Xeno-free cultured mesenchymal stromal cells release extracellular vesicles with a “therapeutic” miRNA cargo ameliorating cartilage inflammation in vitro

Rationale: Mesenchymal stromal cells (MSCs)-derived extracellular vesicles (EVs) emerged as an innovative strategy for the treatment of chronic disorders such as osteoarthritis (OA). Biological activity of EVs is generally driven by their cargo, which might be influenced by microenvironment. Therefore, pre-conditioning strategies, including modifications in culture conditions or oxygen tension could directly impact on MSCs paracrine activity. In this study we selected an appropriate preconditioning system to induce cells to perform the most suitable therapeutic response by EV-encapsulated bioactive factors. Methods: A xeno-free supplement (XFS) was used for isolation and expansion of MSCs and compared to conventional fetal bovine serum (FBS) culture. Bone Marrow-derived MSCs (BMSCs) were pre-conditioned under normoxia (20% O2) or under hypoxia (1% O2) and EVs production was evaluated. Anti-OA activity was evaluated by using an in vitro inflammatory model. miRNA content was also explored, to select putative miRNA that could be involved in a biological function. Results: Modulation of IL-6, IL-8, COX-2 and PGE2 was evaluated on hACs simultaneously treated with IL-1α and BMSC-derived EVs. FBS-sEVs exerted a blunt inhibitory effect, while a strong anti-inflammatory outcome was achieved by XFS-sEVs. Interestingly, in both cases hypoxia pre-conditioning allowed to increase EVs effectiveness. Analysis of miRNA content showed the upregulation in XFS-hBMSC-derived EVs of miRNA known to have a chondroprotective role, such as let-7b-5p, miR-17, miR-145, miR-21-5p, miR-214-3p, miR-30b-5p, miR-30c-5p. Activated pathways and target genes were investigated in silico and upregulated miRNAs functionally validated in target cells. MiR-145 and miR-214 were found to protect chondrocytes from IL-1α-induced inflammation and to reduce production of pro-inflammatory cytokines. Conclusions: XFS medium was found to be suitable for isolation and expansion of MSCs, secreting EVs with a therapeutic cargo. The application of cells cultured exclusively in XFS overcomes issues of safety associated with serum-containing media and makes ready-to-use clinical therapies more accessible.


Introduction
Osteoarthritis (OA) is a complex and multifactorial chronic joint disease, involving multiple tissues and processes. It compromises millions of people worldwide, representing a major health Ivyspring International Publisher condition especially in elderly populations [1]. Cartilage breakdown, ligament calcification and subchondral bone outgrowths are the pre-eminent pathological changes characterizing OA [2]. Although the specific pathogenesis is uncertain, it is generally acknowledged that a link between OA progression and the production of pro-inflammatory cytokines exists [3]. In this scenario, mesenchymal stromal cells (MSCs)-derived extracellular vesicles (EVs) represent a next generation strategy for OA treatment, thanks to their reparative and anti-inflammatory properties [3][4][5]. MSCs have indeed been reported to mitigate OA through the secretion of both EV-encapsulated trophic factors, which support the endogenous cartilage repair, and anti-inflammatory cytokines [6]. MSC-EVs support the maintenance of the tissue microenvironment, playing a crucial role when the tissue homeostasis is altered by a disease or a lesion [7]. The biological effect of EVs is driven by their cargo, consisting of signalling molecules, such as proteins, lipids, DNA, RNA, miRNAs, surrounded by a phospholipidic bilayer. These miRNAs and proteins are involved in many cellular processes, such as communication, inflammation, metabolism, tissue repair and regeneration [8,9]. Different compositions in miRNA cargo may significantly influence biological activities, being involved in epigenetic modification of cells [3]. Since EV cargo mirrors the pathophysiology of the parental cells, the microenvironment may influence MSC paracrine signaling. Therefore, controlling MSC growth conditions by modifying the growth factor composition, oxygen tension, or local mechanical properties might directly shape MSC paracrine activity [10][11][12].
Hypoxia (i.e., 1% O2) has been reported to enhance EV production in different cell types in vitro [13]. It not only affects EV release, but also the cargo composition and, therefore, their therapeutic potential. In this framework, cell culture methods might be crucially important for EV production, especially for therapeutic purposes. Indeed media, often contains animal derivatives (e.g., FBS), may be a source of contamination in the final product. Moreover, for large-scale production and clinical usage, generating completely animal-free products is essential.
In this study, a xeno-free supplement (XFS) was used for isolation and expansion of MSCs. As previously reported, XFS sustains cell proliferation and differentiation, besides improving EV release and their anti-inflammatory properties [14]. However, the mechanism of action behind this biological effect still needs to be elucidated. Here, Bone Marrow-derived MSCs (BMSCs) were pre-conditioned under normoxia (i.e., 20% O2) or under hypoxia and EV production was evaluated. Anti-osteoarthritic activity was also investigated by an in vitro inflammatory model [14]. In addition, the ability of EVs to modulate pro-inflammatory cytokine production was explored on IL-1α-treated human articular chondrocytes. EV-miRNA cargo related to different pre-conditioning of secreting cells was also profiled and the most up-regulated ones were functionally validated in an osteoarthritic environment.
Human articular chondrocytes (n = 10, average age 81 years, 5 males, 5 females) were obtained from femoral condyles of patients undergoing partial knee arthroplasty, as previously described [14]. For each patient an informed consent was obtained, and the present study has been approved by the Ethical Committee of San Martino Hospital (CER Liguria: 372/2019 Genoa, Italy). hBMSCs pre-conditioning hBMSCs were cultured until 60% confluence. Then, the medium was discarded and, after three PBS 1X washes, replaced with α-MEM-GlutaMax medium with only 100 U/mL of Penicillin-100 mg/mL Streptomycin (basal medium). Cells were maintained under normoxic (20% O2 and 5% CO2 at 37 °C) or hypoxic (1% O2 and 5% CO2 at 37 °C in a hypoxic incubator, Eppendorf, Hamburg, Germany) conditions, for 72 h. Conditioned media (CM) were collected and processed as described below.

Propidium Iodide (PI) and FITC-Anti-annexin-V Apoptosis Detection
Viability of secreting hBMSCs was investigated by flow cytometry (BD FACSAria II, BD Biosciences), with double staining for Annexin-V and Propidium Iodide (PI). The Annexin-V apoptosis assay kit (BD Pharmingen, San Diego, CA, United States) was used, according to the manufacturer's instructions.

Separation of Extracellular Vesicles from hBMSCs
EVs were separated from CM after high-speed differential centrifugations. CM was collected and immediately centrifuged at 300 g for 10 min at 4 °C to eliminate dead cells and debris, and at 2,000 g for 20 min at 4 °C to get rid of apoptotic bodies. Clarified CM was then ultracentrifuged at 10,000 g for 40 min at 4 °C, and finally at 100,000 g for 2 h at 4 °C, to pellet medium-size EVs and small-size EVs (sEVs), respectively. sEV pellets were washed with PBS 1X and finally resuspended in 0.22 μm-filtered PBS 1X to obtain normoxia or hypoxia derived FBS-hBMSC-sEVs (namely Normo FBS-sEVs and Hypo FBS-sEVs, respectively) and normoxia or hypoxia derived XFS-hBMSC-sEVs (namely Normo XFS-sEVs and Hypo XFS-sEVs, respectively). A Beckman Coulter ultracentrifuge (Beckman Coulter Optima XPN-100 ultracentrifuge; Beckman Coulter) was used with swinging bucket rotors type SW41Ti and SW55Ti.
The concentration of membrane-bound protein on the surface of freshly isolated, intact hBMSC-EVs was measured using BiCinchoninic Acid (BCA) assay (Thermo Scientific Pierce, Rockford, IL, USA), following manufacturer's instructions. Protein content was normalized on cell number (1x10 6 ).

Nanoparticle Tracking Analysis
EV samples were characterized and quantified by Zetaview (Particle Metrix GmbH, Germany) Nanoparticle Tracking Analysis (NTA), equipped with a sample cell and two lasers (488 nm and 640 nm). The software used was Zetaview 8.05.14_SP7. After calibration with 100 nm polystyrene beads, samples were diluted in 0.22 μm-filtered PBS 1X and injected using a 1 ml syringe. Size distribution analyses of 11 different positions were performed for each sample on at least three different sEVs preparations.

Non-conventional flow cytometry
For flow cytometry characterization, sEVs were suspended in filtered PBS/EDTA, distributed in flow cytometry tubes (100 μl/tube) and stained with 1 μM CFDA-SE (VybrantTM CFDA SE Cell Tracer Kit, Thermo Fisher Scientific) at room temperature (RT) to visualize intact vesicles. Since CFDA-SE diffuses within vesicles only at RT, a control at 4 °C was also included. A mixture of fluorescent beads with a diameter from 100 to 900 nm (Megamix-Plus FSC and Megamix-Plus SSC, Biocytex, France) was used to discriminate sEV size. Expression of typical EV markers, such as CD9 (APC Mouse Anti-Human CD9, 312108, Biolegend, San Diego, CA, USA), CD63 (PE-CyTM7 Mouse Anti-Human CD63, 561982, BD Biosciences) and CD81 (BV421 Mouse Anti-Human CD81, 740079, BD Biosciences) was evaluated within the CFDA-SE positive events. All the evaluations were performed using a BD FACSAria II (BD Biosciences) and Flow Jo software for post-analyses.

Labeling and Internalization of hBMSC-EVs by hACs
Extracellular vesicle uptake was monitored in vitro by staining FBS-sEVs and XFS-sEVs with PKH67 (Sigma-Aldrich), according to the manufacture's protocol. Staining was stopped by adding an equal volume of 1% bovine serum albumin (BSA) and the sEVs were ultracentrifuged to remove unbound dye. Chondrocytes (OA hACs), plated on cell culture slides were treated with 200 U/mL IL-1α (Peprotech, London, United Kingdom) overnight to mimic an inflammatory condition in vitro. After pre-treatment, 1 μg/mL of stained Normo or Hypo FBS-or XFS-sEVs was added to the cells for 3 h prior to fixation with 3.7% paraformaldehyde (PFA) and staining with 5 U/mL phalloidin-594 (Life-Tech) to visualize the cytoskeleton. Nuclei were stained with DAPI (Sigma-Aldrich) and cover slipped with an aqueous mounting. Slides were observed at different magnifications and images acquired with the Axiovert 200M microscope (Carl Zeiss, Oberkochen, Germany).

Confocal microscopy
Laser scanning confocal microscopy was performed by a Leica SP5 inverted confocal microscope with a 40× (N.A. 1.24) oil immersion objective. Laser beams with 405 nm, 488 nm, and 561 nm excitation wavelengths were used for DAPI, PHK67-FITC, and Alexa Fluor 594 imaging, respectively. The fluorescence emission was collected in the range: 415-470 nm for DAPI, 500-555 nm for PHK67-FITC, and 570-670 nm for Alexa Fluor 594, by means of three photomultiplier tubes. The diameter of the detection pinhole was set to the size of 1 Airy. z-stacks of FBS-sEVs and XFS-sEVs were performed upon acquiring 70 slices of 0.13 µm, each slice being the average of four laser scans. Single confocal sections and z-stack data files were processed and analyzed by the ImageJ software.
In order to evaluate the nucleus-exosomes threedimensional distance (Dxyz), fluorescence z-stacks of each sample type were analyzed by considering a specific volume including a single cell with its vesicles. The three-dimensional distance between the nucleus centre and the exosome centre was measured as: where Dxy is the distance in the x-y plane and hz is the distance on the x-z or y-z plane. Moreover, the exosomes three-dimensional density was evaluated as the ratio between the number of exosomes and the cell volume (mm 3 ) enclosing the nucleus and the vesicles. Five regions of interest were analyzed for each sample type.

Electron Microscopy
FBS-sEV and XFS-sEV preparations were fixed by adding 2% PFA in 0.1 mol/L phosphate buffer (pH 7.4) in the same volume of EV resuspension buffer. Five μl drops of EVs were then adsorbed for 10 min on formwar-carbon-coated copper grids. Subsequently, grids were rinsed in PBS and negatively stained with 2% uranyl acetate for 5 min at room temperature. Stained grids were finally embedded in 2.5% methylcellulose for improving preservation and air-dried before examination. Electron micrographs were acquired using a Hitachi 7800 120 kV electron microscope (Hitachi, Tokyo, Japan) working at 100 kV, equipped with a Megaview G3 digital camera and Radius software (EMSIS).
To investigate the effect of sEVs on OA-hACs, confluent hACs were treated for 16 h with 200 U/mL IL-1α and 1 μg/mL Normo or Hypo FBS-or XFS-sEVs. A negative control without IL-1α or sEVs and a positive control treated only with 200 U/mL IL-1α were also included. Conditioned media were collected, and cells were scraped in RIPA buffer to obtain cell lysates. Western blot was performed by incubating membranes with specific primary antibodies for IL-6 and IL-8 (1:200, Santa Cruz Biotechnology, Santa Cruz, CA, USA) and COX-2 (1:500, Abcam, Cambridge, UK). Images were scanned using the Epson perfection 1260 scanner and band densities quantified using Fiji-Image J software. All the analyses were performed on a minimum of four pools of three primary hACs each.

RNA Extraction and Real-Time Quantitative Reverse Transcription Polymerase Chain Reaction (qRT-PCR)
Confluent hACs were treated for 16 h with 200 U/mL IL-1α and 1 μg/mL Normo or Hypo FBS-or XFS-sEVs. A negative control without IL-1α or sEVs and a positive control treated only with 200 U/mL IL-1α were also included. Cells monolayers were collected in TRIzol Reagent (Thermo Fisher Scientific, Waltham, MA, USA) and total RNA was extracted following manufacturer's instructions. Quality control of RNA concentration and purity were assessed using BioSpectrometer (Eppendorf, Hamburg, Germany). Complementary DNA (cDNA) was synthetized starting from 2 μg of total RNA using SuperScript™ IV VILO™ Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) according to manufacturer's instruction. Transcript levels of IL-6, IL-8, COX-2, COL1A1 (type I collagen), COL2A1 (type II collagen) and COL10A1 (type X collagen) genes were measured by real-time quantitative PCR (qRT-PCR) using BlasTaq™ 2X qPCR MasterMix (abmgood, Richmond, BC, Canada) on 7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA).
The house-keeping gene GAPDH was used as the endogenous control for normalization. The selected human-specific primer sequences are reported in Table 1. Data were analyzed with 2 -ΔΔCt method and reported as fold change/GADPH.

NF-κB Nuclear Translocation
Four different pools of hACs were cultured in 24-well plates on glass coverslips (20,000 cells) and treated for 16 h with 200 U/mL IL-1α and 1 μg/mL Normo or Hypo FBS-or XFS-sEVs. Cells were fixed in 3.7% PFA for 10 min at RT and permeabilized with 0.5% Triton X-100 in PBS for 10 min at 4 °C.
Non-specific binding was prevented by incubation with 20% goat serum (EuroClone) in PBS for 30 min at 4 °C. Slides were incubated with a rabbit anti-NF-κB p65 antibody (1:400; Cell Signaling, Danvers, MA, USA) overnight at 4 °C, followed by a goat anti-rabbit Alexa fluor-488 (Life Technologies, Grand Island, NY, United States) for 30 min at 4 °C. Coverslips were mounted with an aqueous mounting and images acquired using AxioPhot microscope (Carl Zeiss). A quantification was also performed counting cells with positive nuclei on at least five different areas for each experiment. Four independent experiments on pools of three hACs primary cultures were performed.

PGE2 quantification
OA-hACs CM were assayed for PGE2 content using a PGE2-specific competitive EIA kit-Monoclonal (Cayman Chemical, Ann Arbor, MI, USA) according to the manufacturer's instructions. Each sample was measured in triplicate in two dilutions. Data were normalized to the IL-1α positive control.

MiRNA profiling by microarray
For RNA extraction Normo and Hypo FBS-and XFS-EVs were suspended in 1 ml of Qiazol reagent (Qiagen, Limburg, The Netherlands). EV-RNA was extracted with the miRNeasy Micro Kit (Qiagen) according to the manufacturer's instructions and stored at -80 °C. MiRNA content was evaluated by the Qubit® 2.0 Fluorometer using the Qubit® microRNA Assay Kit (Thermo Fisher Scientific) and small RNA quality and size were assessed by Agilent 2100 Bioanalyzer using the Small RNA chip (Agilent Technologies, Santa Clara, CA, USA). The expression of 2,549 EV-miRNAs was investigated by microarray using SurePrint Human miRNA platform (Agilent Technologies; AMADID: 070156). Fifty ng of EV-RNA and Spike-in controls were dephosphorylated and labelled with Cyanine 3-pCp. Then, Cy3-miRNA samples were purified using Micro Bio-Spin P-6 Column (Bio-Rad Laboratories, Inc., Hercules, CA, USA) and hybridized with a synthetic DNA poly-A oligonucleotide 3' labelled pCp-Cy3 (50 amol, TIB Molbiol SRL, Genoa, Italy) on the microarray slide. Fluorescent signals were then acquired by a G2565CA scanner (Agilent Technologies) and data were extracted by Feature Extraction software v.9.5.3.1 (Agilent Technologies). Raw data were processed using the limma R package for microarray analysis. Background correction and between-array normalization were carried out using the normexp method, with an offset = 20, and the quantile method, respectively. A log(fold change)>0.378 was assumed to represent upregulated miRNAs, while miRNAs with log(fold change)<-0.378 were defined as downregulated. MiRNAs with p < 0.01 (-log10=2.0) were considered statistically significant. Raw and processed data are uploaded at the Gene Expression Omnibus repository (http://www.ncbi.nlm.nih.gov/ geo/; GEO ID: GSE209585). The list of upregulated and downregulated miRNAs was loaded on DIANA miRPath 3.0 [16], to perform Gene Ontology analysis on target genes. Pathway involvement was firstly investigated by using the algorithm for KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway database examination. Target genes were predicted in silico by using miRsystem [17].
To perform the miRNA functional assay, after transfection for 48 h, medium was removed, and cells were stimulated with 200 U/mL IL-1α for 16 h and then total RNA prepared from chondrocytes was used to check the expressions of IL-6, IL-8, COX-2, COL1A1, COL2A1, COL10A1, SOX9 (SRY-box transcription factor 9) and RUNX2 (Runt-related transcription factor 2). Sequences of selected primers are listed in Table 1.

Statistical analysis
Data were analyzed with GraphPad Prism 9.3 software (GraphPad Software, Inc., San Diego, CA, USA). One-way ANOVA test was performed for western blot on IL-6, IL-8, and COX-2, real time-PCR on IL-6, IL-8, COX-2, COL1A1, COL2A1, COL10A1 and SOX9, NF-κB nuclear translocation, PGE2 ELISA quantification and miRNA functional tests. Two-way ANOVA was used for statistical analysis of cell viability, BCA protein assessment, NTA analysis, non-conventional flow cytometry on sEVs (MFI evaluation) and miRNA comparison. Level of significance was set at p < 0.05 (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). The number of data used for the statistical analyses is reported in the figure legends and corresponds to independent experiments. Data are shown as mean ± SD.

Hypoxic pre-conditioning of hBMSCs enhances sEV protein enrichment
Changes in oxygen concentrations might affect stem and progenitor cells [18,19] and this can be mirrored in the secreted EVs. On this basis, we evaluated whether hypoxic conditioning of human BMSCs could influence their EV secretion. Sub-confluent primary hBMSCs cultured in FBS-or XFS-containing medium were maintained for 72 h in basal medium in an either normoxic or hypoxic environment. To exclude any negative effect on cell viability, early and late apoptotic cells were evaluated. In both normoxia and hypoxia, XFS-hBMSCs showed a higher percentage of live cells (89.9 ± 3.64% and 92.0 ± 4.61%, respectively) together with a decrease in early apoptotic cells (5.8 ± 0.40% and 4.94 ± 2.84), compared to FBS-hBMSCs (live cells: 67.16 ± 3.57% and 67.5 ± 7.99; early apoptotic cells: 26.8 ± 5.07% and 23.63 ± 4.30%) (p < 0.0001 for all the comparisons) ( Figure 1A).

Xeno-free culture enhanced EV secretion of hBMSCs
To detect particle concentration and size distribution, FBS-and XFS-sEVs both in normoxia and hypoxia were further analyzed via NTA using the Zetaview. No difference was observed in size distribution between the different experimental groups with particles showing an average size of 140-160 nm throughout ( Figure 1C). The analysis also showed a reasonably high number of particles/cell in all the four conditions with an almost two-fold increase in XFS-sEVs, both in normoxia (303 ± 57.94 particle/cell) and hypoxia (328 ± 23.64 particle/cell) compared to FBS-sEVs (150 ± 3.77 particle/cell and 195 ± 19.03 particle/cell, respectively), Figure 1D. Contrary to BCA results, no significant difference was found in particle number between normoxic and hypoxic sEVs in both culture conditions (FBS and XFS). The ratio between the number of particles and protein content underlined that in both FBS-and XFS-sEVs, the hypoxic pre-conditioning did not induce a higher secretion of EVs, but rather a protein enrichment ( Figure 1E).
EVs were also investigated by TEM, to confirm the efficient isolation. They exhibited the typical morphology of vesicle surrounded by a lipidic bilayer (Figure 2A), with a low level of clustering in all the samples. EV characterization by Western blot revealed that both FBS-and XFS-sEVs expressed the specific EV markers, such as CD81 and CD63. However, Hypo FBS-sEVs showed a higher expression of both proteins compared to their normoxic counterpart, whereas in XFS-sEVs this increase was less prominent ( Figure 2B). No CD9 expression was found in all the samples analyzed. Syntenin-1 and flotillin-1 were found to be abundant in both FBS-sEVs and XFS-sEVs. To rule out any contamination from other cellular compartments, the expression of the chaperon calnexin, an endoplasmic reticulum trans-membrane protein [20], was also analyzed. However, it was detected in cell lysate controls only ( Figure 2B).

XFS-sEVs are more efficiently internalized from OA-hACs
The development of OA is strictly related to the upregulation of pro-inflammatory cytokines [21] and MSC-derived EVs have recently been reported as cell-free anti-OA therapeutic strategy [22][23][24][25].
To evaluate the role exerted by FBS-and XFS-sEVs in cartilage repair and OA, the interactions of sEVs with recipient cells was preliminarily investigated. Human articular chondrocytes (hACs) were treated with IL-1α to induce inflammation and mimic the OA environment. OA-hACs were subsequently incubated with PKH67-stained FBS-or XFS-sEVs derived from both normoxic and hypoxic hBMSCs. hACs efficiently internalized both FBS-and XFS-sEVs in normoxia and hypoxia within their cytoplasm ( Figure 4A).
A further analysis by confocal microscopy was performed, to verify the effective uptake of sEVs and to exclude any event of sEV adhesion on the cell surface ( Figure 4B). The three-dimensional recons-truction ( Figure 4C) evidenced that hACs internalized EVs, localizing them around the nucleus, with a comparable average distance in all the conditions (Table 3). However, the calculated volumetric density (sEVs/mm 3 ) showed that XFS-sEVs derived from hypoxic cells were internalized by hACs to a larger extent than FBS-hBMSC-EVs (Table 3). Table 3. Quantitative analysis based on confocal microscopy.
Quantitative PCR on IL-6, IL-8 and COX-2 genes confirmed the strong anti-inflammatory effect of Hypo FBS-sEVs and Normo and Hypo XFS-sEVs, especially on weakening the IL-6 (Hypo FBS sEVs and Normo XFS sEVs: p < 0.001; Hypo XFS sEVs: p < 0.0001) and COX-2 expression (p < 0.01 in all the conditions, Figure 5C). Expression of collagens genes was also investigated ( Figure 5C). Interestingly, COL1A1 and COL2A1 expressions were not significantly affected by IL-1α and sEVs treatment, compared to negative control. On the other hand, COL10A1 showed a significant increase after IL-1α treatment, compared to untreated cells (p < 0.001). Normo FBS-sEVs exerted a weak not significant inhibitory effect on COL10A1 expression, while hypoxia pre-conditioning increased FBS-sEVs effectiveness (p < 0.01), reflecting the biological effect observed on pro-inflammatory cytokines expression. Similarly, both Normo and Hypo XFS-sEVs showed a strong ability to revert COL10A1 expression compared to IL-1α treatment (Normo: p < 0.01; Hypo: p < 0.001). Taken together, these results could suggest a beneficial role of sEVs in OA treatment.
PGE2 expression was also evaluated after treatment with IL-1α ± FBS-or XFS-sEVs ( Figure 6C). Both Normo and Hypo FBS-sEVs were able to blunt PGE2 production (p < 0.05 vs. IL-1α), without significant differences between the two groups. Analogously to NF-κB modulation, XFS-sEVs both in normoxia and hypoxia exhibited a robust inhibitory effect on PGE2 secretion (p < 0.001 vs. IL-1α). No differences were observed between normoxia and hypoxia pre-conditioning.
These results suggest that both FBS-and XFS-sEVs could potentially affect OA-related inflammation.

FBS-and XFS-sEVs exhibited a different miRNA cargo
EVs effect their biological function through the combined activity of various molecules, including miRNAs, which mediate EV effects on recipient cells regulating the expression of target genes [29]. To investigate whether the preconditioning strategies had affected the miRNA cargo within both FBS and XFS-sEV, the expression profiles of 2,549 EV-miRNAs were evaluated by microarray. Specifically, we compared how Normo or Hypo pre-conditioning, as well as the culture supplements (i.e., FBS and XFS), could influence sEV-miRNA enrichment (Figure 7).
Conversely, fewer EV-miRNAs were found differentially expressed in FBS and XFS-sEVs, after hypoxic preconditioning ( Figure 7A, in green). Only three upregulated and three downregulated miRNAs were detected ( Figure 7C-D). This data led to the hypothesis that hypoxic pre-conditioning lowered differences between XFS-and FBS-sEVs and may explain the dissimilar anti-inflammatory effect of Hypo and Normo FBS-sEVs described above.

XFS-sEVs selected a miRNA cargo involved in cartilage homeostasis
To better understand the role of differentially expressed miRNAs in sEV biological activity and their role in anti-inflammatory activity, a gene ontology analysis based on in silico predicted target genes, was carried out ( Figure 8A). Eight of the ten miRNAs downregulated in Normo XFS-sEVs and Hypo FBS-sEVs compared to Normo FBS-sEVs (miR-1202, miR-1236-5p, miR-1299, miR-3137, miR-4646-5p, miR-6793-5p, miR-6870-5p, miR-760) were commonly involved in mucin type O-glycan biosynthesis, regulation of pluripotency of stem cells and the Wnt signaling pathway.
On the other hand, miR-21, miR-145, miR-214, miR-301a-5p and let-7b-5p, which were upregulated in XFS-sEVs compared to FBS-sEVs in normoxia, were predominantly implicated in TGF-beta and PI3K-Akt signaling pathways. Both miR-30c-5p and miR-30b-5p were found to be involved in all pathways mentioned above and in glycosaminoglycan biosynthesis to a limited extent.
Finally, the expression levels of collagen genes were also investigated, showing significant deregulations after miRNA transfections. Interestingly, COL2A1 (type II collagen), a specific marker in both hyalin and articular cartilage, was significantly downregulated under the inflammatory stimulus compared to non-transfected negative control (p < 0.05). On the contrary, COL1A1 (type I collagen) did not show significant variations after transfections with selected miRNA. Transfection with miR-301a or miR-21 was able to protect chondrocytes by IL-1α-induced COL2A1 expression (p < 0.05 in both conditions). On the other hand, miR-301a also determined a sustained increase in COL10A1 (type X collagen, p < 0.0001), a typical marker of hypertrophic chondrocytes.
Taken all together, these results suggested that XFS-sEVs anti-inflammatory properties are not attributable to a single miRNA, but rather to an orchestra of EV-encapsulated miRNAs exerting a synergic effect, by targeting different pathways.

Discussion
Recent studies have shown the effectiveness of MSCs in cartilage repair and OA-related symptom relief [1]. However, limited evidence demonstrated that injected MSCs directly contribute to joint tissue restoration. Notwithstanding, they could influence the surrounding environment and mediate endogenous tissue repair by paracrine activity, fulfilled through the secretion of soluble and EV-encapsulated bioactive factors [38][39][40].
EVs are small, membrane-enclosed particles released by nearly all type of cells and generated either from cell membrane budding (microvesicles) or endosomal compartments (exosomes) [41]. They incorporate biologically active signaling molecules, acting as innovate players in intercellular communication [42]. The cargo and function of EVs are recurrently influenced by the pathophysiological state of secreting cells and the microenvironment [41]. In this scenario, an appropriate preconditioning may prompt cells to execute the most suitable therapeutic response. Several methods, as low oxygen tension, starvation or inflammatory stimuli have been reported as strategies to mimic the typical environment of injury sites [12]. Hypoxic conditioning has been proven to enhance proliferation of MSCs without loss of multi-lineage differentiation capabilities [18]. It has also been reported that hypoxia can impacts on EV secretion in adipose tissue-derived MSCs [43], as well as in breast cancer cells like MCF-7 [44] and in endothelial cells [45].
This study focused on paracrine functionality of hBMSCs cultured in a xeno-free culture system, in pursuance of a clinical application. This would allow to obtain a clean, safe and ready-to-use clinical grade product. EVs were separated and characterized and as previously described [14], XFS allows the isolation of cells secreting a higher amount of EVs, although the typical vesicular marker did not significantly vary compared to the control culture condition. Here hypoxic pre-conditioning did not affect the number of EVs secreted, but rather the protein amount, in line with previous data reported for amniotic fluid stem cells derived EVs [46].
Hypoxia was also recently reported to enhance the cartilage repair effect of MSC-EVs [47,48]. According to these recent findings, the anti-inflammatory effect of FBS-and XFS-sEVs was evaluated in an in vitro model, using IL-1α to simulate the inflammation occurring during OA. Hypo FBS-sEVs showed a superior inhibitory effect on the production of pro-inflammatory cytokines in OA-chondrocytes, compared to Normo FBS-sEVs, confirming previous evidence. On the other hand, the XFS culture condition selected MSCs secreting EVs with strong anti-inflammatory properties, which are not enhanced by hypoxic pre-conditioning. The anti-inflammatory effect is also supported by the observation that XFS-sEVs are more efficiently internalized from OA-hACs. This different behavior could be related to a distinctive cargo selection. MiRNA content of hypoxic FBS-sEVs showed indeed a similar profile to XFS-sEVs, both exhibiting a downregulation of miRNAs involved in mucin type-O glycan biosynthesis. It has been indeed reported that the glycosylation signature of chondrocytes and extracellular matrix composition are altered during OA progression [49]. In healthy cartilage, a complicated balance exists between new matrix deposition and its digestion. In OA establishment, articular cartilage degradation includes an anabolic phase initially where chondrocytes struggle to repair the damage, depositing an impaired extracellular matrix. A subsequent catabolic phase by matrix digestion leads to cartilage erosion [50].
According to our results on chondrocytes, miR-214 has been reported to inhibit inflammatory factors (TNF-α, IL-1β, IL-6 and IL-18) in TNF-α-treated vascular endothelial cells through down-regulation of PTEN and NF-κB protein expression [52]. Among miR-214 target genes, the skeletal key players SOX9 and RUNX2 were particularly relevant. SOX9 is unmistakably recognized as a master transcriptional factor of cartilage differentiation and development [31], while RUNX2 is essential for osteoblast differentiation and chondrocyte maturation. Different studies have indeed confirmed that an abnormal expression of miR-214 could determine bone and cartilage development disorders and lead to the onset of OA [53]. These data suggest a dual role of miR-214 both in OA prevention and chondrogenesis.
Additionally, our results also demonstrated that transfection of chondrocytes with miR-21 mimic induced SOX9 and RUNX2 expression and preserved cells by IL-1α-induced COL2A1 downregulation, suggesting a protective role in OA pathogenesis. It has been indeed already reported that up-regulation of miR-21-5p in IL-1β-treated chondrocytes increased COL2A1 expression and decreased MMP13 and ADAMTS5 levels [54]. Nevertheless, miR-21 overexpression also led to an enhanced expression of pro-inflammatory cytokines, such as IL6. Further studies are needed to better elucidate the role of miR-21 in OA.
Expression of COL2A1, together with COL10A1, was also upregulated by miR-301a. Type II collagen is massively synthesised by resting chondrocytes in the growth plate, whereas type X collagen is produced by hypertrophic chondrocytes before mineralisation of the ECM occurs [55]. Alterations in expression of COL2A1 or COL10A1 could be related to impairment and degradation of hyalin cartilage or altered subchondral bone formation, respectively. These results could indicate a putative role of miR-301a in joint homeostasis and maintenance.
Interestingly, let-7b-5p did not show modulations on analysed genes. However, these analyses unveil some target genes of let-7b-5p, such as ADAMTS5, ADAMTS8, TGFBR1, WNT1 and some fibrotic collagens (e.g. COL19A1, COL15A1). In detail, ADAMTS5 is unequivocally implicated in OA pathogenesis through degradative activity on aggrecan, one of the major structural components of articular cartilage. Several ADAMTS proteases cleave the aggrecan core protein, compromising joint mechanical properties and exposing other structural molecules, such as type II collagen, to proteolysis [56].
Finally, miR-30b-5p and miR-30c-5p emerged for their target genes profiles. Both were shown to target SOX9, inducing its upregulation. Interestingly, miR-30b and miR-30c also target ATG5 and BECN1, master regulators of autophagy. Autophagy is a cellular homeostasis mechanism, supporting cell survival under stress-induced conditions such as nutrient deprivation, hypoxia, and injury. This catabolic process regulates energy and nutrients through the removal of damaged or dysfunctional proteins and organelles. In articular cartilage tissue, the resident chondrocyte population cannot be replenished via vasculature. Autophagy is thus essential for maintaining cell and tissue homeostasis [57,58].
Taken together, presented data indicated that XFS culture medium is a relevant tool to obtain clean and ready-to-use cells and cellular products, amenable for clinical treatment of diseases such as OA, thanks to their miRNAs content. This study showed for the first time a potential role of EV-encapsulated miR-145 and miR-214 in blunting OA-related inflammation. However, the absence of an in vivo validation represents a limitation for the study and further experiments are ongoing to fill the research gap and confirm the presented data in an osteoarthritic mouse model.