HLSCs were generated from one donor liver belonging to the category of standard risk. Anemocyte International (Gerenzano, Italy) obtained the ethical approval and consent to participate of the donor to generate the master cell bank by the National Transplant Center in accordance with Italian legislation. HLSCs were cultured as previously described to generate a master cell bank [4, 5, 20]. The phenotypic characterization of HLSCs is provided by Table S1. For this study, cryopreserved cells in fetal bovine serum (Euroclone, Pero, Milano, Italy) in the presence of 10% dimethyl sulfoxide (Sigma-Aldrich^®) were thawed and expanded in T75 flasks. For EVs collection, sub-confluent HLSCs were cultured in serum-free Minimum Essential Medium Eagle-Alpha Modification (α-MEM) (Euroclone) for 18 h, in HYPERFlask^® (Corning^®, VWR International, Milan, Italy). Cell debris and apoptotic bodies were removed by centrifugation at 3,000 g for 20 min and by microfiltration with 0.22 µm filters. Then, the supernatant was ultracentrifuged at 100,000 g for 2 h at 4°C (Beckman Coulter Optima™ L-90K, Brea, CA, USA), and the EV pellet was resuspended in serum-free culture medium, in the presence of 1% of dimethyl sulfoxide. HLSC-EVs size and concentration were determined using the NanoSight NS300 system (NanoSight Ltd., Amesbury, UK) supplied with a 405 nm laser, and the analysis was carried out by the Nanoparticle Tracking Analysis (NTA) 3.2 software [5]. EVs were stored at −80°C until use for successive experiments.

Cytofluorimetric analysis of EV was performed as previously described [5]. The analysis of 37 different exosomal surface proteins was conducted using a bead-based multiplex analysis kit (MACSPlex Exosome Kit, human, Miltenyi Biotec, Bergisch Gladbach, Germany), according to the manufacturer’s instructions [5, 21]. Briefly, EV samples were mixed with MACSPlex Exosome Capture Beads and kept overnight on an orbital shaker at room temperature, protected from light. Then, the resulting EVs-beads complexes were co-incubated with the anti-tetraspanins detection antibodies and rocked for 1 h, in the dark and at room temperature. After intensive washing with MACSPlex buffer to remove the non-specific binding, the CytoFLEX Flow Cytometer (Beckman Coulter) was employed to acquire at least 5,000 single-bead events per sample. Each capture bead subset population was detected and gated based on its respective fluorescence intensity. The measurement of the median fluorescence intensity (MFI) was performed using the CytExpert Software. The MFI value of a blank control (MACSPlex buffer + capture beads + detection antibodies) was subtracted from the MFI value of each capture bead subset, in order to remove background fluorescence intensity.

Transmission electron microscopy (TEM) analysis was carried out as previously described [5, 22]. EVs were spotted onto 200 mesh nickel formvar carbon-coated grids (Electron Microscopy Science, Hatfield, PA, USA). Once adhered to the grid, EVs were fixed using a solution of 2.5% glutaraldehyde and 2% sucrose, and then washed in distilled water. The negative staining was performed by incubating samples with Nano-W® and NanoVan® (Nanoprobes, Yaphank, NY, USA). Images were acquired using a Jeol JEM-1400 flash electron microscope (Jeol, Tokyo, Japan).

For western blotting analysis, protein lysates obtained from cells and EVs were processed as previously described [9]. Protein concentration was determined by bicinchoninic acid (BCA) Protein Assay Kit (Pierce™ Thermo Fisher Scientific, Waltham, MA, USA). Mini-PROTEAN^® TGX™ Precast Protein Gels (4–20% gradient, Bio-Rad, Hercules, CA, USA) were loaded with protein samples (20 µg), under reducing conditions. Using the Trans-Blot^® Turbo™ Transfer System (Bio-Rad), protein bands were transferred onto 0.2 µm nitrocellulose membranes, which were saturated in phosphate-buffered saline supplemented with 0.1% Tween-20 (PBS-T) and 5% non-fat milk. Then, membranes were probed at 4°C over-night with specific primary antibodies for exosomal markers, all used with a dilution of 1:1,000 in PBS-T supplemented with 5% bovine serum albumin: rabbit anti-CD81 (Cell Signaling Technology, Danvers, MA, USA), mouse anti-CD63 (Santa Cruz Biotechnology, Dallas, TX, USA), mouse anti-TSG101, and rabbit anti-CD9 (Abcam, Cambridge, UK). Primary antibody rabbit anti-GM130 (Abcam) was employed to exclude the presence of cell contaminants in EVs. After repeated washing, membranes were probed for 1 h at room temperature with anti-mouse or anti-rabbit peroxidase-conjugated secondary antibodies (Thermo Fisher Scientific), both diluted 1:5,000 in PBS-T. After incubation with the enhanced chemiluminescence substrate (SuperSignal™ West Femto Maximum Sensitivity Substrate, Thermo Fisher Scientific), the chemiluminescent signal was acquired using the ChemiDoc System (Bio-Rad).

In vivo experiments were carried out in accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals, following the guidelines of Animal Research: Reporting of In Vivo Experiments (ARRIVE) and the mice were randomly assigned to the different experimental groups. All procedures were approved by the Italian Health Ministry (authorization number: 419/2016-PR). NASH was induced by continuously feeding the mice with a methionine and choline deficient (MCD) diet, as previously reported [4, 5, 23–26]. Ten weeks old severe combined immunodeficient (Charles River Laboratories, Wilmington, MA, USA) male mice were accustomed to a MCD diet (MP Biomedicals, Eschwege, Germany) by feeding a mixture of standard and MCD diet. After 1 week of mixed diet, the full MCD diet was administered to the mice. Starting at week 2 of the diet, 2.5 × 10⁹ HLSC-EVs were intravenously injected twice a week (Figure 1A). Each mouse in the EV-treated group received a total of four EV injections (n = 8 mice), while mice in the NASH group were injected with vehicle alone (phosphate-buffered saline, n = 10 mice). Healthy mice (n = 8) were fed with the standard diet. All animals were sacrificed by cervical dislocation after deep sedation, at week 4 and liver tissue was recovered and maintained in RNA later at –20°C for subsequent RNA extraction and molecular analyses, as described below.

The human hepatic immortalized stellate cell line (LX-2) [27] (Sigma-Aldrich^®), was cultured in Dulbecco’s Modified Eagle’s Medium high glucose (4.5 g/L, Euroclone) in the presence of 2 nmol/L L-Glutamine (Lonza, Basel, Switzerland) and 2% fetal bovine serum, and used up to passage 6. After seeding at a density of 15,000 cells/cm² in 6-well plates (Corning), LX-2 was maintained in serum-deprived medium over-night, with the addition of 0.2% of bovine serum albumin (Sigma-Aldrich^®), to synchronize the cell culture. Incubation with transforming growth factor-beta 1 (TGF-β1, 10 ng/mL, purchased by Sigma-Aldrich^®) for 6 h was performed to induce the activation of LX-2. Then, activated LX-2 was stimulated using a single administration of 50,000 (50k) EVs per cell, or a lower dose of 10,000 (10k) EVs per cell repeatedly administered once a day for 3 days, as described [9]. Cells were lysed after 24 h and 72 h of incubation with EVs to perform molecular analysis.

For in vitro experiments, total RNA was collected from LX-2 using the commercially available miRNeasy mini kit (Qiagen, Valencia, CA, USA). For in vivo experiments, total RNA was obtained from liver tissue of healthy mice, NASH mice treated and not treated with HLSC-EVs using TRIzol™ reagent (Ambion, Thermo Fisher Scientific), and homogenized in a Bullet Blender instrument (Next Advance Inc., New York, NY, USA), as previously described [5]. The resulting RNA was diluted in RNase-free water and quantified using mySPEC spectrophotometer (VWR, Radnor, PA, USA).

To perform the lncRNA screening, 1 µg of RNA isolated from the liver was converted into complementary DNA (cDNA) with RT² First Strand Kit and lncRNA expression was analyzed using RT² lncRNA PCR Array Mouse Inflammatory Response & Autoimmunity (LAMM-004ZC, Qiagen), following the manufacturer’s protocol. The QuantStudio™ 12K Flex Real-Time PCR instrument (Thermo Fisher Scientific) performed quantitative real-time polymerase chain reaction (qRT-PCR). The expression of 84 inflammation-related lncRNAs (Table S2) was compared among healthy mice (n = 3), NASH mice treated (n = 4) and not treated (n = 5) with HLSC-EVs, using GeneGlobe software (Qiagen). CT cut-off was set to 35 and the fold-regulation threshold was set to 2. The arithmetic mean of two housekeeping genes (B2m and Rn7sk) was used to normalize array data. Fold regulation expression with respect to the control group was calculated for all samples using the ΔΔCt method.

To confirm the expression of specific lncRNAs, qRT-PCR was carried out on the RNA of healthy mice (n = 8), NASH mice treated (n = 8) and not treated (n = 10) with HLSC-EVs. For each sample, 2 µg of RNA were converted into cDNA and analyzed using specific primers (RT² lncRNA PCR Assays provided by Qiagen and listed in Table S3) for murine lncRNAs selected from the array screening. Data analysis was performed using ExpressionSuite software (Applied Biosystems, Waltham, MA, USA). The arithmetic mean of two housekeeping genes (B2m and Rn7sk) was used to normalize qRT-PCR data.

To confirm the expression of lncRNA Meg3 in vitro, qRT-PCR was carried out as previously described [9]. Briefly, 5 ng of RNA from LX-2 were converted into cDNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems), were mixed with Power SYBR™ Green PCR Master Mix (Applied Biosystems) and specific oligonucleotide primers for human Meg3 (forward, 5’-TCCATGCTGAGCTGCTGCCAAG-3’; and reverse, 5’-AGTCGACAAAGACTGACACCC-3’) from previous publication [28] (100 nmol/L, obtained from MWG-Biotech, Eurofins Scientific, Brussels, Belgium). To normalize qRT-PCR data, the TATA binding protein (TBP) gene (primers: forward, 5’-TGTGCACAGGAGCCAAGAGT-3’; and reverse, 5’-ATTTTCTTGCTGCCAGTCTGG-3’) was used as housekeeping. For all samples, fold change expression with respect to control was measured using the ΔΔCt method.

Statistical analyses were conducted using the GraphPad Prism software v.8.0 and the results were displayed as mean ± standard deviation (SD). ANOVA with Dunnett’s multiple-comparisons test was applied for data analysis and P < 0.05 was considered significant.

The therapeutic effect of HLSC-EVs on liver inflammation and fibrosis has been previously demonstrated in a murine model of NASH [5] (Figure 1A). Based on the positive results obtained in this MCD diet model, the expression of 84 inflammation-related lncRNAs was investigated in NASH mice subjected to treatment with EVs or with vehicle alone. The results of the lncRNA screening are summarized in Table S4 and Venn’s diagram is shown in Figure 1B.

By comparing the NASH group with the healthy group, 20 lncRNAs were found modulated in the NASH group (17 up-regulated and 3 down-regulated, as shown in Table 1). When the EV-treated group was compared with the NASH group, the expression of 27 lncRNAs (26 down-regulated and 1 up-regulated, as shown in Table 2) was found modulated by the EV treatment. Interestingly, among the lncRNAs differentially expressed in NASH mice, the expression of 10 lncRNAs (Tables 1 and 2) was partially restored to control levels by the treatment with EVs.

It has been reported in the literature that some lncRNAs included in the screening are involved in chronic liver fibrosis [10]. In particular, the expression of Meg3 and Neat1 was up-regulated in NASH mice and significantly down-regulated by EV treatment. Instead, the expression levels of metastasis associated in lung adenocarcinoma transcript 1 (Malat1) and growth arrest-specific transcript 5 (Gas5) were slightly increased in NASH mice compared to healthy mice, but not affected by EV treatment (Table S4).

To confirm the results obtained from the lncRNA screening, the expression of the 14 lncRNAs reported in Table 3 was validated on larger samples. In particular, 7 lncRNAs dysregulated by NASH and partially restored by treatment with EVs (5830432E09Rik, Meg3, Gm16754, 1810019N24Rik, 4933427G23Rik, D930015M05Rik, Neat1), 3 lncRNAs dysregulated only in NASH group (Gm5602, 4921504A21Rik, Gm6410), and 4 lncRNAs modulated only by treatment with EVs (Atxn7l1os2, Jpx, Gm16998, Gm16575) were selected.

Among the lncRNAs selected by the array screening, the expression levels of 9 lncRNAs were up-regulated by NASH and down-regulated by EV treatment (Figure 1C), while the expression level of 1 lncRNA was down-regulated by NASH and up-regulated by EV administration (Figure 1C). In addition, the expression levels of 4 lncRNAs were not modulated by NASH but were strongly reduced by treatment with EVs (Figure 1D). These results confirmed what has been previously observed in the lncRNA screening on NASH mice treated and not treated with EVs.

Based on the results obtained by the lncRNA screening on NASH mice, subsequent investigation focused on Meg3, a lncRNA whose expression is dysregulated in NASH and restored to control levels by EV treatment. Meg3 expression was evaluated in an in vitro fibrosis model established using TGF-β1-activated LX-2 [9].

HLSC-EVs were characterized according to the International Society of EVs guidelines [29]. EVs expressed classical exosomal markers TSG101, CD9, CD63, and CD81, as shown by western blot analysis (Figure 2A) and cytofluorimetric analysis (Figure 2B). The mesenchymal origin of HLSC-EVs was confirmed by the expression of CD29, CD44, CD105, CD146, and the adhesion molecule melanoma-associated chondroitin sulfate proteoglycan (MCSP) (Figure 2B). Furthermore, the expression of epithelial markers EpCAM/CD326 and CD24, hematopoietic (CD45) and endothelial (CD31) markers, and human leukocyte antigen (HLA) class I and II was not detected on EVs (Figure 2B). NTA analysis showed that the average size of EVs was 173.3 nm ± 5.3 nm (Figure 2C) and TEM analysis confirmed that EV preparations contained intact nano-sized particles with double membrane (Figure 2D).

The setup for the in vitro experiments is shown in Figure 3A. The activation of LX-2 was induced by stimulation with TGF-β1. The effects of different doses and regimens of EV administration were tested on activated LX-2. The up-regulation of Meg3 expression levels was observed in TGF-β1-activated LX-2, compared to non-activated cells. Furthermore, the stimulation of activated LX-2 with EVs down-regulated the expression levels of Meg3 after 24 h (Figure 3B). Interestingly, the down-regulation of Meg3 is maintained after 72 h of incubation with the single EV administration (50k EVs/cell) (Figure 3C).