Extracellular membrane vesicles from umbilical cord blood-derived MSC protect against ischemic acute kidney injury, a feature that is lost after inflammatory conditioning

Background Mesenchymal stromal cells (MSC) are shown to have a great therapeutic potential in many immunological disorders. Currently the therapeutic effect of MSCs is considered to be mediated via paracrine interactions with immune cells. Umbilical cord blood is an attractive but still less studied source of MSCs. We investigated the production of extracellular membrane vesicles (MVs) from human umbilical cord blood derived MSCs (hUCBMSC) in the presence (MVstim) or absence (MVctrl) of inflammatory stimulus. Methods hUCBMSCs were cultured in serum free media with or without IFN-γ and MVs were collected from conditioned media by ultracentrifugation. The protein content of MVs were analyzed by mass spectrometry. Hypoxia induced acute kidney injury rat model was used to analyze the in vivo therapeutic potential of MVs and T-cell proliferation and induction of regulatory T cells were analyzed by co-culture assays. Results Both MVstim and MVctrl showed similar T-cell modulation activity in vitro, but only MVctrls were able to protect rat kidneys from reperfusion injury in vivo. To clarify this difference in functionality we made a comparative mass spectrometric analysis of the MV protein contents. The IFN-γ stimulation induced dramatic changes in the protein content of the MVs. Complement factors (C3, C4A, C5) and lipid binding proteins (i.e apolipoproteins) were only found in the MVctrls, whereas the MVstim contained tetraspanins (CD9, CD63, CD81) and more complete proteasome complex accompanied with MHCI. We further discovered that differently produced MV pools contained specific Rab proteins suggesting that same cells, depending on external signals, produce vesicles originating from different intracellular locations. Conclusions We demonstrate by both in vitro and in vivo models accompanied with a detailed analysis of molecular characteristics that inflammatory conditioning of MSCs influence on the protein content and functional properties of MVs revealing the complexity of the MSC paracrine regulation.

E xtracellular membrane vesicles (MVs) were first identified from platelets as ''platelet dust'' 50 years ago (1) and their importance in platelet function was largely ignored for decades. Currently, secreted MVs are well established and their role in cellular communication has been confirmed by many studies, in which MVs have been shown to transfer surface receptors, mRNA, miRNA and signalling molecules, such as bioactive lipids (2Á6). As suggested by Thery et al. (7), we use the term extracellular MVs to include all secreted cellular MVs such as exosomes, microvesicles and apoptotic bodies.
Mesenchymal stromal cells (MSCs) have been shown to have therapeutic potential in many immunological disorders including graft-versus-host disease (8), Crohn's disease (9) and rheumatoid arthritis (10). An increasing number of in vitro studies have shown that MSCs have the capability to affect both innate and adaptive immune response. Specifically, MSCs are able to directly inhibit T-cell proliferation, change the T-helper lymphocyte balance and induce regulatory T-cells (Tregs) (11Á13). The mechanisms by which MSCs exert their immunomodulatory effects are still largely unknown, but there is an increasing amount of evidence suggesting that the therapeutic effects are mediated by paracrine factors, such as the tryptophan-degrading enzyme indoleamine 2,3-dioxygenase (IDO) and prostaglandin E 2 (PGE 2 ) (11,14,15). Recent reports show that cell-derived extracellular MVs are responsible for at least part of the therapeutic paracrine function of these cells (16Á18).
Ischemia/reperfusion injury (IRI) is a major cause of clinical acute kidney injury (AKI). The pathophysiology consists of alterations in the renal hemodynamics, inflammatory response, kidney endothelial and tubular cell injuries followed by a repair process (19,20). Several studies have shown that T-cells are the key mediators in renal IRI (21). However, it has previously been shown that immunosuppressive Tregs mediate at least in part, the renoprotective effects of ischemic preconditioning (22) and participate in the repair of kidney IRI (23). MSCs as such are described to be a new therapeutic tool for AKI (24). In addition to MSCs, MVs secreted by MSCs are also shown to have a protective effect against the ischemia-induced AKI, mediated by transferred mRNA (6,17,18,25). In the site of the inflammation in injured tissues, there is an abundant amount of different cytokine molecules. The surrounding microenvironment plays a major role in regulating the MSC immunoregulatory function. Several studies have demonstrated that cytokines such as interferon-gamma (IFN-g) and tumour necrosis factor alpha (TNF-a) regulate the production of MSC-derived soluble factors (26Á29). Moreover, priming with cytokines has been suggested to be essential for both in vitro and in vivo immunomodulatory activity of MSCs (30Á32).
It has been reported that MSC origin affects their immunomodulatory properties (33Á35) and response to the cytokine stimulus (36). In the present study, our aim was to examine the effect of IFN-g stimulus on the secretion and function of MVs originating from umbilical cord blood-derived MSCs. The protein content of the produced MVs was analyzed in relation to their therapeutic effect in ischemia-induced AKI in vivo model. Furthermore, the ability of MVs to induce regulatory T-cells and suppress the proliferation of T-cells was assessed.

Materials and methods
Cord blood derived mesenchymal stromal cells Cord blood units were collected at the Helsinki University Central Hospital, Department of Obstetrics and Gynaecology, and Helsinki Maternity Hospital. All donors gave informed consent and the ethical review board of Helsinki University Central Hospital and the Finnish Red Cross Blood Service approved the study protocol. Human umbilical cord blood-derived MSC (hUCBMSCs) were established with protocol designed in our lab and described elsewhere (37). Briefly, cord blood mononuclear cells were isolated by Ficoll-Paque Plus (GE Healthcare, Uppsala, Sweden) gradient centrifugation. 1 )10 6 /cm 2 mononuclear cells were plated on fibronectin-coated (Sigma-Aldrich, St. Louis, MO) tissue culture plates (Nunc, Roskilde, Denmark) in Minimum Essential Alpha-Medium (a-MEM) with Glutamax TM (Gibco, Paisley, UK), supplemented with 10% foetal bovine serum (Gibco), 10 ng/mL epidermal growth factor (EGF, Sigma-Aldrich), 10 ng/mL recombinant plateletderived growth factor (rhPDGF-BB R&D Systems, Minneapolis, MN), 50 nM dexamethasone (Sigma-Aldrich), 100 U/mL penicillin (Gibco) and 100 mg/mL streptomycin (Gibco). The cells were allowed to adhere overnight, and non-adherent cells were washed with medium changes. The initial hUCBMSC establishment was performed under hypoxic conditions (5% CO 2 , 3% O 2 at 378C). For further experiments, hUCBMSCs were cultured in normoxic conditions (5% CO 2 and 20% O 2 at 378C) and proliferation media was renewed twice a week. The expression of characteristic human MSC markers and adipogenic, chondrogenic and osteogenic differentiation capacity was analyzed (38). See Supplementary file for detailed methods.
To collect MVs, the conditioned medium was first centrifuged at 2,000)g for 20 min at '48C. The supernatant was further ultracentrifuged at 100,000 )g for 1Á2 hours at '48C (Sorwall WX Ultra 80, Thermo Fisher Scientific Inc., Waltham, MA), washed with phosphatebuffered saline (PBS) and submitted to a second 1Á2 hours ultracentrifugation under the same conditions. MVs were suspended in a small volume of PBS and stored at (808C. The protein concentration was measured with NanoDrop (Thermo Fisher Scientific Inc.) or with Pierce BCA Protein Assay kit (Thermo Fisher Scientific Inc.).
Electron microscopy MV specimen preparation for scanning electron microscopy (SEM) and for negative staining was done by the personnel of the Electron Microscopy Unit, Institute of Biotechnology, and University of Helsinki. SEM was performed with FEI Quanta 250 Field Emission Gun Scanning Electron Microscope and negative stainings with FEI Tecnai 12 Transmission Electron Microscope (Philips Electron Optics, Holland).

Nanoparticle tracking analysis
After centrifugation, MVs were resuspended with PBS, and analyzed with nanoparticle tracking analysis (NTA) instrument LM14C with blue laser (405 nm, 60 mW, NanoSight Technology, London, UK) and CMOS camera (Hamamatsu Photonics K.K., Hamamatsu City, Japan) to determine the vesicle size distributions and concentrations. Samples were injected manually and data acquisition was done at ambient temperature. Samples were run as triplicates. Settings for data acquisition were: basic, camera level 13, autosettings off, polydispersity and reproducibility high with particles per image 40Á100 (acquisition time 90 seconds). Calibration was done by using 0.2-mm Fluoresbrite † Multifluorescent Microspheres (Polysciences, Inc., Warrington, PA). Data were analyzed with NTA 2.3 software with settings expert, background extraction/auto blur/auto-minimum track length on and minimum expected particle size 50 nm.

MS analysis
Gel-based proteome analysis of MVs was performed with the liquid chromatographyÁmass spectrometry (LC-MS) of tryptic peptides. Proteins from collected MVs were run in a 4Á20% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gel (Bio-Rad Laboratories, Hercules, CA). The gel was silver stained (39) and sliced into pieces. Each gel piece was processed using an in-gel reduction, alkylation and trypsin digestion protocol as previously described (40). Peptides were loaded to precolumn (Protecol Guard C18, 150 mm, 10 mm, 3 mm; SGE Analytical Science Pty Ltd, Melbourne, Australia) and separated in a reversed-phase analytical column (Acclaim PepMap100 C18, 75 mm, 150 mm, 3 mm; Thermo Fisher Scientific Inc.) with a linear gradient of acetonitrile. Ultimate 3000 LC instrument (Thermo Fisher Scientific Inc.) was operated in nano scale with the flow rate of 0.3 mL/min. Eluted peptides were introduced to an LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific Inc.) via ESI Chip interface (Advion BioSciences Inc., Ithaca, NY) in a positive ion mode. Data files from MS were processed with Mascot Distiller (Matrix Science Ltd., London, UK, version 2.3.2). The processed data were searched with Mascot Server (Matrix Science Ltd., version 2.3) against human proteins in UniProtKB database (release 2012_08). Subcellular locations for identified proteins were collected from UniProt database (as is 24.9.2012). Gene ontology (GO) enrichment analysis of identified proteins was performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID, version 6.7) (41) using default settings Co-culture assays Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats from healthy, anonymous blood donors (Finnish Red Cross Blood Service) by density gradient centrifugation (Ficoll-Pague plus, GE Healthcare) and cryo-preserved for later use. hUCBMSCs or MVs were co-cultured with 1.5 )10 6 CFSE [5(6)-carboxyfluorescein diacetate N-succinimidyl ester] solution (Molecular Probes, Eugene, OR) labelled PBMCs. After the collection of medium, 1.5 )10 5 trypsinized hUCBMSCs were resuspended in RPMI growth medium (RPMI, 5% FBS, 100 U/ mL penicillin and 100 mg/mL streptomycin) and plated (in triplicates) on a 48-well plate. To activate the T-cell proliferation, 100 ng/mL of the antihuman CD3 antibody clone Hit3a (BioLegend, San Diego, CA) was added to the co-culture. MVs were resuspended in RPMI medium and added in 50 mL in volume (in triplicates) into the co-culture just before adding the stimulant. T-cell proliferation was recorded after 4 days of incubation as a dilution of fluorescent dye by flow cytometry (FACSAria, BD) and data were analyzed using the FlowJo (7.6.5, Treestar, Asland, OR).
To analyze the effect of hUCBMSCs and MVs on Treg induction, PBMCs were co-cultured with MVs or hUCBMSCs. After 7 days of incubation, non-adherent PBMCs were harvested and labelled with fluorogenic antibodies, CD4-APC-Cy7 (Biolegend), CD25-AlexaFluor † 647 (Biolegend), FOXP3-PerCP5.5 (eBioscience, San Diego, CA) to evaluate the proportion of CD4 ' CD25 ' FOXP3 ' Tregs. Cells were fixed and permeabilized with FOXP3 Staining Buffer Set (eBioscience) according to manufacturer's instructions and including the blocking step with 2% rat serum (eBioscience). PBMCs cultured without MSCs or MVs were used as a control. Appropriate isotype controls were used. Flow cytometry was performed with FACSAria (BD) and data were analyzed using the FlowJo 7.6.5 software (Treestar).

Kidney ischemia-reperfusion (I/R) injury model
Eighteen 6-to 7-week-old male Sprague Dawley rats were purchased from Charles River Laboratories (Research Models and Services, Sulzfeld, Germany). The protocols were approved by the Animal Experimentation Committee of the University of Helsinki, Finland, and the Provincial State Office of Southern Finland (approval number STH059A), whose standards correspond to those of the American Physiological Society. The rats were kept under 12-hour light/12-hour dark cycle and were divided into 3 groups: (1) I/R group (n 08), (2) I/R group' control MV (n 05), (3) I/R group'IFN-g-treated MV (n05). An established model of kidney I/R injury was used (42). The rats were anesthetized with isoflurane (anaesthesia induction in a chamber with 4Á5% isoflurane at the flow rate of 1.5 L/min), intubated, and 1.5% isoflurane at the rate of 1.5 L/min was used to maintain anaesthesia. Abdominal incisions were made and the renal pedicles were bluntly dissected. Bilateral renal ischemia was induced by clamping renal pedicles for 40 minutes with microvascular clamps. After reperfusion, MVs dissolved in 0.5 mL PBS were infused into the rats via the left carotid artery. Controls received the same amount of vehicle. The rats were hydrated with warm saline during the operation and the body temperature was maintained constantly at 378C by using a heating pad until the rats were awake. The wounds were sutured after removing the clips, and the animals were allowed to recover. Buprenorphine (0.1 mg/kg s.c.) was used as postoperative analgesia. Twenty-four hours after the operation, blood samples were collected from the tail vein under short isoflurane anaesthesia. Terminal samples were harvested 48 hours after the operation; the rats were anesthetized with isoflurane, and blood samples were collected from the inferior vena cava with a 5 mL syringe and 22G needle for biochemical measurements. The kidneys were excised, washed with ice-cold saline, blotted dry and weighed. Left kidney was used for histological examinations. Tissue samples for histology were fixed in 10% formaline and processed to paraffin with routine methodology.

Kidney histology
For histological examination, 4-mm thick paraffin sections were cut and stained with hematoxylin-eosin (n 09Á10). Renal samples were visually examined by a pathologist (P. F.) with Leica DMR microscope (Leica Microsystems AG, Heerbrugg, Switzerland) and morphological changes from the whole cross-sectional area of cortex and medulla were assessed according to the acute tubular necrosis (ATN)scoring system (42) (Magnification )200, ]20 fields per kidney section quantified using the ATN-scoring system). The evaluation of histopathological changes included the loss of tubular brush border, tubular dilatation, cast formation and cell lysis. Tissue damage was quantified in a blinded manner and scored according to the percentage of damaged tubules in the sample: 0, no damage; 1, less than 12.5% damage; 2, 12.5Á25% damage; 3, 25Á50% damage; 4, 50Á75% damage; and 5, more than 75% damage.

Biochemical determinations
Serum creatinine, electrolytes, lipids and liver enzymes were measured by routine laboratory techniques (ADVIA 1650 Chemistry System, Siemens Healthcare Diagnostics Inc., Deerfield, IL).

Statistical analysis
Data are presented as the mean9standard error of the mean (SEM). Statistically significant differences in mean values were tested by one-way analysis of variance (ANOVA) and the Tukey's post-hoc test using statistical programming software R (version 2.13.1). The differences were considered significant when p B0.05.

Results
hUCBMSCs produce MVs which increase the percentage of CD25 ' FOXP3 ' T-cells and suppress T-cell proliferation hUCBMSCs were induced to produce MVs by culturing them in serum-free conditions with (MVstim) or without (MVctrl) IFN-g-stimulation for 24Á48 hours. Cell viability in both culture conditions was always !95% (n010). Electron microscopy analysis of the vesicles showed variation in size, with the smallest being around 20 nm and the largest !500 nm ( Fig. 1A and B). NTA confirmed the wide size range seen by electron microscopy analysis (Fig. 1C). Also, the presence of very small vesicles (B50 nm) was confirmed. There was no significant difference in the size distribution between IFN-g stimulated and control conditions (Fig. 1D). Based on the analysis of protein amount, IFN-g stimulus gave only slightly better MV yield (data not shown). The protein yield was on average 16 mg of protein in a vesicle fraction for every 10 )10 6 cells after 24-h production. When hUCBMSCs or MVs derived from hUCBMSCs were co-cultured with PBMCs for 7 days, we were able to show that MVs alone as well as MSCs were able to induce the formation of T-cells with Treg phenotype (CD4 ' CD25 ' FOXP3 ' ). The percentage of Tregs was increased from 3.3% to 4.5% when PBMCs were cocultured with MSCs (Fig. 1E). However, we were not able to see any remarkable differences in Treg induction when PBMCs were co-cultured with either MVctrl (4.5%) or MVstim (4.0%).
Next, we analyzed the ability of hUCBMSCs or the MVs to suppress T-cell proliferation. CFSE labelled PBMCs were cultured with MSCs or MVs. T-cells were activated with antiCD3 monoclonal antibody and the proliferation was analyzed by flow cytometer after 4 days of incubation. We show that both hUCBMSCs and UCBMSC MVs were able to suppress the proliferation of stimulated T-cells, but the MVs alone suppressed T-cell proliferation to a lesser extent (Fig. 1F).
Protein content of MVs was markedly different after IFN-g stimulation To analyze the protein content of the MVs produced by hUCBMSCs, we performed a gel-based MS analysis of the proteome of secreted MVs with or without IFNg stimulation. As demonstrated in a Venn diagram ( Fig. 2A), 246 of the proteins were common for both groups, 220 proteins were found only in MVctrl and 448 proteins in MVstim.
MVs contained proteins which originated from different cellular compartments as illustrated in Fig. 2B. Roughly 25% of identified proteins were cytoplasmic, 15% were membrane proteins, and 10% were from the nucleus. Proteins located in endosomes, endoplasmic reticulum and mitochondrion were also identified (Fig. 2B). Altogether, 5Á8% of all identified proteins were classified as secreted proteins. The distribution of subcellular localizations of identified proteins from MVctrl and MVstim had only minor differences.
The systematic functional GO enrichment analysis was performed by using the DAVID software. Common proteins were excluded from the analysis. Enriched GO biological processes (GO-BP) for MVctrl contained a response to wounding, lipid transport and acute inflammatory response (Fig. 3A). Enriched GO-BP terms unique for MVstim included homeostatic processes, chromatin organization and cell motion (Fig. 3B). Even though proteins subjected to the enrichment analysis were different in both groups, the same GO terms, such as intracellular signalling cascade, regulation of apoptosis and homeostatic processes, were enriched in both groups. This phenomenon was also seen when we analyzed GO molecular functions (GO-MF) (Fig. 1C). GO-MF terms, such as nucleotide binding, GTPase activity and unfolded protein binding, were enriched in both MVs studied.
The detailed MS analysis revealed several differences in the protein content of MVctrl and MVstim. The result of MS analysis was compared to the most often reported proteins from exosomes [Exocarta database as is 16. 4. 2013 (43)]. Altogether, 22 proteins of the top 25 proteins were identified from MVstim compared with only 13 proteins from MVctrl.
According to MS analysis, MVstim ( Fig. 4B and C) contained tetraspanins (CD9, CD63 and CD81), which have previously been used as exosome markers (6,18). Interestingly, MVstim contained also the HLA-A (MHCI) molecule and both a and b units of the proteasome complex ( Fig. 4B and C) required for the antigen presentation and activation of T-cells. The MS detection of HLA molecules was supported by flowcytometric analysis on hUCBMSCs. When hUCBMSCs were stimulated with IFN-g for 24 hours, the expression of HLA-I was clearly induced. In contrast, the HLA-II expression was induced only after 48 hours of stimulation (Supplementary file).
In general, both MVstim and MVctrl (Fig. 4C) contained proteins involved in T-cell regulation (Galectin-1 and -3) and the complement pathway (CD59). MSC markers CD73, CD90 were found in both MVs. Proteins identified only with 1 peptide were excluded from the analysis, including MSC marker CD105. On a    protein-family level, several apoptosis regulating proteins (caspases, annexins and histones), cytoskeleton proteins, adhesion proteins (integrins), vesicle trafficking proteins (Ras-related proteins, Rabs and ADP ribosylating factors, ARFs) and ribosome subunits were identified in both groups. Interestingly, the individual proteins were largely different in MVctrl in comparison to MVstim (Fig. 4B). In particular, the number of histones and ribosome subunits was increased in stimulated MVs (Fig. 4C).
Effects of MV on kidney function and morphology in I/R injury Acute kidney I/R injury was associated with an 8.4-fold increase in serum creatinine concentration as compared to pre-operative baseline values (Fig. 5A). The concentrations of serum urea also had an 8-fold increase (Fig. 5B) and liver enzyme aspartate aminotransferase (ASAT) had a 6.6-fold increase (Fig. 5C) and gamma glutamyltransferase (g-GT) had an 18.9-fold increase ( Fig. 5D) in 24 hours after operation. Histopathological analysis of the kidneys harvested 48 hours after I/R showed marked injury of the renal parenchyma comprising vast necrosis of the tubuloepithelial cells, tubular dilatation and cast formation ( Fig. 6A and D). MVctrl treatment ameliorated I/R injury-induced kidney dysfunction (Fig. 5) and ATN (Fig. 6B and D), whereas MVstim did not influence the development of kidney I/R injury (Fig. 5 and 6C and D).

Discussion
The therapeutic potential of MSCs is currently explained by their ability to regulate immune cells through paracrine interactions. The key players in immune response and immunological diseases are T-cells. It has been established that MSCs suppress T lymphocyte activation and proliferation and alter the T-cell population towards an anti-inflammatory profile by increasing the number of regulatory T-cells (44Á47) and repressing the differentiation of Th17 cells (48). Inflammatory cytokines, such as IFN-g have been shown to induce the production of IDO and PGE 2 , the two best described mediators of MSC immunomodulation (49,50  are capable of modulating T-cell functions and repairing injured tissues (17,51Á54). Although the effects of cytokine stimulus have been in the focus of several studies performed to elucidate the molecular mechanisms of MSCs, the effects of IFN-g on the MV production have been completely overlooked.
In the present study, we show that hUCBMSCs secrete constitutively extracellular membrane vesicles (MVctrl) capable of significantly attenuating ischemia/reperfusion kidney injury in rats. When MVctrl were administered immediately after reperfusion, the drastic increases in serum creatinine and urea levels caused by hypoxia and concomitant reperfusion were significantly decreased. The protective effect of MVctrl was also seen in the histopathological analysis of the kidneys after 48 hours, where vast necrosis of the tubuloepithelial cells, tubular dilatation and cast formation was mostly absent in MVctrl-treated rats. Our findings are thus in very good agreement with previous studies (16,17,25,55,56) demonstrating the renoprotective effects of MV during the early phase of AKI. Our results are also in line with a recent paper by Zhou et al., where MVs derived from umbilical cord stem cells protected kidneys from cisplatin-induced injury (18). To investigate the consequence of the inflammatory signals in MV production, we stimulated hUCBMSCs with IFN-g and compared the therapeutic effect of the MVs produced by the stimulus to the constitutively produced MVs. In comparison to MVctrls, the MVs produced after IFN-g stimulation (MVstim) showed no protective potential, which was indicated by the similar elevation of serum creatinine and urea levels to those seen in PBS control rats and further demonstrated by the histopathological analysis. Such a difference was not seen in vitro potency assays since the co-culture of PBMCs with both MVs gave equal results; both MVstims and MVctrls were able to deliver immunosuppressive effects by enhancing the proliferation of regulatory CD4 ' CD25 ' FOXP3 ' T-cell population (Fig. 1E), a phenomenon also shown by recent studies on bone marrow and cord blood-derived mice stromal cells (53).
In the AKI, activated immune cells accumulating in the kidney can either repair or further aggravate the injury (57). The number of activated T-cells and antiinflammatory Tregs correlates with the severity of the tissue damage after hypoxia/reperfusion. A decreased number of Tregs leads to an increase in cell death, infiltration of inflammatory cells and a reduction in kidney function (20). Preclinical studies demonstrate that MSCs can either prevent or reverse the kidney injury (58Á64). Both MSCs and MVs derived from them have shown to protect kidneys by promoting epithelial cell survival in the damaged tissue (19,55,56). Gandolfo and coworkers (23) demonstrated recently using a murine model of kidney I/R injury that Tregs infiltrate into the kidneys starting at 3 days after the initial ischemic injury. This finding suggests that the anti-inflammatory Tregs promote repair during the healing process in the late phase of AKI rather than acting in the acute phase. Innate immune cells including neutrophils, dendritic cells, monocytes/macrophages and natural killer (NK) cells are responsible for the early response to injury. MSCs are known to modulate the innate immunity by affecting the differentiation of dendritic cells and polarization of M1 macrophages to anti-inflammatory M2 macrophages. MSCs also inhibit the proliferation of NK cells (65). Our findings thus support the notion that the protective effects of unstimulated MVs against kidney I/R injury in the early phase are likely to mediated through innate immune system instead of Tregs-dependent mechanisms.
The clear discrepancy between in vivo and in vitro results encouraged us to analyze the protein content of MVs for possible explanations. The MS analysis verified the presence of several previously described proteins for extracellular vesicles. Common proteins for both MVctrls and MVstims included several cytoskeletal proteins, ribosomal subunits and integrins. As previously described by others, the therapeutic effect of MVs can be mediated by the functional mRNAs and miRNAs (17,55,56). We found 34 common ribosomal proteins in both MV pools; in addition to the common proteins, both pools contained several unique ribosomal proteins. The additional common proteins were functionally interesting, such as Galectin-1 and -3, CD90 and CD73. Moreover, both galectin-1 and -3 have proved to be mediators of MSC T-cell immunosuppression (66Á68). Galectin-1 is an important regulator of Tregs (69), whereas galectin-3 has been reported to regulate T-cell function by participating in the cell surface receptor activity (66,67). Two mesenchymal stromal/stem cell markers, CD73 and CD90 have also been associated with the immunosuppressive capacity (70,71). The expression of CD73 and its ectonuclease activity producing adenosine have been proposed to be important in the functionality of Tregs (72) and recently the role of adenosine signalling was demonstrated to have a beneficial effect in AKI, as it was shown that the pre-treatment of dendritic cells with adenosine receptor 2 (A2R) agonist protected kidneys from ischemia/reperfusion induced injury in mice (73). The role of CD73 in immune regulation has been further demonstrated in mice deficient in CD73 with more severe graft-versus-host disease than the wild-type controls (74). There is on-going scientific debate over the cellular origin, characteristics and nomenclature of extracellular vesicles. In this study, we made a conscious decision not to separate the different sizes or densities MVs by ultracentrifugation or filtration processes. Instead, we wanted to collect the whole pool of secreted MVs from the supernatant or hUCBMSCs culture milieu and study the effect of the different external signals on the composition of the MVs. One of the most interesting findings in our study was the distinct set of Rab proteins found in the different MV pools produced. It is well established that Rab GTPases, members of the Ras superfamily, are important regulators of membrane transport, vesicle formation, movement and fusion and can be considered as labels defining vesicle routing (75), with also their role in extracellular vesicle formation and transport being studied (76,77). This superfamily is known to consist of at least 60 members in humans (78). Altogether 24 Rab proteins were identified in our study. The MS analysis revealed that IFN-g induced production of MVs richer in Rab proteins. Altogether 19 Rab proteins were detected in MVstim, 17 of which were unique. Instead, in MVctrls, only 7 Rab proteins were identified, 5 of which were unique. The Rab proteins found in MVstims were distinctive for exocytosis and deeper endosomal recycling, extending from ER all through the exocytotic pathway (Rab 1, 2, 6 and 8) as illustrated in Fig. 7. In contrast, MVctrls contained Rab proteins more restricted to the early endosomal recycling near plasma membrane (such as Rab 3, 5, 14 and 34). Our interpretation is that this strikingly different and specific subcellular localization of the Rab proteins indicates that for the first time, we have been able to make a distinction between two separate routes of extracellular vesicles from MSCs: cytokineinduced deeper loop and constantly produced rapid loop of MVs. The presence of MHCI molecules in MVstims accompanied by the more complete proteasome complex than in MVctrls further proves the effective IFN-g activation and the induction of the delivery of MHCI molecules to of the cell surface and consequently into MVstims. Already in 2003, Willem Stoorvogel's group described proteomic analysis of B-cell-derived exosomes (82). They demonstrated that B-cell-derived exosomes contained major histocompatibility Class I and II (MHCI and II) as major components in addition to integrins and many other proteins also seen in MSC-derived MVs. On the contrary to B-cells, MSCs do not express MHCII molecules on their cell surface. Furthermore, MHCI molecules are expressed at low levels. Our results show that after 24 hours of IFN-g activation of hUCBMSCs, the surface expression of MHCI is increased but not MHCII for which 48 hours of activation is required. Interestingly, we found that MSC-derived MVctrls do not contain either of these molecules but after IFN-g activation, MHCI and also a complete set of proteasome complex were detected in MVstims.
The presence of the tetraspanin complex (CD9, CD63 and CD81), previously described as a marker of exosomes, in MVstims possibly indicates that this pool of MVs is enriched with extracellular vesicles previously described as exosomes. The functional consequence of the self-antigen presentation (MHCI) in MVstims may entirely or partly explain the loss of function in MVstims in AKI model. It is possible that the presence of the MHCI on the surface of MVstims triggers the innate immune response in rats and although otherwise functional (as seen in human-cell-based co-culture assays), the MVstims might be cleared faster from the circulation and never reach their therapeutic target in adequate numbers. Another explanation might be found in the different content of MVs. We did discover unique and plausible candidates in MVctrls that could explain their enhanced therapeutic potential in vivo. Only the MVctrls contained complement activation-related proteins CD59, C5, C3 and C4A, which on their own or via their degradation products are key molecules in inflammatory responses and the initiation of the classical complement pathway (83), which as such is speculated to be important in the clearance of apoptotic and necrotic cells by phagocytic macrophages (84,85) in damaged tissue.
Further MVctrls contained anti-inflammatory and anti-oxidative apolipoprotein A-I (ApoA1). ApoA1 or high-density lipoproteins (HDL), which major apoliprotein is apoA1, has been shown to protect vascular endothelium in many pathological conditions including atherosclerosis and stroke (86). Recently, the ApoA1 mimetic compound has been reported to have therapeutic effect in kidney injury (87). Furthermore, other apolipoproteins (ApoA2, ApoA4 and ApoC3) and lipid-binding proteins (SCP2 and FABP6) were present only in MVctrls, indicating their likely protective role in AKI. Since it is well known that different apoliproteins and their corresponding lipoprotein particles, containing proteins and lipids, are important in regulation of inflammation and tissue repair, the presence of apolipoproteins only in in vivo functional MVctrls may be of special importance.

Conclusions
Our study compares for the first time the T-cell regulation, therapeutic effect and protein content of differently produced MVs from the same MSC population. The results demonstrate that different external conditions lead to the production of MVs, which originate from very distinct internal vesicle routes, have completely different therapeutic potential and contain unique proteins expected to play a critical role in their function. As MSCs have diverse immunological functions, our results indicate that membrane trafficking routes previously considered as intracellular might actually be an efficient way for the cell to separate diverse extracellular signals and secrete the specific responses as MVs. Further studies are needed to fully elucidate the functional role of the numerous MV proteins identified in this study.