All experiments were performed in compliance with the Guide for the Care and Use of Laboratory Animals, French regulations for animal experiments (Ministry of Agriculture Order No. B92-032-01, 2006) and European Directives (86/609/CEE) and approved by the Institute’s local ethics committee (permit number: D92-032-01, APAFIS#6503-2016082311257373v2).
In a first batch of the experiment, 4 groups of mice were submitted to different doses (15, 13, 12 and 10 Gy) of whole-body irradiation (WBI) and one group was not irradiated (control group). Experiments were performed to assess the survival rate of mice after WBI (between 15 and 31 animals per groups). Three days after WBI, intestinal barrier integrity was also studied by measuring in vivo intestinal permeability (between 10 and 20 animals per groups) and assessing the histological data (between 12 and 18 animals per group). In a second batch of the experiment, 3 groups of mice were submitted to 10 Gy WBI and one group was not irradiated (control group). The irradiated mice were then treated with PBS, MSC or MSC-derived EVs and the non-irradiated mice received only PBS. All mice received the treatment intravenously, through the retro-orbital sinus, 6h, 24h and 48h after irradiation for MSC-derived EVs (200mg for each injection) and 6h after irradiation for MSC (5 millions in one injection). Like the first batch of the experiment, three days after WBI, intestinal barrier integrity was also studied by measuring in vivo intestinal permeability (between 6 and 36 animals per group) and assessing the histological data (between 18 and 24 animals per group). Experiments to assess the survival rate of mice after 10 Gy WBI with or without therapy (MSC or MSC-derived EVs) were also performed (between 11 and 19 animals per group).
In a third batch of the experiment, 3 groups of mice were submitted to 10 Gy WBI and one group was not irradiated (control group). The mice received treatment as described above in the second batch of the experiment. Mice were then sacrificed 24h after irradiation (at this time the mice received a single injection of MSC or a single injection of MSC-derived EVs), 48h after irradiation (at this time the mice received a single injection of MSC or 2 injections of MSC-derived EVs) or 72h after irradiation (at this time the mice received a single injection of MSC or 3 injections of MSC-derived EVs) as shown in figure 3. In this experiment, the apoptosis and proliferation of epithelial crypt cells were assessed on small intestinal sections by TUNEL assay or immunostaining by KI67 antibody respectively. Between 4 and 24 animals per group was used, the two first batch of experiments were realized three times, the third one only one time.
Male NUDE mice (Janvier SA, Le Genest St Isle, France) 6/8 weeks old were received and housed in a temperature-controlled room (21 ± 1°C). They were allowed free access to water and fed standard pellets. The mice were anesthetized by a 2:1 (v/v) ketamine and xylazine mixture diluted in 0.9% NaCl and injected at 0.1 mL/g, and a single WBI dose was delivered by a medical accelerator (Alphée). Alphée is an accelerator-type radiation source (maximum energy 4 MeV with an average energy of about 1.5 MeV; 30 kA). The doses used were 15, 13, 12 and 10 Gy.
Preparation and administration of human MSC
BM cells were obtained after receiving the informed consent of patients undergoing total hip replacement surgery and these were used in accordance with the procedures approved by the human experimentation and ethics committees of Hospital St-Antoine (France). Between 10 and 20 ml of BM cells were harvested in a-MEM (Invitrogen, Cergy, France) supplemented with heparin. Total cells were then isolated from any bone fragments. Nucleated cells were plated at 50000 cells/cm2 in a-MEM supplemented with 10% foetal calf serum (research grade FCS, Hyclone, Perbio, France), 1% L-glutamine, 1% penicillin streptomycin and 1 ng/ml b-FGF (Sigma-Aldrich Chimie SARL, Lyon, France) as used in clinics .
Culture ﬂasks were incubated at 37°C with 5% CO2 in a humidiﬁed atmosphere. After 72h, noncompliant cells were removed, and the medium was replaced twice a week until 90% conﬂuence was reached. Samples of MSC from different donors were collected at passage 2 for transplantation.
At the time of infusion, the MSC were characterized by their expression of CD73 (SH3) and CD105 (SH2) and the lack of their expression of CD45 using FACS analysis and by their potential for osteogenic and adipogenic differentiation 
Five millions of human MSC were intravenously administered in one injection 6h after 10 Gy WBI. The controls received only the vehicle.
Preparation and administration of MSC-derived EVs
All isolations and characterizations were performed as previously described [40, 41], but with some modifications. Briefly, immortalized E1-MYC 16.3 human embryonic stem cell-derived MSCs were cultured in DMEM (GE Healthcare, USA) with 10% foetal bovine serum (FBS) (Thermofisher Scientific, Waltham, MA, USA). To obtain the EVs, 80% confluent cells were grown in a chemically defined medium for 3 days and the conditioned medium was harvested as previously described (see PMID 17565974). The conditioned medium was cleared of cell debris, fractionated and concentrated 50X by tangential flow filtration using a membrane with a molecular weight cut-off (MWCO) of 100kDa (Sartorius, Gottingen, Germany). Its EV yield was assayed by protein concentration using a NanoOrange Protein Quantification Kit (Thermofisher Scientific). Each batch of EV preparation was qualified for its particle size distribution (See below: Nanoparticle tracking analysis in supplementary data 1) and presence of exosome-associated markers (see Transmission electron microscopy in Supplementary data 1).
The MSC-derived EVs were lyophilized by Paracrine Therapeutics using a proprietary technique, stored at −20°C and re-constituted at a concentration of 1mg/ml with water for use. A total of 600 µg of MSC-derived EVs was intravenously administered in three 200 µg injections 6h, 24h and 48h after receiving 10 Gy WBI. The controls received only the vehicle.
Nanoparticle tracking analysis
The Exosome was diluted 3000x with 0.22 µm filtered PBS. The exosome size distribution was then measured and analyzed by Zetaview® (Particle Metrix GmbH, Meerbusch, Germany) according to the manufacturer’s protocol.
Transmission on electron microscopy
Glow-discharged EM grids coated with Formvar/Carbon (EMS) were floated on a 20-µL drop of purified exosome fraction. The excess fluid was blotted away using filter paper and the exosomes adhering to the grid surface were immune-labeled with mouse monoclonal anti-CD81 antibody (Santa Cruz Sc-7637) followed by goat anti-mouse secondary antibody coupled with 6 nm gold (EMS). Finally, the grids were fixed with 1% glutaraldehyde (EMS), washed and embedded in a thin film of uranyl acetate-methylcellulose (a 4% uranyl acetate and 2% methylcellulose mixture in a 1:9 ratio) using the wire loop technique. Samples were analyzed under a JEOL transmission electron microscope (JEM-1010) operating at 80 kV and equipped with an SIA model 12C 4K CCD camera.
Survival curve analysis
NUDE mice were exposed to 15 Gy, 13 Gy, 12 Gy or 10 Gy lethal doses of WBI. The therapeutic effect of MSC-derived EVs was only assessed in NUDE mice subjected to 10 Gy WBI. Animal survival was monitored every 12 h.
Formalin-fixed, paraffin-embedded small intestines were cut at 5 mm on a rotary microtome (Leica Microsystems AG, Wetzlar, Germany) and mounted on polysine slides. The sections were deparaffinized in xylene and rehydrated using ethanol baths and PBS. The hydrated sections were then stained with hematoxylin, eosin and saffron (HES). The sections were studied for histological changes in the mucosa of the small intestine and morphometric analyses were performed. We evaluated surviving crypts as a percentage of crypt containing 10 or more adjacent chromophilic cells and a lumen. We also assessed the villus height of the small intestine (in μm). For each section of the small intestine between 30 and 100 measurements (depending on the severity of the lesion) were performed using image analysis software (Visiolab, Biocom, France).
The hydrated sections were dipped into a permeabilization solution consisting of 0.1% Triton X-100 in PBS and rinsed in a distilled water bath for 5 min. Endogenous enzymes were then blocked using 3% hydrogen peroxide (H2O2) in methanol for 10 min and washed again in a 50 mM Tris buffer containing 9 g/L NaCl (TBS). To expose masked epitopes, the tissues were incubated for 30 min in 10 mmol/L buffered citrate, (pH6.0). Non-specific antibody binding was minimized by incubating the sections with a protein-block solution (DakoCytomation X0909, DakoCytomation, Courtaboeuf, France) for 30 min. The tissues were incubated in the presence of the primary antibody, Ki67 polyclonal rabbit anti-rat antibody (Abcam ab66155) or claudin-3 polyclonal rabbit anti-rat (Thermofisher, PA-16867) at a dilution of 1:1000 and 1:100, respectively, in Dako antibody diluent for 60 min at 37°C in a humidified chamber. The sections were then rinsed in PBS buffer and then incubated with Envision kit anti-mouse HRP (K4002, DakoCytomation) for 30 min at RT. Staining was developed with Histogreen substrate (E109, Eurobio, Les Ulis, France). The sections were then counterstained with Nuclear Fast Red (H3403, VectorLabs, Burlingame, CA, USA), dehydrated and mounted. Isotype control antibodies were used as negative controls. Proliferating cells were determined using image analysis software (Visiolab). A minimum of 30 crypts per section was measured. For each crypt, we evaluated the number of KI67-positive epithelial cells and expressed the results as the percentage of positive cells per crypt.
The hydrated sections were incubated in a citrate buffer (pH 6) for 3 three 5 min cycles in a 600W microwave oven, with 3 min between each followed by 5 min in tap water. The slides were saturated for 30 min with PBS BSA 1%, rinsed in PBS three times and stained with the In-Situ Cell Death Detection Kit (Roche Diagnostics) according to the manufacturer's guidelines. They were then mounted with DAPI (Vector Laboratories, USA). Apoptotic cells were determined using image analysis software (Visiolab). A minimum of 30 crypts per section was measured. For each crypt, we evaluated the number of TUNEL-positive epithelial cells and expressed the result as the percentage of positive cells per crypt.
In vivo intestinal permeability assay
In vivo intestinal permeability was assessed using fluorescein dextran (FITC-Dextran 4, Sigma-Aldrich) as previously described . The mice were orally gavaged with 0.75 mg/g body weight of 4 kDa FITC-labeled dextran and blood samples were obtained from the retro-orbital venous plexus 5 h after this administration. Blood samples were centrifuged for 10 min at 5000 rpm and plasma was taken and frozen at -20°C and analyzed the following day. Intestinal permeability to 4 kDa FITC-labeled dextran was determined by measuring the fluorescence intensity in plasma at 485 nm/525 nm using an automatic Infinite M200 microplate reader (Tecan, Lyon, France).
Data are given as the mean ± S.E.M. (standard error of the mean). The results were compared between groups by one-way ANOVA followed by a Tukey test using GraphPad 7.0 software (GraphPad, San Diego, CA). The results were also compared between groups using normal or Poisson regression models according to the nature of the parameter of interest, continuous response (villus height and intestinal permeability) or count data (apoptosis and proliferation), respectively.
The mouse survival curves were calculated by using the Kaplan–Meier method and the P-value was determined by a log-rank test possibly adjusted for multiple comparisons. The Cox survival model was used to assess the association between the MSC-derived EV therapy and the risk of death . The coefficients in a Cox regression relate to hazard, which quantifies an increase in the risk or a protective effect depending on the sign of the fitted coefficients (positive or negative respectively). The significance analyses were set at ***p≤0.0001, ** p≤0.001, *p≤0.05 vs the non-irradiated group, at §§§p≤0.0001, §§p≤0.001, §p≤0.05 vs the 10 Gy WBI group and ###p≤0.0001, ##p≤0.001, #p≤0.05 irradiated mice treated with MSC vs irradiated mice treated with MSC-derived EVs. The regression and survival analyses were conducted using MATLAB Version 188.8.131.521 (R2013b) and the graphical representations were generated using GraphPad 7.0 software (GraphPad, San Diego, CA).