Exosome Adherence and Internalization by Hepatic Stellate Cells Triggers Sphingosine 1-Phosphate-dependent Migration*

Exosomes are cell-derived extracellular vesicles thought to promote intercellular communication by delivering specific content to target cells. The aim of this study was to determine whether endothelial cell (EC)-derived exosomes could regulate the phenotype of hepatic stellate cells (HSCs). Initial microarray studies showed that fibroblast growth factor 2 induced a 2.4-fold increase in mRNA levels of sphingosine kinase 1 (SK1). Exosomes derived from an SK1-overexpressing EC line increased HSC migration 3.2-fold. Migration was not conferred by the dominant negative SK1 exosome. Incubation of HSCs with exosomes was also associated with an 8.3-fold increase in phosphorylation of AKT and 2.5-fold increase in migration. Exosomes were found to express the matrix protein and integrin ligand fibronectin (FN) by Western blot analysis and transmission electron microscopy. Blockade of the FN-integrin interaction with a CD29 neutralizing antibody or the RGD peptide attenuated exosome-induced HSC AKT phosphorylation and migration. Inhibition of endocytosis with transfection of dynamin siRNA, the dominant negative dynamin GTPase construct Dyn2K44A, or the pharmacological inhibitor Dynasore significantly attenuated exosome-induced AKT phosphorylation. SK1 levels were increased in serum exosomes derived from mice with experimental liver fibrosis, and SK1 mRNA levels were up-regulated 2.5-fold in human liver cirrhosis patient samples. Finally, S1PR2 inhibition protected mice from CCl4-induced liver fibrosis. Therefore, EC-derived SK1-containing exosomes regulate HSC signaling and migration through FN-integrin-dependent exosome adherence and dynamin-dependent exosome internalization. These findings advance our understanding of EC/HSC cross-talk and identify exosomes as a potential target to attenuate pathobiology signals.

Liver fibrosis is the excessive accumulation of extracellular matrix proteins, including collagen I, that occurs in various forms of chronic liver diseases. Advanced liver fibrosis results in cirrhosis, liver failure, and portal hypertension and often requires liver transplantation (1). Although the pathogenesis of liver fibrosis is not fully defined, the activation of hepatic stellate cells (HSCs) 3 into a myofibroblastic phenotype is recognized as a cardinal pathogenic step in the development of this disease (2)(3)(4)(5). One key component of HSC activation is enhanced migration, which allows HSCs to navigate to sites of fibrogenesis. In many circumstances, HSC migration is governed by molecules released from neighboring endothelial cells (ECs). The molecular mechanisms that define this process remain incompletely defined (6,7).
Exosomes are extracellular membrane vesicles that are produced in multivesicular bodies or at the plasma membrane. They have an average diameter of 40 -100 nm, a characteristic density (1.13-1.19 g/ml), and a cup-shaped morphology and sediment at 100,000 ϫ g (8). Exosomes are enriched with endosome-associated proteins (e.g. Rabs, GTPases, SNAREs, and Annexins), multivesicular endosomes, (e.g. Alix and TSG101), and tetraspanins, (e.g. CD63 and CD81) (9 -12). Recent studies have explored potential roles for exosomes in the pathogenesis of liver inflammation, fibrosis, and portal hypertension (13). An increase in this extracellular vesicle subtype has been postulated in patients with cirrhosis (14). However, the mechanisms by which exosomes achieve their effects on target cells are not known, especially in the context of EC regulation of HSC migration.
This study was conducted to test the hypothesis that ECderived exosomes regulate pathological HSC migration during liver fibrosis. The experimental results from this study provide evidence that exosome-induced HSC migration is dependent first on exosome adhesion, which is mediated by exosome fibronectin (FN) binding with ␣5␤1-integrin on target cells. Secondly, adhesion facilitates exosome entry into the target cell through dynamin-dependent endocytosis. These steps are requisite for signal activation and ensuing migration. At the molecular level, we identified the lipid enzyme sphingosine kinase 1 (SK1) as a critical mediator of exosome actions on HSCs. Both SK1 and its product sphingosine 1-phosphate (S1P) are present within the exosome and are required for chemotactic effects. The experimental results extend our understanding of the mechanisms controlling exosome regulation of HSCs. More broadly, the work extends our understanding of paracrine signal transduction and also lays the theoretical foundation for therapies targeting exosomes in the treatment of liver pathobiology such as fibrosis.

Experimental Procedures
Cell Culture and Viral Transfection-The LX-2 human HSC cell line, the primary human HSC cell line (hHSC), and immortalized liver sinusoidal ECs (TSECs) (6) were grown in DMEM containing 10% fetal bovine serum and 1% penicillin/streptomycin (Gibco). Human umbilical vein endothelial cells were grown with endothelial culture media with 10% serum and 1% endothelial growth supplement. Liver hepatocellular carcinoma cells (HepG2) were routinely maintained in DMEM supplemented with 10% fetal bovine serum and 1% penicillin/ streptomycin. The human macrophage line (THP-1) was cultured in RPMI 1640 medium with 10% fetal bovine serum and 1% penicillin/streptomycin. The culture medium used for exosome isolation was prepared with exosome free-FBS as described previously (8). Adenoviral vectors were generated through the Iowa Vector Core and encoded dynamin-2 K44A or a LacZ control. Cells were incubated for 12 h with 0.1% albumin/PBS with adenoviruses (multiplicity of infection of 50), which achieved a transduction efficiency approximating 90% with minimal toxicity. Lentivirus was generated by using 293T cells. Adenoviral and lentiviral transduction were performed as described previously (4,8). All cell lines were maintained under standard tissue culture conditions (37°C, 5% CO 2 incubator).
Isolation of Mouse Liver Endothelial Cells-Liver endothelial cells were isolated from whole livers of healthy mice and mice subjected to CCl 4 -induced fibrosis by mechanical disruption, enzymatic digestion, and immunomagnetic bead separation as described previously but with some minor modifications (15)(16)(17). Briefly, liver tissue was perfused, harvested, dissected, minced, and digested in a collagenase buffer and incubated with immunomagnetic Dynabeads (Dynal) coated with rat antimouse CD146 (BD Biosciences), an endothelial marker, for 1 h at room temperature (18,19). Cells were separated with a magnet and plated on collagen I-coated dishes. Viability was Ͼ90% by trypan blue staining, and purity was Ͼ95% by staining for CD31. Cells were grown in EC growth medium containing 5% fetal bovine serum, 2% endothelial cell growth supplement, and 1% penicillin/streptomycin (ScienCell, San Diego, CA) and maintained under standard tissue culture conditions (37°C, 5% CO 2 incubator).
Site-directed Mutagenesis and Generation of Stable Cell Lines-Site-directed mutagenesis was performed according to the protocol of the manufacturer (Agilent Technology, Santa Clara, CA). cDNA encoding full-length wild-type hSK1-pCW45 was purchased from Addgene (Cambridge, MA). SK1 dominant negative and constitutively active mutants with an Asp-81 to Ala (D81A) and Gly-113 to Asp (G113A) (20, 21) mutation, respectively, were generated by performing PCR using the respective oligonucleotides 5Ј-CTCTGGTGGTCA-TGTCTGGAAACGGGCTGATGCACGAGGTG-3Ј and 5Ј-CACCTCGTGCATCAGCCCGTTTCCAGACATGACCAC-CAGAG-3Ј (for the D81A mutant) and 5Ј-GTAGCCTCCCA-GCAGGCTCTGCCAACGCGCTGGCAGCTTCCTT-3Ј and 5Ј-AAGGAAGCTGCCAGCGCGTTGGCAGAGCCTGCTG-GGAGGCTAC-3Ј (for the G113A mutant). Mutations were verified by DNA sequencing. Lentivirus was generated using 293T cells and used to transduce TSECs and establish stable cell lines that expressed SK1 wild-type and SK1 D81A and G113A mutants. Exosomes were isolated from the conditioned media generated from cell lines.
Exosome Purification and Fluorescent Labeling-Exosomes from de-identified patient serum, from cell culture supernatant of TSECs, and from a pool of murine serum (ϳ300 l) were purified as described previously (6,8,22). The protein concentration contained in each exosome pellet was quantified using a Bradford assay (Bio-Rad) (9,23). Experiments were performed using 50 g of exosomes unless indicated otherwise. Isolated exosomes were characterized by Western blot (WB), nanoparticle tracking analysis, and electron microscopy immunogold labeling. Precipitated exosomes were labeled in some experiments with the green fluorescent linker PKH67 (Sigma-Aldrich) according to the instructions of the manufacturer.
Immunofluorescence, Confocal Microscopy, and Image Quantification-PKH67-labled exosomes were incubated with HSCs for different periods of time. HSCs were first fixed with 4% paraformaldehyde followed by permeabilization with 0.1% Triton X-100. Cells were incubated with primary antibodies (early endosome antigen (EEA), 1:200, catalog no. 610456; LAMP1, 1:200, catalog no. sc20011) overnight at 4°C, and the appropriate Alexa Fluor-conjugated second antibodies (Invitrogen, Thermo Fisher Scientific) were used for secondary detection. Cell nuclei were counterstained with DAPI (1:2000). Fluorescence confocal microscopy was performed using an LSM 510 laser-scanning microscope (Carl Zeiss, Jena, Germany) using a ϫ63 or ϫ40 lens. ImageJ (National Institutes of Health) was used to analyze the raw images. The images used in the figures are representative of trends observed in all images obtained. Co-localization images were created using the ImageJ co-localization plugin. The quantification shows the Pearson coefficients of co-localization between the green channel for the biotin exosome and red channel for EEA or LAMP1. More than six representative pictures in each group were selected for quantification and statistical analysis.
Nanoparticle Tracking Analysis-The presence, size distribution, and concentration of vesicles were assessed by nanoparticle tracking analysis using a NanoSight NS300 instrument (NanoSight Ltd., Amesbury, UK). Exosome samples were diluted with PBS at a range of concentration of 4 -8 ϫ 10 8 particles/ml. Each sample was loaded into a flow cell top plate using a syringe pump, and three videos of 30 s were recorded and analyzed by NanoSight software (NTA 2.3.5 B16) (11,24,25).
Western Blot Analysis-HSCs or liver tissue were lysed and prepared for WB analysis as described previously (26,27). Immunoblot analysis was performed according to the protocol recommended for individual antibodies as listed in supplemental Table  1. Immunoreactive bands were visualized using horseradish peroxidase-conjugated secondary antibody and an enhanced chemiluminescent system (Santa Cruz Biotechnology, Inc.). All experiments were performed in triplicate, and quantitation was done by densitometry.
PCR Studies-The RNeasy kit (Qiagen) was used to extract total RNA from cells and mouse tissue according to the instructions of the manufacturer. 10 g of mRNA was used for cDNA synthesis with dNTP and oligo primer using the SuperScript TM III (Invitrogen) first strand synthesis system for RT-PCR according to the protocol of the manufacturer. Real-Time PCR was performed with the same amount of cDNA in a total 25-lvolume reaction using IQ SYBR Green Mix (Bio-Rad) and the 7500 real-time PCR system (Applied Biosystems) according to the instructions of the manufacturer. Amplification of GAPDH and ␤-actin was performed in the same reaction for the respective samples as an internal control. Each experiment was done in triplicate. Primer sequences are listed in supplemental Table 2.
Transwell Migration Assays-Exosomes, compounds (Dynasore, 10 M; dimethylsphingosine (DMS), 10 nM; FTY720, 10 nM; and AKT inhibitor VIII, 25 nM), and FN-integrin binding neutralizing reagents (RGD, 20 nM; CD29, 0.2 g/ml) were used to assess HSC migration as measured by Transwell assay. Briefly, 8-m-pore polycarbonate filters (catalog no. 3422, Costar) were coated with 1% (w/v) collagen I (Sigma-Aldrich, St. Louis, MO). Cells (1 ϫ 10 4 cells in 100 l/well) were suspended in serum-free DMEM. Exosomes and/or reagents were added to the lower chamber in the same culture medium. After incubation at 37°C for 24 h, migrated cells were stained by crystal violet (8,28). The number of cells that migrated to the bottom side of the chamber was determined by counting the number of cell per field by light microscopy (ϫ10 magnification).
Transmission Electron Microscopy and Immunogold Labeling-EM and immuno-EM analysis of vesicles were performed as described previously (8,29). Briefly, the specimens were resuspended and fixed in 4% paraformaldehyde for 1 h. Subsequently, the samples were deposited onto formvar/carboncoated EM grids (5 l on each grid) and dried for 25 min. The vesicle-coated grids were washed twice with PBS (15 min each) and twice with PBS/50 mM glycine (15 min) and blocked with 10% FBS/PBS (15 min). For the immunogold labeling with antibodies, blocked grids were transferred to a drop of the antibody (listed in supplemental Table 1) and incubated for 1 h. The grids were then washed with 2% FBS/PBS for 5 ϫ 3 min, incubated with specific gold-labeled secondary antibody (10 nm gold, goat-anti-rabbit IgG 10 nm gold, rabbit anti goat IgG 10 nm), diluted with 10% FBS/PBS in a ratio of 1:30 for 1 h, washed 5 ϫ 3 min in 100-l drops of 2% FBS/PBS, and then post-fixed with 1% glutaraldehyde for 5 min. After washing in 8 drops of distilled water, the grids were stained with a mixture of 4% uranyl acetate and 2% methylcellulose (1:9) and viewed for transmission electron microscopy using a JEOL 1400.
Scanning Electron Microscopy-Scanning electron microscopy was performed as described previously (8,14). Samples were fixed in Trump fixative for 1 h or overnight at 4°C. After washing three times with 0.1 M phosphate buffer, the specimens were post-fixed in 1% OsO 4 /0.1 M phosphate buffer for 1 h. After washing with distilled water three times, they were dehydrated with a series of ethanol (50%, 70%, 95%, 100%, 100%, and 100%), dried with critical point drying, mounted and gold-palladium sputter-coated, and then viewed by Hitachi 4700.
Biotinylation Assay for Plasma Membrane Integrin Activity-Biotin-based assays were performed as described previously (22,30). HSCs were fasted overnight and treated with exosome for 15 min. Subsequently, the cells were placed on ice and washed once with cold Dulbecco's phosphate-buffered saline. Cell surface proteins were labeled with 1 mg/ml of EZ-Link cleavable Sulfo-NHS-SS-Biotin (catalog no. 21331, Thermo Scientific) in Dulbecco's phosphate-buffered saline (catalog no. 14040, Invitrogen) for 15 min at 4°C. The cells were lysed by scraping in radioimmune precipitation assay lysis buffer with protease inhibitor mixtures (Roche) and incubated at 4°C for 20 min. Cell extracts were cleared by centrifugation (15,000 ϫ g, 10 min, 4°C). Cell lysates were subjected to protein G-Sepharose beads (catalog no. 17-0618-01, GE Healthcare) pulldown, followed by WB for HUTS4 and CD29 as the loading controls. The integrated densities of the protein bands were quantified by ImageJ (version 1.43u).
Animals and Procedures-C57BL/6 mice (6 -8 weeks old) were purchased from The Jackson Laboratory (Bar Harbor, ME), maintained in a temperature-and light-controlled facility, and permitted ad libitum consumption of water and standard pellet chow. Mice were given carbon tetrachloride (CCl 4 ) intraperitoneally at a dose of 1 l/g of body weight twice a week for 6 weeks, as described previously (2,8). Mice were sacrificed 1 day after the final dose of CCl 4 . For the S1PR2 inhibition experiments, JTE-013 (Cayman Chemical Co., Ann Arbor, MI) or DMSO control was administrated during the CCl 4 -induced liver fibrosis model. C57BL/6 mice were randomized to olive oil and DMSO, CCl 4 (1 l/g of body weight) and DMSO, or CCl 4 and JTE-013 (10 mg/kg of body weight) (n ϭ 6/group). JTE-013 was administered intraperitoneally 1 week before CCl 4 treatment and 1 day before each subsequent CCl 4 injection. CCl 4 was administered for 3 weeks rather than 4 weeks to minimize the toxicity of JTE-013 observed in pilot studies. In a different experiment, mice were subjected to bile duct ligation (BDL) or sham operation for 3 weeks to induce fibrosis, as described previously (27,31). Blood and liver were harvested to evaluate liver fibrosis. All experimental procedures were performed under Mayo Institutional Animal Care and Use Committee oversight.
Sirius Red Staining-Livers samples were fixed in 10% phosphate-buffered formalin and embedded in paraffin. Five-micrometer sections were stained with picrosirius red (Sigma) and counterstained with fast green (Sigma). The proportion of tissue stained with picrosirius red content was quantified with ImageJ. Collagen I was stained with Sirius Red and quantitated in 10 randomly chosen sections per sample.
Serum Analysis of Liver Function-Serum alanine aminotransferase (ALT) and aspartate aminotransferase activity was determined from 5 animals/group as described previously (North Carolina Animal Lab Test Center) (9).
Human Subject-derived Samples-The patient-derived serum was approved by the Ethics Commission of the Mayo Clinic (Institution Review Board no. 13-002715). Samples derived from study subjects were used for isolation, quantification of the exosome, and detection of the S1P level.
Statistical Analysis-Results are expressed as the mean of three or more independent experiments. Two-tailed Student's t test or analysis of variance was used to test the statistical significance between groups as appropriate. p Ͻ 0.05 was considered statistically significant.

SK1 Is an EC-derived Exosome Protein-Because
ECs generate a significant percentage of the serum pool of exosomes (32) and are anatomically situated near HSCs, we focused on ECs and their exosome production as a paracrine mechanism of HSC migration (33). First, we performed angiogenesis pathway-specific microarray studies to identify specific candidate EC genes for further study. Studies were performed with the FGF2 ligand because this molecule has been implicated in both exosome release and liver fibrosis (23,34). The results showed that FGF2 induced a 2.4-fold increase in SK1 mRNA levels in ECs (Fig. 1A). To confirm the array results in vivo, we isolated ECs from control and cirrhotic mice and showed that SK1 mRNA was elevated during liver fibrosis (Fig. 1B). In vitro, we treated ECs with TGF␤, PDGF, or FGF2 to emulate liver fibrosis conditions and showed that EC SK1 mRNA was up-regulated (Fig. 1C). SK1 is an enzyme that produces the HSC chemotactic factor S1P. Additionally, active SK1 enzyme has been detected in the extracellular environment, although not specifically within an exosome fraction (24,25,27,(35)(36)(37). We therefore focused on SK1-derived S1P as a potential pathway relevant to our model of EC-HSC interaction. A, human umbilical vein endothelial cells were stimulated with FGF2 for 12 h, and an angiogenesis microarray was performed. SK1 up-regulation is highlighted (circle). B, isolated ECs from CTRL or cirrhotic mice were subjected to real-time PCR for SK1 quantification (n ϭ 3; *, p Ͻ 0.05). C, TSECs were stimulated with the fibrotic growth factors FGF2, TGF␤, or PDGF for 12 h. mRNA from the CTRL and treatment groups was subjected to real-time PCR for SK1 quantification (n ϭ 3; *, p Ͻ 0.05). D, isolated TSEC-derived and serum-derived exosomes were subjected to nanotracking analyses. Exosome tracking analyses show the mean diameters of 89 nm and 78 nm for TSECs and serum, respectively. E, immunogold staining for SK1 and the exosome markers CD81, TfR, and CD63 was performed. Scale bars ϭ 200 nm. F, WB analysis showing the expression of TSG101, TfR, CD63, and SK1 in serum-derived exosomes. G, TSECs were transfected with control or SK1-overexpressing plasmids. mRNA from CTRL and SK1-transduced TSEC were subjected to real-time PCR for SK1 quantification (n ϭ 3; *, p Ͻ 0.05). H, exosomes were isolated, and WB for SK1, TSG101, and CD63 was performed. I, S1P ELISA was performed with exosomes from TSECs transfected with control and SK1-overexpressing plasmids and from human serum to measure S1P levels (n ϭ 3; *, p Ͻ 0.05).
As an initial step to explore the hypothesis that SK1 might be delivered by exosomes, we isolated exosomes by differential centrifugation from two models: in vitro ECs and human serum, both of which provide abundant sources of exosomes with potential pathophysiological relevance (14,38). The presence of exosomes from these models was confirmed using several criteria, including the mode of diameter approximating 84 Ϯ 5 nm in serum-derived exosomes and 92 Ϯ 4 nm in TSECderived exosomes, as assessed by nanoparticle tracking analysis (Fig. 1D); a classic cup shape and double membrane morphology, as assessed by transmission electron microscopy (Fig. 1E); immunogold positivity for the exosome markers CD81, TfR (26,39), CD63, and SK1 (Fig. 1E); and WB detection of additional well characterized markers of exosomes (TSG101, TfR, CD63, and SK1) (Fig. 1F).
To explore the functionality of exosomal SK1, exosomes were first isolated from the conditioned media of TSECs after lentiviral overexpression of SK1 (6). SK1 mRNA levels in TSECs were up-regulated 13-fold after SK1 transduction compared with control transduction (Fig. 1G). Additionally, SK1 protein was increased significantly in exosomes derived from SK1overexpressing TSECs compared with the control (Fig. 1H). We then examined the S1P level in isolated exosomes to determine whether SK1 in exosomes could produce S1P. SK1-overexpressing exosomes contained 2-fold more S1P compared with control exosomes. Human serum-derived exosomes contained SK1 as well (Fig. 1I). With these well characterized exosome systems in hand, we next explored the effects of these exosomes on HSC activation in an in vitro model system.
Exosomal SK1 Promotes AKT Activation and HSC Migration-Because SK1/S1P have been implicated previously in cirrhosis (29,40), we examined the effect of SK1-containing exosomes on HSC migration, a key feature of HSC activation. Treatment of HSCs with serum-derived exosomes induced a prominent activation of AKT signaling within 15 min ( Fig. 2A). This time point is consistent with previous publications (41), prompting us to use a 15-min time point for the remainder of our experiments examining AKT phosphorylation. In conjunction with the pAKT analysis, HSC migration was analyzed using a Transwell assay. Treatment of HSCs with serum-derived exosomes induced robust HSC migration compared with bovine serum albumin controls (Fig. 2B). We obtained similar results of exosome-induced pAKT signaling and migration when we performed the same sets of experiments using a different hHSC cell line (Fig. 2, C and D). Two other cell types in liver, hepatocytes (HepG2) and macrophages (THP-1), were treated with exosomes. However, no significant up-regulation of pAKT was detected (Fig. 2, E and F). These results provide evidence that exosomes preferentially induce HSC pAKT signaling in the liver. Furthermore, to confirm the role of AKT in exosomeinduced HSC signaling and migration, we performed WB and Transwell assays with the presence of AKT inhibitor VIII, which inhibited exosome-induced AKT phosphorylation and migration (Fig. 2, G and H).
To further establish the role of SK1 activity in exosome function, exosomes were isolated from SK1-mutant, lentivirus-stable, transduced TSEC-conditioned media. SK1 mRNA levels in TSECs were up-regulated after SK1 WT, D81A dominant negative, or G113A constitutively active mutant construct transduction (Fig. 2M). CTRL, SK1, D81A, and G113A exosomes were isolated, and SK1 overexpression was confirmed by Western blot analysis (Fig. 2N). SK1 activity of WT and mutant exosomes was also confirmed by ELISA, showing that the active mutant increased S1P (Fig. 2O). An HSC Transwell assay was performed using exosomes from CTRL-, SK1-, D81A-, and G113A-transduced TSECs. Exosomes from G113A cells induced more HSC migration compared with SK1, whereas exosomes from D81A cells showed decreased HSC migration compared with SK1 WT cells (Fig. 2P). These results implicate exosomal SK1 as a mediator of HSC signal transduction and migration through the delivery of S1P.

FN-Integrin Interactions Contribute to Exosome Adhesion and Are Required for Exosome-induced HSC Signaling and
Migration-Our finding that SK1-overexpressing exosomes promote HSC migration motivated further mechanistic evaluation. It has been postulated that the delivery of exosome components may require exosome adhesion and/or endocytosis (3,5,10,24,25,33,34). The adhesion of exosomes was characterized by morphology, size, and number. We hypothesized that adhesion between exosomes and HSC may require ligand/receptor interactions to facilitate chemotactic signaling. FN is a glycoprotein of the extracellular matrix that is secreted by EC during liver fibrosis (27,(35)(36)(37)42). Indeed, exosomes contained FN, as detected by FN immunogold labeling (Fig. 3A), consistent with previous reports (40,43). To confirm that the observed fibronectin was derived from the exosome-producing cell and not a contaminant from plasma, an antibody against cellular fibronectin was used. Immunoblotting of serum-derived exosomes revealed that the FN observed was indeed cellular FN rather than plasma FN (Fig. 3B). To examine whether FN on exosomes was able to activate integrin on HSCs, biotin was used to label the surface protein of HSC, and streptavidinbeads were used to pull down the protein labeled by biotin. ␤1-integrin was activated after exosome treatment, implicating exosomal FN in the process of integrin activation and adhesion of exosomes to HSCs (Fig. 3C). Other extracellular matrix (ECM) proteins have been reported within exosomes (44,45). To examine whether other ECM proteins could be responsible for integrin activation by exosomes, we blotted lysed exosomes for collagen I, which was not detected in our serum-derived exosomes. This result was consistent with plasma exosome proteomic profiling reports, in which FN is the predominant ECM protein expressed on exosomes that could contribute to FN-integrin interactions (46,47). CD29 or RGD, both of which block integrin-FN binding, decreased the cell surface binding of exosomes (Fig. 3D). We utilized scanning electron microscopy, which provided us with direct visualization of adhesion on the basis of exosome morphology, size, and number. Exosome-induced AKT signaling and migration were also abolished by CD29 and RGD treatment (Fig. 3, E and F). These findings implicate exosome adhesion as a requisite step for exosomeinduced signaling and HSC migration.
We next examined potential links between exosome adhesion and endocytosis because exosome internalization has been implicated previously in exosome-mediated signaling (48). PKH67-labeled exosomes were incubated with HSCs in the presence of CD29 or RGD for 4 h. In the presence of CD29 or RGD, exosome endocytosis was decreased by 30% and 50%, respectively (Fig. 3G). These findings suggest that exosome adhesion is a critical early step required for exosome endocytosis. These insights then led us examine whether the endocytosis events ultimately mediate exosome-induced signaling and migration as well. The cell lysates were subjected to WB, and pAKT, tAKT, and GAPDH expression was analyzed (n ϭ 3; *, p Ͻ 0.05; ns, not significant). B, the effect of serum-derived exosomes in HSC migration was studied using a Transwell assay. Quantification of data in terms of number of cells per field is shown (n ϭ 3; *, p Ͻ 0.05). C, another HSC clonal line (hHSC) was used to repeat the previous experiments, where hHSCs were incubated with exosomes for 15 min and subjected to WB. pAKT, tAKT, and GAPDH expression was analyzed (n ϭ 3; *, p Ͻ 0.05). D, a Transwell assay was performed from exosome-stimulated hHSCs. Quantification of the number of cells per field is shown (n ϭ 3; *, p Ͻ 0.05). E and F, HepG2 and THP-1 cells were stimulated with exosomes for 15 min, and lysates were subjected to WB. pAKT, tAKT, and GAPDH expression was analyzed (n ϭ 3; *, p Ͻ 0.05). G, HSCs were pretreated with AKT inhibitor VIII with or without exosomes for 15 min. Cell lysates were subjected to WB for pAKT, tAKT, and GAPDH (n ϭ 3; *, p Ͻ 0.05). H, exosomes with AKT inhibitor VIII were incubated with HSCs to measure migration by Transwell assay (n ϭ 3; *, p Ͻ 0.05). I, HSCs were incubated with control and SK1-overexpressing exosomes for 15 min and then subjected to WB. SK1-overexpressing exosomes increased the expression of pAKT compared with control exosomes (n ϭ 3; *, p Ͻ 0.05). J, SK1-overexpressing exosomes increased HSC migration, as evaluated by Transwell assay. Quantification of data in terms of number of cells per field is shown (n ϭ 3; *, p Ͻ 0.05). K, conditioned medium from control and SK1-overexpressing exosomes incubated with HSCs was collected to measure S1P levels by S1P ELISA (n ϭ 3; *, p Ͻ 0.05). L, DMSO (control), DMS, or FTY720 (FTY) were delivered to serum-derived exosomes. Control, DMS, or FTY720 exosomes were incubated with HSCs to measure migration by Transwell assay (n ϭ 3; *, p Ͻ 0.05). M, TSECs were transfected with control or SK1 WT or mutant lentivirus. mRNA from CTRL, SK1-, D81A-, and G113A-transduced TSEC were subjected to real-time PCR for SK1 quantification (n ϭ 3; *, p Ͻ 0.05). N, exosomes were isolated from transfected TSECs and subjected to WB for SK1, TSG101, and TfR. O, exosomes were isolated from transfected TSECs and subjected to S1P ELISA for SK1 activity (n ϭ 3; *, p Ͻ 0.05). P, CTRL, SK1, D81A, and G113A exosomes were incubated with HSCs to measure migration by Transwell assay (n ϭ 3; *, p Ͻ 0.05). AKT In, AKT inhibitor; Exo, exosomes. DECEMBER 25, 2015 • VOLUME 290 • NUMBER 52

JOURNAL OF BIOLOGICAL CHEMISTRY 30689
Exosome Endocytosis and Signaling Are Dependent on Dynamin-2-We next sought to determine whether signaling and migration of HSCs induced by exosomes required their internalization. The endocytic pathway involves multiple mechanisms, including both clathrin-and caveolin-coated vesicles, both of which are mediated by dynamin-2 GTPase activity, which is implicated in vesicle scission (38). We therefore perturbed dynamin function and assessed the effects on our exosome signaling model. After transfection of siRNA against dynamin-2, exosome-induced increases in the pAKT/tAKT ratio and HSC migration were both attenuated markedly (Fig. 4,  A and B). We confirmed that exosome endocytosis was decreased by ϳ50% in the dynamin-2 siRNA-transfected cells using a PKH labeling technique (Fig. 4C).
We subsequently focused on disruption of dynamin-2 function using additional approaches. HSCs were transduced with the dynamin-2 K44A dominant negative construct using an adenoviral construct containing a point mutation in the GTP binding element that prevents GTP hydrolysis (49). In control cells, exosomes activated AKT after just 5 min. However, in dynamin-2 K44A cells, activation was delayed, occurring at 30 min (Fig. 4D). Consistent with this observation, a pharmacologic approach using the dynamin-2 inhibitor Dynasore also attenuated exosome-mediated AKT activation. Furthermore, PKH labeling studies confirmed that the dynamin-2 K44A construct inhibited exosome internalization by 50% compared with LacZ control transduction. Pharmacologic inhibition using Dynasore also blocked 60% of exosome internalization compared with control cells (Fig. 4E). To assess the destination of internalized exosomes, we labeled exosomes with PKH67 and detected the subcellular signal by immunofluorescence. We found colocalization of exosomes with early endosomes and, subsequently, their presence in lysosomes on the basis of double immunofluorescence staining using the early endosome  FN (cFN), collagen I, and TSG101 in three different batches of serum-derived exosomes (Exo). HSCs treated with TGF␤ served as a positive control. C, HSCs were subjected to a biotinylation assay of activated ␤1-integrin, followed streptavidin pulldown and WB analysis for HUTS4 and CD29, from the input portion, as a loading control) (n ϭ 3; *, p Ͻ 0.05). D, scanning electron microscopy was performed to evaluate the effect of FN-integrin-blocking reagents (CD29 and RGD) on exosome uptake by HSCs. Exosome adhesion to HSCs is highlighted. Quantification of exosome number per cell is shown (n ϭ 6; *, p Ͻ 0.05; ns, not significant). E, HSC pretreated with RGD or CD29 were treated with exosomes for different times. Cell lysates were subjected to WB for pAKT and tAKT analyses (n ϭ 3; *, p Ͻ 0.05). F, exosomes treated with RGD or CD29 were incubated with HSCs to measure migration by Transwell assay (n ϭ 3; *, p Ͻ 0.05). Veh, vehicle. G, exosomes were stained with PKH67 for quantification analyses. Stained exosomes were quantified after incubation with HSCs pretreated with vehicle, CD29, or RGD (n ϭ 3; *, p Ͻ 0.05).
marker EEA and the lysosome marker LAMP1 (Pearson coefficients of colocalization, 0.32 and 0.36, respectively) (Fig. 4, F  and G). Therefore, these data support a model in which exosome-dependent migration of HSC requires dynamin-2 GTPase-dependent endocytosis.
Exosomal SK1 Contributes to Liver Fibrosis-SK1/S1P pathway has been implicated previously in liver fibrogenesis. This prompted us to explore SK1 expression and S1P production in an in vivo murine model of liver fibrosis with a focus on exosome mechanisms. SK1 was expressed in normal liver and was up-regulated markedly in cirrhotic liver samples from humans (2.5-fold, p Ͻ 0.05) (Fig. 5A). Exosomes isolated from alcoholic hepatitis patient serum contained 1.5-fold more S1P compared with healthy donor serum (Fig. 5B). The same effect was dem-onstrated following chronic liver injury induced in mice by administration of the fibrogenic hepatotoxin CCl 4 or by BDL. SK1 mRNA was increased markedly in both CCl 4 and BDL models of fibrosis (Ͼ20-fold, p Ͻ 0.05, Fig. 5C). Levels of the HSC activation and fibrosis markers ␣-smooth muscle actin and collagen I, respectively, were up-regulated in the CCl 4 mouse model compared with mice administered olive oil (Fig.  5D). Aspartate aminotransferase and ALT levels were significantly greater in CCl 4 mice compared with olive oil-treated control mice (Fig. 5E). Sirius Red, a marker of collagen I deposition, was elevated in CCl 4 mice as well (Fig. 5F). Furthermore, the serum S1P level was also increased 1.5-fold in CCl 4 -treated mice (Fig. 5G). We then isolated exosomes from the serum pool of olive oil-and CCl 4 -treated mice or SHAM and BDL mice by FIGURE 4. Exosome endocytosis and signaling are dependent on dynamin-2. A, HSCs were transfected with control and dynamin siRNA, and transfection efficiency was tested by WB with dynamin-2 antibody. Left panel, Dynamin-2 knockdown cells were treated with exosomes and analyzed for dynamin-2, pAKT, tAKT and GAPDH expression by WB. Right panel, WB densitometry (n ϭ 5; *, p Ͻ 0.05). B, cells treated with control and dynamin-2 siRNA were incubated with exosomes in a Transwell assay. Quantification of number of cells per field is shown (n ϭ 3; *, p Ͻ 0.05; ns, not significant). C, PKH67-labeled exosomes were incubated with HSCs and then imaged to quantify endocytosis ability (n ϭ 3; *, p Ͻ 0.05). D, WB for pAKT and tAKT. Right panel, HSCs transfected with LacZ or dynamin-2 K44A were incubated with exosomes for different times (n ϭ 8; *, p Ͻ 0.05). Left panel, HSCs pretreated with the control (DMSO) or Dynasore were incubated with exosomes for different times (n ϭ 8, *p Ͻ 0.05). Densitometry results for both analyses are shown under the WB. E, PKH67-labeled exosomes were incubated with Dynasore-treated HSCs or LacZ/dynamin-2 K44A-transfected cells and subjected to FACS. The -fold change of positive PKH67 cells is shown (n ϭ 3; *, p Ͻ 0.05). F and G, left panels, HSCs were incubated with PKH67-labeled exosomes for 1, 4, and 12 h and subjected to double immunofluorescence for PKH-67-labeled exosomes (green) and EEA (red in F), or LAMP1 (red in G). Cell nuclei were stained by Hoechst (blue). Right panels, Pearson coefficients of colocalization between exosomes and EEA or LAMP1 were analyzed by ImageJ (n ϭ 3; *, p Ͻ 0.05). The comparison is between 0 h and different time points. DECEMBER 25, 2015 • VOLUME 290 • NUMBER 52 ultracentrifugation. S1P levels were up-regulated in exosomes from CCl 4 -treated mice and BDL mice exosomes compared with olive oil-treated mice or SHAM-treated mice (1.4-fold and 2.5-fold, respectively) (Fig. 5H). The WB results indicated that SK1 was up-regulated in exosomes derived from CCl 4 -treated mice compared with the olive oil group (normalized by TSG101 and CD63 expression, Fig. 5I). Nanoparticle tracking analyses further demonstrated that CCl 4 -treated mouse serum contained higher number of exosomes than serum from olive oiltreated mice (Fig. 5J). Exosomes derived from CCl 4 -treated murine serum induced more HSC migration than those derived from olive oil-treated mouse serum (Fig. 5K). Overall, these findings indicate that exosomal SK1 may contribute to liver fibrosis in vivo.

Exosome and HSC S1P-induced Migration
Blockade of S1PR2 Attenuates Liver Fibrosis in Vivo-Because exosomes contained SK1 and were up-regulated in cirrhosis, we next examined whether administration of the S1PR2 inhibitor (JTE-013) resulted in attenuated fibrosis in the CCl 4induced fibrosis mice model. JTE-013 was administered intraperitoneally twice a week prior to each CCl 4 injection. We then examined the effects of JTE-013 on fibrotic makers. After 3 weeks of CCl 4 treatment, serum ALT levels (Fig. 6A), exosome number (Fig. 6B), collagen I mRNA expression (Fig. 6C), SK1 mRNA (Fig. 6D), Sirius Red staining (Fig. 6E), ␣-smooth muscle actin protein expression, and collagen I protein expression (Fig.  6F) in liver tissue were up-regulated markedly. However, JTE-013 administration resulted in a significant attenuation in these parameters compared with mice treated with CCl 4 alone. These FIGURE 5. Exosomal SK1 contributes to liver cirrhosis. The target gene mRNA level was quantified by real-time PCR. A, normal and cirrhotic liver patient samples were subject to real-time PCR for SK1 quantification (n ϭ 3; *, p Ͻ 0.05). B, exosomes isolated from healthy donor and alcoholic hepatitis sera were subjected to S1P ELISA. S1P levels are shown (n ϭ 3; *, p Ͻ 0.05). C, liver samples from the distinct liver fibrosis mouse models (CCl 4 -induced and BDL) were subjected to real-time PCR for SK1 quantification (n ϭ 3; *, p Ͻ 0.05). D, CCl 4 was used to generate liver cirrhosis in mice. The levels of the HSC activation and fibrosis markers ␣-smooth muscle actin (␣SMA) and collagen I␣1, respectively, were analyzed by real-time PCR (n ϭ 5; *, p Ͻ 0.05). E, the levels of aspartate aminotransferase (AST) and ALT were measured by serum biochemical analysis in control and CCl 4 -treated mice (n ϭ 6; *, p Ͻ 0.05). F, Sirius Red staining was performed in livers of olive oil-and CCl 4 -treated mice to indicate collagen I production. The -fold change of collagen I-positive areas is shown (n ϭ 6; *, p Ͻ 0.05). G, serum was collected from olive oil-treated control and CCl 4 -treated mice for the S1P ELISA assay. S1P production is shown (n ϭ 20; *, p Ͻ 0.05). H, exosomes isolated from olive oil-treated and CCl 4 -treated mouse serum and SHAM and BDL mouse serum were subjected to S1P ELISA. S1P levels are shown (n ϭ 6; *, p Ͻ 0.05). I, exosomes isolated from olive oil-treated and CCl 4 -treated mouse serum were subjected to SK1, TfR, and TSG101 analyses by WB (n ϭ 5; *, p Ͻ 0.05). J, nanoparticle tracking analysis was used to analyze the total number of exosomes from the sera of olive oil-and CCl 4 -treated mice. K, HSCs were incubated with exosomes isolated from olive oil-and CCl 4 -treated mouse serum to measure migration by Transwell assay. The number of migrated cells per field is shown (n ϭ 3; *, p Ͻ 0.05).
results further support the key role of S1P/SK1 and exosomes in liver fibrosis in vivo.

Discussion
Although the pathogenesis of liver fibrosis is not fully defined, the activation of HSCs into myofibroblasts is recognized as a sentinel step in the development of this disease (3,5). This study includes novel observations that advance our understanding of how exosomes can contribute to this process, including the following: exosomes transduce a SK1/S1P pathway that mediates HSC migration, exosomal FN engages HSC integrins to mediate exosome adhesion, and HSC dynamin-2 mediates exosome entry into HSCs and eventual signaling. These findings extend our current understanding of both exosome and HSC biology.
Exosomes are increasingly implicated in a variety of pathobiological conditions, and some prior studies of liver cirrhosis have been variable in outcome. For example, serum exosome numbers are increased in patients with liver disease (14). In vitro studies indicate that exosomes may up-regulate the expression of fibrolytic matrix metalloproteinase genes in HSCs (42,43). In other studies, exosomes were capable of transducing Hedgehog signals (13). Microparticles have also been implicated in cirrhosis and portal hypertension, with evidence that such particles may contain the angiogenic molecule VEGF and other studies providing evidence that these particles can mediate hemodynamic disturbances associated with cirrhosis (50). Our study adds to these existing models not only by showing a role for exosomes in paracrine signaling relevant to cir-FIGURE 6. Blockade of S1PR2 attenuates liver fibrosis in vivo. C57BL/6 mice were subjected to olive oil or CCl 4 for 3 weeks. An S1PR2 inhibitor (JTE-013) was administered 1 week before CCl 4 injection and 1 day before each CCL4 injection. A, levels of ALT were measured by serum biochemical analysis in olive oil control-, CCl 4 -, and CCl 4 ϩ JTE-013-treated mice (n ϭ 6; *, p Ͻ 0.05). B, nanotracking analysis was used to analyze the total number of exosomes from the serum of olive oil-, CCl 4 -, and CCl 4 ϩ JTE-013-treated mice (n ϭ 6; **, p Ͻ 0.05; ns, not significant). C and D, levels of HSC activation and fibrosis markers (collagen I and SK1) were analyzed by real-time PCR (n ϭ 6; *, p Ͻ 0.05). E, liver samples from olive oil control-, CCl 4 -, and CCl 4 ϩ JTE-013-treated mice were subjected to WB for ␣-smooth muscle actin (␣SMA) and collagen I quantification (n ϭ 6; *, p Ͻ 0.05). F, Sirius Red staining of liver sections was performed in olive oil control-, CCl 4 -, and CCl 4 ϩ JTE-013-treated mice to assess tissue collagen levels. The -fold change of collagen I-positive areas is shown, and a representative micrograph is also shown (n ϭ 6; *, p Ͻ 0.05). G, proposed model of exosome function in HSC migration. Liver ECs release SK1-containing exosomes. Exosomes engage with HSCs via FN-integrin-dependent adhesion and dynamin-dependent endocytosis. SK1 is delivered by exosomes to induce HSC signaling and migration. rhosis but also through mechanistic studies that elucidate how exosomes act on target cells.
The SK1-S1P pathway has been implicated previously in liver fibrosis. Our in vivo studies are consistent with previous reports showing that inhibition of the SK1-S1P axis protects mice from liver fibrosis. Co-treatment with an SK1 inhibitor, FTY720, JTE-013 (S1PR2 inhibitor), or VPC23019 (S1P R1/3 inhibitor) protected mice from BDL and CCl 4 -induced liver fibrosis (51,52). S1PR2 KO mice were protected from BDL-induced and CCl 4 -induced liver fibrosis (53,54), although a number of salient details are lacking regarding how this effect is achieved. Fortuitously, we identified that SK1 was up-regulated in ECs after FGF2 stimulation using an unbiased microarray-based approach. Additionally, prior studies have shown that active SK1 could be released from cells, although the mechanism and vehicle of release were not identified (1,24,25,29). Conceptually, the packaging of SK1 into exosomes could protect the enzyme from degradation and maintain its stability until reaching its target cell of action. Indeed, we were able to confirm such a model in this study because pharmacologic and molecular inhibition of exosomal SK1 failed to induce HSC migration. Furthermore, we showed that SK1 was increased in serum-derived exosomes of cirrhotic mice, linking our exosome biology to a pathobiological process. Therefore, our results uncover several new pieces of information regarding the mechanisms of how exosomal SK1 could contribute to paracrine S1P signaling and fibrogenesis.
The mechanisms by which exosomes reach and attach to target cells are not well understood. Here we show that exosomes contain FN, a canonical ligand for specific cell surface integrins, and that this interaction can mediate exosome adhesion to HSCs. Adhesion appears to be a prerequisite step for exosome signaling in target cells and for endocytic entry of exosomes into target cells. Interestingly, integrin recycling and activation may be regulated by engagement with exosomal FN. In a recent study (3,55), human macrophage-derived exosomes suppressed EC migration and abolished collagen I-induced ERK signaling pathways by blocking internalized integrin recycling. Therefore, sequential and reciprocal interactions between exosome adhesion and internalization may occur through the integrin endocytic recycling pathway.
Endocytosis may occur through multiple pathways, including phagocytosis, macropinocytosis, and clathrin-and/or caveola-mediated endocytosis (8,56). In prior studies, the mechanism of exosome uptake has varied, and data exists to support each of these pathways (10,12,14,(57)(58)(59)(60). In this study, exosome endocytosis was mediated by dynamin because knockdown or inhibition of this protein attenuated exosome uptake and associated signaling. Therefore, this work supports the hypothesis that exosome endocytosis is required for its signaling activity.
In summary, exosomal SK1 regulates HSC signaling and migration. FN-integrin induced adhesion between exosomes and HSC as well as subsequent dynamin-2 dependent exosome endocytosis are both required for signaling (Fig. 6G). Murine and human studies also support this model under pathobiological conditions. In total, these results enhance our understand-ing of exosome biology, paracrine vascular signaling, and fibrogenesis.