miR-130a and Tgfβ Content in Extracellular Vesicles Derived from the Serum of Subjects at High Cardiovascular Risk Predicts their In-Vivo Angiogenic Potential

Serum-derived extracellular vesicles (sEV) from healthy donors display in-vivo pro-angiogenic properties. To identify patients that may benefit from autologous sEV administration for pro-angiogenic purposes, sEV angiogenic capability has been evaluated in type 2 diabetic (T2DM) subjects (D), in obese individuals with (OD) and without (O) T2DM, and in subjects with ischemic disease (IC) (9 patients/group). sEV display different angiogenic properties in such cluster of individuals. miRNomic profile and TGFβ content in sEV were evaluated. We found that miR-130a and TGFβ content correlates with sEV in-vitro and in-vivo angiogenic properties, particularly in T2DM patients. Ingenuity Pathway Analysis (IPA) identified a number of genes as among the most significant miR-130a interactors. Gain-of-function experiments recognized homeoboxA5 (HOXA5) as a miR-130a specific target. Finally, ROC curve analyses revealed that sEV ineffectiveness could be predicted (Likelihood Ratio+ (LH+) = 3.3 IC 95% from 2.6 to 3.9) by comparing miR-130a and TGFβ content ‘in Series’. We demonstrate that sEV from high cardiovascular risk patients have different angiogenic properties and that miR-130a and TGFβ sEV content predicts ‘true ineffective sEVs’. These results provide the rationale for the use of these assays to identify patients that may benefit from autologous sEV administration to boost the angiogenetic process.

functional miRNAs into vascular cells using EVs 15 . Similarly, EVs released from cytokine-induced inflamed ECs impact on both ECs and monocytes, and regulate the vascular inflammatory response via the expression of the intercellular adhesion molecule 1, ICAM-1, and the recruitment of inflammatory cells 16 . Indeed, changes in circulating EV cargo have been associated with endothelial 17 and smooth muscle cell dysfunction in type 2 diabetes (T2DM) 18 .
There is a growing body of evidence to suggest that EVs can act as potential therapeutic, diagnostic and prognostic tools 19,20 . We have previously developed an in-vitro test of potency to predict healthy subject-derived serum-EV (sEV) pro-angiogenic properties 21 . Since angiogenesis, particularly in the heart, peripheral arteries, and kidney is impaired in patients with a high cardiovascular risk profile 22,23 , the aims of the present study are: (i) to investigate in-vitro and in-vivo sEV proangiogenic capability by assessing endothelial cell proliferation/ vessel-like structure formation in-vitro and neoangiogenesis in male severe combined immunodeficiency (SCID) mice; (ii) to explore the possibility that a specific sEV cargo could mark effective and ineffective sEV in these high-risk patients making relevant their more safe and feasible autologous clinical application. Additionally, particular attention has been devoted to the analysis of sEV TGFβ content as a pre-selected target, and miRNAs by miRNome profiling. Results sEV characterisation. 9 sEV samples derived from healthy individuals (H) and 36 derived from patients were examined for their number and size. Patient clinical data are reported in Table 1. The size distributions of sEV from healthy individuals and patients did not show any significant differences (Fig. 1A,B). The average particle size was around 138 nm. A significantly higher sEV concentrations were detected in obese (O) and ischemic patients (IC) (Fig. 1C). The Nanoparticle Tracking Analysis (NTA) quantification values for each subject are reported in Supplementary Table S1. Characterisation by Guava FACS analyses revealed that sEV expressed CD144 and CD42b, as is typical of endothelial and platelet markers, respectively. VEGFR3+ sEVs were also found ( Fig. S1).
Patient derived sEV pro-angiogenic activity. sEV from the different patient groups were evaluated for their angiogenic potential in-vitro according to our angiogenesis potency test 21 . As shown in Supplementary  Table S2, we were able to distinguish effective and inefficient sEV in all patient groups. Unlike in healthy subjects, almost all obese/T2DM (OD) patients are characterised by the presence of sEV with proangiogenic capability. Individually detailed results are shown in Fig. 2A. An in-vivo angiogenic assay was performed using effective and ineffective sEV from the different patient groups to validate our in-vitro data in patients. As shown in Fig. 2B,C, we were able to demonstrate the pro-angiogenic capability of patient-derived sEV in-vivo. No statistical differences in the expression of CD144, CD42b and VEGFR3 were found when sEV with or without angiogenic potential from both healthy subjects and patients were compared. tGfβ EV content and sEV angiogenic potential. We have previously shown that TGFβ is crucial for sEV-mediated angiogenic effects in healthy subjects 21 . ELISA was therefore performed to analyse the TGFβ content in sEV from our patients and results were compared to those of healthy subjects. As shown in Fig. 3A, TGFβ sEV content in obese and T2DM patients was significantly associated with their angiogenic activity. No significant differences in TGFβ content were detected in OD patients (Fig. 3A,B), in accordance with their functional activity. Similar results were found in IC patients. However, the TGFβ sEV content in the IC-patient group correlated with their angiogenic properties (Fig. 3B, Supplementary Table S3), suggesting that, unlike in OD patients, TGFβ cargo could be a relevant driver of their angiogenic potential. Accordingly, sEV from patient OD13 were enriched in TGFβ (Fig. 3A) even though they were ineffective. sEV miRNome profile and target validation. The role of miRNAs in angiogenic processes has been extensively documented 24,25 . The possibility that differential miRNA content in sEV could also contribute to their functional capability was therefore investigated. A miRNome analysis (Fig. 4A and Supplementary Table S4), performed using healthy effective and ineffective sEV (3 samples/each), led to the identification of 8 angio-miRNAs as the most differentially expressed: miR-126, miR-21, miR-296-3p, miR-210, miR-130a, miR-27a, miR-29a and miR-191 ( Fig. 4B  while miR-21, miR-29a, miR-191 and miR-210 were down-regulated in sEV with proangiogenic capability (Fig. 4A). The expression level of differentially expressed miRNAs were evaluated in all patients and healthy subjects and reported in Fig. 4B. This miRNA-list was submitted to DIANA-mirPath analysis. As shown in Fig. 4C, the TGFβ signalling pathway (p-value: 9.11E-08) was found to be significantly enriched. To investigate whether this difference was also associated with patient-sEV functional activity, their expression, evaluated using real-time PCR (cut-off Ct value 30), was compared to sEV angiogenic potential. We found an enrichment of miR-210 in almost patient-derived sEV (Fig. 4B), in accordance with results from Shalaby et al. 26 . However, no significant correlation with biological activity was detected. Interestingly, we were able to demonstrate that, unlike miR-210, miR-130a showed the highest level of correlation between Ct distribution and the angiogenesis potency test (3 out of 9 in O patients, 5 out of 9 in OD and 4 out of 9 in IC patients), and in all T2DM (D) patients ( Fig. 4D and Table 2). These results indicate that in the D group both TGFβ and miR-130a sEV content strongly correlated with their angiogenic properties (Figs. 3A and 4D). DIANA-mirPath analyses were interrogated, using miR-130a, by selecting pathways involving at least 15 genes. Once again, a significant enrichment in genes involved in the TGFβ pathway, among others, was detected (Fig. 4E). The network predicted by IPA for miR-130a target genes identified several genes (including: KDR, HOXA5, ROCK1, EPHB6) that are clearly related to the angiogenic process ( Fig. 5A) [27][28][29][30] . Moreover, TGFβ and TGFBR1 genes were found among the miR-130a interactors (Fig. 5A). These results support the miR-130a contribution to the sEV-mediated mechanisms of action. To validate predicted miR-130a target(s), ECs were either transfected with mimic-miR-130a or stimulated with sEV enriched in miR-130a and evaluated by RT-PCR and Western blot. Scramble miRNA and sEV carrying low miR-130a served as controls. Cell miR-130a enrichment was demonstrated by RT-PCR ( Fig. 5B upper panel). The results of protein expression reported in Fig. 5B (lower panel) show that, among the predicted miR-130a targets, only homeoboxA5 (HOXA5) was downregulated in ECs   miR-130a and TGFβ sEV content are valuable predictive markers for the identification of 'true' ineffective sEVs. ROC curves were interrogated to identify a simple method to discriminate sEV functional capability in all patient's groups, as reported by C. Gai et al. 31 . In detail, ROC analyses assessed whether miR-130a and TGFβ sEV content have the predictive capacity to identify sEV that display proangiogenic capability from potency test findings. The results demonstrated that both miR-130a and TGFβ can be considered good predictive measures of the 'true ineffective sEV' that were identified by RS, with statistically significant AUC values (Fig. 5B,C). Both measures displayed good sensitivity in identifying as 'ineffective' the 'true ineffective sEV' identified by RS. This was particularly evident and further underlined by the LH+ = 1.88 (IC95% from 1.49 to 2.27), for miR-130a (Se = 0.94 IC95% from 0.73 to 0.99) and for TGFβ (Se = 0.88 IC95% from 0.66 to 0.97). However, both measurements had a low specificity value (miR-130a: Sp = 0.50; TGFβ Sp = 0.64).

Discussion
The present study demonstrates that a varied cluster of chronic diseases that are characterised by high cardiovascular risk are associated with different sEV angiogenic properties. We also provide evidence to show that, as in healthy subjects, the high TGFβ content of sEV in T2DM and obese patients predicts their proangiogenic capability (Fig. 3A). Moreover, miRnomic analyses demonstrated that miR-130a is the most relevant sEV miRNA, as it can predict their angiogenic properties. This is particularly true in T2DM patients (Fig. 5D). Finally, the ROC curve that combines data from miR-130a and TGFβ sEV content 'in series' demonstrated that this easy and simple assay is able to predict 'true ineffective sEV.
The increased risk of cardiovascular events is a common feature in T2DM and obese patients 32 . Impaired vessel formation is still considered a relevant mechanism in abnormal vascular remodelling in these clinical settings 33 . Moreover, (micro)vessel dysfunction and vessel regression/rarefaction combined with poor neovascularization contribute to pathophysiology of systemic metabolic syndrome/diabetic heart failure 34 . Therefore, boosting the neovascularisation of damaged tissues remains necessary for the improvement of patient outcomes.
Several therapeutic approaches to improve vascular remodelling in patients at high cardiovascular risk have been proposed 35 . In particular, angiogenic growth factors, and cell-based therapy exploiting bone marrow, peripheral blood, mesenchymal, and adipose tissues as stem/progenitor cell sources have been evaluated in clinical trials to improve long-term patient outcomes 36 . Unfortunately, they have failed to provide real benefits 37 , meaning that novel treatment options are still required. In this paper, the angiogenic properties of sEV collected from patients at high risk of cardiovascular events have been thoroughly investigated, as has the role of their cargo in promoting vessel formation.
EVs are released by many cell types and exert their biological effects via the transfer of mRNAs, proteins, lipids and a wide range of non-coding RNA, such as miRNAs 38 . EVs, and in particular their composite cargos, have been consistently demonstrated to contribute to the development of cardiovascular diseases 39 . However, the therapeutic delivery of stem cell-and serum-derived EVs have been proposed as cell-free approaches for cardioprotection in models of cardiac ischemic injury 8,20 . We have recently demonstrated that the sera of healthy donors could contain pro-angiogenic EVs, which principally express endothelial and platelet markers. Moreover, we found that, when effective in promoting vessel formation in-vivo, healthy subjects-derived sEV also prevent skeletal-muscle damage in animals with lower limb ischemia 21 . This indicates that sEV angiogenic properties can be simply provided by the in-vivo angiogenic assay. Since angiogenesis is dysfunctional in obese, obese/ T2DM, IC, and more importantly in T2DM patients, sEV angiogenic capability was investigated as part of the search for novel therapeutic options. We first noticed that patients with higher cardiovascular risk have a higher number of circulating sEV than healthy subjects. Endothelial and platelet antigens were expressed on these sEV. VEGFR3-expressing sEV were also found, indicating that EVs from lymphatic vessels can be also detected in circulation. VEGFR3 is expressed by lymphatic vessels and drives lymphangiogenesis by binding to VEGF-C 40 . The impairment of VEGF-C-induced lymphangiogenesis via the epsin-mediated degradation of VEGFR3 has recently been reported in T2DM 41 . No differences in the number of VEGFR3 + sEV were detected. The possibility that the release of VEGFR3 + EVs may be a protective mechanism against VEGFR3 degradation can therefore be postulated.
The role of sEV in driving vascular complications in diabetes and obesity have been extensively documented 42,43 . Moreover, increases in the numbers of circulating EVs 44-46 that display platelet 47 , monocyte 48 and endothelium markers 49 have been reported in individuals with cardiovascular risk factors. Fittingly, we have demonstrated that IC and obese individuals showed a significantly increased number of sEV. Increases in the number of circulating EVs that display adipose tissue markers have also been reported 50 . Although our data are in line with the study by Murakami et al. 51 , the possibility that adipose tissue-derived EVs may be a relevant source of sEV in obese patients cannot be ignored.
The role of sEV in boosting the activation of the coagulation cascade has been proposed as an explanation for the increased risk of cardiovascular events in obese and T2DM patients 52 . By contrast, whether and how sEV, obtained from sera of patients with cardiovascular-risk factors, display proangiogenic properties is still matter of study. The analysis of patient-derived sEV showed that obese/T2DM subjects had a much more active sEV population than healthy donors, while T2DM had the lowest activity. These data were also confirmed by the IC group that includes patients with diabetes (IC26, IC28), obesity (IC24), and both diseases (IC25, IC27, IC30). The pro-angiogenic potential of patient-derived sEV that was validated in-vivo further supported the in-vitro results. No significant correlation between sEV number and angiogenic activity was detected in patient's groups. This was particularly true in IC patients (see Figs. 1C and 3C). This suggests that sEV angiogenic properties may depend on their specific cargo rather than their absolute number. The possibility that sEV from OD patients are much more enriched with pro-angiogenic factors should be considered.
The role of TGFβ during embryogenesis and in adult-vessel homeostasis has been highlighted by gene inactivation 53 . We have previously shown that sEV TGFβ content was a relevant mediator of sEV proangiogenic activity 21 . We wondered whether sEV TGFβ content could also account for the angiogenic potential of patient-derived sEV. The ELISA assay demonstrated that sEV TGFβ content correlated with their angiogenic potential in only two patient groups: T2DM and obese. This relationship was not significant in the other two subject groups, possibly because of the small number of recruited patients. In all patient group, TGFβ sEV content correlated with the proangiogenic activity when it was compared to sEV angiogenic capability patient-by-patient except for the obese/T2DM group. Therefore, the enrichment of proangiogenic mediators, which are different from TGFβ, can be proposed to explain this difference. Further studies are required to validate this possibility.
However, the pro-fibrotic properties of TGFβ should not be neglected. Indeed, besides acting on ECs to induce the angiogenic switch, TGFβ is a crucial mediator of fibrosis in different organs 54,55 . This would be relevant in all patient groups analysed in this study, since additional kidney, liver, and heart scarring and thickening would eventually lead to permanent organ failure. Therefore, being such effect particularly relevant for sEV clinical www.nature.com/scientificreports www.nature.com/scientificreports/ application, additional studies should be performed to investigate sEV pro-fibrotic effect in the heart, kidney, and liver before exploiting sEV to improve angiogenesis in patients at high cardiovascular risk.
Besides EV protein cargo, EV miRNA content has been reported to impact in their mechanism of action 56 . Indeed, differentially expressed angiogenic miRNAs were detected in sEVs from healthy donors that display different biological responses (with or without angiogenic potential). Interestingly, DIANA-mirPath analyses that included the indicated miRNAs (miR-126, miR-130a, miR-27a, miR-296-3p, miR-21, miR-29a, miR-191 and miR-210) demonstrated a significant enrichment in genes that are involved in the TGFβ-signaling pathways. High miR-210 levels have been proposed as a novel diagnostic biomarker for patients with cardiovascular complications 26 . Our finding that sEV from almost all patients (Fig. 5B) were enriched in miR-210 sustains the possibility that it can also be considered a biomarker for increased cardiovascular risk. This is consistent with the finding that an extensive spectrum of miR-210 targets that mainly regulate mitochondria metabolism, angiogenesis, DNA repair and cell survival have been reported 49 . Interestingly, we noticed that miR-130a content correlated with sEV angiogenic properties in all T2DM patients and in diabetic IC patients (IC26, IC28). Several genes involved in the angiogenic process, such as KDR, HOXA5, ROCK1 and EPHB6 [27][28][29][30] , were identified among the most relevant miR-130a interactors by IPA. Indeed, miR-130a overexpression or EC treatment with sEV enriched in miR-130a led to HOXA5 downregulation. This may be a proof of concept for its contribution to the angiogenic process. Moreover, the finding that TGFβ and TGFBR1 were also found among the genes that are under the control of miR-130a further confirms the cooperation between miR-130a and TGFβ in driving the proangiogenic activity of biologically active sEV.
In general, EVs are considered main paracrine and endocrine players, and their application as therapeutic agents would allow a safe regenerative approach. We focused on the angiogenic switch as a showcase model for the characterisation, efficacy testing and potential clinical translation of sEV as therapeutics. We provide evidence that sEV package can theoretically act on different tissues requiring new vascularization via TGFβ and post-transcriptional regulation of gene involved in new vessel formation. However, the therapeutic effects of the off-the-shelf sEV "secretome" largely depend on the complex interaction of all miRNAs and proteins taken together. This implies that, besides TGFβ and/or miR-130a, other proteins or genetic information transferred by sEV may drive a diverse range of biological processes in different target cells, dictating their fate.
Various biofluids and, more importantly, sera have been investigated as promising biomarkers for many diseases, including diabetes 57 . This would be particularly relevant if predictions of the response to treatment could be obtained. We have herein identified in T2DM patients criteria to distinguish between subjects with effective and ineffective sEV by analysing their TGFβ and miR-130a content. We have also provided evidence that TGFβ and miR-130a sEV content can predict sEV ineffectiveness in all patient's groups. In particular, combining data from TGFβ and miR-130a 'in series' has demonstrated that the test possesses good specificity (75% with IC: 95% vs 50%) and sensitivity (82.4% with IC: 95% vs 94%) in identifying individuals with 'true ineffective sEV' . Our study therefore suggests that TGFβ and miR-130a content may be a promising non-invasive detection method for the recognition of individuals, especially those with T2DM, that may benefit from autologous sEV treatment. A larger study is required to reinforce this observation.
Of note, all obese and 5 out of 9 obese/diabetic patients displayed low sEV-miR-130a content. Low miR-130 and high peroxisome proliferator-activated receptor gamma (PPARγ) were detected in adipose tissue from obese, but not lean women 58 . This suggests that miR-130a sEV content may predict adipose tissue perturbation, and thereby may be considered a specific signature of adipose tissue dysfunction in obese subjects. Finally, a significant negative correlation between PPARγ and circulating miR-130 level was reported in obese patients with colorectal cancer 59 . Therefore, to exploit miR-130 sEV content as a potential biomarker of colorectal cancer susceptibility in obese patients should be a further future challenge.
One limitation of the study is the small clinical sample, requiring external validation in cohorts from other/ multiple centres. Nevertheless, this is the first study which reports a sensitive and specific assay with which to evaluate sEV effectiveness in patients with different sEV angiogenic properties. Moreover, our results provides the rational to explore, in different cluster of patients at high cardiovascular risk, and in particular in T2DM patients, autologous sEV administration to improve angiogenesis. Since EV immunogenicity, isolation and scalability are still a matter of debate, to move from preclinical to clinical studies, autologous samples would avoid concerns on EV safety, isolation and storage.

Methods
Patients. Thirty-six patients at high cardiovascular risk and nine sex-matched healthy volunteers were included in the study. In detail, nine T2DM patients (D:n = 9), nine obese individuals (O:n = 9), nine obese T2DM patients (OD:n = 9) with no previous cardiovascular events, and nine ischemic patients, some of whom display diabetes and/or obesity while others do not, that were undergoing surgical treatment for hind limb ischemia in our Clinic (IC:n = 9) were herein examined. No T2DM patients received insulin treatment. No T2DM, obese, obese/T2DM patients were under statin treatment, whereas all IC patients received statins. Patient clinical data are reported in Table 1. All human experiments were performed in accordance with European Guidelines and policies, and were approved by the Ethical Committee of the University of Turin. Serum from all patients was obtained after admission to our Clinics (D, O, OD) and before surgery, for ischemic patients (IC). Written informed consent was obtained from all patients. Human serum from healthy donors (n = 9) was provided by the Blood Bank of 'Città della Salute e della Scienza di Torino' , after informed consent and approval by the internal Review Board of the Blood Bank. Approval was obtained from Ethical Committee of the of 'Città della Salute e della Scienza di Torino' (protocol no: 0007814). Isolation of EVs from sera. Blood was obtained via venipuncture, with 9 ml serum/each participant being collected and stored at −80 °C. After thawing, total EVs were isolated and purified by Ultracentrifugation at 100,000 × g for 2 h, which was preceded by centrifugation at 3000 g to remove debris. Pellets were washed once with PBS and centrifuged at 100,000 × g, 4 °C for 1 h. Samples were either used fresh, or thawed after being stored at −80°2 1 .

Study approval.
Nanoparticle-tracking analyses. EVs underwent nanoparticle tracking analyses (NTA) that were performed using the NanoSight LM10 system (NanoSight Ltd., Amesbury, UK), to delineate their size and profile. All acquisitions were carried out at a Camera level setting of 14, and three videos of 30 s duration were recorded for each sample. sEV were diluted (1: 1000) in 1 ml of vesicle-free physiological solution (Fresenius Kabi, Runcorn, UK) and analysed as previously described 21 . FACS analysis. sEV FACS analyses were performed using GUAVA (Guava easyCyte ™ Flow Cytometer, Millipore, Germany) as previously described 60 . The detection of nanoparticles was facilitated by the use of flow cytometry beads (Aldehyde/Sulfate latex 4% w/v 4 µm, Invitrogen, Germany) and APC-, PE-and FITC-conjugated antibodies directed to CD42b, VEGFR3 and CD144, (Dako Denmark A/S, Copenhagen, Denmark). FITC and PE mouse non-immune Isotypic IgG (Beckton Dickinson, Franklin Lakes, NJ) served as controls. EVs were incubated with each antibody, or isotype control antibody at 4 °C in 100 μl PBS containing 0.1% bovine serum albumin, and analyzed. EVs were detected mainly below the forward scatter signal corresponding to 1 mm beads (data not shown). All samples were gated on the basis of negative controls and compensated appropriately prior to analyses. Population percentages/numbers were generated for gated populations from each experiment using Guava Incyte software (Millipore). sEV angiogenic assay. Primary macrovascular endothelial cells (ECs) and microvascular endothelial cells (HMEC) were purchased from Lonza (Basel, Switzerland) and cultured as indicated in the manufacturer's instructions. In-vitro tests of potency and in-vivo angiogenesis were performed as previously described 21 . Briefly, all the in-vitro study were performed administering 5 × 10 4 sEV/target cells. BrdU and in-vitro tubulogenesis assays were used to assess single sample sEV pro-angiogenic activity. EVs from of all the analysed groups were classified as effective or ineffective sEV (% cut-off value = 50). In-vivo angiogenesis was assessed by measuring the growth of blood vessels as previously described 60,61 . Briefly, ECs (1 × 10 6 cells/injection) were incubated overnight with sEV (5 × 10 10 EVs per 1 × 10 6 ECs). After incubation, sEV-treated ECs were mixed with Matrigel immediately before their subcutaneous injection. Male severe combined immunodeficiency (SCID) mice (6 weeks old/24 g) (3/each group) were then injected subcutaneously. Since the female estrus cycle could introduce unexpected variables, only male mice have been used in this study. An equal number of non-stimulated ECs were used as a negative control. The Matrigel plugs, containing wheals, were recovered on day 7 and then fixed and stained using the trichrome stain method. The vessel lumen area was determined as previously described 21 . Two different blind observers performed the in-vivo analysis. Animals have been used after in-vitro studies which have help to obtain the preliminary and key information to select the most effective therapeutic tools to be validated in-vivo. For animal experiments three mice have been placed in each cage. Housing of animals have been standardised to the following procedures: Room temperature: 22+/− 1 °C; U.R: 55+/− 10%; 12/12 light /dark cycle. Caging condition: SPF animal facilities. Group housing. Environmental Enrichment. Cage size 350 cm 2 minimum, 100 cm 2 per animal minimum. Animals have been anesthetised by gas anaesthesia prior to any "surgical" procedure. Animals have been sacrificed when they show signs of distress. We were careful in selecting a number of animals that was sufficient to provide statistically significant results while not unnecessary wasting them. tGfβ ELISA assay. sEV (2.5 × 10 8 particles) were lysed from all healthy subjects and patients to evaluate TGFβ EV content (performed in triplicate). A solid phase sandwich Enzyme Linked-Immuno-Sorbent Assay (ELISA, Invitrogen Multispecies TGF-β1 kit, Germany) was used, according to manufacturer's instructions. miR Screening. sEV miRNome was assessed via the RT-PCR of 1140 microRNA, using miRNome microRNA Profilers QuantiMir (SBI, System Biosciences). The kit includes assays in pre-formatted plates for human microRNAs with three endogenous reference RNA controls (U6, RNU43 and Hm/Ms/Rt U1) on each plate. We profiled the miRNome of effective (n = 3) and ineffective sEV (n = 3) 21 collected from the serum of healthy subjects. miRNA Ct values were extrapolated for each EV sample analysed. The Ct averages from the different samples (n = 3) of both EV populations were normalised to endogenous controls and converted into 2 −(∆Ct) values. miRNA validation was performed using the miScript SYBR Green PCR Kit (Qiagen, Valencia, CA, USA). 100 ng of input RNA were reverse transcribed and the cDNA used to detect and quantify miRNAs of interest. Experiments were run in triplicate using 3 ng of cDNA/each reaction as described in the protocol (Qiagen). The following miRNAs were screened in all patient-derived EVs: miR-126, miR-21, miR-296-3p, miR-210, miR-130a, miR-27a, miR-29a and miR-191. qRT-PCR data were normalised using RNU6B and RNU43 as internal controls.
Pathways and miR target prediction analysis of EV content. The web-based program DIANA-mirPath was used for EV miRNA target prediction and biological pathway enrichment analyses, as previously reported 62 . The algorithm microT-CDS was chosen to predict EV-derived miRNA targets using a default threshold (microT = 0.8). Only biological pathways with a P value < 0.01 to all known Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathways were considered as significantly enriched. IPA was used to predict the target genes for miR-130a. We set up the miRNA Target Filter tool on IPA (Qiagen: http://www. qiagen-bioinformatics.com/products/ingenuity-pathway-analysis/) to associate miR-130a with predicted mRNA targets.

miR-130a target(s) validation.
Transfection of mimic-miR-130a (10 nM) was performed on ECs using HiPerfect Transfection method (Qiagen, Valencia, USA) 63 . Total RNA from untreated ECs, ECs transfected with mimic-miR-130a, treated with sEV with high (D17, D18, D20) or low miR-130a content (D1, D2, D4) was extracted using All-in-One Purification kit (Norgen, Thorold, ON Canada). miR-Scramble (Scr) was used as transfection reference sample. The expression of miR-130a was evaluated by RT-PCR 63 . miR-130a targets were analyzed by Western Blot analysis. Proteins were obtained using RIPA buffer, as previously described 63  Statistical analysis. Data were analysed using the GraphPad Prism 6.0 Demo program. Results are expressed as means ± SD or ±SEM, unless otherwise reported. 1-way ANOVA, followed by Tukey's post hoc or multiple comparison, Student t tests for 2-group comparison and Newman-Keuls Multiple Comparison Test where appropriate were used. The cut-off for statistical significance was set at p < 0.05 (*p < 0.05, **p < 0.01, ***p < 0.001). Our data passed normality and equal variance tests. According to our previous data, the minimum sample size that has permitted us to detect a 40% difference between the experimental groups, with 90% power and a probability level of 0.05 in a two-tailed hypothesis, was n = 3 mice/group. ROC analysis: Principal data are presented as means, standard deviations (SD), median and 95% confidence intervals for the two investigated groups -'True effective sEV'/'True ineffective sEV' considered as the Reference Standard (RS). The achievement of RS was evaluated using ROC curves 31,64-67 to evaluate predictability for miR-130a and TGFβ. sEV were classified into the following categories according to the Ct value of miR-130a and pg/ ml of TGFβ: 1. sEV displaying a miR-130a Ct value ≥ 30 were considered to be ineffective sEV; 2. sEV displaying TGFβ content < 1100 pg/ml were considered to be ineffective sEV.
Finally, we evaluated the 'goodness' of the cut-off score used for the two measures to predict 'True ineffective sEV' as defined by the RS. The predictive capacity was evaluated both for each of the two measures individually and by combining the two measures 'in series' (sEV ineffective for both measures were considered 'ineffective sEV'). The analysis was based on Sensitivity (Se), Specificity (Sp), positive Likelihood ratio (LH+) (Probability of identifying a 'true ineffective sEV' as an 'ineffective sEV' compared to that of identifying a 'true effective sEV' as an 'ineffective sEV') and relative 95% Confidence Intervals values 68,69 .