Persistent elevation of hepatocyte growth factor activator inhibitors in cholangiopathies affects liver fibrosis and differentiation


  • Potential conflict of interest: Nothing to report.

  • Supported by grants from the NHRI-EX99-9909BC, NHRI-EX100-9909BC, NSC97-2320-B-002-052-MY3, NSC100-2628-B-002-004-MY4, the Frontier and Innovative Research Project of National Taiwan University (99R71427) (to M.S.L.), and NSC 97-2321-B-002-020-MY3 (to H.P.H.).


Alteration of cell surface proteolysis has been proposed to play a role in liver fibrosis, a grave complication of biliary atresia (BA). In this study we investigated the roles of hepatocyte growth factor activator inhibitor (HAI)-1 and -2 in the progression of BA. The expression levels of HAI-1 and -2 were significantly increased in BA livers compared with those in neonatal hepatitis and correlated with disease progression. In BA livers, HAI-1 and -2 were coexpressed in cells involved in ductular reactions. In other selective cholangiopathies, ductular cells positive for HAI-1 or HAI-2 also increased in number. Inflammatory cytokines, growth factors, and bile acids differentially up-regulated expression of HAI-1 and -2 transcripts in fetal liver cells and this induction could be antagonized by a cyclooxygenase-2 inhibitor. Conditioned media from cell lines stably overexpressing HAI-1 or HAI-2 enhanced the fibrogenic activity of portal fibroblasts and stellate cells, suggesting that both proteins might be involved in liver fibrosis. Because HAI-1 and -2 colocalized in ductular reactions sharing similar features to those observed during normal liver development, we sought to investigate the role of HAI-1 and -2 in cholangiopathies by exploring their functions in fetal liver cells. Knockdown of HAI-1 or HAI-2 promoted bidirectional differentiation of hepatoblast-derived cells. In addition, we showed that the hepatocyte growth factor activator, mitogen-activated protein kinase kinase 1, and phosphatidylinositol 3-kinase signaling pathways were involved in hepatic differentiation enhanced by HAI-2 knockdown. Conclusion: HAI-1 and -2 are overexpressed in the liver in cholangiopathies with ductular reactions and are possibly involved in liver fibrosis and hepatic differentiation; they could be investigated as disease markers and potential therapeutic targets. (Hepatology 2012)

Biliary atresia (BA) is one of the most important causes of hepatic fibrosis in children.1 Liver fibrosis is a complex process that involves extensive extracellular matrix (ECM) remodeling and proteolysis.2 Using microarray analysis, Chen et al.3 identified many molecular markers associated with BA of which at least one-third are related to ECM remodeling. Recently, pericellular proteolysis by secreted or membrane-anchored proteases and their inhibitors have been strongly implicated in fibrosis.4, 5 In BA, hepatocyte growth factor (HGF) has been shown to be significantly elevated in the serum of patients who required liver transplantation.6 The proteolytic maturation of HGF can be mediated by several proteases including HGF activator (HGFA),5 hepsin,7 and matriptase.8 The activity of these proteases can be modulated by two transmembrane serine protease inhibitors: HGFA inhibitor (HAI)-1 and HAI-2 (Supporting Fig. 1).5, 7, 9, 10 HAI-1 is expressed in many epithelial-derived tissues including bile duct.11 HAI-2 is an isoform of HAI-1, which is colocalized with HAI-1 in most epithelia, and also detected in nonepithelial cells of the brain and lymph nodes,9 suggesting that HAI-2 may have a nonredundant role. HAI-1−/− or HAI-2−/− mice are not viable beyond embryonic day (E) 10.5 and the gastrulation stage, respectively,12, 13 suggesting that HAI-1 and HAI-2 are critical in early development. In adult tissues, HAI-1 is known to be involved in the progression of pulmonary fibrosis5 and augmented expression of HAI-1 is also found in the small bile ducts in primary biliary cirrhosis.14 These observations led us to explore the possible roles of HAI-1 and -2 in BA or other cholangiopathies, as well as identify the protease(s) on which they act that might be involved in BA- or other cholangiopathy-associated fibrosis.

In BA livers it is known that periductular fibrosis frequently follows the ductular reaction which consists of proliferative bile ductules, bile ducts, and hepatic stem cells (HSCs).15 Striking similarities have been reported between proliferative bile ductules and developing bile ducts in human fetus.16 Because there are increased numbers of both proliferating ductular cells and newly regenerating hepatocytes in BA livers,17 it has been suggested that both types of cells may differentiate from HSCs activated in BA livers.17 Based on these observations, here we explored the mechanisms controlling the activation and differentiation of cells in ductular reactions and their possible relationship to HAI-1 and -2-related ECM remodeling and fibrosis progression in livers with BA or other cholangiopathies.


BA, biliary atresia; BRIC, benign recurrent intrahepatic cholangitis; COX-2, cyclooxygenase-2; ECM, extracellular matrix; GFP, green fluorescent protein; HAI, hepatocyte growth factor activator inhibitor; HGF, hepatocyte growth factor; HGFA, hepatocyte growth factor activator; HSCs, hepatic stem cells; IF, immunofluorescence; IHC, immunohistochemistry; LT, liver transplantation; MEK1, mitogen-activated protein kinase kinase 1; NH, neonatal hepatitis; PBC, primary biliary cirrhosis; PFIC, progressive familial intrahepatic cholestasis; PFs, portal fibroblasts; PI3K, phosphatidylinositol 3-kinases; PSC, primary sclerosing cholangitis; Q-PCR, quantitative real-time polymerase chain reaction.

Materials and Methods

Patients and Ethical Considerations.

This study included 15 BA patients and five control patients with neonatal hepatitis (NH) who were followed at National Taiwan University Hospital (NTUH). Their ages and gender are listed in Supporting Table 1. The diagnosis of BA or NH in patients was based on the pathologists' reports on biopsies. Needle biopsy samples were obtained from five infants (Supporting Table 1, patients 1-5) with NH without metabolic or transport defects. BA liver tissues were taken by wedge biopsy from nine infants (Supporting Table 1, patients 6-14) who underwent the Kasai operation for BA. They were further divided into two groups retrospectively according to the clinical features at the end of 12-month follow-up after the Kasai operation: three patients (patients 6-8) who were jaundice free and with good bile flow (BA1) and six patients (patients 9-14) who had jaundice and finally underwent liver transplantation (BA2). In addition, liver tissues were also obtained from six children (Supporting Table 1, patients 15-20) with advanced-stage BA at the time of liver transplantation (LT). Near-normal liver tissues were the nontumor parts of surgically removed liver tissues from two patients with colon cancer metastasized to the liver. Two human fetal liver samples were obtained after legal termination of pregnancy (gestational age: ≈18 weeks). For liver tissues of other cholangiopathies (Supporting Table 2), archived paraffin tissues were obtained from the Department of Pathology, NTUH. All liver tissues in this study were obtained after acquiring written informed consent from parents. The protocol for this study was approved by the Ethics Committee of the Institutional Review Board (IRB) of NTUH. All pathological samples were stored in liquid nitrogen prior to use and handled according to the approved IRB protocol.

Reagents, RNA Extraction, and Quantitative Real-Time PCR, Western Blot, RNAi Knockdown, Immunofluorescence Microscopy, MTT and Transwell Assay.

Details are provided in the Supporting Materials and Methods section. Primers for quantitative real-time polymerase chain reaction (Q-PCR) are listed in Supporting Table 3. All Q-PCR reactions were performed in triplicate unless noted otherwise, normalized to control, and presented as mean ± standard deviation (SD). Antibodies and dilutions used are listed in Supporting Table 4.

Mouse Bile Duct Ligation.

The protocol used to generate the mouse model has been reported.18 Details are provided in the Supporting Materials and Methods. Animal procedures in this study were performed in accordance with protocols approved by the Institutional Laboratory Animal Care and Use Committee of NTUH. Animals received humane care according to the guide published by the National Institutes of Health (NIH), USA.

Isolation, Maintenance, and Differentiation of Mouse Fetal Liver Cells and Hepatoblast-Derived Cells.

The mouse fetal liver cell line, Hepo-2, which consists of mostly differentiated hepatocytes and cholangiocytes, and the hepatoblast-derived cell line N8, which consists of mostly bipotential cells, were established from the liver of C57BL6 mice at E14.5 using a reported protocol.19 The detailed procedures for isolation, maintenance, and induction of N8 cells are provided in the Supporting Materials and Methods.

Isolation of Stellate Cells and Portal Fibroblasts.

Rat stellate cells, a gift from Dr. Hsuan-Shu Lee at NTUH, were characterized as described.20 Mouse portal fibroblasts were freshly isolated from C57BL6 mice according to a published protocol.21 Cells were used at low-passage numbers (less than 10 and 5 for stellate cells and portal fibroblasts, respectively).

Statistical Analysis.

Student's t test was used to evaluate the differences between two different parameters. A two-sided P-value of less than 0.05 was considered statistically significant.


Increased Expression of HAI-1 and HAI-2 in BA Livers Is Correlated with Disease Progression.

To establish whether HAI-1 and HAI-2 expression is changed in BA, we performed Q-PCR to analyze their messenger RNA (mRNA) levels in the livers of BA patients, including two groups of patients who received the Kasai operation: patients without disease progression (BA1 group, average age at biopsy: 59.0 days), patients with disease progression in need of LT (BA2 group, average age at biopsy: 52.2 days), and a third group of patients receiving LT (LT group, average age at surgery: 330.7 days) due to endstage BA. Liver samples with NH (NH group, average age at biopsy: 51.6 days, no significant difference in the age of biopsy between NH, BA1, and BA2 groups) and from the nontumor (near-normal) part of patients with metastasized liver tumors were also included as controls. The results showed that HAI-1 expression was significantly increased in the livers of BA patients compared with that in the NH group (Fig. 1A) (BA2 versus NH, P < 0.05; BA1 and BA2 versus NH, P < 0.01). Moreover, the extent of HAI-1 up-regulation was increased in endstage BA (LT versus BA1, P < 0.05; LT versus BA2, P < 0.01; LT versus BA1 and BA2, P < 0.01). Similarly, HAI-2 expression was significantly increased in the livers of the LT group compared with NH and BA1 groups (Fig. 1A; LT versus NH, P < 0.01; LT versus BA1, P < 0.05). In addition, immunofluorescence (IF) microscopy showed that in normal liver both HAIs were selectively expressed in the bile duct (Fig. 1B), whereas in BA livers HAI-1- or HAI-2-positive cells were found in the ductular reactions of portal areas, including bile duct- and ductule-like structures, cell clusters, and even in single cells (Fig. 1B).

Figure 1.

Expression of HAI-1 and HAI-2 in livers of BA and bile duct-ligated mice. (A) Liver mRNAs were obtained from near-normal tissue (N), neonatal hepatitis (NH), BA patients with (BA2) or without progression (BA1), and endstage BA with LT, and analyzed by Q-PCR for HAI levels. *P < 0.05, **P < 0.01. (B) HAI-1/-2 localization in normal and BA livers examined by IF microscopy. Scale bars = 50 μm. (C) Analysis of HAI expression in the liver of 2-week bile duct-ligated (BDL) and sham mice (Ctrl). Liver tissues were examined by silver staining to detect collagens (left panel/black). Scale bars = 20 μm. Q-PCR (right panel, n = 3) showed an increased expression of both HAIs. *P < 0.05, **P < 0.005.

Up-regulation of HAI-1 and -2 Expression in Mouse Livers with Experimental Obstructive Cholestasis and Other Human Cholangiopathies.

To further determine whether the up-regulation of both HAIs in human BA livers was caused by obstructive jaundice, we employed two widely used murine models of obstructive cholestasis, in which the pathology is induced by bile-duct ligation18 or rotavirus infection.22 In the first model we surgically ligated the mouse bile duct for 2 weeks to induce obstructive cholestasis and portal fibrosis; control mice received sham operations. Silver staining, used to highlight collagen (Fig. 1C), showed that the bile-duct ligation successfully induced portal fibrosis and ductal proliferation. In the second model we infected newborn mice with rotaviruses and euthanized them after 2 weeks. Whereas control mice had normal liver histology, the rotavirus-infected mice showed bile-duct proliferation in the livers (Supporting Fig. 2A). In both models Q-PCR (Fig. 1C and Supporting Fig. 2A, right panels) showed that the mRNA levels of both HAIs were significantly up-regulated in the livers of mice receiving bile-duct ligation or rotavirus infection, a phenomenon similar to observations in human BA. Using immunohistochemistry (IHC), we also evaluated the expression of both HAIs in the livers of other human cholangiopathies, including primary biliary cirrhosis (PBC), primary sclerosing cholangitis (PSC), and intrahepatic cholestasis in children, such as progressive familial intrahepatic cholestasis (PFIC) and benign recurrent intrahepatic cholestasis (BRIC). An increased number of ductular cells positive for CK19, HAI-1, and HAI-2 (Fig. 2A for PBC and PSC) was found in the livers of cholangiopathies except BRIC, apparently with different incidence (Fig. 2B). Because cryopreserved human tissues were not available, we instead employed a xenobiotic-induced PBC mouse model to assay HAI expression23 and found that the expression of both HAIs was also elevated in PBC mice (Supporting Fig. 2B). In contrast, we could not detect any increase in HAI expression in the only two cryopreserved liver biopsies of type-II PFIC available (Supporting Fig. 2C) compared with those in a near-normal liver and an NH liver. This was consistent with the IHC result that showed no increase of CK19-positive ductular cells in type-II PFIC (Supporting Fig. 2D). In a type-III PFIC liver (Supporting Fig. 2E,F), however, the ductular reactions were seen with increased expression of CK19, HAI-1, and HAI-2.

HAI-1 and -2 Are Coexpressed in Cells of Cholangiocyte Lineage and Human Hepatic Stem Cells.

To further identify which types of cells expressed abundant HAI-1 and -2 in BA livers, we performed colocalization studies and showed that both HAI-1 (Fig. 2C, left) and HAI-2 (Fig. 2C, middle) were coexpressed with CK19, a well-known marker for HSCs and cells of the cholangiocyte lineage.24 Moreover, HAI-1 was also coexpressed with epithelial cell adhesion molecule (EpCAM), Gli-2, and OV6, additional biomarkers for HSCs (Supporting Fig. 3A-C). Because the majority (>90%) of cells expressing HAI-1 also coexpressed HAI-2 in BA livers (Fig. 2C, right), we assumed that HAI-2 was also coexpressed in most EpCAM-, Gli-2-, and OV6-expressing cells, although colocalization studies could not be performed as these antibodies were raised in the same animal species. In addition, HAI-1 was occasionally found in the cells expressing α-fetoprotein (AFP), a marker for hepatoblasts (Supporting Fig. 3D). Therefore, HAI-1 and -2 were expressed mostly in cells of cholangiocyte lineage and HSCs.

Figure 2.

Immunostaining of HAI-1, HAI-2 and CK19 in PBC, PSC, PFIC, BRIC, and BA livers. (A) Analysis of CK19, HAI-1 and HAI-2 in PBC and PSC livers by IHC. The adjacent sections of representative PBC and PSC livers were stained with anti-CK19, anti-HAI-1 and anti-HAI-2 antibodies (brown color). Scale bars = 20 μm (B) Summary of IHC results. The cholangiopathy tissues available (4 PSC, 12 PBC, 4 PFIC and 2 BRIC cases) were stained with the above antibodies. The positive staining of each protein was shown as a percentage. Parenthesis, the ratios of positive staining numbers to total patient numbers. dz; disease. (C) Coexpression of HAI-1/CK19 (left), HAI-2/CK19 (middle), and HAI-1/HAI-2 (right) in BA livers was analyzed by double IF microscopy. Arrows indicate the coexpression of HAI-1 (left) and HAI-2 (middle) with CK19 in small cell clusters or single cells in ductular reactions. All IF studies were performed on cryostat sections of three BA patients with similar results. Nuclei were counterstained with DAPI. Scale bars = 25 μm.

Regulation of Expression of Both HAIs by Cytokines, Growth Factors, Bile Acids, and a cyclooxygenase-2 (COX-2) Inhibitor.

Because several proinflammatory cytokines,25 growth factors,26 and bile acids27 have been found elevated in the serum and/or liver of BA patients, we next determined whether these factors are involved in activating HAI expression in BA livers. Due to the difficulty of culturing HSCs or hepatoblasts from BA livers, we instead established fetal liver cells derived from E14.5 mouse livers, which are comprised of several different liver cell lineages (Supporting Fig. 4A-C).19 These cells, named Hepo-2, exhibited low HAI expression (Supporting Fig. 4A,C). Using these Hepo-2 cells we found that interleukin (IL)-β, tumor necrosis factor alpha (TNF-α), and transforming growth factor beta (TGF-β)-1 stimulated HAI-1 expression, whereas TNF-α and HGF marginally induced HAI-2 expression (Fig. 3A). In addition, the expression of an inflammation-related enzyme activated in BA liver,28 COX-2, was significantly elevated in the cells treated with the above factors (Supporting Fig. 4D), and a COX-2 inhibitor, celecoxib, efficiently blocked these stimulatory effects on both HAIs (Fig. 3B). Furthermore, among various bile acids, deoxycholic and lithocholic acids, but not cholic or chenodeoxycholic acid, significantly stimulated HAI-1 expression but not HAI-2 (Fig. 3C). Deoxycholic acid induced the HAI-1 expression in a dose-dependent manner (Fig. 3D). Taken together, these data indicate that selective factors enriched in BA livers activate HAI expression, possibly through the increased expression of COX-2.

Figure 3.

Analysis of inflammatory factors, celecoxib and bile acids, on expression of the HAIs in Hepo-2 cells by Q-PCR. (A) Effects of IL-1β, TNF-α, TGF-β1, and HGF on the expression of HAI-1 and -2 were examined after 24-hour treatment. (B) Effect of celecoxib on expression of the HAIs induced by IL-1β, TGF-β1 and TNF-α was analyzed by Q-PCR after treatment with IL-1β, TNF-α and TGF-β1 alone or in combination with celecoxib for 24 hours. (C) Effects of bile acids (cholic acid [CA], chenodeoxycholic acid [CDCA], deoxycholic acid [DeoxyCA], and lithocholic acid [LithoCA]) on HAI expression were examined after 24-hour treatment. (D) Effect of DeoxyCA on HAI expression was analyzed 24 hours after the treatment with different concentrations of DeoxyCA. RNA levels were normalized to control (PBS or DMSO). *P < 0.05.

Conditioned Media from Cells Overexpressing HAI-1 or HAI-2 Enhance Fibrogenic Activity.

To test whether increased HAI-1 or HAI-2 expression plays a role in the fibrosis process in BA livers, portal fibroblasts (PFs) (Fig. 4A) and stellate cells,20 two major cell types responsible for cholestasis-related fibrosis,29, 30 were treated with conditioned media from Hep3B cells stably overexpressing HAI-1 or HAI-2 (Fig. 4B) or control media (green fluorescent protein [GFP] or vector) to assay their effects on the fibrogenic activity. Conditioned media containing HAI-1 or HAI-2 significantly increased mRNA levels of collagen I and IV, two common types of collagen expressed in the ductular reaction in BA livers,31 in both cells compared with controls (Fig. 4C). Western blot analysis further confirmed that the conditioned media containing HAI-1 or HAI-2 enhanced the protein levels of collagen I in PFs (Supporting Fig. 5A,B). We also found that conditioned media containing either HAI-1 or HAI-2 significantly increased PF cell migration, but only HAI-2 significantly increased stellate cell migration (Supporting Fig. 5C). Moreover, colorimetric cell viability (MTT) assays revealed that the HAI-1-rich, and probably HAI-2-rich (from one of two clones) conditioned media, significantly increased the survival of both fibroblasts (Supporting Fig. 5D). Furthermore, recombinant HAI-2 protein (Fig. 4D) significantly up-regulated the expression of Co11a1 and Col4a1 in PFs and the Co11a1 expression in stellate cells.

Figure 4.

Examination of the conditioned media with secreted HAI-1 or HAI-2 protein, and a recombinant HAI-2 protein on fibrogenic activity. (A) Establishment of mouse portal fibroblasts (PFs). PFs were grown from portal connective tissues, as shown in a phase-contrast picture (left). The isolated cells were characterized as PFs by RT-PCR (right; PF marker genes: IL6, Eln, NTPD2, and Fbln2). RT-N ctrl, RT-negative control. (B) HAI-1/-2 protein levels in the conditioned media (CMs, 48-hour collection) from Hep3B cells stably overexpressing HAI-1/-2 and control cells were analyzed by immunoblotting with anti-HAI-1 and anti-HAI-2 antibodies. (C) Effects of CMs on the expression of collagen I (Col1a1) and IV (Col4a1) in PFs and stellate cells were analyzed by Q-PCR after 24-hour treatment. Control cells were treated with control media (vector and GFP). *P < 0.05, **P < 0.005. (D) Effect of a recombinant HAI-2 (rHAI-2) protein on the expression of collagen I and IV in PFs and stellate cells. The rHAI-2 protein was purified as the eluted fractions 2∼4 (left panel) from the CMs from HEK293T cells transfected with secretory plasmids, using an Ni-NTA affinity column as described in Supporting Methods. Cells were treated with 7 nM rHAI-2 for 24 hours and the expression of Col1a1 and Col4a1 was analyzed by Q-PCR (right panel). *P < 0.05. CMs, conditioned media; FT, flow through fraction; W, washing fractions; E, elution fractions.

Abundant Expression of HAI-1 and HAI-2 in the Human Fetal Liver.

Both HAIs were also expressed in BA livers in cell clusters or single cells with much CK19 and little or no AFP expression (Fig. 2C, arrows; Supporting Fig. 3D), which probably represent a subset of HSCs.15, 24 Thus, we hypothesized that HAI-1 and/or -2 may also have functions in HSCs. To prove this, we examined HAI expression in the developing livers of legally aborted fetuses. IF microscopy showed that HAI-1 expression was mainly found in the ductal plate (CK19-positive, Fig. 5A), whereas HAI-2 was widely expressed in the area surrounding the ductal plate and in other fetal liver cells (Fig. 5B,C). There was, however, a small population of cells coexpressing both HAIs (Fig. 5C, arrows), suggesting that HAI-1 and -2 are expressed in two different populations of fetal liver cells with overlapping expression in a subset of cells. It is likely that HAI-1 is mainly expressed in human HSCs based on its expression pattern in human BA livers and fetal livers. During mouse development, mRNA levels for both HAIs were higher in the liver at E13.5 than at E15.5 (Fig. 6A); E13.5 is a stage when the liver is enriched with bipotential progenitor cells.32 Taken together, these data suggest that HAI-1 and/or HAI-2 might exhibit potential functions in fetal liver cells.

Figure 5.

Immunofluorescence staining of HAI-1, HAI-2 and CK19 in human fetal livers. Expression of HAI-1/CK19 (A), HAI-2/CK19 (B), and HAI-1/HAI-2 (C) in the liver of aborted human fetuses was analyzed by double IF microscopy. Some cells around or in the ductal plate coexpressed HAI-2 and CK19 (B, right panel, arrows). Both HAIs were coexpressed in some cells near the ductal plate (right panel, arrows). Nuclei were DAPI-counterstained. Scale bars = 50 μm for A, 25 μm for B,C.

Figure 6.

HAI-1/-2 expression in mouse embryonic liver and their roles in hepatoblast differentiation. (A) Expression levels of HAI-1 and -2 in mouse livers at E13.5 and E15.5 was analyzed by Q-PCR. (B) Mouse hepatoblast-derived cells were isolated from E14.5 liver using a protocol published by Tanimizu et al.19 and named N8 cells (phase contrast, upper left). Cell differentiation was analyzed by culture on Matrigel for hepatocytes (upper middle) and in collagen gel for cholangiocytes (upper right). Afp, albumin, and CK19 in N8 cells were visualized by IF microscopy (lower panels). Scale bars = 25 μm for lower panels, 125 μm for upper panels. N8 cell differentiation toward hepatocytes (C) and cholangiocytes (D) was determined by Q-PCR analyzing the expression of hepatocyte/cholangiocyte-enriched genes, and HAIs levels (right panels in C,D). (E) Effect of HAI-1 knockdown on N8 cell differentiation characterized by Q-PCR. (F) Effect of HAI-2 knockdown on N8 cell differentiation assayed by Q-PCR. Knockdown of HAI-1, HAI-2, or luciferase (control) was performed using small interfering RNAs (siRNAs) or lentiviral short hairpin RNAs (shRNAs). *P < 0.05.

Knockdown of HAI-1 or HAI-2 Affects Hepatic Differentiation.

To elucidate the possible functions of HAI-1 and HAI-2 in hepatic differentiation, we established a hepatoblast-derived cell line, named N8, from E14.5 mouse embryos (Fig. 6B, upper left; Supporting Fig. 6), using a protocol that can generate bipotential progenitor cells.19 N8 cells indeed expressed HSC and hepatoblast markers including AFP, albumin (Alb), CK19, and EpCAM, but not the genes found in hepatocytes (TAT) or cholangiocytes (Aqp1) (Supporting Fig. 6B). IF microscopy showed that N8 cells homogenously expressed AFP, albumin, and CK19 (Fig. 6B). Flow cytometry studies further confirmed that more than 90% of gated N8 cells expressed AFP or CK19 (Supporting Fig. 6F). Under conditions to induce differentiation,19 N8 cells were capable of undergoing bi-lineage differentiation into hepatocytes (Fig. 6B, upper middle) or cholangiocytes (Fig. 6B, upper right). According to the literature,19 N8 cells behaved very similarly to hepatoblast-derived bipotential cells (also called hepatic progenitor cells). The results further showed that the majority of N8 cells expressed both HAIs (Supporting Fig. 6B-E; Fig. 8A), the expression of which decreased significantly after differentiation (right panels in Fig. 6C,D), suggesting potential roles of both HAIs in hepatic differentiation. Interestingly, knockdown of HAI-1 resulted in partial N8 cell differentiation, as evidenced by increased expression of two hepatocyte marker genes (Tat and Cps1) and two cholangiocyte markers (Aqp1 and Notch 1) (Fig. 6E), whereas knockdown of HAI-2 caused a more general induction of hepatic differentiation in which most tested genes of both lineages were up-regulated (Fig. 6F). These results suggest a role for both HAIs, especially HAI-2, in maintaining the undifferentiated status of fetal liver cells.

Involvement of HGFA, MEK1, and PI3K Signaling Pathways in HAI-2 Knockdown-Mediated Hepatocyte Differentiation.

To further investigate the molecular mechanism responsible for the above findings, we first aimed to identify potential target protease(s) upon which HAI-2 might act in BA livers. Comparing the distribution patterns of HAI-2 and three reported target proteases,5, 7, 9 matriptase, hepsin, and HGFA in BA livers using IF and IHC, we found that the expression pattern of HAI-2 was different from those of matriptase or hepsin, because matriptase was mainly expressed in the periportal region (Fig. 7A), whereas hepsin was predominantly found in hepatocytes (Fig. 7B). In contrast, the HGFA expression pattern almost overlapped with that of HAI-2 (Fig. 7C,D). Similarly, the majority of N8 cells were found to also coexpress HAI-2 and HGFA (Fig. 8A). Coimmunoprecipitation confirmed that HAI-2 interacts with HGFA in N8 cells (Fig. 8B). Furthermore, we evaluated the knockdown effect of HGFA and/or HAI-2 on N8 cell differentiation. Knockdown of HGFA alone decreased the expression of the majority of hepatocyte markers, but increased the expression of cholangiocyte marker genes Aqp1 and Notch 1 (Supporting Fig. 7A). Remarkably, HGFA knockdown significantly decreased the effect of HAI-2 knockdown on hepatocyte differentiation compared with HAI-2 knockdown alone (Fig. 8C). On the contrary, knockdown of HGFA enhanced the effect of HAI-2 knockdown on inducing cholangiocyte differentiation (Fig. 8C). To further dissect the possible pathway(s) that mediated the signals involved in HAI-2 knockdown-induced hepatic differentiation, we examined whether PD98059, a MEK1 inhibitor, and LY294002, a PI3K inhibitor, could alter the impact of HAI-2 knockdown on hepatic differentiation. PD98059 partly blocked the effects produced by HAI-2 knockdown, resulting in decreased expression of three out of four hepatocyte markers and three out of five cholangiocyte markers assayed (Supporting Fig. 7B), whereas LY294002 efficiently antagonized HAI-2 knockdown-induced expression of all but one of these genes (Supporting Fig. 7B). Taken together, our results suggest that HGFA is the most likely target protease for HAI-2 to modulate hepatic differentiation into hepatocytes, but not cholangiocytes; both PI3K and MEK1 pathways may mediate some effect of HAI-2 knockdown on bi-lineage differentiation of N8 cells. The hypothetic effects of persistent overexpression of both HAIs in livers with cholangiopathies are summarized in Fig. 8D.

Figure 7.

Localization of HGFA, matriptase, hepsin, and HAI-2 in BA livers. BA liver tissues were analyzed by IHC and Co-IF microscopy. (A) An IF image of HAI-2 and matriptase in BA livers. Dashed line indicates the junction of the portal area and a hepatic lobule. (B) An IF image of hepsin and HAI-2 in BA livers. (C) An IHC image of HGFA in BA livers. (D) IF images of HGFA and HAI-2 in BA livers. HGFA (left/green), HAI-2 (middle/red), merged (right). Scale bars = 25 μm for (B), 50 μm for (A) and (D), and 250 μm for (C).

Figure 8.

Involvement of HGFA in HAI-2 knockdown-induced N8 cell differentiation. (A) The expression of HGFA and HAI-2 in N8 cells was examined by IF microscopy. Scale bars = 25 μm. (B) Interaction of HAI-2 and HGFA in N8 cells was examined by immunoprecipitation with an anti-HAI-2 antibody and immunoblotting (IB) with anti-HGFA and anti-HAI-2 antibodies. (C) Role of HGFA in HAI-2 knockdown-induced N8 cell differentiation was determined by analyzing the expression changes of hepatocyte- or cholangiocyte-enriched genes (Alb, Tat, Cps1, Aqp1, itgb4, jagged-1, and Notch-1) using Q-PCR, after knockdown of HAI-2, HAI-2/HGFA, or luciferase control. *P < 0.05, between shHAI-2 and shHAI-2_shHGFA. (D) Scheme of possible effects of persistent HAI overexpression in the livers with cholangiopathies on liver fibrosis and hepatoblast (HB) differentiation. PF, portal fibroblasts; SC, stellate cells.


Our present study has established that HAI-1 and HAI-2 expression is up-regulated in cholangiocyte precursors and probably HSCs in BA livers and that this up-regulation is correlated with disease progression. Furthermore, we propose that elevation of HAI-1 and -2 in livers with BA or other cholangiopathies may recapitulate some of their functions in early liver development, but their persistent overexpression may be unfavorable for hepatocyte differentiation and enhance fibrosis.

We showed that both HAIs are involved in enhancing the fibrogenic activity of PFs and stellate cells. Because both HAIs were highly expressed in areas of ductular reaction in the livers of BA or other cholangiopathies and might enhance the activity of adjacent fibroblasts, our results may explain the observations of previous studies,33 which showed that periductular fibrosis frequently followed ductular reactions in livers with cholangiopathies. Two possible mechanisms have been proposed to explain how ductular reactions promote liver fibrosis33: (1) by secreting profibrogenic factors, and (2) by promoting epithelial mesenchymal transition.34 In this study, we have shown that conditioned media from cells overexpressing HAIs, and recombinant HAI-2, stimulated fibroblasts to express collagens, so HAI-1 and -2 might serve as profibrogenic factors in ductular reactions. Such profibrogenic effects, however, may be direct or indirect, because conditioned media might contain not only HAI-1 or HAI-2, but also other factors that are possibly processed by both HAIs. The possible role of both HAIs in epithelial mesenchymal transition remains to be determined.

Ductular reactions have been demonstrated to recapitulate some of the differentiation processes involved in normal liver development,15 and so to better understand the role of HAI-1 and -2 in BA or other cholangiopathies, we sought to examine their functions in liver development. We found that both HAIs were highly expressed in mouse hepatoblast-derived bipotential cells and probably expressed in human HSCs in BA livers, whereas in human fetal liver, HAI-1 and -2 were differentially expressed in HSCs and hepatoblasts, respectively, according to the definition and staging of HSCs and hepatoblasts proposed by Dr. Lola M. Reid and colleagues.24 Thus, both HAIs might function as regulators keeping hepatic precursor cells in a less-differentiated status prior to undergoing differentiation. To link the roles of the HAIs in hepatic differentiation and fibrosis, two seemingly unrelated phenomena, we propose that the key discriminative factor may be the action dose and duration of HAI expression. At higher expression levels and longer durations, the HAIs may shift from being favorable physiological regulators to demons with pathological roles in BA or other cholangiopathies. For example, an initial moderate increase in HAI expression in BA livers may indicate the role of HAIs in participating in a compensatory activation of HSCs for bidirectional differentiation, which is accompanied by down-regulation of HAI expression after this process (right panels, Fig. 6C,D). However, the persistent and extremely high levels of HAI-1 and -2, as seen in advanced BA, might block differentiation of hepatic cells, induce fibrogenic activity in adjacent fibroblasts, and contribute to fibrosis. This hypothesis may be supported by our observations showing that fibrosis frequently accompanies persistent ductular reactions rich in HAI-positive cells in other cholangiopathies. It is possible, therefore, that an intervention to down-regulate the extremely high levels of HAI-1 and/or HAI-2 in BA livers may slow disease progression by generating more differentiated cells with less fibrosis.

In summary, our results reveal novel potential functions of HAI-1 and -2 in the liver in the context of cholangiopathies, especially in the regulation of fibrosis and hepatic differentiation. The molecular mechanism we propose deserves further elucidation and may provide insights valuable to the development of new therapeutic strategies for BA and other cholangiopathies complicated by fibrosis.


The authors thank Dr. Chen-Yong Lin, Department of Biochemistry and Molecular Biology, University of Maryland, for providing anti-HAI-1 and anti-matriptase antibodies; Dr. Yen-Hsuan Ni for providing study materials; Dr. Hsuan-Shu Lee and Dr. Wei-Hsuan Yu for providing rat stellate cells; Dr. Jun-Tai Wu for confocal microscopy techniques; Dr. Hurng-Yi Wang for statistics consultation; Dr. Shu-Wha Lin for providing animal experimental facilities, the National RNAi Core Facility in Academia Sinica (NSC 97-3112-B-001-016) for lentiviral shRNA clones; Dr. Ming-Jer Tsai for critical advice; and Dr. Michael D. Johnson, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, for editing.