HLSCs were obtained and characterized as previously described (Herrera et al., 2006; Bruno et al., 2019; Spada et al., 2020). HLSCs were expanded in the presence of Minimal Essential Medium (α-MEM, Lonza, Basel, Switzerland) supplemented with 10% Fetal Calf Serum (Gibco/Cambrex, Invitrogen, Carlsbad, CA, United States), 10 ng/mL of human recombinant Epidermal Growth Factor, 10 ng/mL of human recombinant Fibroblast Growth Factor basic (Miltenyi Biotec, Bergisch Gladbach, Germany) and used until passage 8 (Bruno et al., 2019; Spada et al., 2020). Human bone marrow-derived mesenchymal stromal cells (MSCs) were purchased by Lonza (Basel, Switzerland), maintained in a MSC basal medium bullet kit (Lonza) and used until passage 6. Human hepatic stellate cells LX-2 (Xu et al., 2005) (Sigma Aldrich, St. Louis, MO, United States), were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) high glucose (4.5 g/L, Euroclone, Pero, MI, Italy) supplemented with 2% Fetal Calf Serum and 2 nM L-Glutamine (Lonza) and used until passage 8. All cells were maintained in a humidified 5% CO2 incubator at 37°C.

For EV collection, HLSCs were seeded in hyperflasks (Corning, VWR International, Milano, Italy) at the density of 3,000 cells/cm², while MSCs were seeded in T150 flasks (Corning) at the density of 2,000 cells/cm². Both cells were cultured in standard conditions until they reach 70–80% of confluence. EVs were isolated from supernatants of sub-confluent HLSCs cultured in serum-free α-MEM (Lonza) and sub-confluent MSCs cultured in serum-free RPMI 1640 (Euroclone) for 18 h. The next day, the cell supernatant was collected, then subjected to centrifugation at 3,000 g for 15 min at 4°C and microfiltrated with 0.22 μm filters to remove cell debris and apoptotic bodies, transferred to polycarbonate centrifuge bottles (Beckman Instruments) and then ultracentrifuged at 100,000 g for 2 h at 4°C (Beckman Coulter Optima L-100K, Fullerton, CA, United States). The pellet obtained was resuspended in RPMI supplemented with 1% dimethyl sulfoxide (DMSO, Sigma-Aldrich) and stored at −80°C until use for subsequent experiments.

In selected experiments, an iodixanol floating separation protocol was used to further purify the EV preparations from free floating contaminating RNAs and proteins (Kowal et al., 2016; Ranghino et al., 2017). After ultracentrifugation of the EV samples, pellets obtained were resuspended in 60% iodixanol (Optiprep from Sigma-Aldrich), mixed with 0.25 M sucrose and transferred to polycarbonate centrifuge bottles (Beckman Instruments). Subsequent gradients of iodixanol at 30, 15, and 5% were then layered on top of the EV-containing 60% iodixanol-sucrose preparation. The tubes were then ultracentrifuged without braking at 350,000 g for 1 h at 4°C (Beckman Coulter Optima L-100K). The 15 and 30% fractions were then recovered, diluted with phosphate-buffered saline (PBS), and ultracentrifuged again at 100,000 g for 1 h. The pellet obtained was resuspended in RPMI supplemented with 1% DMSO and stored at –80°C for subsequent studies.

EVs diluted at least 1:200 in sterile saline solution were examined using the NanoSight LM-10 instrument (NanoSight Ltd., Amesbury, United Kingdom), equipped with a 405 nm laser. For each EV preparation, the recordings of three 60-s videos were analyzed by the Nanoparticle Tracking Analysis Software (NTA version 3.4), in order to assess EV concentration.

The phenotype of EVs was characterized by cytofluorimetric analysis. As previously described (Koliha et al., 2016; Bruno et al., 2020b), approximately 2 × 10⁹ EVs of 7 batches of HLSC-EVs and 7 batches of MSC-EVs were processed using a bead-based multiplex analysis system (MACSPlex Exosome Kit, human, Miltenyi Biotec). Briefly, EV samples were diluted with MACSPlex buffer to a final volume of 120 μL and incubated with MACSPlex Exosome Capture Beads (containing 39 different antibody-coated bead subsets) for 18 h at 450 rpm. Counterstaining was performed by incubating the EVs bound by capture beads with the APC-conjugated anti-CD9, anti-CD63, and anti-CD81 detection antibodies for 1 h at 450 rpm, protected from light. Washing steps were performed following manufacturer’s instructions. Approximately 5,000 single bead events were recorded per sample using a Cytoflex (Beckman Coulter, Brea, CA, United States). Following manufacturer’s instructions, all bead populations were identified and gated based on their respective fluorescence intensity and the median fluorescence intensity (MFI) for each capture bead subset was calculated using the CytExpert Software. To remove the background fluorescence intensity, the MFI value of a blank control that was processed exactly like the EV-containing samples (medium + capture beads + detection antibodies) was subtracted from the MFI of all 39 capture bead subsets.

Transmission electron microscopy was performed to evaluate the size and the integrity of EVs. Approximately 3 × 10⁹ EVs were placed on 200 mesh nickel formvar carbon-coated grids (Electron Microscopy Science, Hatfield, PA, United States) and left to adhere for 20 min, as previously described (Deregibus et al., 2016). Then, after washing in PBS the grids were incubated with 2.5% glutaraldehyde containing 2% sucrose and extensively washed in distilled water. Finally, the EVs were negatively stained with NanoVan (Nanoprobes, Yaphank, NK, United States) and observed using a Jeol JEM 1400 Flash electron microscope (Jeol, Tokyo, Japan).

For protein analyses, cells and EVs were lysed at 4°C for 30 min in RIPA buffer (20 nM Tris-HCl, 150 nM NaCl, 1% deoxycholate, 0.1% SDS, 1% Triton X-100, pH 7.8) supplemented with 1% phenylmethylsulfonyl fluoride, 1% protease and phosphatase inhibitors cocktail (all purchased from Sigma-Aldrich). BCA Protein Assay Kit (Pierce^TM Thermo Fisher Scientific, Waltham, MA, United States) was used to assess protein concentration of EV and cells lysates. Then, 10 μg of protein samples were loaded on 4–20% gradient Mini-PROTEAN TGX precast electrophoresis gels (Bio-Rad, Hercules, CA, United States), separated under reducing conditions and electroblotted onto 0.2-μm nitrocellulose membranes using the Trans-Blot Turbo Transfer System (Bio-Rad). Membranes were blocked with 5% bovine serum albumin in PBS supplemented with 0.1% Tween-20 before incubation for 18 h at 4°C with primary antibodies. The following primary antibodies were used: mouse anti-human vimentin, mouse anti-human alpha-smooth muscle actin (both from Sigma-Aldrich), rabbit anti-collagen type I-alpha 1 (Cell Signaling Technology, Danvers, MA, United States), mouse anti-CD63, mouse anti-Alix (both from Santa Cruz Biotechnology, CA, United States), mouse anti-TSG101 and rabbit anti-GM130 (both from Abcam, Cambridge, United Kingdom). After intensive washing, membranes were incubated for 1 h with appropriate peroxidase conjugated secondary antibodies (Thermo Fisher Scientific). Chemiluminescent signals were detected by the Chemidoc system using the enhanced chemiluminescence substrate (Bio-Rad).

Human Pro-Collagen I alpha 1 SimpleStep ELISA^® Kit was used to quantify the release of pro-collagen I α1 (pro-COL1A1) from LX-2 in selected experiments. About 1 mL of cell supernatant was centrifuged at 2,000 × g for 10 min to remove debris and stored at −20°C until the assay was performed following manufacture’s protocol. Briefly, cell supernatant was diluted up to 1:40 into Sample Diluent NS in a 96-well plate with pro-collagen I alpha-1 antibody-coated wells and incubated with a mixture of capture and detector antibodies for 1 h at room temperature. The unbound material was removed by extensive washing. To allow the enzymatic reaction, development solution was added to the plate and incubated for 10 min in the dark. The generated signal was proportional to the amount of bound analyte and its intensity was recorded spectrophotometrically (Bio-Rad) at 450 nm.

Total RNA was isolated from cells and EVs using miRNeasy mini kit (Qiagen, Valencia, CA, United States) according to manufacturer’s instructions, and it was quantified spectrophotometrically (mySPEC, VWR, Radnor, Pennsylvania, United States). RNA was converted into cDNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, United States) to assess the expression of specific genes, while the miScript II RT Kit (Qiagen) was used to evaluate microRNA (miRNA) expression. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed using a 96-well QuantStudio 12K Flex Real-Time PCR system (Thermo Fisher Scientific). Each well was loaded with 20 μL reaction mixture containing 5 ng of sample cDNA, Power SYBR Green PCR Master Mix (Applied Biosystems) for gene expression evaluation and miScript SYBR Green PCR kit (Qiagen) for miRNA expression analysis. Sequence-specific oligonucleotide primers (100 nM, purchased by MWG-Biotech, Eurofins Scientific, Brussels, Belgium) are listed in Table 1 (genes) and Table 2 (miRNAs). To normalize gene expression, TBP was used as housekeeping gene, while miRNA expression was normalized using miR-24-3p. Fold change expression with respect to control was calculated for all samples using ΔΔCt method.

Preliminary experiments on LX-2 incubated for 6–18–24–48 h with 10 ng/mL of transforming growth factor-β1 (TGF-β1, purchased by Sigma-Aldrich) indicated that TGF-β1 induced an activated phenotype in LX-2 at every timing tested. Stimulation with TGF-β1 induced a myofibroblast-like phenotype in LX-2 (Supplementary Figure 1A). An increased expression of α-SMA and COL1A1 was observed by qRT-PCR analysis (Supplementary Figure 1B) and the increment of the two pro-fibrotic markers was confirmed at protein level by western blot analysis (Supplementary Figures 1C,D). Furthermore, the release of pro-COL1A1 in LX-2 supernatant increased after stimulation with TGF-β1 (Supplementary Figure 1E). A 6-h incubation with 10 ng/mL of TGF-β1 was chosen to activate LX-2 in subsequent experiments.

LX-2 were seeded at a density of 15,000 cells/cm² and synchronized over-night in serum-deprived medium, supplemented with 0.2% of bovine serum albumin. To induce the activated phenotype, LX-2 were incubated with TGF-β1 (10 ng/mL) for 6 h. In preliminary experiments, activated LX-2 were stimulated for 24 h with different doses of HLSC-EVs and MSC-EVs: 1,000 (1 k)–10,000 (10 k)–50,000 (50 k)–100,000 (100 k) particles per cell. Further experiments were performed up to 72 h using a single administration of 50 k EVs per cell or a lower dose of 10 k EVs per cell repeatedly administered once a day for 3 days. In selected experiments, LX-2 were stimulated for 18 h with EVs in the presence of α-amanitin (10 μg/mL, purchased by Sigma-Aldrich) to inhibit cell transcription. Cells were lysed at 24–48–72 h for RNA extraction and at 72 h for protein isolation.

Prediction of miRNAs targeting α-SMA and COL1A1 transcript was performed by miRWalk 3.0, considering those miRNAs with a p < 0.05 and targeting the 3′ or the 5′ untranslated region (3′UTR, 5′UTR), or the coding sequence (CDS). Comparison analysis between miRNAs vehicled by HLSC EVs (Grange et al., 2019) and predicted miRNAs was performed with Funrich V3 software (Pathan et al., 2017).

The day before transfection, LX-2 were seeded into a 6-well plate at a density of 15,000 cells/cm². Once activated with TGF-β1 (10 ng/mL for 6 h), LX-2 were transiently transfected with miR-146a-5p mimic (miRCURY LNA miRNA Mimic, Qiagen) using the HiPerFect Transfection Reagent (Qiagen), according to the manufacturer’s protocol. Briefly, miR-146a-5p mimic at a concentration of 50 nM was mixed with an appropriated volume of HiPerFect Transfection Reagent and incubated for 10 min at room temperature. Same concentration of a negative control mimic (Qiagen) was used as scrambled control. The transfection mix was added dropwise to the cells in FBS-depleted culture medium. Twenty four hours after the transfection, cells were lysed for RNA extraction. Protein lysates were performed 48 h after the transfection.

Data analyses were performed using GraphPad Prism v.8.0. Results are expressed as mean ± SD. Statistical analyses were performed using ANOVA with Dunnett’s multiple-comparisons test. A p-value of < 0.05 was considered significant.

Both EVs from HLSCs and MSCs showed the expression of the classical exosomal markers CD9, CD63, CD81, and the mesenchymal markers CD29, CD44, and CD105. They did not express HLA-1 and HLA-DR, the epithelial markers CD24 and CD326, and the endothelial (CD31) and hematopoietic (CD45) markers (Figures 1A,B). The expression of exosomal markers (ALIX, TSG101, and CD63) was confirmed by western blot analysis (Figure 1C). Transmission electron microscopy analysis showed that HLSC-EVs and MSC-EVs retain a homogeneous pattern of nano-size membrane vesicles (Figures 1D,E).

Preliminary experiments on TGF-β1-activated LX-2 stimulated for 24 h with a single administration of HLSC-EVs (from 1 k to 100 k particles per cell) showed a dose-dependent down-regulation of α-SMA in activated LX-2 that was more evident at higher doses (Figure 2A). The effect of HLSC-EVs on activated LX-2 was comparable to that obtained using MSC-EVs (Figure 2B).

To further evaluate the effect of EVs on the activated phenotype of LX-2 at different timings, activated LX-2 were incubated with HLSC-EVs or MSC-EVs, using a single administration of 50 k EVs per cell or a lower dose of 10 k EVs per cell repeatedly administered once a day for 3 days (Figure 3A). The expression of pro-fibrotic genes α-SMA and COL1A1 was evaluated by qRT-PCR analysis (Figures 3B,C). The stimulation of LX-2 with HLSC-EVs reduced the expression levels of α-SMA and COL1A1 at 24, 48, and 72 h: in particular, the reduction of these two pro-fibrotic genes was significant for LX-2 incubated with the 3 doses of 10 k EVs (Figure 3B). After 24 h of stimulation with MSC-EVs, a significant reduction in the expression of α-SMA and COL1A1 was detectable. However, after 72 h of incubation with MSC-EVs, this effect was no longer evident as the expression levels of α-SMA and COL1A1 were comparable to those of activated LX-2 (Figure 3C). Protein expression of the pro-fibrotic markers was also assessed by western blot analysis at 72 h (Figures 3D–G). The stimulation of LX-2 with the 3 doses of 10 k HLSC-EVs reduced the protein expression of α-SMA, whereas the incubation with a single dose of 50 k HLSC-EVs up-regulated in a significant manner the expression of COL1A1 in LX-2 (Figures 3D,E). The stimulation of LX-2 with MSC-EVs increased the expression of α-SMA and COL1A1 (Figures 3F,G).

These results indicate that a persistent attenuating effect on LX-2 activation state is achieved through incubation with HLSC-EVs, while MSC-EVs induce only a transient attenuating effect.

The amount of pro-COL1A1 released by activated LX-2 was measured after 24, 48, and 72 h of incubation with HLSC-EVs and MSC-EVs. After 72 h of stimulation with HLSC-EVs, the amount of pro-COL1A1 secreted by LX-2 was significantly reduced by the 3 doses of 10 k EVs (Figure 4A). On the contrary, after 72-h incubation with 50 k MSC-EVs, a strong increase in the pro-COL1A1 release in LX-2 cell supernatant was observed (Figure 4B). Moreover, we noticed a significant increase in the amount of pro-COL1A1 after 24-h stimulation with the single dose of 50 k HLSC-EVs and MSC-EVs (Figures 4A,B). To verify if the EVs could be responsible of the increase in pro-COL1A1 secretion by LX-2, we performed western blot analysis on HLSC-EVs and MSC-EVs and we demonstrated that both contain COL1A1 (Figure 4C). In particular, compared to HLSC-EVs, MSC-EVs showed higher levels of COL1A1 (Figure 4D).

Since the presence of COL1A1 in EVs could depend on the ultracentrifugation (UC) method used to purify them from cells, HLSC-EVs and MSC-EVs were also isolated by floating (FL). Transmission electron microscopy analysis showed that FL-HLSC-EVs and FL-MSC-EVs retained a homogeneous pattern of membrane vesicles (Figures 4E,F). Western blot analysis on FL-HLSC-EVs and FL-MSC-EVs demonstrated that only FL-MSC-EVs still contain COL1A1, although at lower levels compared to UC-MSC-EVs (Figures 4C,D).

We demonstrated that the effect of FL-HLSC-EVs on activated LX-2 was comparable with the effect of UC-HLSC-EVs (Figures 4G,H). In fact, a single administration of EVs (both 10 and 50 k) for 24 h downregulated the expression of α-SMA. Moreover, the pro-COL1A1 release in LX-2 cell supernatant was not increased after 24-h incubation with FL-HLSC-EVs (Figure 4I). These results suggest that the presence of COL1A1 protein associated with UC-HLSC-EVs does not alter their biological effect on LX-2.

The miRNA content of HLSC-EVs and MSC-EVs was previously characterized (Grange et al., 2019). Based on these results, we investigated whether HLSC-EVs might contain miRNAs that could possibly target α-SMA and COL1A1 and modulate their expression in LX-2. We considered the 20 most expressed miRNAs in HLSC-EVs (Supplementary Table 1) and we predicted their interaction with the CDS, 3′UTR, 5′UTR regions of α-SMA, and COL1A1 mRNA using miRWalk prediction tool. The predictive analysis showed that 6 miRNAs expressed at high levels in HLSC-EVs could possibly target both α-SMA and COL1A1 mRNA (Table 3). QRT-PCR validation confirmed the presence of these 6 miRNAs in HLSC-EVs and their enrichment compared to the one in MSC-EVs (Figure 5A). Notably, miR-146a-5p resulted the miRNA with the highest expression levels in HLSC-EVs, compared to MSC-EVs. Then, we evaluated the expression of these 6 miRNAs into activated LX-2 incubated with HLSC-EVs for 24 h. We demonstrated that the 24-h treatment with HLSC-EVs up-regulated only the expression of miR-146a-5p (Figure 5B).

To further investigate whether miR-146a-5p is transferred into LX-2 by HLSC-EVs, we treated LX-2 with α-amanitin, a selective inhibitor of RNA polymerase II and III. Then, we evaluated the expression of miR-146a-5p into activated LX-2 stimulated for 24 h with HLSC-EVs. In the presence of α-amanitin, we observed a significant increase in the expression of miR-146a-5p in LX-2 stimulated with 50 k HLSC-EVs, indicating that miR-146a-5p is transported by HLSC-EVs into LX-2 (Figure 5C). However, the increase occurs to a lesser extent than in LX-2 not treated with α-amanitin, suggesting also an induction of the expression of miR-146a-5p in HLSC-EV-stimulated LX-2.

To evaluate whether the induction of miR-146a-5p may interfere with the activated state of LX-2, we transfected activated LX-2 with miR-146a-5p mimic. LX-2 transfected with a negative control mimic (scramble) were used as reference control. We observed that miR-146a-5p mimic efficiently up-regulated the expression of miR-146a-5p in activated LX-2 (Figure 6A). Besides, miR-146a-5p mimic reduced COL1A1 protein expression in LX-2 (Figures 6B,C), thus indicating an attenuation in the activated state of LX-2.