**4. Discussion**

EVs have attracted considerable attention for the role they likely play in mediating cell-cell communication throughout the body. In the liver, EVs are proposed to regulate hepatic homeostasis or to contribute to pathophysiogical processes such as viral spread, non-alcoholic steatohepatitis, alcoholic liver disease, and cancer [16,17]. A large body of research has shown that hepatocytes infected with hepatitis B or C viruses or that have been exposed to agents such as alcohol, carbon tetrachloride or palmitate produce EVs that stimulate macrophage activation and immune function which are common features of numerous liver diseases and often associated with fibrotic pathology [30,35–40]. HSC fibrogenesis is stimulated by these types of hepatocyte-derived EVs [29,32,41], while other phenotypic features of activated HSC such as migration and AKT phosphorylation have been shown to be enhanced by EVs from liver sinusoidal endothelial cells [42]. On the other hand, EVs from healthy hepatocytes [43], various stem cells [44–51] or the serum of healthy mice [52] have the ability to inhibit experimental liver fibrosis, largely by suppressing inflammatory responses and/or pathways of activation or fibrogenesis in HSC. The recognition that HSC are EV targets has highlighted an important new mechanism by which fibrogenic pathways in the liver are modulated and has given a new lead for novel anti-fibrotic therapies based on suppressing the action of pro-fibrotic EVs or harnessing the actions of EVs that are intrinsically anti-fibrotic.

As shown in this report, HSC themselves are also EV producers but relatively little is known about this phenomenon. Whereas the physico-chemical properties of HSC EVs were quite consistent whether their producer cells were activated or not, we found that the HSC EV production rate, biological properties and protein composition were highly dependent on HSC activation status. Specifically, our results showed, firstly, that fibrogenic gene expression was suppressed upon exposure of activated HSC to EVs from relatively quiescent HSC but not to EVs from activated HSC and, secondly, that gene expression in relatively quiescent HSC was stimulated upon exposure of the cells to EVs from activated HSC but not to EVs from relatively quiescent cells. Thus, pathways of EV communication between HSC may be either stimulatory or inhibitory and are manifest principally when the activation status of the EV producer HSC is different to that of the EV recipient HSC. We also showed that HSC activation is associated with a 4.5-fold increase in EV release and that EVs from activated HSC contained considerably more proteomic information than their quiescent counterparts: there were 337 proteins in EVs from P1 mHSC but only 46 proteins in EVs from D4 mHSC. Generally, EVs from D4 mHSC exhibited a high abundance and high proportion of histones and keratins, while EVs from P1 mHSC contained proteins principally associated with extracellular spaces, ECM, proteasome complexes, collagens, ECM, vesicular transport, metabolic enzymes, ribosomes and chaperones. These differences reflected the distinct phenotypes and functions of their respective EV producer cells: quiescent HSC are resting vitamin-A storing cells whereas activated HSC are contractile myofibroblasts that interact with various immune cells, are highly proliferative and migratory, are metabolically very active, have high energy requirements and produce numerous cytokines, chemokines and extracellular matrix components. Even so, it will be important in future studies to evaluate the molecular payload of EVs from HSC that have been activated in vitro by cytokines (e.g., TGF-β) or in vivo due to liver injury as there may be qualitative or quantitative differences that have functional impact as compared to the EVs in this study that were from HSC that had autonomously activated in culture.

It is interesting that FN1 was expressed exclusively in EVs from P1 HSC and was the most abundant protein overall in EVs from activated or quiescent HSC. FN1 exists either as a soluble plasma form that lacks EDA and EDB domains and is produced principally by hepatocytes or as a cell-associated form which contains the EDA and EDB domains and is produced by numerous cell types. Analysis of the FN1 protein in P1 mHSC EVs showed it to be the cell-associated form because sequencing identified a near-complete EDA domain as well as a partial N-terminal EDB domain, consistent with its detection using FN1 antibodies directed to the EDA or EDB regions. We have previously reported that HSC EVs use cell surface integrins as receptors, including integrin α5β1, which is a receptor for FN1 [34]. In activated HSC, FN1-integrin α5β1 interactions are important for the regulation of cell adhesion, survival, cytoskeletal rearrangements or expression of matrix metalloproteases or collagen I [53–57], but this interaction may also underlie the binding of EVs to target HSC, the extent of which will be dependent on activation-associated changes in cellular integrin expression and EV FN1 levels. Importantly, EV FN1 may also directly participate in downstream pro-fibrogenic actions of HSC EVs in light of recent studies showing, firstly, that FN1 in cancer cell microvesicles mediates their ability to confer transformation characteristics on fibroblasts and epithelial cells [58] and, secondly, that exosomal FN1 mediates the mitogenic activity of exosomes from mesenchymal stem cells [59]. Such functions may also be conserved in EVs from hHSC, since we showed FN1 to be the most abundant component of EVs from LX-2 cells in these studies. Indeed, an important feature that emerged from the current investigation was the recognition that EVs from activated hHSC have the same pro-fibrogenic properties and share many of the same proteins as their mouse counterparts resulting in considerable overlap in the components and pathways in which they are involved. The 206 proteins that were common to both species represented approximately 40% of all 524 proteins in EVs from activated hHSC and 61% of all 337 proteins in EVs from activated mHSC. The incorporation of these shared proteins into EVs from activated HSC of mice and humans suggests that their functional roles were under strong selective pressure during evolution.

It remains to be determined what other constituents (protein, RNA or miRNA or combinations thereof) in the EV molecular payload might be relevant to the relative suppressive or stimulatory actions of EVs from, respectively, quiescent or activated HSC. Even so, it is striking that the proteomic payloads of EVs from activated human or mouse HSC correspond to cellular components that are

well characterized for their involvement in HSC activation and/or fibrogenesis (i.e., ECM, proteasome complexes, collagens, metabolic enzymes, ribosomes, chaperones) and it is tempting to speculate that the EV counterparts have similar functions. Although we have previously used a transfection approach to show that HSC-derived EVs can shuttle GFP-CCN2 intercellularly [24], this overexpression system may not have faithfully mimicked native mechanisms because CCN2 was not detected in EVs from P1 mHSC in this study. With respect to the anti-fibrogenic actions of EVs from D4 mHSC, the high prevalence of keratins or histones in the relatively small proteome suggests that structural elements or nucleosomal regulation may underlie the suppressive activities. However, preliminary transfection studies in which activated HSC were transfected with cDNAs encoding single D4 mHSC-specific or -enriched EV histones (H4, H10, H11, H13, H14, H15, H2B1F) did not result in an attenuation of fibrogenic markers even though each over-expressed protein was detected by Western blot (X.L. and D.R.B., unpublished data), suggesting that these proteins are not anti-fibrogenic at least when tested individually, but additional combinatorial testing of these and other candidates (e.g., keratins) must be undertaken in the future. For a variety of differentially expressed mHSC EV proteins, we found that the levels of their corresponding cellular transcripts were quite variable, ranging from concordant to discordant with their respective EV protein levels, suggesting that there is selectivity in the post-transcriptional or post-translational mechanisms by which a given protein is incorporated into the EVs.

Apart from proteins, the EV molecular payload contains miRs and mRNAs that may also contribute to EV-mediated regulation of fibrogenesis. For example, we previously showed that inhibition of pro-fibrogenic CCN2 in quiescent HSC is achieved by targeting of the CCN2 3 untranslated region by Twist-1-miR-199a-miR-214 and that all three components of this axis can be delivered in EVs to activated HSC in which CCN2 expression is then suppressed [25,27,28]. Even so, one must recognize that while these types of reductionist strategies to understand EV bioactivity have been widely pursued in the EV research field in general, they do not take account of EV heterogeneity at the single vesicle and systems level and the importance of adopting a more global view of EV cargo as it relates to functional aspects of EV biology has recently been emphasized [60]. Thus, a holistic approach may be preferable for understanding the combinatorial actions of EV cargo constituents in mediating EV biological actions [60]. We expect that detailed comparative studies between EVs from quiescent versus activated HSC of their respective miRnomes and RNAnomes, together with their proteomes as accomplished in this study, will provide a foundation for the identification of EV components and their corresponding cellular targets that accounts for their distinct biological actions. Whatever the factors involved, it is interesting to speculate that the suppressive actions of EVs from quiescent HSC may help to protect the liver from overt HSC activation, especially in cases of mild or acute injury.

In conclusion, mHSC activation is associated with an increase in EV production, a switch in EV bioactivity whereby EV-mediated suppression of mHSC fibrogenic gene expression gives way to EV-mediated stimulation of the same, and a dramatic increase in the complexity of the EV proteome. Activated hHSC produce EVs that exhibit profibrogenic activities and have similar protein components and functions as EVs from activated mHSC. We thus propose that activation-associated changes in production, function and protein content of EVs from HSC may contribute to the regulation of HSC function in vivo and to the fine-tuning of fibrogenic pathways in the liver.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4409/9/2/290/s1: Figure S1. Fibronectin in EVs from P1 mHSC; Figure S2. String analysis for entire proteome in EVs from D4 mHSC; Figure S3. String analysis for entire proteome in EVs from P1 mHSC; Figure S4. String analysis for entire proteome in EVs from LX-2 hHSC; Figure S5. String analysis for proteins shared in EVs from P1 mHSC and LX-2 hHSC; Table S1. qRT-PCR primers used for mouse gene detection; Table S2. MS datasets for D4 mHSC proteome; Table S3. MS datasets for P1 mHSC proteome; Table S4. MS datasets for LX-2 hHSC proteome.

**Author Contributions:** Conceptualization, X.L. and D.R.B.; Data curation, X.L., R.C. and S.K.; Formal analysis, X.L., R.C. and D.R.B.; Funding acquisition, D.R.B.; Investigation, X.L., R.C. and S.K.; Methodology, X.L., R.C. and S.K.; Project administration, D.R.B.; Writing – original draft, X.L. and D.R.B.; Writing – review and editing, X.L., R.C., S.K. and D.R.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by NIH grants R21 AA025974 and R21 AA023626.

**Acknowledgments:** We thank the staff of the Proteomics Shared Resource of the OSU Campus Chemical Instrument Center for help with Mass Spectrometry which was supported by NIH grant P30 CA016058.

**Conflicts of Interest:** The authors declare no conflict of interest.
