Hepatic stem cell Numb gene is a potential target of Huang Qi Decoction against cholestatic liver fibrosis

Numb is a negative regulator of Notch signal pathway. Previous study has demonstrated that Notch signal pathway activation is required for hepatic progenitor cell (HPC) differentiating into cholangiocytes in cholestatic liver fibrosis (CLF), and Huang Qi Decoction (HQD) could prevent CLF through inhibition of the Notch signal pathway. However, the role of Numb in HQD against CLF is yet unclear. Thus, CLF rats transplanted into rat bone marrow-derived mesenchymal stem cells with knocked down Numb gene (BMSCNumb-KD) were treated with HQD. Simultaneously, Numb gene knockdown was also performed in WB-F344 cell line and then treated with refined HQD in vitro. In vivo study revealed that liver fibrosis was inhibited by HQD plus BMSCNumb-KD treatment, while Hyp content in liver tissue, the gene and protein expression of α-SMA, gene expression of Col I, TNF-α, and TGF-β1 were increased compared to that in HQD group. Furthermore, Notch signal pathway was inhibited by HQD plus BMSCNumb-KD, while the protein expression of Numb was decreased and RBP-Jκ and Hes1 was increased compared to that in HQD group. In vitro, HQD reduced the differentiation of WB-F344 cells into cholangiocyte phenotype, while this effect was attenuated after Numb-knockdown. This study highlights that the absence of hepatic stem cell Numb gene decreases effect of HQD against CLF, which give rise the conclusion that Numb might be a potential target for HQD against CLF.


Materials and methods
Materials. HQD, which contains 30 g of Radix Astragali [Astragalusmembranceus Fisch. (Bge.), root, Huangqi] and 5 g of Radix Glycyrrhizae (Glycyrrhizauralensis Fisch., root and rhizome, Gancao), was obtained from the Shanghai Huayu Herbs Co. Ltd. (Shanghai, China) and was accredited and prepared by pharmacognosist of Shuguang Hospital affiliated to Shanghai University of TCM and was maintained at − 20 °C. HQD fingerprint was obtained by ultra-high performance liquid chromatography Q-Orbitraphigh resolution mass spectrometry (UHPLC-Q-Orbitrap-HRMS), with the results shown in Supplementary Fig. 1.
DAPT (a prototypical gamma-secretase inhibitor, and inhibiting gamma-secretase can prevent Notch receptor cleavage and thereby block Notch signal transduction) from Gene Operation (INO1001-0010MG, Ann Arbor, MI, USA) was used as a positive control drug. The antibodies included in this study were mouse monoclonal anti-α-smooth muscle actin (α-SMA, Clone 1A4; Sigma-Aldrich, St  Adipogenic and Osteogenic induction. To assess the differentiation potential of the BMSCs, Adipogenic and Osteogenic induction referenced to reported method 15 . In briefly, cells were plated on Petri dishes in 15% FBS/DMEM-L. For induction of adipogenesis, when the cells fused 100%, add the lipogenic induction Asolution (Cyagen Biosciences Inc.containing1% penicillin-streptomycin, 1% glutamine, 0.2% insulin, 0.1% 3-isobutyl-1-methyl xanthine, 0.1% rosiglitazone, and 0.1% dexamethasone), replace it with the lipogenic induction B solution (Cyagen Biosciences Inc. containing 1% penicillin-streptomycin, 1% glutamine, and 0.2% insulin) three days later, and replace it with the lipogenic induction A solution one day later. Repeat this process. After 3 weeks of differentiation, cells were fixed and sliced. Adipogenic differentiation was monitored as red droplets after Oil Red O staining. For induction of osteogenic differentiation, when the cells fused 60%, the osteoblast induction solution (Cyagen Biosciences Inc. containing 1% penicillin-streptomycin, 1% glutamine, 0.2% ascorbiate, 0.2% β-glycerophosphate, and 0.01% dexamethasone) was added, and the fluid was changed every 3 days. After 4 weeks of differentiation, cells were fixed and sliced. The osteogenic-induced culture was analyzed by Alizarin Red staining for visualization of calcium deposition.
Animals and experimental protocol. Male SD rats (160-180 g) were purchased from Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China). Animals were maintained in an environment with a constant temperature and supplied with laboratory chow and water ad libitum. The experimental protocols were approved by the Animal Research Committee at Shanghai University of TCM (No. PZSHUTCM18111607).
Bile duct ligation (BDL) with modifications was performed as previously described 19 . In brief, 42 rats were randomly divided into sham group (n = 6) and model group (n = 36). Model rats were anesthetized with pentobarbital sodium and laparotomy was performed with a sterile technique. The common bile duct and the left and right hepatic ducts were isolated. The left and right hepatic ducts and the hepatic portal and duodenal site of the common bile duct were ligated, respectively, and the abdomen was closed. In sham rats, the surgery was identical, except that the bile duct was not ligated. After the BDL, model rats were randomly divided into BDL (n = 6), HQD (n = 6), DAPT (n = 6), BMSC Numb empty virus (negative control, BMSC NC-KD , n = 6), BMSC Numb knockdown (BMSC Numb-KD , n = 6), and HQD plus BMSC Numb-KD (n = 6) groups, and single injection of 1 × 10 6 BMSC NC-KD cells or BMSC Numb-KD cells into livers of corresponding groups. From the second week after BDL, HQD, HQD plus BMSC Numb-KD and DAPT groups were administrated orally at dosages of 0.935 g/kg HQD or 50 mg/kg DAPT respectively for 3 weeks once per day. Sham and BDL rats were given same volume of physiological saline. At the end of 4 weeks, all rats were euthanized with pentobarbital sodium at a dose of 60 mg/kg, and blood and hepatic tissue samples were obtained.

WB-F344 cell line culture and treatment.
In vitro studies were performed using the WB-F344 cell line, a rat hepatic progenitor cell line 23 . The cells were cultured in Ham's F12 medium (Life Technologies) supplemented with 10% fetal calf serum (Gibco). Chemically induced differentiation was obtained by culturing WB-F344 cells on six-well Permanox Lab-Tek culture slides (Nalge Nunc International, Naperville) at a density of 2 × 10 3 cells/cm 2 , starting 24 h after seeding. Cells were divided into normal group (N), SB group (3.75 mM, Sigma, B5887-1G) 24 , SB plus HQD (800 μg/ml) group , SB plus DAPT (50 μM) group, SB plus LV empty virus (negative control, NC) group, SB plus Numb KD group, and SB plus HQD plus Numb KD group (n = 3, respectively). RNAi was performed as described above, but at a MOI = 50. At 6 h post-transfection, the media was replaced with media containing SB, with or without HQD, while positive control sample media contained SB and DAPT without HQD. Media was changed every 2 days, and the culture time was 7 days. Statistical analysis. All data are presented as a mean ± SD. Statistical analyses were performed using an analysis of variance (ANOVA) for multiple comparisons with SPSS 10.0 software and P < 0.05 was considered statistically significant.

Ethics approval and consent to participate. The experimental protocol was approved by the Animal
Research Committee at Shanghai University of Traditional Chinese Medicine (No. PZSHUTCM18111607). All methods were performed in accordance with relevant guideline and regulations in the manuscript.

Results
Lentivirus-mediated knockdown of the BMSC Numb gene. P3 BMSC morphological characteristics were consistent with long fusiform and whirl-like growth. BMSCs were confirmed by flow cytometry, and showed that CD10 (−), CD14 (−), CD29 (+), CD34 (−), CD45 (−), and CD90 (+) (Fig. 1a). Furthermore, the proliferative capacity was evaluated by examining the cell cycle via flow cytometry, and showed that 72.36% of the BMSCs were in the G1 phase (Fig. 1b). Additionally, the BMSCs showed osteogenic and adipogenic abilities following the differentiation assay, with a large number of calcium deposits following osteogenic induction, and a large number of fat droplets following adipogenic induction (Fig. 1c, www.nature.com/scientificreports/ transfection rate of more than 80% at a MOI = 80, and maintained a normal cell morphology (Fig. 1e). The qRT-PCR result showed that Numb gene expression was significantly decreased in the Numb-KD group compared to that in the negative control (NC) group (P < 0.01), about 33% that of the NC group (Fig. 1f). In addition, the Numb protein expression was significantly decreased in the Numb-KD group compared to that in the NC group (P < 0.01) (Fig. 1g,h).
BMSC Numb-KD transplantation reduces the effect of HQD against CLF. H&E staining showed extensive bile duct proliferation in the BDL and BMSC Numb-KD groups, while bile duct proliferation was markedly reduced in the HQD and HQD plus BMSC Numb-KD groups compared to that in the BDL and BMSC Numb-KD groups, respectively. However, bile duct proliferation was markedly increased in the HQD plus BMSC Numb-KD group compared to that in the HQD group (Fig. 2a).
Sirius red staining revealed that proliferated BECs were surrounded by abundant collagen in the BDL and BMSC Numb-KD groups, while collagen deposition was markedly reduced in the HQD and HQD plus BMSC Numb-KD groups compared to that in the BDL and BMSC Numb-KD groups, respectively. However, collagen deposition levels were higher in the HQD plus BMSC Numb-KD group compared to that in the HQD group (Fig. 2b). Moreover, immunostaining showed that α-SMA (myofibroblast marker) expression was localized in the fibrotic septa, which was consistent with that of the Sirius Red staining (Fig. 2c).
Sirius Red stained area was decreased significantly in HQD, DAPT and BMSC NC-KD groups compared to that in BDL group (P < 0.01). Sirius Red stained area in HQD plus BMSC Numb-KD group was decreased compared to www.nature.com/scientificreports/ that in BMSC Numb-KD group but no significant difference (P > 0.05). However, Sirius Red stained area was significantly higher than that in HQD group (P < 0.01) (Fig. 2d).
As show in Fig. 2e, the Hyp content of liver tissues were significantly increased in the BDL and BMSC Numb-KD groups compared to that in the Sham group (P < 0.01) and the BMSC NC group (P < 0.01), respectively, while it was significantly decreased in the HQD and HQD plus BMSC Numb-KD groups compared to that in the BDL group (P < 0.01) and the BMSC Numb-KD group (P < 0.01), respectively. However, Hyp content was higher in the HQD plus BMSC Numb-KD group compared to that in the HQD group (P < 0.01). Moreover, the mRNA and protein levels of α-SMA were consistent with the changes of Sirius Red standing positive area and Hyp content (Fig. 2f,g).
Additionally, the expression levels of TGF-β1, TNF-α, ColI and ColIV mRNA were significantly increased in the BDL and BMSC Numb-KD groups compared to that in the Sham group (P < 0.01) and BMSC NC group (P < 0.05 or P < 0.01), respectively, while the expression levels were significantly reduced in the HQD group compared to that in the BDL group (P < 0.05 or P < 0.01), and α-SMA and ColI expression was significantly reduced in the HQD plus BMSC Numb-KD group compared to that in the BMSC Numb-KD group (P < 0.05). However, the α-SMA, TGF-β1, TNF-α, and ColI levels were significantly increased in the HQD plus BMSC Numb-KD group compared to that in the HQD group, (P < 0.05 or P < 0.01) (Fig. 2h).

BMSC Numb-KD transplantation reduces HQD inhibition on bile duct proliferation. CK7 and CK19
are regarded as a hallmark of BEC 25 , immunostaining showed that CK7 and CK19 were mainly expressed in BECs in the Sham and BMSC NC group. In BDL and BMSC Numb-KD rats, CK7 and CK19 were strongly expressed in proliferated BECs, while they were clearly decreased in the HQD and HQD plus BMSC Numb-KD groups compared www.nature.com/scientificreports/ with the BDL and BMSC Numb-KD groups, respectively. However, in the HQD plus BMSC Numb-KD group, CK7 and CK19 were increased relative to the HQD group (Fig. 3a,b). It is consistent with immunostaining, as shown in Fig. 3d, the positive areas of CK7 and CK19 were increased significantly in the BDL and BMSC Numb-KD groups compared to that in Sham (P < 0.01) and BMSC NC (P < 0.01) groups, respectively, while the positive area of CK19 was decreased significantly in the HQD and the HQD plus BMSC Numb-KD groups compared to that in the BDL (P < 0.01) and BMSC Numb-KD (P < 0.05) groups, respectively; the expression of CK7 were decreased significantly in the HQD group compared with the BDL (P < 0.01), and CK7 showed a decreasing trend in the HQD plus BMSC Numb-KD group compared to that in the BMSC Numb-KD group, with no significant difference (P > 0.05). However, in the HQD plus BMSC Numb-KD group, the positive areas of CK7 and CK19 were significantly increased compared to that in the HQD group (P < 0.01).
In addition, the protein expression levels of CK7 and CK19 were increased significantly in the BDL and BMSC Numb-KD groups compared to that in the Sham (P < 0.05 or P < 0.01) and BMSC NC (P < 0.05 or P < 0.01) groups, respectively, while the protein expression levels were decreased significantly in the HQD group compared to that in the BDL group (P < 0.01). However, they were increased significantly in the HQD plus BMSC Numb-KD group compared to that the HQD group (P < 0.05) (Fig. 3e,f). Additionally, CK7 and CK19 mRNA expression was consistent with their protein expression (Fig. 3g).
HNF4α, a key regulator of hepatocyte differentiation, which also maintains a mature hepatocyte differentiation phenotype during embryonic development 26,27 , was also examined. Immunostaining showed that in the BDL and BMSC Numb-KD groups, HNF4α expression was clearly lower compared to that in the Sham and BMSC NC groups, respectively, while its expression was increased in the HQD and HQD plus BMSC Numb-KD groups compared to that in the BDL and BMSC Numb-KD groups, respectively. However, HNF4α expression was decreased in the HQD plus BMSC Numb-KD group compared to that in the HQD group in immunostaining (Fig. 3c). Moreover, the HNF4α mRNA level was consistent with the immunostaining changes (Fig. 3g).
To further evaluate BMSC Numb-KD differentiation in liver, co-immunostaining with EGFP/CK19 or EGFP/ CK7 was performed. The result showed that no clearly co-expression was noted in the BMSC NC group, while extensive co-expression was noted in proliferating BECs in the BMSC Numb-KD group, and the above pathological change was not clearly reversed in the HQD plus BMSC Numb-KD group (Fig. 3h,i). These results suggest that BMSCs differentiate into BECs after Numb knockdown in liver, and that HQD has no obvious intervention effect on this pathological process. Fig. 4a-c, immunofluorescence staining clearly indicated that Numb was localized in hepatocytes, while RBP-Jκ and Hes1 were associated with proliferating BECs. In the BDL and BMSC Numb-KD groups, Numb expression was markedly decreased, while RBP-Jκ and Hes1 levels were markedly increased compared to that in the Sham and BMSC NC groups, respectively. In the HQD and HQD plus BMSC Numb-KD groups, Numb expression was increased, while RBP-Jκ and Hes1 expressions were decreased compared to that in BDL and BMSC Numb-KD groups, respectively. However, the improvement in HQD group was better than that of HQD plus BMSC Numb-KD group.

BMSC Numb-KD transplantation reduces the effect of HQD inhibition of Notch signaling activation. As shown in
Additionally, Numb, RBP-Jκ and Hes1 protein and mRNA levels were also examined. In the BDL and BMSC Numb-KD groups, Numb was significantly decreased (P < 0.01), while RBP-Jκ and Hes1 were markedly increased compared to that in the Sham and BMSC NC groups (P < 0.01), respectively. In the HQD and HQD plus BMSC Numb-KD groups, Numb protein and mRNA expression levels were significantly increased, while RBP-Jκ and Hes1 expression levels were markedly decreased compared to that in the BDL and BMSC Numb-KD groups (P < 0.05 or P < 0.01), respectively. However, in the HQD group, Numb protein and mRNA expression levels were significantly increased, while RBP-Jκ and Hes1 expression levels were markedly decreased compared to that in the HQD plus BMSC Numb-KD (P < 0.05 or P < 0.01) (Fig. 4d-f).
To further characterize the effect of Numb knockdown in BMSCs, the upstream components of Notch signaling were also examined. The results showed that Notch-2/-3, DLL1, and Jag-1/-2 mRNA expression levels were significantly increased in the BDL group compared to that in the Sham group (P < 0.01), while Notch-2/-3 and Jag-1/-2 were significantly decreased in the HQD group compared to that in the BDL group (P < 0.05 or P < 0.01). In the BMSC Numb-KD group, Notch-2/-3/-4, DLL-1/-4 and Jag-1/-2 were significantly increased relative to the BMSC NC group (P < 0.01). However, only Notch-2 and DLL-4 were significantly decreased in the HQD plus BMSC Numb-KD group compared to that in the BMSC Numb-KD group (P < 0.01) (Fig. 4g,h). Suggesting that Numb knockdown promotes the activation of Notch pathway in BMSCs, while decreases the inhibitory effect of HQD on Notch signal pathway.
Moreover, the mRNA expression of E3 ubiquitin ligase including LNX-1/-2 (promoting proteasome-dependent degradation of Numb 28 ), and ITCH (promoting ubiquitination-dependent proteasomal degradation of the NICD 29 ) were examined, the results showed that in the BDL and BMSC Numb-KD groups, LNX1 was increased significantly, while the LNX2 and ITCH were decreased significantly compared to that in the Sham and BMSC NC groups (P < 0.05 or P < 0.01), respectively. In contrast, the mRNA level of LNX1 was decreased significantly, while ITCH was significantly increased in the HQD group compared to that in the BDL group (P < 0.01). However, they were no change in HQD plus BMSC Numb-KD group compared to that in the BMSC Numb-KD group (P > 0.05). In addition, the mRNA expression of ITCH was significantly decreased in HQD plus BMSC Numb-KD group compared to that in the HQD group (P < 0.05) (Fig. 4i). To   www.nature.com/scientificreports/ could reach more than 80% at a MOI = 50, with a normal cell morphology (Fig. 5a). Furthermore, the qRT-PCR result showed that Numb expression was significantly decreased in the Numb-KD group compared to that in the negative control (NC) group (P < 0.01; Fig. 5b). Immunostaining showed that CK19 expression was markedly increased in the SB and Numb-KD groups compared to that in the N and NC groups, respectively. However, in the HQD group, CK19 expression was reduced compared to that in the SB group, but it was no clearly difference between the HQD plus Numb-KD and Numb-KD groups (Fig. 5d). Additionally, CK19 mRNA expression was significantly increased in the SB group (vs. N group, P < 0.01), while it was significantly decreased after HQD treatment (vs. SB group, P < 0.01). However, in the Numb-KD group, CK19 expression was significantly increased compared to that in the NC group (P < 0.05), but it was no significant difference between the HQD plus Numb-KD and Numb-KD groups. Furthermore, CK19 mRNA level was significantly higher in the HQD plus Numb-KD group compared to that in the HQD group (P < 0.01) (Fig. 5c). These findings suggest that Numb knockdown significantly attenuates the inhibitory effect of HQD on WB-F344 cell differentiation into BECs phenotype. www.nature.com/scientificreports/ Numb gene knockdown reduces the inhibitory effect of HQD on Notch signaling activation of WB-F344 in vitro. As shown in Fig. 6a-c, immunostaining showed that Numb expression was decreased in the SB and Numb-KD groups compared to that in N and NC groups, respectively, with the greatest reduction noted in the Numb-KD group. On the other hand, RBP-Jκ and Hes1 expression were markedly increased in the SB and Numb-KD groups compared to that in the N and NC groups, respectively. In the HQD and HQD plus Numb-KD groups, Numb expression was markedly increased, while RBP-Jκ and Hes1 were markedly decreased relative to the SB and Numb-KD groups, respectively. However, the improvement degree of compared to that in HQD group was better than that of HQD plus Numb-KD group. Immunoblotting showed that the protein expression level of Numb was decreased significantly in the SB and Numb-KD groups compared to that in the N and NC groups (P < 0.01), respectively. Additionally, RBP-Jκ and Hes1 protein levels were significantly increased in the SB and Numb-KD groups compared to that in the N and NC groups (P < 0.05 or P < 0.01), respectively. When compared to that in the SB and Numb-KD groups, Numb protein levels were significantly increased in the HQD group (P < 0.01) and HQD plus Numb-KD group (P < 0.05), respectively, with the HQD group is of significantly higher levels compared to that in the HQD plus Numb-KD group (P < 0.01). Furthermore, RBP-Jκ and Hes1 levels were significantly decreased in the HQD and HQD plus Numb-KD groups compared to that in the SB and Numb-KD groups (P < 0.05 or P < 0.01), respectively, with significantly higher levels noted in the HQD plus Numb-KD group compared to that in the HQD group (P < 0.01) (Fig. 6d,e). Additionally, Numb, RBP-Jκ and Hes1 mRNA expression levels were consistent with their corresponding protein levels (Fig. 6f). These results suggest that Numb knockdown reduces the inhibitory effect of HQD on Notch signaling activation in WB-F344 cells.

Discussion
BMSC Numb-KD transplantation promotes CLF progression. In recent years, Numb, a negative regulator of Notch signaling, has been widely examined in the science, with Numb found to modulate Notch-1 during Notch mediated skeletal muscle regeneration 30 . Furthermore, Numb inhibition of Notch signaling transduction had been implicated as a potential therapeutic target for prostate cancer 31 . Thus, the interaction between Numb www.nature.com/scientificreports/ and Notch have gained increasing interest, with Numb found to modulate Notch signaling by targeting Notch proteolytic degradation and its intracellular domains 32 . Additionally, in mouse hepatocytes, the Notch receptor and ligand are highly expressed during biliary regeneration 33 . In our previous study, inhibiting Notch signaling was found to inhibit CLF progression, with Numb expression found to be significantly reduced in a CLF model induced by BDL 12 . However, at present, no studies have confirmed the role of Numb in CLF. Thus, further characterizing the role of Numb in CLF could facilitate the development of further treatments. Herein, BMSC Numb-KD transplantation was shown to promote CLF development (detailed data will be presented in another article). BMSCs are exogenous hepatic stem cells and easily acquired [34][35][36][37] and have a high degree of plasticity. Under specific conditions in vitro, BMSCs can be induced to differentiate into various functional cells, such as bone cells, hepatocytes and fat cells 38 . Furthermore, BMSCs have a low rejection level, rapid expansion in vitro, easy differentiation induction, and can easily have exogenous genes introduced 39 . In this study, Numb knockdown was performed in BMSCs by using RNAi, with these cells then transplanted into CLF rats induced by BDL. The results showed that BMSC Numb-KD transplantation significantly increased the Hyp content and promotes α-SMA, Col I, TGF-β1, TNF-α, CK7 and CK19 expression in vivo. These findings indicate that BMSC Numb-KD transplantation can promote CLF progression. Additionally, in the BMSC NC group, EGFP was not co-localized with CK7 or CK19, but in BMSC Numb-KD group, they were extensively co-localized in proliferating BECs, suggesting that Numb knockdown promotes the differentiation of BMSCs into BECs and thus promotes CLF progression.
BMSC Numb-KD transplantation reduces the anti-CLF effect of HQD. In our previous studies, HQD was found to inhibit CLF by inhibiting HPCs differentiation into BECs, and HQD was found to promote Numb expression in vivo and in vitro 14 . Therefore, Numb gene of hepatic stem cells might be related to HQD against www.nature.com/scientificreports/ CLF effect. To prove this hypothesis, HQD was administered after BMSC Numb-KD transplantation into rats with CLF. In the BMSC Numb-KD group with HQD treatment, although liver fibrosis has improved compared with the BMSC Numb-KD group, Hyp content and the mRNA levels of α-SMA, TGF-β1, TNF-α, ColI,CK7 and CK19 were significantly increased compared to that in the HQD group, suggesting that BMSC Numb-KD transplantation into the liver promotes CLF by reducing Numb expression, thereby weakening the effect of HQD against CLF. However, in this case, the liver still had Numb expression as a result of not directly knocking out the Numb gene, thus the above experimental results were obtained. To further examine the interaction between Numb of hepatic stem cell and HQD, EGFP (mark BMSC NC-KD or BMSC Numb-KD ) co-expression with CK19 or CK7 was examined. In the BMSC Numb-KD group, EGFP co-stained with CK19, while this phenomenon was not obviously improved after HQD treatment. The same result was seen for EGFP/CK7, suggesting that HQD has no obvious intervention effect in BMSC Numb-KD differentiation into BECs. This result was further substantiated using WB-F344 cells in vitro, with Numb knockdown enhancing CK19 expression and HQD treatment having no significant effect. As confirmed in vivo, these observations further indicated that hepatic stem cell with Numb knockdown reduces the effect of HQD on CLF progression.
In a previous study to examine human embryonic stem cells, Numb was found to be upregulated following Sal B treatment, thus further negatively regulating the Notch pathway and subsequently inhibiting biliary differentiation 40 . Considering this relationship between Numb and Notch signaling, changes in the Notch signaling resulting from Numb modulation were further examined. In our previous study, HQD was found to inhibit Notch signaling pathway activation 14 . Herein, only Notch-2 and DLL-4 mRNA levels were significantly lower in the HQD plus BMSC Numb-KD group compared to that in BMSC Numb-KD group, suggesting that BMSC Numb-KD transplantation reduces the ability of HQD to inhibit Notch signal pathway activation.
When the Notch receptor binds to its ligand, γ-secretase cleaves the TNF-α, thereby converting enzyme and releasing NICD 41 , and subsequent translocation of the NICD to the nucleus to modulate downstream gene expression 42 . In this work, the changes in RBP-Jκ and Hes1 expression were examined. The results showed that in the HQD plus BMSC Numb-KD group, Numb expression was reduced, while RBP-Jκ and Hes1 expressions were enhanced compared to that in HQD group in vivo. Furthermore, in vitro study showed that in the Numb-KD plus HQD group, Numb expression was increased, while RBP-Jκ and Hes1 expression were reduced compared to that in Numb-KD group. However, when compared to that in the HQD group, Numb expression was significantly reduced, while RBP-Jκ and Hes1 expressions were significantly enhanced. Based on these observations, HQD might inhibit Notch signaling pathway by promoting Numb expression and subsequently inhibiting bile duct hyperplasia and improving CLF.
On the other hand, E3 ubiquitin ligase plays an important role in Notch receptor regulation. LNX, as a Numb PTB-binding molecule, it was found to act as a RING finger-type E3 ubiquitin ligase, causes proteasomedependent degradation of Numb and can enhance Notch signalling 25,43 . In contrast, ITCH, an E3 ubiquitin ligase that belongs to the HECT family, negatively regulates Notch signaling and promotes ubiquitination-dependent proteasomal degradation of the NICD. Furthermore, Numb can interact with ITCH to cooperatively enhance Notch ubiquitination and degradation, circumventing its nuclear localization and downstream activation of Notch1 target genes 26,[44][45][46] . Therefore, we also observed the mRNA expression of LNX and ITCH. The results indicated that HQD could significantly inhibit the expression of LNX1 and increase the expression of ITCH. However, when BMSC cells were transplanted after Numb knockdown, the above positive effects of HQD disappeared, which further proved that Numb might be the key target of HQD against CLF.
In summary, in BDL-induced liver fibrosis, the intervention effect of HQD and BMSC was similar to DAPT, but the transplantation of BMSCs with Numb knockdown can significantly reduce the anti-CLF effects of HQD. This may be attributed to the weakening inhibitory effect of HQD on Notch signal pathway after BMSCs Numb knockdown. Therefore, HQD may inhibit the progress of CLF through the hepatic stem cell Numb gene and which will provide a new therapeutic approach for CLF. To be sure, one potential short coming of utilizing RNAi technology is that it is impossible to completely knock out Numb. Therefore, future studies should be focused on knockout the Numb gene and then examining the effect of HQD against CLF. On the other hand, both HQD and BMSC do have rather good effect against CLF, there will be better therapeutic effect if they are used in combination and it is our top issue in the follow-up study.

Data availability
All data generated or analyzed during this study are included in this published article and its supplementary information files.