The wild type (WT) mouse hepatocyte line, AML12 (CRL-2254, American Type Culture Collection (ATCC, Manassas, VA, United States), and its derivatives were cultured in DMEM/F12 medium (Thermo Fisher Scientific, Waltham, MA, United States) supplemented with 10% fetal bovine serum (FBS; Corning Inc., Corning, NY, United States), and 1% penicillin-streptomycin-antifungal containing insulin, transferrin, selenium and dexamethasone (Lonza, Alpharetta, GA, United States). Human hepatocyte HepG2 cells (HB-8065; ATCC) or HEK293T cells (CRL-3216; ATCC) were cultured in DMEM containing 10% FBS. Primary mouse hepatocytes or mouse hepatic stellate cells (mHSC) were isolated from male wild-type Swiss Webster mice (6–8 weeks) by perfusion and digestion of livers, followed by buoyant-density centrifugation as previously described (Li et al., 2019). Animal procedures were approved under protocol #04504AR by the Institutional Animal Use and Care Committee at Nationwide Children’s Hospital (Columbus, OH, United States). Mouse HSC were verified for purity, identity and phenotypic transition into activated pro-fibrogenic myofibrobalsts over the first week in culture as described (Li et al., 2020) and the cells were maintained in DMEM/F12/10% FBS/1% penicillin-streptomycin-antifungal, and split when confluent for use up to passage 6 (P6).

The knock-out oligonucleotides containing the target sequence of guide RNA (gRNA) in the mouse fibronectin (mFN1) genome (GAC TGT ACC TGC ATC GGG GC) were annealed and inserted into lentiviral vector lentiCRISPR-V2 (plasmid # 52961; Addgene, Watertown, MA, United States), which was a gift from Dr. Feng Zhang (Massachusetts Institute of Technology, Cambridge, MA, United States) (Sanjana et al., 2014). The lentiCRISPR-V2-mFN1 gRNA was confirmed by sequencing. Lentiviruses were produced by transfecting HEK293T cells with original lentiCRISPR-V2 or lentiCRISPR-V2-mFN1 gRNA, psPAX2, and pCMV-VSVG at a ratio of 3:2:1, and collecting culture supernatants at 48 and 72 h post-transfection. The supernatants were briefly centrifuged at 1,000 × g for 10 min, clarified by passage through a 0.45 μm filter, and used to transduce AML12 cells which were cultured under selection with 2 μg/ml puromycin (InvivoGen, San Diego, CA, United States). The lentiCRISPR-V2-transduced cells were used as scramble control cells. The positive cell population was collected and subjected to single clone selection. The knock-out cells were validated by immunostaining of mFN1 and genome sequencing. Two positive mFN1 knockout clones, hereafter referred to as ΔFN1 cells, were randomly selected for the experiments.

The knock-down oligonucleotides containing the target sequence of clathrin-1 heavy chain (CLTC, GATTACCAAGTATGGTTATAT), caveolin-1 (CAV1, CGACGTGGTCAAGATTGACTT), or Dynamin-2 (DNM2, GCCCTTGAGAAGAGGCTATAT) were annealed and inserted into lentiviral vector pLKO.1, a kind gift from Dr. Zongdi Feng (Nationwide Children’s Hospital, Columbus, OH, United States). The insertions were justified by restriction enzyme digestion and sequencing. The lentiviral stocks were generated as described above and used to transduce AML12 cells or passaged mHSC. The positive cell population was selected with puromycin and the knock-down efficiency was confirmed by Western blot. The pLKO.1-transduced cells were used as scramble control cells.

Mouse or human hepatocytes were plated in T175 flasks until they reached >90% confluency after which spent medium was removed, and the cells were rinsed twice with Hanks Balanced Salt Solution (Thermo Fisher, Waltham, MA, United States) prior to incubating with serum-free medium overnight, followed by replacement with fresh serum-free medium for 48 h. The supernatants were subjected to sequential centrifugation (300 × g for 10 min, 2,000 × g for 20 min, 10,000 × g for 30 min) and ultracentrifugation (100,000 × g for 70 min at 4°C) in a Type T70i fixed-angle rotor, the pellet from which was resuspended and subjected to the same ultracentrifugation conditions again. The resulting EV pellet was dispersed in PBS and characterized as described below. EVs from wild type or ΔFN1 AML12 cells are hereafter named “EV^WT” or “EV^ΔFN1,” respectively. For some experiments, serum used for tissue culture was depleted of its constituent EVs by ultracentrifugation at 100, 000 × g for overnight (Thery et al., 2018).

Extracellular vesicles were labeled either by membrane staining using PKH26 (MilliporeSigma, St. Louis, MO, United States) as described (Li et al., 2019) or by labeling of their RNA payload by incubation of AML12 producer cells with RNAselect (Thermo Fisher Scientific, Waltham, MA, United States).

Extracellular vesicles were subjected to Nanoparticle Tracking Analysis (NTA) using a Nanosight 300 (Malvern Instruments, Westborough, MA, United States) that had been calibrated with 100 nm polystyrene latex microspheres. Recordings were performed at room temperature with a camera gain of 15 and a shutter speed of 4.13 ms. The detection threshold was set to 6. Each EV sample was analyzed twice for determination of mean particle concentration and size distribution.

An enzyme-linked immunosorbent assay (ELISA) was used to quantify FN1 in culture supernatant, large vesicle pellets recovered from 10,000 × g centrifugation, and EV pellets after 100,000 × g ultracentrifugation, using a commercial kit (cat# LS-F2426, LifeSpan BioSciences, Seattle, WA, United States).

A proteinase K digestion protection assay was performed by treating 20 μg of EV^WT with or without 1% NP40 at 37°C for 15 min, followed by incubation with proteinase K (0, 20, 200 μg/ml) for another 30 min. FN1 and flotillin-1 were detected by Western blot in control or digested samples.

Western blot was used to detect common EV marker proteins and FN1. 10–20 μg protein from EV or cell lysates were subjected to sodium dodecyl polyacrylamide gel electrophoresis (SDS-PAGE). Blots were incubated with primary antibodies to CD63 (1:100; MilliporeSigma, Burlington, MA, United States), flotillin-1 (1:200; BD Biosciences, San Jose, CA, United States), CD9 (1:500; Abcam, Cambridge, MA, United States), HNF4α (1:200, Thermo Fisher Scientific, Waltham, MA, United States), clusterin (Clu, 1:500; Proteintech, Rosemont, IL, United States), major vault protein (MVP; 1:500; Proteintech, Rosemont, IL, United States), albumin (1:500, Abcam, United States), mFN1 (1:500, Abcam, United States), huFN1 (1:500, Sinobiological, Beijing, China), or β-actin (1:1,000; Invitrogen, Carlsbad, CA, United States). Blots were developed using an Odyssey Imaging System (LI-COR Biosciences, Lincoln, NE, United States). CLTC (ab21679, Abcam), CAV1 (ab2910, Abcam), and DNM2 (A303-513A, Bethyl Laboratories, Montgomery, TX, United States) antibodies were used to measure the knockdown efficiency of their respective targets in AML12 or passaged mHSC transduced with lentiviral short-hairpin RNA.

The purified EV samples were subjected to digestion and mass spectrometry as described previously (Li et al., 2020). Briefly, EV pellets were resuspended in 50 mM ammonium bicarbonate containing 0.1% Rapigest (Waters Corp., Milford, MA, United States), homogenized by sonication, and clarified by centrifugation at 13,000 rpm. Protein concentration was determined using a Qubit assay kit (Thermo Fisher Scientific, Waltham, MA, United States), dithiothreitol and iodoacetamide were sequentially added before sequencing grade trypsin (Promega Corp., Madison, WI, United States) was added for digestion for overnight at 37°C. Trifluoroacetic acid was then added to precipitate the Rapigest which was then removed by centrifugation. The clarified supernatant was dried and resuspended in 20 μl 50 mM acetic acid. Peptide concentration was determined at 280 nm using a nanodrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States). Three separate EV^WT preparations were individually prepared for mass spectrometry.

Extracellular vesicles protein identification was performed using nano-liquid chromatography-nanospray tandem mass spectrometry (LC/MS/MS) on a Thermo Scientific Q Exactive mass spectrometer equipped with an EASY-Spray^TM Sources operated in positive ion mode. The MS/MS analysis was programmed for a full scan recorded between m/z 400–1600 and an MS/MS scan to generate product ion spectra to determine amino acid sequence in consecutive scans starting from the most abundant peaks in the spectrum, and then selecting the next nine most abundant peaks.

Sequence information from the MS/MS data was processed by converting the raw files into a merged file using MS convert (ProteoWizard). The resulting mgf files were searched using Mascot Daemon by Matrix Science version 2.6.0 (Boston, MA, United States) and the database searched against Uniprot Mouse database. A decoy database was also searched to determine the false discovery rate (FDR) and peptides were filtered according to the FDR. Proteins with less than 1% FDR as well as a minimal of two significant peptides detected were considered as valid proteins. Proteomics data were summarized in Scaffold 4.9.0 (Proteome Software Inc., Portland, OR, United States) for spectral counting analysis. Complete MS datasets are available in the Supplemental Data.

Gene Ontology (GO¹ and the Kyoto Encyclopedia of Genes and Genomes (KEGG²) analyses of EV proteins were accomplished using the DAVID online program³. Search Tool for the Retrieval of Interacting Genes (STRING⁴ was utilized to determine interactions among EV proteins using a medium interaction score of 0.7, and the Markov Cluster Algorithm method with an inflation parameter of 3 was applied for clustering. These analyses were each performed with a criterion FDR < 0.05.

100 μg of purified AML12 cell EVs were incubated with 4 μg of anti-mFN1 antibody (Abcam) or with 4 μg isotype normal IgG at room temperature for 3 h, followed by addition of 50 μl of Dynabeads (cat#10004D, Thermo Fisher Scientific, Waltham, MA, United States) and incubation overnight at 4°C. Beads were then magnetically separated and the unbound material was retained. The beads were washed with PBS-0.02% Tween four times before resuspending them in PBS. The immunoprecipitated samples and the unbound material were boiled and subjected to SDS-PAGE under reducing conditions, followed by Western blot. CD9, flotillin-1, and FN1 primary antibodies and IP-specific secondary antibodies (ab121366 VeriBlot HRP secondary IgG, ab131368 rat anti-mouse IgG, Abcam) were used for detection.

Mouse HSC or hepatocytes seeded in 96-well plates were cultured in, respectively, 2% exosome-depleted serum-containing medium or serum-free medium overnight. The following day, the plates were placed on ice and cells were treated with PKH26- or RNASelect-labeled EVs in the presence or absence of a panel of endocytosis or macropinocytosis inhibitors to assess uptake pathways for EV internalization. These included clathrin-mediated endocytosis (ClME) inhibitor chlorpromazine (CPZ, cat# HY-B0407A, MCE; Monmouth Junction, NJ, United States); caveolin-mediated endocytosis (CaME) inhibitors including Genistein (cat# HY-14596, MCE), Nystatin (cat# HY-17409, MCE), and Filipin (cat# F4767, MilliporeSigma); macropinocytosis inhibitors EIPA (cat# A3085, MilliporeSigma, Burlington, MA, United States) and LY294002(cat# HY-10108, MCE); and Dynasore (cat# D7693, MilliporeSigma, Burlington, MA, United States) to inhibit both ClME and CaME. As positive controls, pHrod red-transferrin (cat# P35376, Thermo Fisher Scientific, Waltham, MA, United States), Alexa Fluor 488-cholera toxin subunit B (cat# C34775, Thermo Fisher Scientific, Waltham, MA, United States), and Oregon Green 488-dextran (70KD, cat# D7172, Thermo Fisher Scientific, Waltham, MA, United States) were used to confirm the potency of the above inhibitors. Cholesterol absorption inhibitor ezetimibe (cat# HY-17376, MCE), a competitive inhibitor of the lysosomal acid lipase lalistat 2 (cat# SML2053, MilliporeSigma, Burlington, MA, United States), lysosomotropic agents including bafilomycin-A1 (Baf-A1) (cat# HY-100558, MCE), NH₄Cl, and chloroquine (ChQ) (cat# PHR1258, MilliporeSigma, Burlington, MA, United States) were also used to test the internalization pathway of EVs. All test reagents were used at concentrations that were pre-determined to have no cytotoxicity on HSC or hepatocytes; the concentrations used for each reagent are shown in the figure legends. The cells were then shifted to 37°C to initiate the EV uptake process. Echistatin, a potent inhibitor or RGD-binding integrins (cat#E1518, MilliporeSigma, Burlington, MA, United States), or heparin (cat#H4784, MilliporeSigma, Burlington, MA, United States) were used to test the involvement of cell surface integrin or heparin-like molecules in mediating EV uptake. For this, the recipient cells were pretreated with or without echistatin while EVs were pretreated with or without heparin, at 37°C for 1 h before EV incubation with the recipient cells in the presence of echistatin or heparin alone or in combination. At 24 h post-EV addition, the cells were washed extensively with PBS to remove unbound EVs, fixed with 4% paraformaldehyde, counterstained with DAPI, and photographed with an LSM 800 microscope (Carl Zeiss Inc., Thornwood, NY, United States). PKH26 or RNAselect fluorescence intensity (EV uptake) was quantified using ImageJ (NIH, Bethesda, MD). Alternatively, the cells were lysed at the end of the EV uptake assay and a spectrophotometer (Spectra Max M2, VWR, Sunnyvale, CA, United States) was used to measure the PKH26 signal at Ex/Em = 540/580 with cut-off = 570 nm.

Purified AML12 cell EVs were loaded on top of an iodixanol cushion (40, 32, 24, 16, and 8%, Serumwerk Bernburg AG, Germany) and ultracentrifuged at 37,500 rpm in a SW55Ti rotor for 17 h at 4°C. Twenty fractions were collected from the top to bottom and the density of each fraction was measured using an Abbe refractometer (Bausch and Lomb, Rochester, NY, United States). FN1 and flotillin-1 in each fraction were detected by Western blot. In some cases, the EVs were pretreated with 10% NP40 at 37 °C for 15 min prior to ultracentrifugation.

1.5 ml of clarified AML12 cell culture supernatant after low-speed centrifugation or purified EVs from AML12 cells were loaded onto a 10–60% sucrose gradient (0.5 ml each of 60, 40, and 30%; 1 ml of 20%, and 10% sucrose in TNE buffer (10 mM Tris–HCl, pH 8.0, 150 mM NaCl, 2 mM EDTA, kept at 4°C overnight before use), and centrifuged at 42,000 rpm (∼167,000 × g) in a SW55Ti rotor at 4°C for 2 h. Fifteen fractions were manually collected from the top and the distributions of FN1 and flotillin-1 were determined by Western blot.

Wild-type male Swiss Webster mice (4–5 weeks old; n = 5 per group) were injected i.p. with CCl₄ (4 μl in 26 μl olive oil) or corn oil (30 μl) three times per week for 5 weeks as described (Li et al., 2019) using IACUC-approved protocol #04504AR (see above). Some mice received i.p. EV^WT or EV^ΔFN1 (3e + 9 particles/dose) three times per week over the last 2 weeks of the experiment according to our published procedures (Li et al., 2019). Mice were sacrificed and individual liver lobes were harvested and snap-frozen for histology measurement and RNA extraction for RT-PCR to detect transcript expression of multiple genes including extracellular matrix (COL1A1, COL3A1, MMP2, and RELN), and cell cycle (CCNB2, CDC25C, and KIF2C).

Perfused mouse livers were fixed with 4% paraformaldehyde and embedded in paraffin. Sections with 5 μm thickness were cut and stained with H&E. Sections were stained with 0.1% Sirius Red (MilliporeSigma) for collagen detection. Positive signals were quantified by ImageJ analysis.

Total RNA from liver tissues or cultured cells was extracted using a miRNeasy mini kit (Qiagen, Germantown, MD, United States) and reverse transcribed with a miScript II RT kit (Qiagen) according to the manufacturer’s instructions. Transcript expression was evaluated by qPCR using SYBR Green Master Mix (Eppendorf, Enfield, CT, United States) on an Eppendorf Mastercycler System. Primers are shown in Table 1. Each reaction was run in duplicate, and samples were normalized to 18S rRNA.

Experiments were performed at least twice in duplicate or triplicate, with data expressed as mean ± SEM. Fluorescence images were scanned and quantified using ImageJ software (NIH). Data from qRT-PCR and imaging were analyzed by student’s t-test. P-values < 0.05 were considered statistically significant.

Mass spectrometry analysis of three separate AML12 cell EV samples resulted in the identification of 481, 305, and 474 proteins, respectively (Figure 1A and Supplementary Table 1). Subsequent analysis was focused on 455 proteins that were present in at least two EV samples. Of these, the most abundant proteins (quantitative value ∼100–1000) in AML12 cell EVs included FN1 (quantitative value ∼1000), complement 3, histones (Hist1h4a, Hist1h2bf, Hist1h2ab, Hist1h2aa1, and Hist3h2bb), pregnancy-zone protein (PZP), galectin 3 binding protein (LGALS3BP), Clu, and MVP (Figure 1B). The identification of Clu and MVP is consistent with their presence and function in EVs from other systems (Foglio et al., 2015; Teng et al., 2017). When GO analysis was used to group all 455 EV proteins into cell components, the EV proteome was characterized as being highly enriched for components related to extracellular exosomes (350 out of 455; 76.9%) (Figure 1C and Supplementary Table 2). Other enriched components included cytoplasm (295), membrane (257), nucleus (227), cytosol (124), focal adhesion (91) and extracellular space (91) (Figure 1C). KEGG pathway analysis revealed 61 enriched pathways, for which metabolic pathways, ribosome, proteasome, regulation of actin cytoskeleton, and endocytosis were ranked as the top five pathways (Figure 1D). STRING analysis of the proteomic data resulted in a complex interaction network, in which principal nodes contained proteins associated with protein synthesis and degradation, nucleic acid binding, histones, enzymes, actins, ECM, cell adhesion, complements, keratins, cytoskeletons, and tRNA-protein interactions (Figure 1E).

Proteomic analysis of EV^WT showed that FN1 ranked as the most abundant protein (Figure 1B). To validate the presence of FN1 in hepatocyte EVs, conditioned medium from mouse hepatocyte AML12 cells was subjected to differential centrifugation and the distribution of FN1 in each fraction was measured by ELISA. As shown in Figure 2A, high speed (10, 000 × g) centrifugation did not result in FN1 loss in the supernatants, and no FN1 was detected in the pellets after high-speed centrifugation. By contrast, approximately two-thirds of FN1 were present in the supernatant after ultracentriguation (100,000 × g) while the remaining one-third of FN1 was precipitated with EV pellets. Western blot was subsequently used to demonstrate the presence of FN1 in purified preparations of EV^WT which were also positive for EV marker proteins such as flotillin-1, ALIX, and CD9 but negative for the cell-specific marker Calnexin (Figure 2B). FN1 was barely detected in AML12 cell lysates but it was still present in EV-depleted conditioned medium consistent with its release, in part, from the cells as a soluble (non-EV) component (Figure 2B). The structure of EV FN1 was determined from three separate mass spectrometry sequencing analyses in which 1648 to 1760 of the 2477 amino acids in the primary protein sequence (67–71% coverage) were individually identified. Interestingly, the EDB domain was not detected but 26 to 39 amino acids of the 88-residue EDA domain were detected, albeit at a lower coverage rate (29.5–44.3%) than for full-length FN1 (Supplementary Figure 1). To confirm that the FN1 data were not limited to EV^WT from the AML12 mouse hepatocyte line, similar EV preparations were purified from primary mouse hepatocytes or the human hepatocyte HepG2 cell line with the result that FN1 was associated with EVs that were also positive for hepatocyte markers (albumin, HNF-4α) and EV markers (flotillin-1) (Figures 2C,D). We have previously shown that HepG2 EVs are also positive for other EV markers including Alix, CD9, and Tsg101 (Li et al., 2019). Thus, FN1-associated EVs were broadly produced by primary or immortalized hepatocytes of human or mice origin.

The results above suggested that while FN1 was present in hepatocyte conditioned medium in its free form which is consistent with its known properties as a secreted protein, an appreciable quantity of FN1 was also EV-associated. This latter possibility was supported by the observation that when purified EVs were subjected to rate zonal ultracentrifugation, the peak of FN1 immunoreactivity (fractions #4–6) was coincident with the presence of flotillin-1 in the same fractions (Figure 2E). Similarly, iodixanol isopycnic ultracentrifugation of EV^WT that had been labeled with PKH26 membrane dye resulted in co-distribution of the signals for PKH26 or flotillin-1 (Figure 2F). Moreover, destruction of EV structural integrity using NP40 detergent resulted in liberation of free PKH26, disappearance of the flotillin-1 signal, and a concomitant change in FN1 density shown by a shift to the right of the FN1 signal (Figure 2F). The FN1-association with EVs was further supported by the detection of flotillin-1 or CD9 in anti-FN1 immunoprecipitation of purified EVs, a result that was accompanied by correspondingly diminished intensities of the FN1, flotillin-1 and CD9 signals in the residual unbound sample after immunoprecipitation (Figure 2G). Finally, proteinase K digestion dose-dependently degraded FN1 in EV^WT preparations resulting in the production of variably sized fragments (75, 50, 37 kDa) that were resistant to further breakdown (Figure 2H). However, pre-treatment of EV^WT with NP40 resulted in complete digestion of FN1 showing the importance of EV structural integrity for protecting FN1 from proteolysis, as was also observed for flotillin-1 (Figure 2H). The susceptibility of FN1 to digestion by proteinase K in the absence of NP40 suggests that FN1 is associated peripherally, likely on the EV surface, which is fully consistent with the role for FN1 in mediating EV binding to target cells (see below). Since NP40 facilitates proteinase K digestion of FN1, this likely reflects the partial association of FN1 with the EV membrane, possibly by direct anchoring or by being tethered to a binding partner within the EV membrane. In contrast, flotillin-1 is a fully membrane-associated protein (Otto and Nichols, 2011) accounting for its proteinase K resistance unless the membrane is disrupted by NP40. Overall, these various approaches lend strong support for an intimate association between FN1 and hepatocyte EVs.

To investigate the biological function of EV-associated FN1, ΔFN1 AML12 cells were first generated using CRISPR-Cas9. Genome sequencing (Figure 3A) and immunofluorescence assay (Figure 3B) both confirmed the knockout of FN1 in two single clones (ΔFN1-1 and ΔFN1-2). As assessed using alamarBlue reagent, cell growth kinetics under normal (Figure 3C) or serum-free condition (Figure 3D) were not significantly affected by FN1 deficit. When purified EVs were assessed by Western blot, no FN1 was detected in EV^ΔFN1 unlike the substantial FN1 signal in EV^WT (Figure 3E). Interestingly, MVP and Clu (Figure 1B) were present in EV^ΔFN1 at highly reduced levels as compared to EV^WT whereas the signal for proteasome subunit alpha type-6 (PSMA6) and flotillin-1 was comparable between the two types of EVs (Figure 3F).

To evaluate the consequences of FN1 deficit, the EV yield was firstly assessed. WT or ΔFN1 AML12 cells were incubated with serum-free medium for 48 h, followed by EV purification and quantification by NTA. The EV yield was calculated by EV numbers/cell and normalized to WT. As shown in Figure 4A, no significant alterations were detected in either ΔFN1-1 or ΔFN1-2 cells. NTA showed that while the overall profiles of EV^ΔFN1 were similar to that of EV^WT, the average sizes of EV^ΔFN1–1 (123.9 ± 3.2 nm) or EV^ΔFN1–2 (135.4 ± 4.4 nm) were approximately 10–20% greater than EV^WT (101.5 ± 9.2 nm) (Figure 4B). A higher percentage of EV^ΔFN1 with larger size (≥120 nm) was seen compared to EV^WT (Figure 4B) but it remains to be determined if this size difference represents differences in pathways of EV biogenesis or a structural contribution by FN1 to EV ‘compactness.” That said, the buoyant density of the EVs was unaffected by FN1 knockout as shown by the sedimentation of EV^WT, EV^ΔFN1–1, or EV^ΔFN1–2 over the same density range (fractions 12–16) upon iodixanol isopycnic ultracentrifugation (Figure 4C). To assess the efficiency of EV cellular uptake of the EVs after FN1 knockout, equivalent numbers of EV^WT, EV^ΔFN1–1, or EV^ΔFN1–2, all with comparable PKH26 signal, were incubated with WT or ΔFN1 AML12 cells. EV^WT uptake efficiency was comparable in WT or ΔFN1 cells, suggesting that cell-associated FN1 is dispensable for EV uptake (Figures 4D,E). By contrast, uptake of EV^ΔFN1–1 or EV^ΔFN1–2 to WT or ΔFN1 cells was reduced by 50–70% of control values (Figures 4D,E), showing that EV-associated FN1 facilitates (but is not essential for) EV uptake. As compared to EV^WT, EV^ΔFN1 labeled with RNAselect produced a weaker signal in mHSC or AML12 (Figure 4F) showing that EV-mediated RNA delivery was dependent on EV FN1. Echistatin, a potent inhibitor of RGD-binding integrins, significantly inhibited cellular uptake of EV^WT but not of EV^ΔFN1 (Figure 4G), showing that integrin-FN1 interactions are important for EV^WT uptake. Treatment with soluble heparin also dose-dependently inhibited cellular binding of either EV^WT (Figure 4H) or EV^ΔFN1 (Figure 4I) but the heparin-mediated inhibition was less robust for EV^ΔFN1.

To investigate which pathways are involved in EV uptake, a panel of inhibitors to ClME, CaME, and macropinocytosis were used for a small-scale screening. First, the cell cytotoxicity (Supplementary Figure 2) and the ability of the reagents to inhibit endocytosis or macropinocytosis was validated using, respectively, fluorophore-labeled transferrin (for ClME), Cholera enterotoxin subunit B (CtxB, for CaME), or dextran (for macropinocytosis) (Supplementary Figure 3). Next, EV^WT were added to WT AML12 cells pretreated with different reagents for 1 h and the EVs and reagents were then simultaneously incubated with the cells for another 24 h. Inhibitors of endocytosis but not of macropinocytosis reduced the uptake of EV^WT in a dose-dependent manner, with dynasore having the most potent inhibition (Figure 5A). Whereas uptake of PKH26-labeled EV^WT by WT AML12 cells was readily visualized by the presence of PKH26 fluorescence in the cells (Figure 5B), shRNA-mediated knockdown of CLTC, CAV1, or DNM2, which resulted in a significant decrease in expression of each component (Figures 5C,D), caused the cellular binding of EV^WT be impaired by more than 50% as compared to WT cells (Figures 5B,D). Similarly, EV^WT uptake by mHSC was reduced by inhibitors of clathrin- or caveolin-mediated endocytosis, but not of macropinocytosis (Figure 5E) and the fluorescent signal associated with the uptake of PKH26-labeled EV^WT by mHSC was reduced by prior knockdown of CLTC or CAV1 in the mHSC target cells (Figures 5F–H).

To understand if EV^ΔFN1 utilized the same mechanisms to enter cells, mHSC were pretreated with the endocytosis inhibitors and incubated with EV^ΔFN1 in the presence of the inhibitors for 24 h before imaging and quantification. Clathrin- or caveolin-mediated endocytosis was shown to be required for EV^ΔFN1 uptake, as the inhibitors to either pathway significantly reduced the uptake (Figure 5I). Interestingly, two macropinocytosis inhibitors (EIPA and LY49002) individually inhibited the EV^ΔFN1 uptake (Figure 5I), showing EV^ΔFN1 can enter cells through macropinocytosis pathway even though EV^WT did not (Figures 5A,E). Similar with EV^WT, shRNA-mediated knockdown of CLTC or CAV1 significantly reduced EV^ΔFN1 uptake in mHSC (Supplementary Figure 4).

To evaluate if cholesterol, lysosomal acid lipase, or lysosomal pH are involved in EV uptake, mHSC were pretreated with cholesterol absorption antagonist ezetimibe (Chang and Chang, 2008), lysosomal acid lipase inhibitor lalistat 2 (Hamilton et al., 2012), or lysosomotropic agents including Baf-A1, NH₄Cl, or ChQ before EV inoculation. The uptake of EV^WT by mHSC was susceptible to each of these agents (Figure 5J), suggesting the involvement of cholesterol and lysosome, in which acid lipase and the low pH conditions are both required for the following possible membrane fusion step. By contrast, uptake of EV^ΔFN1 was also sensitive to ezetimibe or lalistat 2 treatment but more resistant to lysosomotropic agents (NH₄Cl, Baf-A1) (Figure 5K) suggesting that FN1 may facilitate low pH-mediated EV entry.

We have previously shown that EV^WT are therapeutic for CCl₄-induced liver fibrosis, resulting in attenuated expression of fibrosis-related and cell cycle related genes (Li et al., 2019). To evaluate if there was a functional difference between EV^ΔFN1 and EV^WT, each type of EV was administered to mice over the last 2 weeks of a 6-week course of CCl₄ to induce hepatic fibrosis. CCl₄-treated mice demonstrated excessive hepatic collagen deposition as compared to control mice (Figure 6A). Administration of either EV^WT or EV^ΔFN1 resulted in diminished amounts of collagen deposition as shown by Sirius red staining (Figures 6A,B). The previously reported attenuation by EV^WT of CCl₄-induced extracellular matrix (ECM) genes (COL1A1, COL3A1, MMP2, and RELN) or cell cycle genes (CCNB2, CDC25C, and KIF2C) (Li et al., 2019) was also seen in response to EV^ΔFN1 (Figures 6C,D). Thus EV-associated FN was not required for EV-mediated suppression of collagen deposition or CCl₄-induced genes.