Reversine attenuates cholestatic ductular reaction in rats

Ductular reaction (DR) is usually observed in biliary disorders or various liver disorders, including nonalcoholic fatty liver disease. Few studies have focused on interrupting the DR process in the cholestatic environment. Here, we investigated the impact of reversine on DR in rats that had undergone bile duct ligation (BDL). Cholestatic injury was induced in rats 2 weeks following BDL. DR was assessed with biliary markers by immunohistochemistry. Biliary epithelial cells (BECs) were isolated for the analysis of proliferation and biliary factor gene expression. The effects of reversine on DR and fibrosis were analyzed in vivo via intraperitoneal injection in rats for 2 weeks. Chemically‐induced BEC formation was used to investigate the biliary markers affected by reversine in vitro. DR with increased BEC expansion was identified in cholestatic liver injury, as indicated by CK7, CK19, and EpCAM expression around the portal vein in BDL rats. BDL‐induced DR cells showed the increased expression of genes regulating cell proliferation (Ki67, Foxm1, and Pcna) and biliary markers (Krt7, Krt19, Epcam, Sox9, Cftr, and Asbt). Reversine attenuated cholestatic fibrosis and DR in rats. Reversine affected chemically‐induced BEC formation, with the decreased expression of biliary Krt7, Cftr, and Ggt1 genes in vitro. BDL‐induced Notch activation was attenuated upon reversine treatment in vivo, in part via the Notch/Sox9 pathway. In conclusion, reversine attenuated cholestatic ductular reaction and fibrosis in rats and reduced the bile duct formation associated with Dlk1/Notch/Sox9 signaling. Reversine may be regarded as a potential drug for cholangiopathies for preventing a ductular reaction.

Ductular reaction (DR) is usually observed in biliary disorders or various liver disorders, including nonalcoholic fatty liver disease. Few studies have focused on interrupting the DR process in the cholestatic environment. Here, we investigated the impact of reversine on DR in rats that had undergone bile duct ligation (BDL). Cholestatic injury was induced in rats 2 weeks following BDL. DR was assessed with biliary markers by immunohistochemistry. Biliary epithelial cells (BECs) were isolated for the analysis of proliferation and biliary factor gene expression. The effects of reversine on DR and fibrosis were analyzed in vivo via intraperitoneal injection in rats for 2 weeks. Chemically-induced BEC formation was used to investigate the biliary markers affected by reversine in vitro. DR with increased BEC expansion was identified in cholestatic liver injury, as indicated by CK7, CK19, and EpCAM expression around the portal vein in BDL rats. BDL-induced DR cells showed the increased expression of genes regulating cell proliferation (Ki67, Foxm1, and Pcna) and biliary markers (Krt7, Krt19, Epcam, Sox9, Cftr, and Asbt). Reversine attenuated cholestatic fibrosis and DR in rats. Reversine affected chemically-induced BEC formation, with the decreased expression of biliary Krt7, Cftr, and Ggt1 genes in vitro. BDL-induced Notch activation was attenuated upon reversine treatment in vivo, in part via the Notch/Sox9 pathway. In conclusion, reversine attenuated cholestatic ductular reaction and fibrosis in rats and reduced the bile duct formation associated with Dlk1/Notch/Sox9 signaling. Reversine may be regarded as a potential drug for cholangiopathies for preventing a ductular reaction. Ductular reaction (DR) refers to a process of expansionactivated biliary epithelial cells (BECs) around the portal vein upon injury, in which a transient luminal epithelium is formed, establishing an auxiliary biliary system [1]. Ductular reaction is usually observed in biliary disorders such as primary biliary cholangitis [2], primary sclerosing cholangitis [3,4], and biliary atresia [5], or in various liver disorders, including nonalcoholic fatty liver disease [6,7]. Previous studies showed that expanded BECs during DR transdifferentiated into hepatocytes to support liver regeneration [1,8]. However, BECs also promote inflammation and peribiliary fibrosis, and the overt proliferation of malignant-transformed BECs has been found to result in cholangiocarcinoma formation [9,10]. Few studies have focused on interrupting the DR process in the cholestatic environment. Common bile duct ligation (BDL) causes cholestasis; and a BDL model, therefore, has been used widely to study cholestatic liver injury [11] Abbreviations ALP, alkaline phosphatase; BDL, bile duct ligation; BECs, biliary epithelial cells; CLiPs, chemically-induced liver progenitor cells; DR, ductular reaction; EpCAM, epithelial cell adhesion molecule; HSCs, hepatic stellate cells; T-Bil, total bilirubin; c-GTP, gamma-glutamyl transferase. and fibrogenesis [12]. However, the use of a BDL model for studying DR has seldom been reported.
Reversine, a synthetic 2,6-disubstituted purine analog, reportedly neutralized liver fibrosis by inducing hepatic stellate cell (HSC) apoptosis, restrained cell proliferation, reduced HSC activation, and degraded the extracellular matrix in vitro [13]. Here, we hypothesized that this small molecular compound, reversine, could also attenuate BDL-induced cholestatic injury by affecting the DR process in vivo. Herein, in the present study, we investigated the characteristics of DR caused by BDL and studied the impact of reversine on the BDLinduced DR process. By applying our previously reported method for BEC conversion for bile duct formation [14] from chemically-induced liver progenitor cells (CLiPs) [15], we explored the impact of reversine on chemically-induced BEC formation in vitro. The results showed that reversine attenuated cholestatic DR in rats and reduced bile duct formation in vitro, in part by downregulating Dlk1/Notch/Sox9 signaling. Reversine may be regarded as a potential drug for cholangiopathies for halting the DR.

Animals
Male Lewis rats aged 7 weeks, weighing between 220 AE 20 g, were purchased from Charles River Company (Beijing, China), and all rats were housed in smooth-bottomed plastic cages in a pathogen-free animal room at a controlled temperature (22 AE 2°C), humidity (50 AE 10%), and light (12 h light-dark cycle) with free access to rodent chow and water. To accustom the animals to the laboratory environment, an acclimation period of 1 week was allowed before the initiation of the experiment. This study was approved by the Institutional Animal Care and Use Committee of Guangzhou First People's Hospital, the Second Affiliated Hospital of South China University of Technology (No: k2022047-01).

BDL model
Double ligation of the common bile duct in rats was used as the BDL model [11]. Briefly, the rats were anesthetized with inhaled isoflurane. After an upper midline laparotomy (6 cm), the common bile duct was exposed and ligated twice with non-absorbable 6-0 monofilament sutures. The abdomen was closed with 5-0 braided silk sutures in layers. After 2 weeks, the rats fasted overnight and were sacrificed. Blood samples were collected via the inferior vena cava and centrifuged at 1600 g for 10 min at 4°C, and the collected plasma was stored at À80°C until assayed. Liver tissue samples of the right lobe were collected at 14 days and fixed with a 4% paraformaldehyde phosphate-buffered solution for histological analysis.

Reversine administration
Reversine (CAS NO. 656820-32-5) was purchased from Sigma-Aldrich (Shanghai, China) and dissolved in dimethyl sulfoxide according to the manufacturer's instructions. Reversine was intraperitoneally injected at a dose of 200 lgÁkg À1 body weight at the time of BDL. Then, rats were injected every 3 days for 2 weeks. BDL rats without reversine administration were used as a positive control. The Sham rats served as a normal control group. The rats fasted overnight and then were sacrificed. Blood samples were collected via the inferior vena cava and centrifuged at 1600 g for 10 min at 4°C, and the collected plasma was stored at À80°C until assayed. Liver tissue samples of the right lobe were collected at 14 days and fixed with a 4% paraformaldehyde phosphate-buffered solution for histological analysis.

Blood assays
Levels of alkaline phosphatase (ALP), total bilirubin (T-Bil), and gamma-glutamyl transferase (c-GTP) in the plasma were measured by the laboratory of the Guangzhou First People's Hospital.

Histopathology
For histological review, the fixed liver tissues were collected and embedded in paraffin. Sections (4 lm) were cut, deparaffinized, and stained with hematoxylin and eosin (HE) and aniline blue (Azan) staining in accordance with standard techniques.

Isolation of hepatocytes and biliary cells
Hepatocytes were isolated using a modified two-step collagenase perfusion [16]. After perfusion with a Ca 2+ -free Hank's/ ethylene glycol tetraacetic acid (EGTA) solution through the portal vein, the liver was perfused with 130 mL of Hank's solution containing 130 unitÁmL À1 collagenase at 20-30 mLÁmin À1 . The liver was extracted and mechanically minced with a surgical knife. The minced liver was then filtered twice with a four-layer cotton mesh and 45-lm stainless steel mesh. The suspension was then purified three times in high-glucose Dulbecco's modified Eagle medium (DMEM) by centrifugation at 50 9 g for 2 min at 4°C. The cells were resuspended in a 40% Percoll solution (GE Healthcare, Shanghai, China), and the dead cells were removed via centrifugation at 50 9 g for 20 min. All experiments were conducted using purified hepatocytes of at least 90% viability determined using trypan blue. Hepatocytes were plated on collagen-coated dishes at a density of 6 9 10 4 cellsÁcm À2 , cultured in DMEM/F12 medium containing 2.4 gÁL À1 NaHCO 3 and L-glutamine, and supplemented with 5 mM HEPES, 30 lgÁmL À1 L-proline, 0.5 mgÁmL À1 bovine serum albumin (BSA), 10 ngÁmL À1 epidermal growth factor, insulintransferrin-serine-X, 0.1 lM dexamethasone, 10 mM nicotinamide, 1 mM ascorbic acid-2 phosphate, 100 UÁmL À1 penicillin, and 100 mgÁmL À1 streptomycin.
Biliary cell isolation was conducted using a previously described method with modifications [17]. The filtered residue from the hepatocyte isolation was minced into small fragments by using surgical scissors, suspended in BEC digestion medium in a 50-mL tube, and shaken in a 37°C water bath for 30 min. The digested tissue was then filtered through a 40-lm cell strainer. Gradient density centrifugation was used to separate the BECs from non-parenchymal cell fractions. As reported previously, the best gradient for BEC isolation was about 800 9 g [14,17], and it was 200 9 g for HSCs, 600 9 g for Kupffer cells, and 900 9 g for sinusoidal endothelial cells [18]. The cells were centrifugated three times at 600 9 g for 15 min to wipe off the HSCs and Kupffer cells. Thereafter, cells were centrifuged three times at 800 9 g for 10 min at 4°C, washed, and resuspended in 10 mL of 10% fetal bovine serum (FBS)-DMEM, then centrifuged again at 800 9 g for 5 min at 4°C. The biliary cells were plated on a collagen-I-coated dish containing a BEC culture medium. Compositions of the pre-perfusion buffer, collagenase buffer, isolation buffer, BEC digestion medium, and BEC culture medium are listed in Table S1.

Real-time quantitative polymerase chain reaction (RT-qPCR)
Cell samples were acquired for mRNA extraction by using spin columns in accordance with the manufacturer's instructions (NucleoSpin RNA II; Macherey-Nagel, Duren, Germany). cDNA was synthesized from total RNA by using a high-capacity cDNA reverse transcription kit (Applied Biosystems, Beijing, China). The samples were stored at À30°C until use. PCR was performed on an Applied Biosystems Step One Plus real-time PCR system using a TaqMan Gene Expression Assay kit (Applied Biosystems). Gapdh was used as the internal reference, and mRNA expression levels were determined using the comparative cycle time (DDC t ) method [19]. The TaqMan primers used for PCR are listed in Table S2.

Biliary formation in vitro
Biliary epithelial cells and biliary formation were induced in vitro as previously reported with chemically-induced liver progenitor cells [14,15]. Briefly, 1 day before collecting the CLiPs, commercial mouse embryonic fibroblasts (MEFs; Cat #PMEF-N, Merck Millipore, Beijing, China) were used to form a MEF feeder layer by seeding 5 9 10 4 cells on collagen-coated 12-well plates (1.3 9 10 4 cellsÁcm À2 ) in DMEM containing 10% FBS. Then the dissociated CLiPs were plated onto the MEF feeder layer at a density of 5 9 10 5 cells per well (1.3 9 10 5 cellsÁcm À2 ) in a reprogramming medium supplemented with 5% FBS. Thereafter, the BEC induction medium (BIM) was replaced every 2 days for 6 days, followed by BIM supplemented with 2% growth factor reduced Matrigel (Corning, Beijing, China) for an additional 6 days, to facilitate the maturation of BECs and the formation of the biliary structure. For reversine administration during biliary formation, 0 or 5 lMÁmL À1 of reversine was added into the medium during the last 6 days, and the medium was changed every 2 days.

Statistical analysis
Data were presented as mean AE standard deviation. Statistical analyses were conducted with GRAPHPAD PRISM (GraphPad Software, Inc., San Diego, CA, USA) using a two-sided Student's t test or ANOVA. A probability (P) value < 0.05 was considered statistically significant.

Cholestatic liver-induced ductular reaction in BDL rats
Cholestatic injury was induced after 2 weeks in rats following BDL (A). BDL increased the markers of cholestasis in serum (Fig. S1A), such as T-Bil, ALP, and c-GTP. The injured liver was observed with HE staining (Fig. S1B). The BDL liver showed prominent lobular and portal changes, including periportal and parenchymal fibrillar collagen deposition as shown on Azan staining (Fig. S2A) and inflammatory cell infiltration as evidenced by increasing Tgfb1 and Tnfa expression (Fig. S2B).
To measure the DR induced by BDL, the biliary markers, CK7, CK19, and EpCAM, were assessed in the liver tissues. Data showed that BDL injury increased the positive cells marked with CK7 (Fig. 1B), CK19 (Fig. 1C)., and EpCAM (Fig. 1D), around the portal vein, demonstrating the activated BEC expansion. Therefore, 2 weeks of cholestatic injury caused by BDLinduced DR in rats, characterized as the expansion of activated BECs upon injury.

BDL-induced DR cells showed the increased expression of genes regulating proliferation and biliary formation
To characterize the genetic expression of the BDLinduced DR cells, we isolated activated BECs from the BDL rats (Figs 1A and 2A). The isolated cells were positive to a-tubulin and CK7, showing the biliaryspecific expression of those cells ( Fig. 2A). Compared to the sham rats, the BDL-induced cells showed the increased expression of Ki67, Foxm1, and Pcna genes (Fig. 2B), indicating the increased proliferation-specific gene expression of the BDL-induced DR cells. As shown in Fig. 2C, the BDL-induced DR cells showed increased biliary formation gene expression, such as Krt7, Krt19, Epcam, Sox9, Cftr, and Asbt. These data indicated that the highly proliferative DR cells from BDL could be characterized as being specific to biliary formation.

Reversine attenuated cholestatic liver fibrosis and ductular reaction in rats
We then studied the impact of reversine on liver fibrosis and the DR in BDL-induced cholestatic rats. BDL rats given reversine treatment (BDL + Rev) were compared to BDL rats not given reversine treatment (BDL). Serum biochemistry showed the significantly downregulated expression of T-Bil and c-GTP in BDL + Rev (Fig. S1A); Data showed a moderate, but not statistically significant, downregulation of ALP levels upon reversine treatment (Fig. S1A). Histologically, the administration of reversine attenuated liver injury, as shown by HE staining (Fig. 3A), and decreased the fibrillar collagen deposition, as shown by Azan staining (Fig. 3A). We also assessed two markers of liver fibrosis, a-SMA, and Desmin, via immunostaining in liver tissues (Fig. 3B). The quantitative data showed that reversine significantly decreased the positive areas of a-SMA, and Desmin staining (Fig. 3B).
Because the severity of BDL injury increased BEC expansion marked with CK7-, CK19-, and EpCAMpositive cells (Fig. 1B), we then compared those markers across the sample groups given and not given reversine treatment. The immunostaining and quantitative data showed that levels of the biliary protein markers of CK7 (Fig. 4A), CK19 (Fig. 4B), and EpCAM (Fig. 4C) were significantly decreased upon administration of reversine, demonstrating an attenuated BDL-induced DR by reversine.

Reversine affected BEC formation in vitro
Previously, we reported a method for BEC formation from CLiPs (Fig. S3). Here, to investigate the impact of reversine on BEC formation, we applied this in vitro method and administered 0 or 5 lMÁmL À1 of reversine during BEC differentiation and bile duct (BD) formation, compared the cells without biliary formation treatment (termed as Ctrl in figures). As shown in Fig. 5A, the Ctrl cells showed no BEC differentiation or ductulus or biliary formation. Cells that received biliary induction showed several ductular formations (white arrows). However, the administration of reversine during biliary formation showed decreased ductuli (Fig. 5A), indicating that reversine affected BEC differentiation and BD formation. By analyzing the biliary protein markers of CK7 and CFTR, we found that reversine also repressed the expression of both biliary proteins (Fig. 5B). We also found that reversine decreased the BD gene expression of Krt7, Ggt1, and Cftr (Fig. 5C). Those data clearly established that reversine affected BEC formation in the in vitro model. . Levels of the cell proliferation genes of Ki67, Foxm1, and Pcna were analyzed and compared to those of BECs isolated from BDL and Sham rats using RT-qPCR. Quantitative data showed the increased gene expression of Ki67, Foxm1, and Pcna. (C). Biliary marker genes of Krt7, Krt19, Epcam, Sox9, Asbt, and Cftr were analyzed and compared to BECs isolated from BDL and Sham rats using RT-qPCR. Quantitative data showed the increased gene expression of Krt7, Krt19, Epcam, Sox9, Asbt, and Cftr.Gapdh was used as an internal reference, n = 4; Data were presented as scatter plot with mean AE SD and were compared using the Student's t test, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Reversine affected notch signaling
Previous reports have indicated the importance of Notch signaling in biliary differentiation [20,21]. We analyzed the gene expression of the known Notch ligands (Jag1 and Dlk1), the receptor (Notch1, Notch2, Notch3, and Notch4), and the target genes (Hes1, Hes5, and Hey1). Gene samples were acquired from cells given 0 or 5 lMÁmL À1 reversine during BEC differentiation and bile duct (BD) formation. Compared to the samples not given reversine treatment (i.e., Rev 0 lMÁmL À1 ), data showed the increased gene expression of Notch ligand Dlk1, but not the Jag1, increased levels of Notch receptors Notch1 and Notch2, and decreased expression of Notch3 and Notch4 (Fig. S4). Many of the target genes, such as Hes5, though not Hes1 or Hey1, were showed increased expression upon reversine treatment (Fig. S4). These results established that reversine could affect the Notch signaling pathway in our simulated in vitro environment. We then further evaluated the changes in gene expression in the in vivo setting utilizing the whole liver samples (Fig. 1A). As shown in Fig. 6A, BDL increased the ligand expression of Dlk1 and Jag1. The levels of the receptors of Notch2 and Notch4 significantly increased in BDL liver, and those of Notch1 and Notch3 showed a detectable, but not statistically significant, tendency to increase. The target genes of Hes1 and Hes5 showed no difference in expression between Sham and BDL liver, though the Hey1 gene was significantly downregulated upon BDL injury. Compared with the BDL samples, the reversine treatment decreased the level of the Notch ligand Dlk1, but it did not affect the ligand Jag1 (Fig. 6A). Reversine treatment decreased levels of Notch1, but it did not affect the levels of receptors Notch2, Notch3 or Notch4. However, data showed no significant changes in the expression of the Notch target genes Hes1, Hes5, or Hey1 (Fig. 6A). As per previous reports, Notch partially controls the expression of Sox9, a key player in bile system morphogenesis [21]. We further analyzed the expression of Sox9 as well as Hnf1b, which is a necessary factor in biliary epithelium formation from the onset of the development of the biliary system [22]. The significantly increased levels of Sox9 in BDL liver became attenuated upon reversine treatment The immunostaining images of BEC protein markers CK7 and CFTR (red) in cells with or without reversine administration. The cytoskeleton was stained with F-actin (green), and the nuclei were stained with DAPI (blue). Data showed reversine decreased CK7 and CFTR levels during bile duct formation. Scale bar = 50 lm. n = 3. (C) The gene expression of the biliary markers Krt7, Krt9, and Cftr were compared and analyzed via RT-qPCR. Quantitative data showed reversine decreased the gene levels of Krt7, Krt9, and Cftr during bile duct formation. Gapdh was used as an internal reference. Data were presented as scatter plot with mean AE SD and were compared using ANOVA, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. n = 3. ( Fig. 6B). However, the increases in Hnf1b levels appeared unaffected by reversine in BDL livers (Fig. 6B).
Τhese findings showed that the impact of reversine on livers in vivo is associated with the Dlk1/Notch/Sox9 pathway.

Discussion
Ductular reaction demonstrates histological, cellular, molecular, and tissue diversity, which are believed to cause the diverse outcomes of DR of hepatobiliary regeneration, fibrogenesis, and hepatocarcinogenesis [23]. In the model of BDL here, which caused a cholestatic internal environment, both biliary proliferation and fibrogenesis were present. Thus, DR behaves as a double-edged sword [24]. Indeed, as reported previously, expanded DR cells transdifferentiated into hepatocytes to support liver regeneration [1,8]. Conversely, increased DR also promoted inflammation and peribiliary fibrosis, and overt proliferation of malignant transformed BECs resulted in cholangiocarcinoma formation [9,10]. It should be clear to investigators that focusing on attenuating the DR process may advance our understanding of developing novel therapeutic strategies for curing cholangiopathies.
In this study, we first investigated the details of DR caused by BDL. Data showed that BDL injury increased the cells positive for the markers CK7 and CK19 around the portal vein, demonstrating BEC expansion and the formation of a biliary system. Furthermore, we also found increased EpCAM in the luminal epitheliums forming the biliary system. Previous research demonstrated that EpCAM was highly expressed in hepatic stem cells or progenitor cells, and was absent in mature hepatocytes [25][26][27]. Consistent with these previous studies, EpCAM was associated with the maintenance of the undifferentiated phenotype of embryonic stem cells [28,29]. Therefore, the EpCAM + cell has been regarded as a potential hepatic stem cell residing in the normal and injured liver. From this point of view, in our study, the DR cell was an EpCAM + cell, demonstrating the stem cell properties of the DR cells. This may be one of the reasons for DR cell trans-differentiation into hepatocytes to support liver regeneration under certain conditions. Genetically, we found that isolated BECs showed the increased expression of genes regulating cell proliferation, such as Ki67, Foxm1, and Pcna, and increased biliary markers, such as Krt7, Krt19, Epcam, Sox9, Cftr, and Asbt, indicating cell proliferation and biliary formation in DR cells. However, fibrogenesis also accompanied the DR process. Furthermore, we found that isolated BECs showed increased gene expression levels of the pan-HSC marker Desmin and the Kupffer cell marker CD45 (Fig. S5A,B), but there were no significant changes in the expression of the CD31 gene (Fig. S5C) between the BDL and Sham groups.
Previously, we studied the effects of reversine on the cell cycle, apoptosis, and activation of HSCs in vitro [13] and on liver injury and hepatic fibrosis in vivo. We found that reversine neutralized hepatic fibrosis by inducing HSC apoptosis, restrained cell proliferation, Data were presented as bars with mean AE SD and were compared using ANOVA, *P < 0.05, **P < 0.01, and ****P < 0.0001, ns, no significance.
reduced HSCs activation, and degraded the extracellular matrix [13]. With the hypothesis that reversine attenuated BDL-induced cholestatic injury, we studied the effects of reversine on the BDL-induced DR process. Our results supported this hypothesis by both in vivo and in vitro results. In the in vivo study, serum biochemical data showed restoration of the T-Bil and c-GTP levels. Histologically, reversine significantly decreased the expression of the liver fibrosis markers a-SMA and Desmin, and decreased the expression of biliary protein markers CK7, CK19, and EpCAM, demonstrating reversine's ability to attenuate DR and fibrogenesis. In the in vitro study, we explored the impact of reversine on bile duct formation by applying the previously reported method for bile duct formation [14]. Upon biliary induction, the cells showed several ductular formations, while the administration of reversine showed decreased ductuli during biliary formation, indicating that reversine reduced bile duct formation in vitro.
Notch signaling has been shown to be important in biliary differentiation [20,21,30]. A previous report showed that inhibition of this signaling pathway reduced the differentiation of hepatic progenitor cells into cholangiocytes in biliary atresia [31]. The impairment of Notch signaling was reported to associate with Alagille syndrome, in which individuals do not have enough bile ducts [32]. By analyzing the Notch signal genes in vitro cell samples and in in vivo liver tissue samples, we found that the trends in changes in Notch ligand and receptor levels differed between those two systems. These trends may have been caused by the differences between the in vivo model and the in vitro system. In our in vitro system, we used CLiPs, which are chemically-induced liver progenitor cells [33,34], for bile duct formation [35]. As per our results, BDL injury caused increases in the levels of the ligands Jag1 and Dlk1 in the liver. Jag1, a canonical Notch ligand, functioned as an activator of Notch receptors on adjacent cells. Dlk1, a non-canonical Notch ligand, is a member of the epidermal growth factor-like repeat-containing family of proteins [36]. Increasing evidence indicates that Dlk1 functions as a secreted or transmembrane protein, and interacted with Notch to act as a negative regulator of Notch activation and signaling [37,38]. Dlk1 was also found to be a potentiator of adipogenesis of mesenchymal cells. It functioned as a Notch signaling inhibitor [37]. The simultaneous increases in Jag1 and Dlk1 indicated that the cholestatic environment plays a complicated role in BDL rats. In our results, the levels of the receptors Notch2 and Notch4 significantly increased in BDL livers. Notch1 and Notch3 levels showed a detectable but not statistically significant tendency to increase. These changes in Notch signals may be attributable to BDL-induced DR, which acts as a double-edged sword. With reversine treatment, the levels of the ligand Dlk1 decreased, followed by a decrease in that of Notch1 receptor. Our data, however, showed no statistically significant changes in levels of the Notch target genes of Hes1, Hes5, or Hey1, indicating the involvement of some other target gene during reversine treatment. Previous reports showed Sox9 to be a Notch target regulating morphogenesis in the bile system [21]. Our data indicated that reversine significantly neutralized the elevation of Sox9 expression caused by BDL, indicating the attenuation of biliary formation by reversine, which was regulated in part via Notch/Sox9 signaling in vivo. In other words, reversine may be a potential drug to attenuate biliary formation associated with Dlk1/Notch/ Sox9 signaling. The mode of action of reversine still remains unclear, so it should be further explored in the future.
A major limitation of our in vitro studies relates to the isolation of the primary BECs for gene analysis. Although we used gradient density centrifugation to separate the BECs from non-parenchymal cell fractions, we still observed HSC, Kupffer cell, and endothelial cell populations in our isolated BEC samples. Such contamination may have been avoided through the use of magnetic-activated cell sorting or fluorescence-activated cell sorting. However, HSCs and Kupffer cells were present in both the Sham and BDL groups at the same level, which was acceptable in our setting.
In conclusion, reversine attenuated cholestatic DR and fibrosis in rats and reduced bile duct formation associated with Dlk1/Notch/Sos9 signaling. Our study helps to advance our understanding of BEC biology in BDL-induced DR cells and the development of novel therapeutic strategies to treat cholangiopathies. Reversine may be regarded as a potential drug for cholangiopathies for halting the ductular reaction.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. The serum level of T-Bil, c-GTP, and ALP in the Sham group, the BDL, and the BDL + Rev rats, n = 6 per group. Data were presented as scatter plot with mean AE SD and were compared using ANOVA, *P < 0.05, **P < 0.01, and ***P < 0.001, ns-no significance. (B) Representative images of HE stains in the Sham group, BDL, and BDL + Rev liver tissue. Scale bar = 500 lm. Fig. S2. BDL-induced collagen deposition and inflammatory factors in liver. (A) Images of Azan staining in Sham and BDL rats. Collagen deposition was quantified via positive areas in Azan staining. ****P < 0.0001. Scale bar = 500 lm. n = 6. (B) The gene expression of inflammatory factors of Sox9 and Hnf1b were compared and analyzed via RT-qPCR in Sham and BDL livers. Gapdh served as an internal reference. Data were presented as scatter plot with mean AE SD and were compared using the Student's t test, **P < 0.01 and ****P < 0.0001.  The gene expression levels of Notch signaling ligands (Jag1 and Dlk1), receptors (Notch1, Notch2, Notch3, and Notch4), and the target genes (Hes1, Hes5, and Hey1) were analyzed via RT-qPCR in the in vitro cell samples. Gapdh served as an internal reference. n = 4; Data were presented as bars with mean AE SD and were compared using ANOVA, *P < 0.05, ***P < 0.001, ****P < 0.0001, ns-no significance. Fig. S5. Gene marker of nonparenchymal cells. (A) The gene expression of HSC of Desmin was compared and analyzed via RT-qPCR in BEC samples isolated from the Sham and BDL group. (B) The gene expression of Kupffer cell of CD45 was compared and analyzed via RT-qPCR in BEC samples isolated from the Sham and BDL group. (C) The gene expression of the endothelial cell of CD31 was compared and analyzed via RT-qPCR in BEC samples isolated from the Sham and BDL group. Gapdh served as an internal reference. Data were presented as scatter plot and were compared using the Student's t test, ****P < 0.0001, nsno significance. Table S1. List of key reagents and resources used in Methods. Table S2. List of TaqMan primers used for qRT-PCR.