Male Wistar–Kyoto rats (11–12-week-old) were purchased from a Nursery for laboratory animals “Pushchino,” Branch of IBCH, RAS (Pushchino, Russia). Euthanasia was conducted under isoflurane narcosis by secondary cervical spine dislocation in compliance with national and European Union directives and were approved by the Institutional Ethics Board for Animal Care (National Medical Research Center of Cardiology). Rat MSCs were isolated from subcutaneous adipose tissue. Pieces of tissue were minced to a size of 1–2 mm and digested with a mixture of collagenase I (200 U/ml) (Sigma-Aldrich, United States) and dispase (10 U/ml) (Thermo Fisher Scientific Inc., United States) for 30 min at 37°C with continuous stirring. After enzymatic treatment, the cell suspension was centrifugated and filtered through 100-μm nylon cell strainers (BD Bioscience, United States). Cells were seeded in a complete growth medium of 1 g/L of D-glucose DMEM/10% FBS (Gibco, United States) and cultivated under standard conditions (5% CO₂, 37°C). The next day, the unattached cells were washed out, and the medium was replaced with a fresh one. Then the medium was changed every 2–3 days. Cells were passaged upon reaching 70% confluency using 0.05% trypsin/Versene solution (Paneco, Russia) and subcultured at a 1:4 ratio. Cells were used up including passage 5.

Rat SCF coding sequence (NM_021843.4) was cloned into a VPK-410 expression vector (Cell Biolabs Inc., United States). The pAAV–hrGFP-expressing vector was purchased from Cell Biolabs, Inc., United States). Packaging vectors VPK-420-DJ (AAB- DJ) and pHelper Vector (No. 340202) (Cell Biolabs, Inc., United States), together with expression plasmids, were used to co-transfect HEK293T.

Cells at 80% confluency were transfected using the calcium phosphate method; 48-h post-transfection cells were harvested by centrifugation and subjected to four freeze–thaw cycles (liquid nitrogen/37°C water bath) and then incubated with 25–50 U/ml of Benzonase (Merck, Germany) at 37°C for 30 min. The solution was cleared by centrifugation at 5,000× g (30 min), and a supernatant was used for further purification by an iodixanol density gradient (OptiPrep^TM; Sigma-Aldrich). The 8 ml of 15% iodixanol layer was underlayered in 35 ml of PA Ultracrimp Tube (03989, Thermo Fisher Scientific Inc., United States) with 6 ml of 25%, 5 ml of 40%, and 5 ml of 60%. Cell lysates were then overlaid onto the gradient, the tube was sealed and centrifuged for 90 min at 280,000× g, 16°C in a T-865 rotor (Thermo Fisher Scientific Inc., United States). After centrifugation, the adeno-associated virus vectors (AAV)-containing 40% layer was collected and diluted with the concentration buffer PBS/0.001% Pluronic F-68 (A1288, Applichem).

The AAV product then was concentrated and desalted using the Amicon Ultra-4 (Merck, Germany) centrifugal concentrator devices. Final concentrated product was stored at 4°C.

For further analysis, 5 μl of the AAV sample was treated with 3 U of DNaseI (EN0521, Thermo Fisher Scientific Inc., United States) at 37°C for 1 h, and then DNaseI was inactivated at 75°C for 10 min. One unit of Proteinase K (EO0491, Thermo Fisher Scientific Inc., United States) was added, and the solution was incubated at 60°C for 1 h; after which, it was inactivated at 95°C for 10 min.

Quantitative PCR analysis was performed using the Qiagen Rotor-Gene Q system. PCR reactions were carried out in a final volume of 20 μl including qPCRmix-HS (PK145S, Evrogen, Russia) supplemented with 200 nM of forward (GGAACCCCTAGTGATGGAGTT) and reverse primers (CGGCCTCAGTGAGCGA), 100 nM FAM/BHQ probe (CACTCCCTCTCTGCGGCTCG), and diluted plasmid standard or treated AAV sample. The PCR profile was as follows: denaturation step at 95°C for 10 min, 40 cycles of denaturation at 95°C for 15 s, and annealing or extension at 60°C for 1 min.

Rat MSC up to fifth passages were cultured until 70% confluency (corresponds to 1–1.5 × 10⁶ cells). The media was then replaced with DMEM without antibiotics and serum and AAV stock solution at concentration 4.5 ^∗ 10¹¹ gc/ml, MOI 95000).

Cells were incubated for 3 h with mild stirring every 30 min to ensure even cell distribution at 37°C, 5% CO₂. Afterward, FBS was added, and MSCs were cultivated overnight, then one half of the culture medium was replaced with a fresh one. Cells were passaged 3 days post-transduction.

Extracellular vesicles were isolated from AAV-transduced MSCs conditioned medium at day 7 after transduction or from intact (control) cells. Cells were washed three times from standard culture medium with Hanks balanced salt solution and maintained in DMEM/F-12 with 1% exosome-depleted FBS (System Biosciences, United States) for 48 h. Conditioned media were cleared from cells and debris by serial centrifugations: 400× g for 10 min, 2,000× g for 30 min. The supernatant (sample volume 36 ml) was then centrifuged at 100,000× g and 10°C for 90 min in a SW32 Ti bucket rotor (Beckman Coulter, United States) using ultracentrifuge Optima XE-90 (Beckman Coulter, United States). Pellet was resuspended in DPBS, and the presence of EVs in samples was confirmed by transmission electron microscopy. The protein concentration in EV samples was routinely measured using the Bradford assay.

The size distribution of EVs and their concentration were determined by nanoparticle-tracking analysis (NTA) using NanoSight LM10 HSBF instrument and NTA 2.3 software build 0033 (NanoSight Ltd., United Kingdom). Configuration included 405-nm, 65-mW laser unit with passive temperature readout and high sensitivity camera of EMCCD type. All measurements were performed according to ASTM E2834 (ASTM, 2018), using previously reported camera- and video-processing setups, optimized for EVs (Livshits et al., 2015; Silachev et al., 2019; Evtushenko et al., 2020). Briefly, samples were diluted by particle-free PBS down to the optimal concentration around 1.5 × 10⁸ particles/ml. Fourteen videos at 60 s each were recorded using the fresh portion of a diluted sample for each measurement. Tracks datasets from all videos were merged to obtain joint particle size distribution, mean size and total particle concentration, corrected for the dilution factor.

For TEM imaging, the EV sample was applied to nitrocellulose carbon-coated Cu grids and negatively stained with 2% uranyl acetate solution as described previously (Silachev et al., 2019). The grids were imaged at 80 kV using a JEOL JEM-1011 transmission electron microscope (JEOL, Akishima, Japan) equipped with a digital Orius^TM SC1000 W digital camera (Gatan Inc., Pleasanton, CA, United States).

Intact and transduced MSCs were cultivated in a complete growth medium for 48 h.

The conditioned media on the seventh day after transduction were collected, centrifuged to remove floating cells (400× g) and debris (2,000× g) and stored frozen before analysis. SCF concentrations were measured using SCF Mouse ELISA Kit (Abcam, United States) following the instructions of the manufacturer using Victor^TM X3 Multi Label Plate Reader (Perkin Elmer Inc., United States).

Total RNA was isolated using the miRCURY RNA Isolation kit (Exiqon-Qiagen, United States). The total RNA concentration was determined by measuring the absorbance at 260 nm in a spectrophotometer (NanoDrop Technologies, United States). The rat microRNA content was evaluated using the Affymetrix GeneChip^TM miRNA Arrays 4.0 kit (Thermo Fisher Scientific, United States) containing probes for 728 mature and 490 rat pro-miRNAs.

RNA was biotin-labeled with a FlashTag^TM Biotin HSR RNA Labeling Kit (Thermo Fisher Scientific, United States). Labeled RNA samples were hybridized to the chip in a GeneChip^TM Hybridization Oven 645 (Thermo Fisher Scientific, United States) (18 h at 48°C and 60 rpm).

After hybridization, miRNA arrays were washed and stained using the Affymetrix GeneChip^TM Hybridization Wash and Stain Kit on GeneChip^TM Fluidics Station 450 and were then scanned with the Affymetrix GeneChip^TM Scanner 3000 7G (Thermo Fisher Scientific, United States), according to the protocol of the manufacturer. The data were processed using the Transcriptome Analysis Console software 4.0.1 (Thermo Fisher Scientific, United States).

CEL-files of raw data were produced with the Transcriptome Analysis Console software (version 4.0.1, Thermo Fisher Scientific, United States). miRNAs that fulfilled the criteria (p-value < 0.05, absolute value of fold-change ≥ 2.0) were selected for further analysis.

Hierarchical clustering was performed to build a hierarchy of clusters using the Transcriptome Analysis Console software v4.0.1. The distance metric used between objects is the Euclidean distance. Complete-linkage clustering method was used for measuring the distances between a pair of objects in the two clusters. Results are displayed as a heatmap.

The gene interaction network was built and analyzed using the NetworkX 2.0 package for Python 3. The constructed network consists of “nodes” representing target genes and “edges”—the molecular interactions of their protein products. The resulting network was visualized using the Cytoscape software. All calculations were performed as described earlier (Osmak et al., 2020).

Data displayed in diagrams are either as mean ± SD or as representative single experiments. NTA data are presented as mean ± 95% confidence interval, calculated from individual measurements using Student’s t-distribution.

Unpaired two-tailed Student’s t-test was used to compare groups for RNA content. A value of p > 0.05 equals not significant, ≤ 0.05^∗, ≤ 0.01^∗∗, ≤ 0.001^∗∗∗. Two-way ANOVA was used for microarrays data p-value calculation.

Rat adipose-tissue-derived mesenchymal cells were transduced with highly effective AAV-DJ viral vector, encoding rat KITLG gene or Stem Cell Factor (SCF). For infection efficiency monitoring, we used GFP-transduced the MSC. At day 5 postinfection, MSC culture media were replaced by media supplemented with exosome-depleted FBS, and 48 h after, media

were collected for extracellular vesicles isolation. Purified EVs were characterized phenotypically by TEM and NTA analysis. TEM confirmed the presence of intact vesicular structures with classical morphology such as clearly discernible lipid bilayers and “cup shape” (Figure 1A). We did not observe visible differences between vesicles from transduced (SCF) or intact (control) cells. NTA analysis revealed that cultured MSC-released vesicles with a mean size of 120 nm (Figure 1B), which is within the size range for microvesicles (0.1–1.0 μm) and overall profile exhibits a size distribution of particles mostly less than 200 nm. NTA also revealed that the particle concentration, mean particle size (Table 1), and size distribution (Figure 1B) were not significantly different between transduced or intact MSCs. However, SCF vesicles demonstrate a clear tendency to smaller size and less average protein content, but total RNA content significantly distinguishes between groups (Table 1).

Previously, we examined changes in protein content by mass spectrometry (Zubkova et al., 2019; Dergilev et al., 2020) to determine if rAAV infection could induced alteration in MSC EV cargoes and did not find significant differences between EVs from intact or transduced MSC (less than 5% of unique proteins). In the present study, we focused on microRNA content of EVs. We analyzed the miRNA content of MSC-derived exosomes using the GeneChip^TM miRNA Arrays 4.0 kit. The number of miRNAs detected in cells above background (DABG) was 1,250. Most miRNA were equally present in both the EVs and cells. However, we identified about 300 miRNAs significantly downregulated in EVs compared with cells with threshold settings p-of value ≤ 0.05 and fold change ≥ 2. Only seven to nine miRNAs with unknown targets were enriched in EVs but insufficiently significantly. To analyze the enriched pathways targeted by downregulated miRNAs, we used KEGG or WikiPathways databases and g:Profiler—the web server for functional enrichment analysis. Search in Gene Ontology (GO) biological process categories revealed among the top terms “cellular response to oxygen-containing compound” (GO:1901701), “cell population proliferation” (GO: 0008283), and others and target genes (192) were significantly enriched in a total of 56 KEGG pathways including the “FoxO signaling pathway” (KEGG:04068). According to the WikiPathways database, targets were significantly enriched in 11 pathways (Table 2). Overall, we found that miRNAs were more abundant in parental cells (Figure 2). A heatmap visualized clustering between the cellular and the EV miRNAs (Figure 2). Within this heatmap, clear division in the pattern of miRNA expression is evident between the cells and extracellular vesicle samples.

The differentially expressed miRNAs between exosomes from AAV-transduced or intact (control) cells were analyzed similarly as the distribution of miRNAs between parental cells and EVs. miRNA microarray analysis was performed with the following settings: p-value ≤ 0.05 and fold change absolute value ≥ 2. The Venn diagram depicts the overlapping differentially expressed miRNAs between all samples. According to our results, the difference between EVs from SCF or GFP-transduced MSCs was negligible—3.4% (Figure 3A) and very few miRNAs, which were above fivefold upregulated or downregulated (Figure 3B). That indicates rather intrinsic changes in MSC after viral transduction than the transgene effect by itself. Hierarchical clustering of the detected miRNAs clearly identified EVs from transfected MSCs (SCF and GFP together) and control MSCs as two separate groups (Figure 3C). We uncovered a total of 91 differentially expressed miRNAs and identified two miRNAs significantly upregulated in AAV-transduced MSCs. One of them—rno-miR-344a-2—was previously associated with aging and regulation of a-SMA (Wang et al., 2015) and the other—rno-miR-208a-3p—mediates the myocardial endoglin expression that led to increased myocardial fibrosis (Wang et al., 2014). Eighty-nine miRNAs were downregulated in both EVs from SCF- or GFP-MSC compared with EVs from intact MSCs. Network analysis of genes that are targeted by these downregulated miRNAs were plotted using the String database (Supplementary Material 1). Moreover, we performed pathway enrichment analysis of these genes and found that they were enriched in the categories of PI3-Act and FoxO-signaling pathways (Supplementary Material 2).

We built an interaction network of genes regulated by miRNAs decreased in EVs from AAV-transduced rat MSC using miRNet 2.0—a network-based visual analytic tool (Chang et al., 2020). The algorithm divided miRNAs into 13 separate networks, most of which consist of one or two miRNAs. For further analysis, we chose the largest network (Figure 4A) and the network of rno-mir-128-3p target genes, each of which had at least one edge over another target gene (Figure 4B). The target genes of other microRNAs did not form a network like this.

We identified four “hub” genes: Vim, Sept3, FGF16, and Vsnl1. Vim, Sept3, and Vsnl1 are involved in the regulation of cellular migration, which is consistent with our previous data on the regulation of c-kit cell migration by EVs (Zubkova et al., 2019).

We analyzed the network of rno-mir-128-3p target genes using our previously described method (Osmak et al., 2020) in order to better understand its possible regulatory effects. Target genes of our miRNAs were taken as network nodes according to the miRTarBase and TarBase databases. Any two nod genes were considered linked if their protein products have an interaction score greater than 0.9 (String interaction score) according to the String database. A total of 15 such genes were identified.

The key genes for this network were extracted based on our previously developed algorithm (Osmak et al., 2020). The following genes were identified: RT1-Da, Dctn3, Egfr, Ap2s1, and Hgf (listed by decreasing centrality). To clarify their functional role, a reactome overrepresentation analysis was performed. It turned out that RT1-Da, Dctn3, and Ap2s1 were enriched in the signaling pathway “MHC class II antigen presentation” (p = 0.0058), and Egfr and Ap2s1 were enriched in the “L1CAM interactions” (p = 0.023). The rest of the pathways are overrepresented by no more than one gene from our key gene set, and therefore, we did not consider it for further analysis.