MiRNA Profiles of Extracellular Vesicles Secreted by Mesenchymal Stromal Cells—Can They Predict Potential Off-Target Effects?

The cardioprotective properties of extracellular vesicles (EVs) derived from mesenchymal stromal cells (MSCs) are currently being investigated in preclinical studies. Although microRNAs (miRNAs) encapsulated in EVs have been identified as one component responsible for the cardioprotective effect of MSCs, their potential off-target effects have not been sufficiently characterized. In the present study, we aimed to investigate the miRNA profile of EVs isolated from MSCs that were derived from cord blood (CB) and adipose tissue (AT). The identified miRNAs were then compared to known targets from the literature to discover possible adverse effects prior to clinical use. Our data show that while many cardioprotective miRNAs such as miR-22-3p, miR-26a-5p, miR-29c-3p, and miR-125b-5p were present in CB- and AT-MSC-derived EVs, a large number of known oncogenic and tumor suppressor miRNAs such as miR-16-5p, miR-23a-3p, and miR-191-5p were also detected. These findings highlight the importance of quality assessment for therapeutically applied EV preparations.


Introduction
Mesenchymal stromal cells (MSCs) have been extensively studied in preclinical and clinical trials over the past few decades for their promising capabilities in regenerative medicine [1]. There is consensus that MSCs cannot regenerate damaged human heart tissue. However, preclinical studies showed that MSCs may provide cardioprotective effects after myocardial damage by modulating the immune response, promoting neoangiogenesis, and reducing fibrosis in the myocardial scar [2]. The therapeutic efficacy of MSCs is mainly attributed to their paracrine secretion of various growth factors, chemokines, cytokines, and extracellular vehicles (EVs) [3]. Studies in rodents and pigs showed a reduction in scar size after a single injection of MSCs after myocardial injury [4,5]. In clinical trials, the results regarding the therapeutic effect of MSCs after single treatments in patients with myocardial infarction are more inconsistent [6]. Potential issues associated with the use of MSCs include: (i) the difficulty in generating a consistent source of cells with a stable phenotype, (ii) a significant first-pass effect due to entrapment of large cells in the lung and liver microvasculature, and (iii) patient-specific comorbidities in autologous applications [7].
In addition, less than 2% of the injected human cells remain at the target site after 60 min [8]. In a porcine model of acute myocardial ischemia, intramyocardial injections resulted in a retention rate of just over 10% after 60 min [9]. Furthermore, the same study showed that less than 1% of the engrafted cells were still present four weeks after transplantation. This, in turn, means that the release time of the cardioprotective MSC secretome at the site of injury is significantly shorter than the overall process of myocardial remodeling, which prompted scientists to further investigate the secretome of MSCs, specifically MSC-derived EVs. In general, EVs are membranous nanoparticles produced by cells that are divided into three categories based on their biosynthesis: apoptotic bodies, microvesicles, and exosomes [10]. All of them are considered intercellular messengers that, when stimulated, can transmit biological signals through the blood and lymphatic system to neighboring cells and distant tissues. Proteins, messengerRNAs (mRNAs), and microRNAs (miRNAs) partially encapsulated and protected by the lipid membrane of EVs act as the biological mediators between cells. In fact, gain-of-function and loss-of-function assays have demonstrated that miRNAs transported by EVs are primarily responsible for the cardioprotective effect of MSCs [11]. MiRNAs are short nucleotide sequences of 18-22 base pairs that can bind to the 3 untranslated region of their target mRNAs, either to interfere with their transport to the ribosome or to prevent their translation at the ribosomal site [12]. Because of their short length, miRNAs usually target more than one mRNA, making specific target prediction difficult. To date, more than 150 miRNAs have been identified in MSC-derived EVs [13]. Although there are some differences in the miRNA profile depending on the source of MSCs, a number of cardioprotective miRNAs have been identified that are commonly transported by EVs from various MSC tissue origins [14]. MiRNAs encapsulated in EVs have several functions including regulation of cell physiology, proliferation, cell differentiation, and apoptosis. For example, they can regulate the expression of members of the hypoxia-inducible factor family, which are important for the modulation of vascular sprouting in the setting of hypoxia, via the RNA interference pathway [15]. Furthermore, miRNAs can also target mRNAs that regulate fibrosis and fibroblast activation, such as tissue growth factor-beta (TGF-beta) and members of the SMAD family [16].
Since it was shown that EVs isolated from MSCs can recapitulate the cardioprotective effects of their parent cells, it was hypothesized that the use of EVs may offer significant advantages over their cellular counterparts due to a higher safety profile, lower immunogenicity, and the inability to directly induce tumors [17]. However, whereas many preclinical studies use multiple direct myocardial injections to deliver EVs, this strategy may not be optimal for many patients in clinical practice. Direct access to the heart (i.e., intracoronary or intramyocardial) is achieved either through catheter-based techniques or by cardiovascular surgery, and both methods are associated with a risk of complications. In turn, a single intramyocardial injection may not be sufficient to improve tissue remodeling after

Nanoparticle Tracking Analysis (NTA) and Total Protein Analysis
Particle concentration and size distribution of EV preparations were examined using the ZetaView instrument (Particle Metrix, Inning, Germany). Particles were automatically tracked and sized based on Brownian motion and the diffusion coefficient. The NTA measurement conditions were as follows: temperature = 26.6 ± 2.2 • C, viscosity = 0.87 ± 0.04 cP, frames per second = 30, and measurement time = 75 s. Sample videos were analyzed using NTA software (ZetaView, Particle Metrix, Inning, Germany, version 8.04.02).
Total protein content of EV preparations was determined using the commercially available Bicinchoninic Acid (BCA) Protein Assay Kit with bovine serum albumin as a standard (Thermo Scientific, catalog no. 23227). Briefly, 20 µL of samples or standards were mixed with 200 µL of freshly made BCA working reagent and incubated for 30 min at 50 • C. Absorbance was measured at 560 nm with a Mithras LB940 plate reader (Berthold Technologies, Pforzheim, Germany) and analyzed with MikroWin 2000 software (Mikrotek Laborsysteme, Overath, Germany, version 4.41).

MiRNA Analysis
MiRNA was extracted from 200 µL of isolated EVs using the miRNeasy Mini Kit (Qiagen, Hilden, Germany, catalog no. 74104) according to the manufacturer's instructions. The RNA quantity and purity were assessed with the Agilent 2100 Bioanalyzer system (Agilent Technologies, Waldbroon, Germany). Reverse transcription (RT) was performed using the miRCURY LNA Universal cDNA Synthesis Kit II (Exiqon-Qiagen, Hilden, Germany, catalog no. 203301). RT thermocycling parameters were as follows: 42 • C for 60 min and 95 • C for 5 min. Quantitative polymerase chain reaction (qPCR) was performed using the miRCURY LNA Universal RT microRNA PCR system (Exiqon-Qiagen, catalog no. 339340) with 752 known human miRNAs and 3 interplate calibrators and 1 spike-in miRNA as an internal control. All primer/probe sets for miRNAs were custom designed by the supplier. Three extraction controls and two cDNA synthesis controls were additionally used as indicated by the provider. Two real-time qPCR amplifications were performed for each RT reaction. Reactions were performed according to the manufacturers' instructions using a LightCycler 480 II system (Roche, Rotkreuz, Switzerland). QPCR thermocycling conditions were as follows: 95 • C for 10 min, followed by 40 cycles at 95 • C for 10 s and 60 • C for 1 min. Melt curve analysis was performed between 60 and 95 • C at a ramp rate of 0.11 • C/s. After interpolation calibration, the examined miRNAs were classified into three categories: (i) miRNAs with mean corrected CT (CTcorr) values below 30.00 were considered as detected with certainty, (ii) miRNAs with mean CTcorr values between 30.00 and 32.99 were considered as detected with uncertainty, and (iii) miRNAs with mean CTcorr values equal or greater than 33.00 were considered as not detected.
All analyzed miRNAs and their expression values are listed in Supplementary Materials Table S1. The obtained CT values of miRNAs were normalized using the geNorm method, which calculates a normalization factor based on multiple reference miRNAs [25]. In brief, the arithmetic mean of the CT values of miRNAs that were stably expressed across all samples, namely hsa-miR-1260a, hsa-miR-125b-5p, hsa-miR-21-5p, hsa-miR-23a-3p, hsa-miR-24-3p, hsa-miR-221-3p, hsa-let-7i-5p, hsa-miR-199a-3p, and hsa-miR-100-5p, were subtracted from CTcorr values to calculate delta CT (dCT) values for every sample. In order to plot miRNA expression on heatmaps, Z-scores were determined from logarithmically transformed dCT values for each miRNA. The Z-scores were calculated as a numerical measurement of the mean value group with z = (x − µ)/σ, where x is the raw score, µ is the population mean, and σ is the population standard deviation. Finally, heatmaps of miRNAs were created with the gplots package of RStudio (version 1.3.959).

Literature Search for miRNAs
A systematic literature search was conducted for all miRNAs with a low mean CTcorr value (≤29.99) in both CB-and AT-MSC-derived EVs. Pubmed, Medline, and Scopus were used as search engines with the following search terms: "name of miRNA", "name of miRNA" AND "heart", "name of miRNA" AND "fibrosis", "name of miRNA" AND "cancer", "name of miRNA" AND "fibroblasts", "name of miRNA" AND "endothelial cells", "name of miRNA" AND "angiogenesis", "name of miRNA" AND "immunomodulation", "name of miRNA" AND "macrophages", "name of miRNA" AND "t-cells", and "name of miRNA" AND "immune cells". For published miRNA targets, only studies were considered that confirmed miRNA targets by luciferase reporter assays or gain-and loss-of-function experiments. The findings are summarized in Appendix A Tables A1-A3.

Statistical Analysis
GraphPad Prism (GraphPad Software, San Diego, CA, USA, versions 6.0 and 8.3.0) was used for performing data analysis and generating graphs. The statistical significance of differences in EV particle concentration, total protein amount, and surface marker expression was determined by the Mann-Whitney test; a p-value of less than 0.05 was considered significant. All miRNA data are shown as median with interquartile range, if not indicated otherwise. Data were tested with Shapiro-Wilk test for normal distribution. Statistical differences between two groups with only one variable in paired observations were determined either with the Wilcoxon matched-pairs signed rank test for non-parametric samples or with the unpaired t-test for parametric samples. Results were considered significant with * p < 0.05, ** p < 0.01, and *** p < 0.001.

Characterization of EVs
All EVs were harvested from the supernatants of in vitro-cultured CB-and AT-MSCs, which were derived from tissues of four healthy subjects each. Although isolated from different sources, both MSC lines showed a typical spindle-shaped cell morphology under EV biogenesis conditions ( Figure 1). The mean number of EV particles obtained was 7.1 ± 1.2 × 10 10 per mL for CB-MSC-derived EVs and 5.5 ± 0.5 × 10 10 per mL for AT-MSC-derived EVs (Figure 2A), but this difference was not significant (p = 0.057). Similarly, protein concentrations between EVs from CB-and AT-MSCs were not statistically significant (p = 0.343), with mean values of 27.9 ± 7.4 and 35.0 ± 8.7 µg/mL protein ( Figure 2B). Quantitative analysis of EV diameters demonstrated an asymmetrical distribution, with a mean diameter of 132.7 ± 12.1 nm for EVs from CB-MSCs and a mean diameter of 123.9 ± 6.6 nm for EVs from AT-MSCs ( Figure 2C), indicating the presence of exosomes, which are typically 40 to 150 nm in diameter [26]. Furthermore, both EV variants, which were isolated with the Qiagen kit, exhibited typical cup-like shapes as observed by TEM ( Figure 3A,B). In comparison, EVs isolated by sequential UC showed a similar shape ( Figure 3C,D). However, in contrast to the EVs isolated by UC, the EVs isolated by Qiagen membrane affinity columns were covered by a corona that bound larger amounts of uranyl acetate ( Figure 3A,B, red triangles). EVs isolated by sequential UC have not been further examined because this manuscript focuses on EVs isolated by the Qiagen exoEasy Maxi Kit due to its excellent scalability, which is needed for the production of large EV amounts for clinical application. Next, we analyzed the isolated EV preparations for selected membrane proteins that have been associated with EVs in the past. Regardless of the cell source, it was possible to detect on all EV preparations CD9, CD63, and CD81, with CD9 exhibiting the highest normalized mean fluorescence intensities (MFIs) (Figure 4). Interestingly, all of the aforementioned markers tended to have higher values in AT-MSC-derived EVs than in CB-MSC-derived EVs, while only CD63 levels were significantly higher (p = 0.029). Figure 4 also shows that CD73 was only detected in EVs from AT-MSCs, but not from CB-MSCs. Since it was hypothesized that MSC-derived EVs do not carry human leukocyte antigens (HLAs) and are therefore less immunogenic [23], we also included HLA-ABC and HLA-DR in the flow cytometry analysis. Our data indicate that EVs from CB-MSCs did not exhibit a signal for HLA-ABC and HLA-DR ( Figure 4). For EVs from AT-MSCs, HLA-ABC was also not present, while   Particle number, protein amount and size distribution of EVs isolated from CB-and AT-MSCs. Particle concentration (A) and size distribution (C) of EV preparations were measured by nanoparticle tracking analysis. Protein content (B) was determined by the bicinchoninic acid assay. In (A-C), the results are mean values ± standard deviation (SD) obtained from four different donors per cell type.

MiRNA Profile of CB-and AT-MSC-Derived EVs
Of the 752 miRNAs examined in this study, 117 were detected with certainty according to the guidelines of the Qiagen-Exiqon miRCURY LNA Universal RT microRNA PCR system. Based on these miRNAs, a heatmap was created ( Figure 5). The grouping of donors shows a consistent clustering with only one outlier per group (CB_MSC_4 and AT-MSC_4). Interestingly, the expression profile of EV surface markers for these donors also differed from the other donors in the same group. For further analysis, all miRNAs with mean CTcorr values below 33.00 in at least one group were included. Following this, 205 miRNAs were detected in EV samples, while the majority of miRNAs (547) were not detected ( Figure 6). From our analysis, 76 miRNAs were highly expressed in CB-MSC-derived EVs and 80 miRNAs were strongly expressed in AT-MSC-derived EVs with mean CTcorr values of less than 30.00. Intriguingly, among them, 66 miRNAs were found in EVs from both MSC sources. Only 10 were uniquely highly expressed in CB-MSC-derived EVs, namely let-7d-5p, miR-30a-5p, miR-106b-5p, miR-107, miR-136-5p, miR-140-3p, miR-181b-5p, miR-320b, and miR-320c, and miR-342-3p, and 14 were uniquely highly expressed in AT-MSC-derived EVs, namely miR-10b-5p, miR-29b-3p, miR-138-5p, miR-148a-3p, miR-185-5p, miR-210-3p, miR-424-3p, miR-424-5p, miR-433-3p, miR-484, miR-503-5p, miR-663b, miR-874-3p, and miR-940. Furthermore, 100 and 103 miRNAs in CB-MSC-derived EVs and AT-MSC-derived EVs, respectively, which showed mean CTcorr values of 30.00 to 32.99, were considered to be low expressed. To visualize differential miRNA expression profiles, a heatmap of all miRNAs that were significantly different in expression between CB-and AT-MSC-derived EVs was created, showing a clear clustering of CB-MSC-EV-miRNAs and AT-MSC-EV-miRNAs (Figure 7, 44 miRNAs). Overall, the differences in expression after normalization did not exceed a two-fold increase or decrease for almost all miRNAs, except for miR-10b-5p (8.23-fold higher in AT-MSC-derived EVs), miR-103a-3p (3.35-fold higher in CB-MSC-derived EVs), miR-222-5p (8.28-fold higher in AT-MSC-derived EVs), miR-376a-3p (2.45-fold higher in CB-MSC-derived EVs), miR-663a (7.68-fold higher in AT-MSC-derived EVs), and miR-1260a (2.87-fold higher in AT-MSC-derived EVs). Three miRNAs were only found to be highly expressed in AT-MSC-derived EVs, but were absent in CB-MSC-derived EVs, namely miR-148a-3p, miR-424-3p, miR-503-5p. In sum, CB-and AT-MSC-derived EVs are similar in their miRNA composition, with the exception of a small number of miRNAs.  The data are presented as means ± SD of normalized mean fluorescence intensities (MFIs), which were calculated as the ratio of the geometric MFI of EV samples (beads + EVs + antibodies) to control samples (beads + antibodies). Statistical analysis was performed by the Mann-Whitney test with * p < 0.05. ND indicates not detected. EVs from four different donors per cell type were included.

MiRNA Profile of CB-and AT-MSC-Derived EVs
Of the 752 miRNAs examined in this study, 117 were detected with certainty according to the guidelines of the Qiagen-Exiqon miRCURY LNA Universal RT microRNA PCR system. Based on these miRNAs, a heatmap was created ( Figure 5). The grouping of donors shows a consistent clustering with only one outlier per group (CB_MSC_4 and AT-MSC_4). Interestingly, the expression profile of EV surface markers for these donors also differed from the other donors in the same group. For further analysis, all miRNAs with mean CTcorr values below 33.00 in at least one group were included. Following this, 205 miRNAs were detected in EV samples, while the majority of miRNAs (547) were not detected ( Figure 6). From our analysis, 76 miRNAs were highly expressed in CB-MSC- . CB-and AT-MSC-derived EVs display a distinct surface marker profile. Detection of the surface marker proteins CD9, CD63, CD73, CD81, HLA-ABC, and HLA-DR using flow cytometry on EV preparations. The data are presented as means ± SD of normalized mean fluorescence intensities (MFIs), which were calculated as the ratio of the geometric MFI of EV samples (beads + EVs + antibodies) to control samples (beads + antibodies). Statistical analysis was performed by the Mann-Whitney test with * p < 0.05. ND indicates not detected. EVs from four different donors per cell type were included. 5p (8.28-fold higher in AT-MSC-derived EVs), miR-376a-3p (2.45-fold higher in CB-MSC-derived EVs), miR-663a (7.68-fold higher in AT-MSC-derived EVs), and miR-1260a (2.87-fold higher in AT-MSC-derived EVs). Three miRNAs were only found to be highly expressed in AT-MSC-derived EVs, but were absent in CB-MSC-derived EVs, namely miR-148a-3p, miR-424-3p, miR-503-5p. In sum, CBand AT-MSC-derived EVs are similar in their miRNA composition, with the exception of a small number of miRNAs. Heatmap and dendrograms of all microRNAs (miRNAs) detected with certainty according to the guidelines of the Qiagen-Exiqon miRCURY LNA Universal RT microRNA PCR system. Sample IDs are shown on the x-axis. Samples with similar miRNA expression are clustered together. The heatmap was generated by RStudio and 2 dCT was used for data input. Z-scores of more than zero indicate a higher expression of miRNAs in one sample compared to the others; Z-scores of less than zero indicate the opposite. Heatmap and dendrograms of all microRNAs (miRNAs) detected with certainty according to the guidelines of the Qiagen-Exiqon miRCURY LNA Universal RT microRNA PCR system. Sample IDs are shown on the x-axis. Samples with similar miRNA expression are clustered together. The heatmap was generated by RStudio and 2 dCT was used for data input. Z-scores of more than zero indicate a higher expression of miRNAs in one sample compared to the others; Z-scores of less than zero indicate the opposite.

Classification of miRNAs: Tumor Suppressor miRNAs, Oncogenic miRNAs, and Cardioprotective miRNAs
We then conducted a literature research ( Figure 8) to group all 66 miRNAs found at high levels in both CB-and AT-MSC-derived EVs based on their function. As indicated in Figure 9, the majority of identified miRNAs have a well-known role as tumor suppressor. We also found many miRNAs, such as miR-103a-3p, miR-151a-5p, and miR-191-5p, which are known oncogenic miRNAs (oncomiRs). Interestingly, we also identified a large number of miRNAs (26) known to act both as oncomiRs and as tumor suppressor. The EV samples examined in this study also showed positive hits for well-known cardioprotective miRNAs, such as miR-21-3p, miR-22-3p, miR-26a-5p, and miR-125b-5p. While having cardioprotective properties, most of them are also associated with oncogenic and tumor suppressor properties. In summary, these data indicate that both CB-and AT-MSC-derived EVs not only transfer a certain set of miRNAs that are involved in one particular mechanism, but rather a multitude of miRNAs that are linked to several biochemical processes, including tumor suppression, tumorigenesis, and cardioprotection.

Classification of miRNAs: Tumor Suppressor miRNAs, Oncogenic miRNAs, and Cardioprotective miRNAs
We then conducted a literature research ( Figure 8) to group all 66 miRNAs found at high levels in both CB-and AT-MSC-derived EVs based on their function. As indicated in Figure 9, the majority of identified miRNAs have a well-known role as tumor suppressor. We also found many miRNAs, such as miR-103a-3p, miR-151a-5p, and miR-191-5p, which are known oncogenic miRNAs (oncomiRs). Interestingly, we also identified a large number of miRNAs (26) known to act both as oncomiRs and as tumor suppressor. The EV samples examined in this study also showed positive hits for well-known cardioprotective miRNAs, such as miR-21-3p, miR-22-3p, miR-26a-5p, and miR-125b-5p. While having cardioprotective properties, most of them are also associated with oncogenic and tumor suppressor properties. In summary, these data indicate that both CB-and AT-MSCderived EVs not only transfer a certain set of miRNAs that are involved in one particular mechanism, but rather a multitude of miRNAs that are linked to several biochemical processes, including tumor suppression, tumorigenesis, and cardioprotection. .99) in both CB-and AT-MSC-derived EVs. Search terms were "name of miRNA", "name of miRNA" AND "heart", "name of miRNA" AND "cancer", "name of miRNA" AND "fibrosis", "name of miRNA" AND "endothelial cells", "name of miRNA" AND "angiogenesis", "name of miRNA" AND "immunomodulation", "name of miRNA" AND "macrophages", "name of miRNA" AND "tcells", and "name of miRNA" AND "immune cells". Search terms were "name of miRNA", "name of miRNA" AND "heart", "name of miRNA" AND "cancer", "name of miRNA" AND "fibrosis", "name of miRNA" AND "endothelial cells", "name of miRNA" AND "angiogenesis", "name of miRNA" AND "immunomodulation", "name of miRNA" AND "macrophages", "name of miRNA" AND "t-cells", and "name of miRNA" AND "immune cells".

EV Phenotype
Overall, the EVs analyzed in our study showed the expected proteins to be present in both CBand AT-MSCs, such as the tetraspanins CD9, CD63, and CD81. The latter was present in significantly lower amounts in EVs from CB-MSCs than in EVs from AT-MSCs, an observation that was not made in other comparative studies before. The phenomenon that EVs from MSCs have only little or no HLAs present on their surface and therefore have a low immunogenicity [23] was confirmed in our study, since HLA-ABC was not found in both CB-and AT-MSC-derived EVs. Furthermore, HLA-DR was not detected in CB-MSC-derived EVs and it was only slightly above the detection level for the flow cytometry assay in AT-MSC-derived EVs. Consequently, the phenotype of the EVs might reflect the low expression of HLA molecules of the parent CB-and AT-MSCs.
It is known that the isolation method can significantly influence the composition of miRNAs in EV preparations [26,27]. To date, there is a multitude of different EVs isolation protocols available [28], and an ideal isolation method for clinical use remains to be determined. In this study, EVs were isolated using a commercially available EV isolation kit from Qiagen. In contrast to protocols using sequential UC to isolate EVs, this kit is more appropriate for scaling up the production of EVs. Initially, we performed side-to-side comparisons for the isolation of EVs using sequential UC and

EV Phenotype
Overall, the EVs analyzed in our study showed the expected proteins to be present in both CBand AT-MSCs, such as the tetraspanins CD9, CD63, and CD81. The latter was present in significantly lower amounts in EVs from CB-MSCs than in EVs from AT-MSCs, an observation that was not made in other comparative studies before. The phenomenon that EVs from MSCs have only little or no HLAs present on their surface and therefore have a low immunogenicity [23] was confirmed in our study, since HLA-ABC was not found in both CB-and AT-MSC-derived EVs. Furthermore, HLA-DR was not detected in CB-MSC-derived EVs and it was only slightly above the detection level for the flow cytometry assay in AT-MSC-derived EVs. Consequently, the phenotype of the EVs might reflect the low expression of HLA molecules of the parent CB-and AT-MSCs.
It is known that the isolation method can significantly influence the composition of miRNAs in EV preparations [26,27]. To date, there is a multitude of different EVs isolation protocols available [28], and an ideal isolation method for clinical use remains to be determined. In this study, EVs were isolated using a commercially available EV isolation kit from Qiagen. In contrast to protocols using sequential UC to isolate EVs, this kit is more appropriate for scaling up the production of EVs. Initially, we performed side-to-side comparisons for the isolation of EVs using sequential UC and Qiagen membrane affinity columns. A similar comparison reported by the group of Streanska et al. [29] demonstrated that both methods lead to EVs with encapsulated miRNAs. However, they found differences in EV size and surface protein expression depending on the isolation method. While in their study, they were not able to detect the tetraspanins CD63 and CD81 using the Qiagen kit for EV isolation, we were able to detect tetraspanins such as CD9, CD63 and CD81, considered as typical EV markers. It should be noted, however, that we performed flow cytometry analysis, whereas the others used the Western blot. Furthermore, TEM analysis revealed that EVs isolated by the Qiagen kit were coated with either proteins or nucleic acids. For this experiment, EVs were incubated with uranyl acetate to stain phosphate groups of the lipid membrane. However, the presence of phosphate-rich proteins or nucleic acids in the so-called EV corona can also result in strong staining. We therefore hypothesize that the structures surrounding the EVs are most likely a mixture of proteins and nucleic acids. In line with this, the group of Varga et al. [30] has recently shown that EVs in vivo are also surrounded by a variety of different proteins that are not integrated in their own membrane. Furthermore, Jeppesen et al. [26] were able to separate a protein fraction from a pure vesicle fraction and they demonstrated that different EV isolation methods impact the EV-miRNA composition. Our data suggest that the Qiagen membrane affinity method produces EVs with an intact corona, indicating that miRNAs may also be bound to proteins in the corona. However, it cannot be conclusively determined whether the analyzed miRNAs were encapsulated, bound to co-isolated proteins, or bound within the EV protein corona. Studies that have so far investigated the therapeutic potential of EVs did not purify the EVs in their in vivo models prior to injection. Therefore, regardless of the isolation method, co-purified miRNAs will be injected together with the EV fraction. However, when EVs are used clinically, it is expected that additionally administered miRNAs could also play a role in the cardioprotective mechanism. It is therefore of importance to validate the miRNA profiles for each isolation method before conducting downstream experiments or even clinical studies. A more in depth analysis of the isolated EVs might have answered this question, but would be beyond the scope of this project.

MiRNA Profile
As mentioned above, miRNA analysis of EVs derived from CB-and AT-MSCs showed that a large number of detected miRNAs play an important role in tumor biology. Due to the multiple targets a miRNA can have, it is difficult to predict all possible targets of each miRNA. In this study, we therefore only reviewed targets that were confirmed already by other groups through in vitro assays. Since the PCR array used in our experiment focused on cellular miRNAs, which play a well-known role in cancer biology, it is not surprising that most miRNAs (547 out of 752) were not detected in the EV samples. Our data show a biological variability that is expected from human-derived samples [31]: EV samples derived from both CB-and AT-MSCs contain one outlier in terms of their surface marker configuration and their miRNA profiles (Figures 4 and 7). Due to the small number of donors examined in this study, the effect of donor-specific confounding factors (e.g., gender, age, or race) on the miRNA profiles cannot be determined. Our literature search revealed that most studies focused on the therapeutic aspect of miRNAs of MSC-derived EVs. Only a few studies made the data of their miRNA arrays publicly available [32][33][34]. Additionally, the role of MSC-derived miRNAs in cancer biology has been discussed and investigated by other groups. Even here, however, only few groups made all collected data available for secondary analysis. In the case of AT-MSCs, one group has investigated the role of AT-MSC-derived EVs in the development and treatment of osteoarthritis [32,33]. In both publications, the raw data of the miRNA array were made available by the authors. A side-to-side comparison revealed that 71.0% and 73.3% of the 65 highest expressed miRNAs in both data sets were identical to the miRNAs found in our EV samples. The discrepancy could be explained by the difference in treatment of AT-MSCs at time of isolation and the isolation method itself.

Anti-fibrotic Signaling via Suppression of the TGF-Beta Pathway
MiRNAs were initially examined in the context of cancer biology. Target search was therefore biased and provided a greater number of miRNAs related to cancer than, for instance, to cardioprotection. However, some miRNAs with cardioprotective properties often interfere with proteins that are also regulated in cancer cells. For instance, miRNAs that advantageously modulate fibrosis and activation of fibroblasts usually target either the mRNA of proteins in the TGF-beta/SMAD-axis or promoter and receptor mRNAs that modulate cell cycle activation. Typically, miRNA-mediated suppression of TGF-beta signaling leads to decreased fibrosis in different tissues [35]. Both CB-and AT-MSC-derived EVs contain sets of miRNAs that target TGF-beta receptors directly or downstream signaling proteins such as SMAD proteins. In the context of TGF-beta signaling, SMAD2, 3, and 4 are the downstream promoters that can activate pro-fibrotic gene expression in multiple tissues including the heart [36]. MiR-16-5p (−1.03-fold change, p = 0.84), miR-23a-3p (−1.14-fold change, p = 0.99), and miR-130a-3p (−1.11-fold change, p = 0.75), which showed no difference in relative amounts for the comparison of CB-MSC-derived EVs to AT-MSC-derived EVs, all target the SMAD mRNA directly and exhibit an anti-fibrotic, and in most cancers, a tumor suppressor effect [37][38][39]. At the same time, miR-130a-3p can also act as an oncomiR in esophageal cancer by inhibiting the expression of SMAD4 [40], which incidentally leads to a tumor suppressor effect in hepatoma cells [38]. This dual role of miRNAs in cancer biology is well known and shows the complexity of gene expression regulation via RNA interference [41]. Similarly, while miR-130a-3p suppresses fibrosis in hepatic steatosis by suppressing the TGF-beta receptors 1 and 2 [37], the suppression of TGF-beta receptor 3 by miR-23b-3p and miR-27b-3p in atrial fibroblasts leads to increased fibrosis in the context of atrial fibrillation [42]. This underlines that, similar to the effect of miRNAs in cancer, a dual role of miRNAs and thus potential off-target effects can be hypothesized. It also highlights that adverse effects, such as increased fibrosis, may depend on the presence of miRNA clusters. For the EV samples investigated in the present study, both miR-23b-3p and miR-27b-3p were found with mean CTcorr values of 25.2 ± 1.4 and 25.5 ± 1.1 versus 27.4 ± 0.9 and 27.2 ± 1.1 in CB-and AT-MSC-derived EVs, respectively.

Role of miRNA-Mediated Mammalian Target of Rapamycin (mTOR) Suppression
The miRNA target analysis also revealed that some miRNAs found in CB-and AT-MSC-derived EVs target mTOR or mTOR-associated proteins, including miR-99b/a, miR-100-5p, miR-143-3p, miR-199a-5p/3p, and miR-199b-5p. MTOR is a protein kinase that regulates cell growth, autophagy, and cell survival [43]. Since activation of mTOR plays a crucial role in maintaining growth and inducing metastasis in many cancers, it has been intensively studied as a potential target for cancer therapy [44]. For all of the miRNAs mentioned, overexpression in cancer cell lines led to the induction of apoptosis and autophagy. Interestingly, miR-100-5p can also suppress angiogenesis by preventing cell proliferation in vascular smooth muscle cells, an effect that could counteract a potential cardioprotective effect [45]. Similarly, both miR-143-3p and miR-199a-3p can increase apoptosis during hypoxic or inflammatory injury in kidney and synovial cells, respectively [46,47]. One could therefore postulate that miRNAs that inhibit mTOR signaling are unproblematic in the context of promoting preexisting tumors at the time of EV therapy. However, further studies are needed to elucidate whether MSC-derived EVs suppress mTOR signaling and how this affects the injured heart. There is some evidence that mTOR plays an important role in the activation of cell autophagy in myocardial injuries, which can prevent cell apoptosis and necrosis in the myocardial scar [48]. In the EV samples examined in this work, at least six miRNAs were found that can target mTOR or mTOR signaling related protein mRNAs (Tables A1-A3). A prolonged exposure to EVs containing these miRNAs may therefore either aggravate myocardial injury by increasing apoptosis in the early stages of myocardial infarction or improve wound healing and remodeling via autophagy.

Conclusions
The administration of MSC-derived EVs containing miRNAs offers a promising therapeutic approach for cardiovascular disease due to their proposed cardioprotective effects. In the present work, we have isolated EVs from two clinically relevant MSC sources, i.e., CB and AT, using membrane affinity columns and analyzed their miRNA cargo by qRT-PCR. Our data show that EVs from CBand AT-MSCs are similar in their miRNA composition. Although a large number of miRNAs found in EVs from both MSC sources have been associated with cardioprotective properties, our literature research for known miRNA targets has revealed that they may also play a critical role in the tumor biology of various cancers. Given that EVs and miRNAs have a half-life of less than 24 h, a single administration of EVs may not be sufficient to improve tissue remodeling after a myocardial injury and multiple EV administrations would be required. However, this procedure, in turn, could lead to the accumulation of miRNAs in patients with early-stage cancers that may not have been recognized prior to treatment. Therefore, careful screening of patients for preexisting neoplasms prior to EV administration is important to reduce the risk of potential side effects that could facilitate or even worsen existing tumors. Further reports and functional studies are needed to evaluate both the therapeutic and adverse effects of EVs and their transported miRNAs, depending on the dose and duration of treatment.