinylated-EVs for 60 min (Supplementary Fig. 9a and Fig. 4c). We found that the endosome-enriched fraction (EEF) was highly enriched in biotinylated proteins as compared with the lysosome-enriched fraction (LEF), suggesting that the EV components taken up by the cells preferentially are retained in endosomes, where they may evade lysosomal degradation at the time of activating the cell (Fig. 4c). The observed prolonged EV-TGFβ-1 mediated activation of cells may therefore at least partly be explained by EV lysosomal evasion.
To confirm the endosomal localization of the EV components, two different approaches were taken. First, EEF and LEF samples were incubated with organelle-specific antibodies to pull down the endosomal and lysosomal fractions (Fig. 4d), again showing enrichment of biotinylated proteins in the EEA1 specific pull-down (as a marker of endosomes) compared to the LAMP1-specific pull-down (as a marker of lysosomes) (Fig. 4f).
To confirm this finding in human MSCs we traced the biotinylated EVs by performing a similar organelle specific pull-down of EEA1 and LAMP1 from crude cytoplasmic preparation of MSCs. As for the HEK-293T cells, also MSCs showed a higher biotinalylation signal in the endosomal compartments compared to the lysosomal compartments. Secondly, we looked for endosomal traces of streptavidin captured EV-biotin to validate the finding.
Cells were incubated with biotinylated EVs and then we used streptavidin-coated beads to pull down the cellular compartments containing biotinylated proteins where the EVassociated biotinylated proteins are present, as described in supplementary Fig. 9b (Left panels). As expected, the total amounts of proteins binding to the beads were significantly higher in the EEF compared to the LEF sample ( Supplementary Fig. 9b (i)). Furthermore, a higher percentage of the proteins were bound to the streptavidin-coated beads in the EEF samples, while a higher percent of the proteins were unbound to the beads in the LEF sample, indicating that more biotinylated proteins are present in the endosome enriched fraction. To further confirm that the bead-bound material (biotinylated proteins) is originating from the organelles of interest, we identified the early endosomal protein, EEA1 (here used as a marker for the early endosome) and LAMP1 (here used as a marker for the lysosome) (Supplementary Fig. 9b (ii)) in the bead-bound samples. In summary, the results of these collective experiments strongly argue that the EVs remain longer in the early endosomes, suggesting early avoidance of traffic of the EVs to the lysosomes, thus allowing for prolonged TGFβ-1 signaling in the recipient cell.  Supplementary Fig. 10). The distribution of the MSC-derived luciferase activity in the mice was examined at four different time points (10, 30, 60 and 120 mins). The injected MSCs primarily localized to the lung, but administration of EV-treated Luc-GFP-MSCs resulted in relatively higher luciferase signal in lungs compared to the untreated Luc-GFP-MSCs, especially at the early time points (Fig. 5a). Moreover, we detected increased number of EVtreated Luc-GFP-MSCs in the lung parenchyma using fluorescence microscopy, compared to non-treated Luc-GFP-MSC, as seen by the detected GFP signal at 72 h post injection (Fig.   5b). We could also observe that EV-treated MSCs significantly reduced eosinophilia in the bronchiolar alveolar lavage fluid (BALF) in OVA-exposed mice, as compared to mice given the same number of untreated MSCs (Fig. 5c). These results support the hypothesis that mast cell-derived EVs enhance the migratory activity of MSCs to the inflamed lung, which results in an immune-regulatory function resulting in reduced allergic inflammation in vivo.

DISCUSSION:
Human MSCs have the capacity to regulate multiple types of inflammation, and are being tested clinically in multiple inflammatory diseases. We here show that mast cell-derived EVs in a dose-dependent manner induce an enhanced migratory MSC phenotype, and this effect is associated to SMAD2 signaling in the cell. This response is induced by TGFβ-1 present on the surface of the mast cell EVs, as shown by both TGFβ-1 antibody blocking experiments, as well as fluorescent correlation spectroscopy with single-EV resolution. The biological effect of the EV-associated TGFβ-1 is more potent and extended vs. that of free TGFβ-1. We further observed that EVs traffic primarily the endosomal compartment of the recipient cells, and less so to the lysosomal compartment, at a time point when SMAD2 phosphorylation was significant. Lastly, the MSCs treated with mast cell EVs had a greater tendency to home to allergen-exposed mouse lung than untreated MSCs, and this increased homing was associated with reduced eosinophilic inflammation in the lung.
In response to injury or inflammation, MSCs have been reported to provide repair and immune-suppressive functions, which putatively could be further enhanced by any local inflammation (39). We here discovered that mast cell derived EVs change the phenotype of MSCs to a more migratory phenotype, which could alter their functionality at a site of inflammation. The increased migration of the MSCs was associated with increased expression of both matrix metalloproteinase 2 and 6, which also suggests that the cells could become not only migratory, but also more invasive. This was indeed confirmed by a Matrigel invasion assay, in which the MSCs were dose-dependently more migratory. We have previously suggested that bioactive molecules on EV surfaces may influence recipient cells, specifically the stem cell growth factor receptor c-kit, which may enhance recipient lung cancer cell growth (6). In the current study we utilized a stringent EV membrane proteomics approach to identify the low-abundance TGFβ-1 and its receptor (Supplementary Table 1). Validation of the membrane proteomics data, utilizing single-vesicle resolution, confirmed the presence of TGFβ-1 on the EV surfaces (Fig. 2b), which may suggest that this highly bioactive cytokine could mediate immediate biological responses when interacting with a cell, via TGFβ-1 receptor activation, partly explaining the observed change in MSC phenotype. Previous studies have shown that free signaling factors like PDGF, TGF-β, and FGF can induce growth and differentiation of MSCs and change the fate of MSC function, but our current study suggests that the cytokines actually are more potent if associated to the EV membrane. It is possible that this enhanced potency depends on the three dimensional association of the TGFβ -1, facilitating its interaction with its receptor on the recipient cell. . Importantly, blocking EV-associated TGFβ-1 with an antibody was sufficient to significantly reduce TGFβ signaling, further supporting the importance of this bioactive molecule for the observed biological effects in MSCs.
A detailed evaluation of the EV-associated TGF-β-1 suggested that approximately 40% is in active vs 60% inactive forms of TGFβ-1 ( Supplementary Fig. 7). This distribution of TGFβ-1 was different from that seen in a previous study by Webber et al (2010) where they suggested that only ∼2% of the TGFβ-1 from mesothelioma or prostate cancer cell linederived EVs was in active form (40). This further emphasizes that mast cell-derived EVs may have a much higher proportion of active TGFβ-1, than EVs derived from other cells, which potentially could favor mast cell EVs to influence MSC function, putatively to enhance the therapeutic effects of these cells.
Importantly, it has been shown that free TGFβ-1 and EVs from tumor cells (prostate and breast cancer cells) can induce differentiation of MSCs into osteogenic and myogenic lineages respectively (41,42). The effects of free TGFβ-1 in driving differentiation into cellular lineages has been argued to take place in a context-dependent manner, and that other niche-specific may be important to direct the MSC fate (43). This feature of commitment to differentiation is not desirable in the clinical setting if MSCs are being administered to patients, as they could give rise to unwanted mesenchymal or otherwise harmful phenotypes.
Importantly, using linage specific medium for adipocytes and osteocytes, we could demonstrate that MSCs treated with TGF-β-1 positive EVs retain multipotency, despite the observed phenotypic changes induced by the TGF-β-1-EVs. It is important to note that TGFβ-1present on the EVs was not able to induce lineage commitment in the MSCs, despite the previously reported effects of free TGF-β-1, which may suggest that that other cargo components in the EVs such as for example IL-9, insulin and TNF receptor (all which were identified in our membrane proteomics), could have a counteracting role in attenuating an undesirable differentiation of MSCs (40).
In all biological systems free biomolecules such as cytokines interacts with surface receptor on cells, resulting in downstream cellular signaling (44,45). It may be counterintuitive to observe that the low dosages (picograms) of TGFβ-1 on the surface of EVs could contribute so significantly to downstream signaling, for example phosphorylation of SMAD2 (Fig. 3e). It is possible that exosomes present molecules to recipient cells in a very unique context, which could explain the increased potency of EV-associated TGFβ-1 compared to free molecules. For example, it has been shown that EVs enhance the oligomerization of prions proteins (non-toxic PrP c to toxic PrP sc ) by generating high local concentrations of template and PrP c proteins and bringing them to close proximity (44). Thus, emerging evidence suggests that much lower concentrations of bioactive molecules can induce a stronger response when associated to the EVs.
When comparing the effects of EV-associated TGFβ-1 with free TGFβ-1, we could see both enhanced and prolonged effects of the EV-associated cytokine, both in relation to migratory function and pSMAD signaling in MSCs at a 60 minutes time point. It has previously been shown TGFβ-1 alone induces signaling by the activation of the TFGβ-1 receptor, and this complex is rapidly internalized and is transferred to lysosomes for degradation (46,47). The fate of EVs in recipient cells, which is known to involve endosomal translocation (48) (49), could at least partially explain the longer signaling half-life of EVbound TGFβ-1 over free TGFβ-1 observed in the current experiments. To investigate this, we isolated early endosomal and lysosomal compartments, and observed that EVs were preferentially located within the early endosomal compartment (EEA1-enriched), with a small amount of EV proteins in the lysosomal compartment ( Fig. 4d-e) (50) . Therefore, the presence of TGFβ-1 on the surface of EVs seems to allow its prolonged signaling, probably by avoiding degradation in the lysosomes. This could indeed be a novel strategy for long-term signaling by low abundance bioactive molecules, and putatively expression of other bioactive molecules on the surface of EVs could be important pathway by which EVs change recipient cell phenotype.

The increased migratory phenotype of mast cell-derived EV-treated MSCs in vitro,
and that EVs enhance MSC homing towards inflammatory cues in mouse lungs in vivo that leads to a reduction in eosinophil recruitment, a signature of reduced lung inflammation. EVtreated MSCs may thus be useful for treating lung disorders in which inflammation plays a significant role, such as asthma (39), idiopathic pulmonary fibrosis (51), and acute lung injury (52). Previous approaches to treating different diseases with MSCs include pre-treatments of MSCs with cytokines such as TNF-α, IFN-γ, and IL-1β, and these have resulted in some success. However, these approaches have suffered from problems like differentiation, loss of immune-suppressive function, and limited target usage in certain treatments (53). Unlike other methods, MSCs pre-treated with EVs maintain their multi-potency ( Supplementary Fig. 4) and immuno-suppressive effects, which makes them interesting therapeutic candidates in diverse diseases. We here observed an enhanced protective effect of MSCs pre-treated with EVs compared to untreated MSCs in an OVA-induced allergen inflammatory model (54  after the final OVA exposure as described previously (55).

Isolation of EVs
EVs were isolated from conditioned cell medium by differential centrifugation with a filtration step. Briefly, 3-4 days culture medium was centrifuged at 300 × g for 10 minutes to remove cells. The supernatant was further centrifuged at 16,500 × g for 20 minutes. Then the supernatant was centrifugation at 120,000 × g for 3 hours (Type 45 Ti rotor, Beckman Coulter). The pellet was dissolved in PBS and protein concentration was measured by using BCA Protein assay kit (Pierce™, Thermo Fisher Scientific, Waltham, MA, USA).

Reverse migration and invasion assay
The migration capacity and invasiveness of MSCs were evaluated using a 48-well Boyden chamber (Neuroprobe, Gaithersburg, MD, USA). In some experiments, MSCs were preincubated with mast cell-derived EVs for 48 hours before seeding and referred to as EV- hours.

Gelatin zymography
The supernatant from MSCs, cultured with or without mast cell-derived EVs, were collected at 24 and 48 hour and electrophoresed onto zymogram precasted gels containing 10% gelatin (BioRad Laboratories, Hercules, CA, USA) with 5 × non-reducing loading buffers (Sigma Aldrich). Gels were renatured with 2.5% Triton X-100 (Sigma Aldrich) for 1 hour at room temperature and then incubated in development solution (50 mM Tris (pH 7.4), 5 mM CaCl 2 , 200 mM NaCl) at 37°C overnight. Gels were stained (coomassie brilliant blue) and destained (30% methanol and 10% acetic acid) until the clear bands appeared. Finally, the gel was incubated with stop solution (2 % acetic acid). Bands intensity was quantified by using ImageJ software.

Sample digestion and nanoLC-MS analysis of identify membrane proteins on EVs
Proteomic analyses were performed at The Proteomics Core Facility at the Sahlgrenska

Transmission electron microscopy
To describe the nano-structures in EVs derived from primary human mast cells, we performed transmission electron microscopy (TEM). Isolated EVs described in section "Isolation of EVs", were bottom loaded and floated on optiprep density gradient (0, 20,22,24,26,28,30 and 50%) and centrifuged at 182,300 × g for 16 hours (SW40-Ti Rotor) to separate them from free proteins. Nine different optiprep fractions were collected and samples from fraction no.2 was subjected for TEM as described in our previous study.

Labeling of EVs and uptake
EVs was obtained as described in "Isolation of EVs" section and were labeled with PKH67

Fluorescence Correlation Spectroscopy
Freshly isolated EVs from HMC1.

Particle number and size measurements
The size and concentration of EVs were measured using ZetaView ® PMX110 (Particle Metrix, GmbH, Meerbusch, Germany). Each EV sample were diluted in PBS in a range of 1:1000~1:5000 and were injected into the instrument. The chamber temperature was maintained automatically. Measurements were obtained in triplicate and each individual data was obtained from two stationary layers with five times measurement in each layer.
Sensitivity of camera was fixed at 70 in all measurements. Data were analyzed using ZetaView analysis software version 8.2.30.1 with a minimum size of 5 nm, a maximum size of 1000 nm, and a minimum brightness of 20.

Lineage differentiation of MSCs
MSCs treated with 100 μg/ml EVs for 48 hours were cultured in lineage differentiation medium for 15 days as per manufacturer instruction (Human mesenchymal stem cell functional identification kit, SC006, R&D Systems). Lineage specific marker for Adipocytes (FABP4, Oil red O) and Osteocytes (Osteocalcin, Alizarin Red) was probed and analyzed in respective cells as described in previous published guidelines (58).

Localization of EVs and isolation of organelles
In order to locate the biotinylated-EVs in these fractions (LEF and EEF) two complementary approach were used. Firstly, LEF and EEF were both incubated separately with streptavidincoated beads and probed for total bounded protein or EEA1 and LAMP1 antibody (Fig. 4C).
In second approach, LEF and EEF fractions were both incubated separately either with EEA1 or LAMP1-coated beads and probed to detect biotinylated proteins in respective fraction.     (ii).

ACKNOWLEDGEMENT:
We thank the Proteomics Core Facility at Sahlgrenska Academy, Gothenburg University for performing the proteomic analysis. We thank Madeleine Rådinger (Krefting Research Centre,