Ferulic Acid Induces Bone Marrow Mesenchymal Stem Cells to Alleviate the Activation of Hepatic Stellate Cells and Liver Fibrosis via Cytoskeletal Rearrangement Inhibition and Mir-19b-3p Transfer

Background: Bone marrow mesenchymal stem cells (BMSCs) are effective for treating brotic liver. BMSCs contain a variety of proteins and RNAs, which have functions similar to their derived cells, but the specic mechanism is unclear. In a recent study, ferulic acid (FA) was highly effective in treating liver brosis. Therefore, we combined BMSCs and FA to treat CCl 4 -induced brosis models. Methods: First, we used BMSCs and FA to treat CCl 4 -induced brosis models and observed their therapeutic effect, investigated the specic mechanism of this combination therapy in liver brosis. Second, we created a BMSC/hepatic stellate cell (HSC) co-culture system and used FA to treat activated HSCs. We next used cytochalasin D and angiotensin II to investigate whether BMSCs and FA inactivate HSCs through cytoskeletal rearrangement. MiR-19b-3p was enriched in BMSCs and targeted TGF-β receptor II (TGF-βR2). We transfected miR-19b-3p into HSCs and BMSCs separately and detected whether BMSCs transferred miR-19b-3p to HSCs or inactivated HSCs. Results: We used BMSCs and FA to treat CCl 4 -induced brosis models and found that the combination therapy had better effects than FA or BMSCs alone. The expression of the probrotic markers α-SMA and COL1-A1 was signicantly decreased in HSCs co-cultured with BMSCs and FA treatment. Cytoskeletal rearrangement in HSCs was inhibited, and RhoA/ROCK pathway gene expression was decreased. With angiotensin II treatment, COL1-A1 and a-SMA expression increased, while with cytochalasin D treatment, probrotic gene expression decreased in HSCs. COL1-A1, α-SMA and RhoA/ROCK pathway genes were decreased in activated HSCs treated with a miR-19b-3p mimic, indicating that miR-19b-3p inactivated HSCs by suppressing RhoA/ROCK signalling. In contrast, probrotic genes were signicantly decreased in BMSCs treated with the miR-19b-3p mimic or a miR-19b-3p inhibitor and FA compared with BMSCs treated with the miR-19b-3p mimic alone. Conclusion: BMSCs attenuated HSC activation and liver brosis by inhibiting cytoskeletal rearrangement and delivering miR-19b-3p to activated HSCs, inactivating RhoA/ROCK signaling. FA-based combination therapy showed better inhibitory effects on HSC activation, suggesting that BMSCs and their miRNAs combined with FA are novel antibrotic therapeutics for treating chronic liver disease. UTR of TGF- -βR2 were transformed into Escherichia coli, and then plasmid DNA was extracted from well-transformed, ampicillin-resistant E. coli, using an AccuPrep plasmid mini extraction kit (Tiangen, China). The sequences of the miR-19b-3p-binding sites of the 3’UTR of TGF-βR2 were conrmed by sequencing analysis (KeyGEN, China). Mutant vectors lacking the miR-19b-3p-binding site were manufactured by KeyGEN (KeyGEN, China). For the luciferase reporter assay, HSC-T6 cells were seeded in 24-well culture plates in culture medium without P/S 1 day before transfection. Using an electronic transient transfection machine (BTX Genimi X2, USA), cells were transfected with a mixture of pmirGLO vector construct and either 20 ug miR-19b-3p mimic (KeyGEN) or scrambled miRNA (NC-d; KeyGEN) as an NC. At 48 h after transfection, cells were harvested and tested with the Dual-Luciferase reporter assay system (Beyotime, China) according to the manufacturer’s protocol. All rey luciferase activity data are normalized to Renilla luciferase activity and presented as the mean ± SEM of values from at least three repetitive experiments. These results demonstrate that miR-19b-3p plays a critical role in mediating the effect of BMSCs on liver brosis. Taken together, these results suggest that FA combined with BMSC treatment can induce greater release of miR-19b-3p, which inhibits the activation of HSCs by suppressing the RhoA/ROCK1/SRF and RhoA/ROCK1/LIMK1 pathways, contributing to alleviating liver brosis. to HSCs and that the cytoskeleton was similar to that of inactivated HSCs; BMSCs transfected with the inhibitors released less miR-19b-3p to HSCs, and the cytoskeleton was similar to that of activated HSCs. The protein expression of RhoA, ROCK, SRF, LIMK1, COL1-A1 and α-SMA showed that BMSCs could release miR-19b-3p to inhibit HSC activation, and BMSCs combined with FA treatment showed better HSC inhibition. These results support the notion that combining BMSCs with FA could be a promising therapeutic strategy for the treatment of liver disease.


Introduction
Although the aetiologies of liver diseases are varied, a common pathological feature of most chronic liver diseases is liver brosis, which is characterized by the progressive replacement of functional hepatic tissue with extracellular matrix (ECM) [1,2] . Hepatic stellate cells (HSCs) are a key brogenic cell type that contributes to liver brosis. Upon liver injury, Kupffer cells and endothelial cells in the liver release cytokines, such as transforming growth factor-β (TGF-β), to activate quiescent HSCs. Activated HSCs transdifferentiate into myo broblastic HSCs [3] . Then, the cytoskeleton in these HSCs undergoes rearrangement, the cell dynamics increase, and these cells migrate to the site of liver injury and produce substantial amounts of ECM components, such as α-smooth actin (α-SMA) and collagens. Fired-Man proved that activated HSCs are the main source of extracellular collagen in the liver. Hence, regulating HSC activation and migration could be a potential anti brotic therapeutic strategy.
Bone marrow mesenchymal stem cells (BMSCs) have become an emerging therapeutic agent for liver brosis in recent years due to their advantages of proliferative potential, low immunogenicity, abundant sources, and tissue repair, chemotactic and homing abilities [4] . Many studies have reported that the therapeutic effects of mesenchymal stem cells (MSCs) against liver brosis/cirrhosis are related to the capacity of these cells to undergo hepatocyte-like differentiation, exert immunomodulatory activities, and secrete paracrine factors. Infusion of BMSCs was reported to ameliorate CCl 4 -induced liver damage and brosis in rats via the FGF2-Dlk1-Notch1 pathway [5] . Due to the cytoskeletal changes that occur when HSCs are activated, we hypothesized that BMSCs could ameliorate liver brosis by inhibiting cytoskeletal rearrangement. In addition, many recent studies have proven that BMSCs can produce microRNAs (miRNAs) [6] . Feng et al. demonstrated that MSC-derived miRNAs reduced infarct size and cardiac brosis by inhibiting apoptosis in mice with myocardial infarction. miR-125b contained in exosomes derived from chorionic plate-derived MSCs (CP-MSCs) ameliorated hepatic brosis by suppressing hedgehog (Hh) signalling, which is an essential regulator in liver brosis [7,8] . These ndings suggest that miRNAs play important roles in MSC-mediated tissue repair and regeneration.
Ferulic acid (FA), a derivative of cinnamic acid, has therapeutic activity against a variety of diseases. The antioxidant effects of FA have also been demonstrated in previous studies. The hepatoprotective effects of FA have been demonstrated in studies using CCl 4 as an initiator of liver brosis [9][10][11] .
The protective mechanisms of BMSCs in liver brosis have been studied, but few reports have studied the inhibition of cytoskeletal rearrangement in liver brosis. Of the studies performed, single BMSC treatment for liver diseases has shown inconsistent e cacy. Therefore, the mechanisms underlying the anti brotic effects of BMSCs in combination with FA were investigated in the current study.

Elisa Assay
We extracted the culture medium and centrifuged (10000 g, 20min) to remove the cell debris. Then we detected the content of COL1-A1 and α-SMA in culture medium with enzyme linked immunosorbent assay kits (mlbio, China) according to the manufacturer's protocol. All data are presented as the mean ± SEM of values from at least three repetitive experiments.

Cloning of vector constructs and luciferase reporter assay
For cloning of vector constructs, target genes of Rat miR-19b-3p were predicted by bioinformatics analysis using the online database: http://mirwalk.umm.uniheidelberg.de. The 3' UTR of rat TGF-βR2 from genomic DNA, containing binding sites for rat miR-19b-3p, was ampli ed by PCR. The primers sequences used for vector construction were as follows: forward, 5'TTCTAGTTGTTTAAACGAGCTCGCTAGCCTCGAGCTTTTTCTGGGCAGGCTGGGCCAAGACTCCG3',reverse,5'GCAGCCGGATCAGCTTGCATGCCTGCAGGTCGACA 3'. The PCR product was puri ed using an AccuPrep PCR puri cation kit (Tiangen), cut by the restriction enzymes Xho1 and Sall, and then cloned into the pmirGLO vector (Promega, WI, USA). The vector constructs with 3' UTR of TGF--βR2 were transformed into Escherichia coli, and then plasmid DNA was extracted from well-transformed, ampicillin-resistant E. coli, using an AccuPrep plasmid mini extraction kit (Tiangen, China). The sequences of the miR-19b-3pbinding sites of the 3'UTR of TGF-βR2 were con rmed by sequencing analysis (KeyGEN, China). Mutant vectors lacking the miR-19b-3p-binding site were manufactured by KeyGEN (KeyGEN, China). For the luciferase reporter assay, HSC-T6 cells were seeded in 24-well culture plates in culture medium without P/S 1 day before transfection. Using an electronic transient transfection machine (BTX Genimi X2, USA), cells were transfected with a mixture of pmirGLO vector construct and either 20 ug miR-19b-3p mimic (KeyGEN) or scrambled miRNA (NC-d; KeyGEN) as an NC. At 48 h after transfection, cells were harvested and tested with the Dual-Luciferase reporter assay system (Beyotime, China) according to the manufacturer's protocol. All re y luciferase activity data are normalized to Renilla luciferase activity and presented as the mean ± SEM of values from at least three repetitive experiments.

Experimental animal models
Male Sprague-Dawley rats were treated as per the National Institute of Health Guidelines for the Care and Use of Laboratory Animals. All experiments were approved and supervised by the Animal Care and Use Committee of Capital Medical University (approval number: SCXK 2016-0006), housed with a 12-h light/12-h dark cycle and allowed free access to normal food and water. Eight-week-old rats received 0.2 mL/kg of body weight CCl 4 (Jin Chemical Pharmaceutical, Seoul, Korea) dissolved in olive oil by intraperitoneal injection, were then injected with 0.1 mL/kg of body weight CCl 4 for 2 weeks and were nally injected with 0.05 mL/kg of body weight CCl 4 for 2 weeks. All injections were administered three times per week (n = 30). As a control (NC group), mice were injected with an equal volume of olive oil (n = 6). Next, we randomly chose 2 CCl 4 groups to inject with 5 10 6 BMSCs (stained with CMSE, meilunbio, China). After that, we chose a CCl 4 group and a CCl 4 + BMSC group to treat with FA (10 mg/kg) and another CCl 4 group to treat with colchicine (0.1 mg/kg) as a positive control. The other groups were treated with equal amounts of normal saline. After treatment for 2 weeks, all mice were sacri ced, and the serum and liver tissues were obtained. The groups were as follows: normal control group (NC), model group (model), positive control group (positive), ferulic acid group (FA), BMSC group (BMSCs), BMSC+ ferulic acid group (FC).

Measurement of aspartate aminotransferase and alanine aminotransferase
We collected 1 mL of blood from each animal. After storage at room temperature for 4 h, the serum was separated by centrifugation (1500× g, 1 min). Then, 200-μL serum samples were taken for analysis. The activities of serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were determined using commercial kits obtained from Konkahongyuan Co. (Beijing, China) according to the manufacturer's instructions.

Histopathological evaluation
Fresh liver tissue samples were xed in 10% formaldehyde at 4°C for 12 h and embedded in para n wax for histological evaluation. Sections were stained with haematoxylin and eosin (HE), and the severity of histological changes was evaluated. We obtained images by using a microscope equipped with a LI-COR Odyssey Infrared Imaging System (LI-COR, Inc., Lincoln, NE, USA).

Masson trichrome staining
For staining, 4 to 6 um-thick sections were cut from para n-embedded blocks and stained with Masson trichrome (MT) and FANCM. For Masson trichrome staining, the Trichrome III Green Staining Kit, a modi ed version of the Masson trichrome stain, was used with Bouin's Solution application to intensify the nal coloration.

BMSCs tracking in vitro
Liver tissue (selected from BMSCs group) was soaked into 4% PFA in 4°C for 24 h, then liver was soaked into 30% sucrose solution for 48h. After washed with PBS for 3 times, the tissue was frozen into liquid nitrogen and sliced into 4 um sections. Sections were observed and captured using confocal laser scanning microscope (Zeiss SP8-STEAD, Germany).

Tissue Clearing
Liver tissue (selected from BMSCs group) was soaked into 4% PFA in 4°C for 24 h, then washed with PBS for 2 h at RT for 3 times. After tissue xation, the tissue was soaked into 50% Cubic-L (TCI, Tokyo Japan) for 24 h, then soaked into Cubic-L for 48h at 37°C. After washed with PBS for 3 times, the tissue was soaked into 50% Cubic-R+ (TCI, Tokyo, Japan) for 24 h and then soaked into Cubic-R+ for 48 h. The tissue was observed with lightsheet microscopy (Abberior instruments, USA).

Quantitative real-time PCR analysis
We extracted total RNA from frozen liver tissues using a Trizol reagent kit (Tiangen Biotech, Beijing, China) and then reverse transcribed the isolated RNA into cDNA using a reverse transcription kit (Tiangen Biotech, Beijing, China). To assay target gene expression, we used quantitative polymerase chain reaction (qRT-PCR) with a SYBR Green Taq kit (KAPA Biotechnology) on a 7500HT fast real-time PCR system (ABI, Foster City, CA, USA). The primers were designed using Primer 5 ( Table 1). The PCR conditions were as follows: (1) 95°C for 30 s and (2) 40 cycles of 95°C for 20 s, 60°C for 20 s, and 72°C for 20 s. The experiment was repeated three times. Table 1 Primer sequences Primer name Sequences 5'-3'

Cytoskeleton staining
HSCs were treated with TGFβ1 and then co-cultured with BMSCs. After the co-culture period, the cells were washed twice with PBS, xed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100 in PBS for 20 min at room temperature. Then, the cells were stained with phalloidin (1:200) for 30 min. After phalloidin staining, the cell nuclei were stained with DAPI (1:100) for 10 min. Cells were observed and captured using confocal laser scanning microscope (Zeiss SP8, Germany).

Western blot analysis
Cells and liver tissues were lysed in RIPA buffer (Lablead, Beijing, China), and total protein was extracted. The protein concentration was measured using a BCA protein assay kit (Beyotime, Nanjing, China). Protein probing by Western blot analysis was performed routinely; membranes were incubated with primary

Immunohistochemistry analysis
All tissue samples were routinely xed in 4% formalin and embedded in para n. The para n sections were dewaxed in xylol and rehydrated in a series of ethanol solutions (95%, 80%, and 70%). Antigen retrieval was performed in a microwave oven with citrate buffer at pH 6.0 for 20 min and then cooled for 20 min. Endogenous peroxidases were blocked using solutions provided with the Reveal Kit (Biogen Cambridge, MA). After this procedure, we used 10% normal goat serum to block the sections for 1 h at 37°C. The blocked sections were then incubated with the primary antibody overnight at 4 ℃. After incubation with primary antibodies against α-SMA (1:500), the sections were incubated with Reveal buffer. After 10 min, the sections were incubated with the secondary antibody for 1 h. The reaction was stained with diaminobenzidine (DAB) and counterstained with hematoxylin. The sections were then hydrated in distilled water and dehydrated in a series of alcohol solutions and xylol. The slides were then covered with coverslips with the aid of Entellan (Merck, Germany). The experiment was repeated three times.

Statistical analysis
Statistical analysis results are expressed as the mean ± SEM. Statistical differences were analyzed by one-way and two-way ANOVA (SPSS 20.0, Chicago, IL, USA). P values <0.05 were considered statistically signi cant.

Results
FA promoted the ability of BMSCs to decrease serum AST and ALT levels in liver brosis models As shown in Figure 1a, we found that the liver was dark red, the liver surface was rough, and adhesions appeared between the liver lobes in the CCl 4 -treated group compared to the normal group. In contrast, with BMSC and FA treatment, the liver changed substantially. The color of the liver was returned to light red, the liver surface was smooth, and adhesions were reduced in the BMSC+FA group compared to the model group. Rat weight also showed differences among the groups. Compared to that of normal control rats, the weight of model rats was signi cantly decreased (p<0.001). Weight was signi cantly increased with FA and BMSC treatment (p<0.001, vs. the model group) (Figure 1b). The serum ALT and AST levels in model rats were signi cantly increased (p<0.05, vs. control group), indicating successful establishment of the liver brosis rat model. FA, BMSC or FA&BMSC administration signi cantly decreased serum AST and ALT levels (p< 0.05, vs. the model group) (Figure 1c-d).

Histopathological Assessment Of Liver Damage
To further con rm the anti brotic effects of BMSCs, FA and their combination in an in vivo rat liver brosis model, we performed HE and Masson staining. As shown in Figure 2a and 2b, in ammatory cell in ltration, fragmented hepatic nuclei, and collagenous bre formation were observed in the model group.
Damage caused by CCl 4 was signi cantly attenuated by BMSCs, FA, and their combination. In addition, the combination treatment achieved better reductions in collagenous bres compare to FA and BMSCs single group.

Tracking Bmscs In Vivo
BMSCs were pre-labelled with CMSE (green uorescence, 492 nm) before injection. To identify successful BMSC injection and BMSC localization, we extracted liver tissue for cryo-sectioning and tissue clearing. As shown in Figure 3a, liver sections contained clearly visible CMSE-labelled BMSCs throughout. As shown in Figure 3b, the CMSE-labelled BMSCs were mostly located around intrahepatic vessels.
FA promotes the ability of BMSCs to improve liver brosis via the RhoA/ROCK pathway We identi ed that FA could promote the ability of BMSCs to reduce liver brosis via the RhoA/ROCK pathway. Compared with the normal group, the protein expression of the model group showed signi cant increases in the pro brotic markers α-SMA and COL1-A1, while in the FA and BMSC-treated group, their expression was decreased. In addition, FA combined with BMSCs showed better effects than the other two treatments. The expression patterns of RhoA and ROCK showed the same trends (Figure 4a-d).
These results clearly indicated that BMSCs combined with FA could be a novel treatment for liver brosis (Figure 4e).

FA promotes the ability of BMSCs to inhibit cytoskeletal rearrangement in HSCs
Whether FA promotes the ability of BMSCs to inhibit cytoskeletal rearrangement in HSCs was investigated. Co-cultured HSCs were evaluated by uorescence and confocal microscopy and demonstrated successful uptake. Compared to those in the normal group, activated HSCs exhibited F-Actin remodelling to form a large number of thick stress bres across the whole cell. In the FA-treated group, F-Actin was mainly distributed in peripheral cells, and a few ne stress bres were found in these cells ( Figure 5). Western blot analysis con rmed that FA could promote BMSC-mediated inhibition of HSC activation, as indicated by assessment of HSC activation markers, including a-smooth muscle actin (a-SMA) and collagen type 1 a 1 chain (COL1-A1). The protein levels of a-SMA and COL1-A1 in the model group were signi cantly increased (p<0.05 vs. the NC group). However, the expression of a-SMA and COL1-A1 was signi cantly decreased after FA treatment, and 1 mg/mL FA showed better effects (p<0.05 vs. the model group). We also detected a-SMA and COL1-A1 in extracellular medium by elisa kits and found that the content of them was decreased (Figure 6c-d). After con rming the effect of FA on HSC inactivation, we further assessed the direct pathway between BMSCs and HSCs. The cytoskeleton changed when HSCs were activated. We chose the RhoA/ROCK/SRF and RhoA/ROCK/LIMK1 pathways. The protein levels of RhoA, ROCK, SRF, LIMK1 were signi cantly increased in the model group compared to the NC group (p<0.05), and with 1 mg/mL FA treatment, the expression of RhoA, ROCK, SRF, LIMK1 was signi cantly decreased. The other two dose groups showed no differences. The content of COL1-A1 and a-SMA in the ECM also decreased with 1 mg/mL FA treatment (Figure 6f-g). The mRNA levels showed the same trends (Figure 6e).

BMSCs regulate the activation of HSCs by inhibiting cytoskeletal rearrangement
To further prove that the cytoskeleton pathway acts as the direct interaction between BMSCs and HSCs that allows BMSCs to inhibit HSC activation, we added a cytoskeleton inhibitor (cytochalasin D) or cytoskeleton agonist (angiotensin II) when BMSCs were co-cultured with activated HSCs. Compared to the model group, the agonist group showed signi cantly increased protein expression of COL1-A1, a-SMA, RhoA, ROCK, SRF, and LIMK1 (p<0.05), while the expression levels of these proteins were signi cantly decreased in the inhibitor group (p<0.05). The content of COL1-A1 and a-SMA in the ECM was also decreased with cytochalasin D treatment (Figure 7a-b). The content of COL1-A1 and a-SMA in the ECM also decreased with 1 mg/mL FA treatment (Figure 7d-e). The RNA levels of COL1-A1, a-SMA, RhoA, ROCK, SRF, and LIMK1 were reduced in the cytochalasin D groups (Figure 7c). These results indicate that FA promotes the ability of BMSCs to in uence HSC activation. The cytoskeleton-related RhoA/ROCK pathway acts as the direct interaction between BMSCs and HSCs.
miR-19b-3p derived from BMSCs inactivates HSCs by suppressing the expression of its target TGF-βR2 miRNAs are found in BMSCs, and miRNAs transferred into target cells can impact cell behaviors by regulating the expression of target genes. Because miRNAs derived from BMSCs in uenced HSC activation, we assessed which miRNAs contained in BMSCs regulated HSC activation. In previous studies, 20 miRNAs were found to be relatively highly expressed in BMSCs. Among them, miR-19b was found to be potentially linked to brosis. Bioinformatic analysis using TargetScan and mirwalk predicted TGF-βR2, a receptor for TGF-β1, as a putative target of miR-19b-3p, and a luciferase reporter assay revealed that miR-19b-3p directly bound to the 3' untranslated region (UTR) of the TGF-βR2 mRNA transcript (Figure 8a-b).
To examine whether miR-19b-3p impacted HSC activation, HSCs were transfected with a miR-19b-3p mimic, meaningless fragment (negative control). Additionally, we chose inactivated HSCs and activated HSCs as the normal and model controls, respectively. Although miR-19b-3p was rarely expressed in activated HSCs, its expression was greatly elevated in miR-19b-3p-transfected HSCs (Figure 8c). In addition, the pro brotic markers a-SMA, COL1-A1, RhoA, and ROCK1 were downregulated in HSCs transfected with the miR-19b-3p mimics compared with those in the model and inhibitor groups (p<0.05) (Figure 8de). The content of a-SMA and COL1-A1 in the ECM was also showed decreased trend in the mimic group (Figure 8f-g). Western blot assays con rmed the RNA data, showing reductions in TGF-βR2 and pro brotic marker expression in HSCs transfected with miR-19b-3p. These results suggest that miR-19b-3p inhibits HSC activation by directly targeting TF-βR2.
We also transfected the miR-19b-3p mimic, miR-19b-3p inhibitor or negative control with the uorescent Cy3 gene into activated HSCs. With confocal scanning microscopy, we observed red uorescence in the miR-19b-3p mimic and negative control groups. Compared to the model group, the group transfected with the miR-19b-3p mimics showed F-Actin mainly distributed in peripheral cells, and a few ne stress bres were found in these cells. The cytoskeleton of cells in the negative control groups was similar to that of cells in the model group (Figure 8h).

FA induces BMSCs to release more miR-19b-3p to HSCs and inhibit the activation of HSCs
To determine whether BMSCs transfer miR-19b-3p to HSCs, we transfected the miR-19b-3p mimic or miR-19b-3p inhibitor with the uorescent Cy3 gene into BMSCs and co-cultured the transfected cells with activated HSCs. With confocal scanning microscopy, we observed red uorescence in the miR-19b-3p mimic group, and F-Actin was mainly distributed in peripheral cells. A few ne stress bres were found in these cells compared to those in the model group. In addition, in the miR-19b-3p inhibitor group, the cell shape was similar to that in the model group, and F-Actin in activated HSCs was remodelled to form a large number of thick stress bres across the whole cell. While with FA treated, we found more red uorescence in the miR-19b-3p mimic groups (Figure 9a).
To assess protein levels, the BMSC/HSC co-culture system was divided into 8 groups: the normal, model, mimics, inhibitor, mimics+FA, inhibitor+FA, and negative groups. We detected a-SMA, COL1-A1, RhoA, ROCK1 SRF, and LIMK1 and found that their expression was signi cantly downregulated in the mimic, mimic+FA, and inhibitor+FA groups compared to the model group. The content of a-SMA and COL1-A1 in the ECM was also decreased in these 3 groups (Figure 9b-c). The content of a-SMA and COL1-A1 in the ECM was also showed decreased trend in the mimic&FA group (Figure 9d-e). The mRNA levels also showed the same trends (Figure 9f).
These results demonstrate that miR-19b-3p plays a critical role in mediating the effect of BMSCs on liver brosis. Taken together, these results suggest that FA combined with BMSC treatment can induce greater release of miR-19b-3p, which inhibits the activation of HSCs by suppressing the RhoA/ROCK1/SRF and RhoA/ROCK1/LIMK1 pathways, contributing to alleviating liver brosis.

Discussion
In this study, we identi ed a new mechanism by which FA and BMSCs treat liver brosis. We used BMSCs, FA or combination therapy to treat rats with hepatic brosis. By evaluation of AST, ALT, HE and Masson staining results, we observed that liver brosis was decreased to different degrees and that the combination therapy showed better effects than FA or BMSC monotherapy. In addition, we detected the protein expression of COL1-A1, α-SMA, ROCK1 and RhoA, and the results for the combination therapy also showed better effects. Therefore, these results suggest that FA can promote the ability of BMSCs to improve liver brosis by targeting the RhoA/ROCK pathway.
The RhoA/ROCK signaling pathway is ubiquitously related to intracellular cytosis, colonization, migration, apoptosis, gene expression and other behaviors [12,13] . The RhoA/ROCK pathway also affects the formation of brosis. It promotes collagen and ECM production in broblasts and increases the transformation of broblasts into myo broblasts expressing α-SMA and other activated factors [14,15] .
To verify the mechanism by which FA promotes BMSCs, we established a BMSC/HSC-T6 co-culture system. Activated HSCs were co-cultured with an equal amount of BMSCs. We observed the cellular structure of HSC-T6 cells and found that with FA treatment, the deformed cell structure was restored, F-Actin was mainly distributed in peripheral cells, and a few ne stress bres were found in these cells. In addition to assessing cellular structure, we detected the migration of HSCs. The number of migration cells was signi cantly decreased with FA and BMSC treatment. The protein expression of RhoA, ROCK, SRF, LIMK1, COL1-A1 and α-SMA was signi cantly decreased. These results all proved that FA promoted the ability of BMSCs to decrease HSC migration, meanwhile, inhibit activation.
The RhoA/ROCK pathway is related to both brosis development and cytoskeletal rearrangement. RhoA/ROCK/SRF directly in uences collagen and α-SMA production, while RhoA/ROCK/LIMK1 in uences cytoskeletal rearrangement, indirectly affecting HSC migration. Activated HSCs migrate to sites of damage and release collagen. Therefore, inhibiting cytoskeletal rearrangement is also an effective strategy for reducing brosis [16][17][18] .
To verify the mechanism by which BMSCs inhibit HSC activation, we established a BMSC/HSC co-culture system with angiotensin II and cytochalasin D used as an agonist and inhibitor, respectively. The decreased protein expression of RhoA, ROCK, SRF, LIMK1, COL1-A1 and α-SMA also indicated that BMSCs could reduce HSC activation by inhibiting cytoskeletal rearrangement.
In a recent study, BMSCs were found to release miRNAs, targeting speci c proteins to reduce brosis [19,20] . The results for the BMSC genomic sequence showed that BMSCs contain a variety of miRNAs [21,22] . Among them, miR-19b-3p was shown to be related to inhibiting brosis [23][24][25][26] . We used the TargetScan website to predict the target gene of miR-19b-3p and found TGF-βR2 to be a putative target. The luciferase reporter assay revealed that miR-19b-3p directly bound to TGF-βR2 in HSCs.
To identify whether miR-19b-3p can inhibit HSC activation, we transfected miR-19b-3p mimics or inhibitors with the Cy3 gene into HSCs and BMSCs. We observed the cytoskeleton of HSCs transfected with the miR-19b-3p mimics or inhibitors. With miR-19b-3p mimic transfection, the cytoskeleton of activated HSCs was not changed considerably, while with miR-19b-3p inhibitor transfection, the cytoskeleton of activated HSCs was still rearranged. The decreased expression of RhoA, ROCK, SRF, LIMK1, COL1-A1 and α-SMA showed that miR-19b-3p could affect HSC activation.
We also observed that BMSCs transfected with the mimics released more miR-19b-3p to HSCs and that the cytoskeleton was similar to that of inactivated HSCs; BMSCs transfected with the inhibitors released less miR-19b-3p to HSCs, and the cytoskeleton was similar to that of activated HSCs. The protein expression of RhoA, ROCK, SRF, LIMK1, COL1-A1 and α-SMA showed that BMSCs could release miR-19b-3p to inhibit HSC activation, and BMSCs combined with FA treatment showed better HSC inhibition. These results support the notion that combining BMSCs with FA could be a promising therapeutic strategy for the treatment of liver disease.
Bone marrow has been the most widely used source of MSCs. However, the use of these cells in clinical applications has been limited due to the instability of BMSCs [27][28][29] . FA is a derivative of cinnamic acid and has therapeutic activity against a variety of diseases. It is also an effective ingredient in some traditional Chinese medicines, such as Angelica Sinensis. In addition, FA may protect against brotic deterioration in rats with hepatic brosis [30] . Therefore, we chose FA combined with BMSCs to treat hepatic brosis in rats, and found that their combination showed better effects on treatment.

Conclusion
Collectively, we demonstrated that FA promoted the ability of BMSCs to suppress HSC activation and liver brosis. MiR-19b-3p derived from BMSCs inactivated HSCs by suppressing the RhoA/ROCK/SRF and RhoA/ROCK/LIMK1 signaling pathways. Therefore, our ndings suggest that BMSCs combined with a traditional Chinese medicine have great potential as an effective anti brotic therapeutic agent for treating chronic liver disease.           A dual-luciferase assay was performed to verify binding interaction between miR-19b-3p and TGF-βR2 mRNA. Cells co-transfected with pmir-GLO basic vector containing either wild-type (WT) or mutant (mut) target sites plus either the miR-19b-3p mimic or scrambled (Scr-) miR (control); (c) qRT-PCR analysis for a-SMA, COL1-A1, ROCK1, RhoA in HSCs. Data are presented as mean ± SEM of experiments performed in triplicate (n=3 independent experiments, two-way ANOVA; *p < 0.05 versus normal; #p<0.05 versus model); (d-e) Western blot bands and cumulative densitometric analyses of each group (Normal, Model, mimic,NC-d). Data are presented as the mean ± SEM (n=3 independent experiments, two-way ANOVA; *p < 0.05, versus normal; #p<0.05, versus model); (f-g) The protein content of α-SMA and COL1-A1 in extracellular medium. Data are presented as mean ± SEM of experiments performed in triplicate (n=3 independent experiments, two-way ANOVA; *p < 0.05 versus normal; #p<0.05 versus model); (h) Representative images of the cytoskeleton (green) in HSCs treated with miR19b-3p mimics or NC-d. DAPI was used to stain cell nuclei (original magni cation, 40×; scale bars: 25 um).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. WBbands.rar