Hepatic biliary epithelial cells acquire epithelial integrity but lose plasticity to differentiate into hepatocytes in vitro during development

During development, tissue stem/progenitor cells differentiate to multiple types of epithelial cells, which establish various tissue structures, including alveoli in the lung, renal tubules in the kidney, and hepatic cords and bile ducts in the liver. Given that organs need to perform their physiological functions after the birth, epithelial tissue structures may be mostly developed at the birth or soon after. However, it is unknown whether epithelial cells are fixed in their lineage and fully functional in neonatal organs.

The liver contains two types of epithelial cells, named hepatocytes and cholangiocytes, which originate from hepatoblasts (fetal liver stem/progenitor cells) during development (Oertel et al., 2003; Tanimizu et al., 2003). Cholangiocytes are biliary epithelial cells forming bile duct tubules. Bile ducts connect the liver to the intestine to drain the bile secreted by hepatocytes. It can be assumed that cholangiocytes acquire epithelial characteristics including secretory and barrier functions when they establish the tubular structure, since it is physiologically important to modulate the composition of bile and avoid any leakage of bile during drainage. However, it is unknown whether cholangiocytes in the neonatal liver have similar epithelial characteristics as those in the adult liver.

In the adult liver, there are at least three possible sources of hepatocytes and cholangiocytes: self-duplication of mature cells, the stem cell system, and lineage conversion. The self-duplication of hepatocytes and cholangiocytes to replace aged or damaged cells is the simplest way, which may be the case in normal and in acutely injured livers (Michalopoulos, 2007; Malato et al., 2011). In contrast, after severe chronic liver injury, the duplication ability of the epithelial cells may be exhausted and stem or progenitor cells may be activated to supply hepatocytes and cholangiocytes (Espanol-Suner et al., 2012). In addition to the self-duplication and stem/progenitor cell systems, lineage conversion should be taken into consideration (Michalopoulos, 2011). It has been shown that mature hepatocytes (MHs) have the potential to transdifferentiate into cholangiocyte-like cells (Nishikawa et al., 2005; Zong et al., 2009). In contrast to hepatocytes, it remains unclear whether cholangiocytes have the ability to convert into hepatocytes.

In this work we examined the differentiation potential of cholangiocytes in neonatal and adult mouse liver. We found that neonatal, but not adult, epithelial cell adhesion molecule (EpCAM)⁺ cholangiocytes expressed hepatocytic transcription factors and converted into hepatocytes in vitro that were structurally and functionally similar to MHs. Interestingly, neonatal cholangiocytes are still immature compared with adult ones even though they have already established tubular structures in vivo. Our results indicate that neonatal cholangiocytes possess plasticity to convert into hepatocytes but lose this ability during maturation of bile ducts. We further demonstrated that a transcription factor implicated in epithelial maturation limited lineage plasticity of cholangiocytes.

Because the number of cholangiocytes isolated from the liver is limited and not enough to examine their differentiation potential, we first established a primary culture in which cholangiocytes keep the original characteristics and efficiently proliferate. To isolate mature cholangiocytes from 6-week-old (6W) mouse liver, two-step collagenase perfusion was performed and the remaining tissue containing Glisson's capsules was further digested. EpCAM⁺ cholangiocytes were enriched by magnetic-activated cell sorting (MACS; supplementary material Fig. S1). They were plated on culture wells coated with type I collagen (Col-I) or a thin layer of Matrigel (MG), or covered with Col-I gel or MG gel (Fig. 1A). On wells coated with Col-I, only a very small number of cells survived and proliferated. On MG-coated or MG gel wells, 2 or 3 days after plating, cells began to proliferate slowly. On Col-I gel, cells proliferated very efficiently. In all four conditions, cells survived and proliferated after replating at day 7 of primary culture. Importantly, on Col-I gel, as well as MG gel, expression of EpCAM was retained on cholangiocytes but disappeared when grown in wells coated with Col-I (Fig. 1B; supplementary material Fig. S2). During the culture on Col-I gel, cholangiocytes maintained expression of cholangiocyte markers [osteopontin (OPN), SRY-related HMG box transcription factor 9 (Sox9) and hepatocyte nuclear factor (Hnf)1β], but did not express the hepatocyte marker, Hnf4α (Fig. 1C), and kept epithelial characteristics, such as the ability to form cystic structures in 3D culture; about 1% of cells formed cysts during 10 days in culture ever after the fourth passage (Fig. 1D; supplementary material Fig. S2). We further confirmed that, like mouse cholangiocytes, human EpCAM⁺ cholangiocytes proliferated and retained the expression of EpCAM on Col-I gel (supplementary material Fig. S3). In the following experiments, we examined differentiation potential of cholangiocytes after expansion in Col-I gel culture.

To examine the hepatocytic differentiation potential of cholangiocytes, EpCAM⁺ cells derived from 6–8W mouse livers were cultured on Col-I gel for 5 days and then replated onto dishes coated with gelatin. To induce hepatocytic differentiation, oncostatin M (OSM) was added to the culture medium after the cells reached confluency. On day 9 in culture, cells were overlaid with 5% MG (Fig. 2A). Dense cytoplasm and clear cell–cell contacts were observed after sequential treatment with OSM and MG (Fig. 2B). However, as shown in Fig. 2B, the cells barely expressed hepatocyte markers including albumin, carbamoylphosphate synthetase I (CPSI), phosphoenolpyruvate carboxykinase (PEPCK), and tryptophan 2,3-dioxygenase (Tdo2). Thus, hepatocytic characteristics could not be induced in adult cholangiocytes.

To investigate whether cholangiocytes have the potential to differentiate into hepatocytes during the early stage of bile duct formation, we applied the same culture conditions to neonatal cholangiocytes isolated from 1W liver. Similar to adult cholangiocytes, neonatal cholangiocytes continued to express cholangiocyte markers during culture on Col-I gel (supplementary material Fig. S4). As shown in Fig. 2A, neonatal cells cultured on gelatin proliferated and formed a monolayer in which the cells were in close contact with each other. After addition of OSM to the medium on day 5, the cells altered their morphology, developing round nuclei and dense cytoplasm. When the cells were overlaid with MG, cytoplasmic granularity increased. Furthermore, bile canaliculus (BC)-like structures were observed between the cells. During the sequential treatment of OSM and MG, there was increased expression of the genes for albumin, metabolic enzymes including glucose 6-phosphatase (G6Pase), PEPCK, tyrosine aminotransferase (TAT), Tdo2, CPSI and cytochrome P450 proteins (Cyps) (Fig. 2B). We also examined expression of cholangiocyte markers including cytokeratin (CK) 7, CK19 and EpCAM, and found that CK7 and EpCAM were downregulated during hepatocytic differentiation (Fig. 2B and supplementary material Fig. S5). Immunocytochemical analysis showed that albumin and CPSI proteins, which were not expressed in neonatal cholangiocytes at the beginning of the culture period, were expressed in the cytoplasm after inducing hepatocytic differentiation (Fig. 2C3; Figs 4, 7, 8), whereas both proteins were not induced in adult cholangiocytes (Fig. 2C1; Figs 2, 5, 6). However, EpCAM was not downregulated in adult cholangiocytes but was in neonatal ones during culture (Fig. 2C9–12). To examine whether cells treated with MG acquired differentiated functions, ammonium chloride was added to the culture medium. The concentration of ammonium ions in the medium gradually decreased with the time in the wells of cultured neonatal cholangiocytes but not in those of adult cells (Fig. 2D). Finally, to confirm whether BC-like structures were generated, we added fluorescein diacetate (FDA) to the culture medium after augmenting formation of BC-like structures in the presence of taurocholate (Fu et al., 2011). We found that metabolized fluorescein was excreted into BC-like structures (Fig. 2E). These data indicate that cholangiocytes possessed the ability to convert into functional hepatocytes during the neonatal period.

Transcription factors have been shown to determine and convert the lineages of many types of cells. At the time when hepatoblasts are committed to cholangiocytes, transcription factors related to hepatocytic differentiation, including HNF4α and CCAAT/enhancer binding protein α (C/EBPα), are suppressed, whereas those related to cholangiocytic differentiation are upregulated (Tanimizu and Miyajima, 2004; Yamasaki et al., 2006). Therefore, we tested the possibility that the expression patterns of these transcription factors differ between neonatal and mature cholangiocytes. We focused on HNF4α and C/EBPα, because both of these are crucial for the differentiation and/or maturation of hepatocytes (Parviz et al., 2003; Mackey and Darlington, 2004). Using quantitative PCR, we examined the expression of HNF4α and C/EBPα in neonatal and mature cholangiocytes during culture for hepatocytic differentiation. We also examined the expression of FoxA1 (HNF3α), which has been shown to be a crucial factor conferring hepatocytic characteristics on multipotent as well as somatic cells (Sekiya et al., 2009; Sekiya and Suzuki, 2011). HNF4α and C/EBPα genes were clearly induced in neonatal but not in mature cholangiocytes, whereas FoxA1 was expressed in both cell types (Fig. 3A). These results suggest that the efficient induction of HNF4α and C/EBPα is necessary for cholangiocytes to convert into hepatocytes. Immunofluorescence analysis further confirmed that HNF4α and C/EBPα were induced in neonatal cholangiocytes but not in adult ones after inducing hepatocytic differentiation (Fig. 3B).

To examine whether HNF4α and C/EBPα could induce hepatocytic characteristics, we introduced their cDNAs into mature cholangiocytes using retroviral vectors. Cholangiocytes induced with HNF4α or C/EBPα were sequentially treated with OSM and MG. Both HNF4α and C/EBPα slightly increased expression of albumin, whereas only C/EBPα upregulated CPSI (Fig. 3C).

Because the Notch signaling pathway has been implicated in cholangiocyte differentiation of hepatoblasts and hepatocytes (Tanimizu and Miyajima, 2004; Zong et al., 2009), we considered a possibility that constitutive activation of the pathway might inhibit hepatocytic differentiation of adult cholangiocytes. Therefore, we also examined whether inhibition of the Notch pathway by adding 3,5-difluorophenylacetyl-L-alanyl-L-2-phenylglycine t-butyl ester (DAPT), a γ-secretase inhibitor that potentially blocks the Notch signaling pathway (Sastre et al., 2001), could induce the hepatocytic differentiation of mature cholangiocytes. The DAPT treatment slightly decreased expression of Hes1, one of major targets of the Notch pathway, whereas expression of albumin was significantly increased in the presence of DAPT (supplementary material Fig. S6).

Next, we examined whether overexpression of C/EBPα and inhibition of the Notch pathway have an additive effect on hepatocytic differentiation. As shown in Fig. 3D, albumin and CPSI were induced to a greater extent by a combination of DAPT and C/EBPα expression than by the treatment of either of them alone. Tdo2 and Cyp2d10 were slightly induced by the combination of DAPT with C/EBPα. Although the level of expression of hepatocyte markers was much lower than in MHs, C/EBPα expression affected the differentiation status of mature cholangiocytes.

Overexpression of C/EBPα only slightly promoted hepatocytic differentiation. Therefore, we assumed that molecular machinery strongly stabilizing the cholangiocyte lineage might exist in adult cholangiocytes. As candidates of inhibitory factors, we examined expression of cholangiocytic transcription factors including Sox9, hairy-enhance of slit 1 (Hes1), Hey1 and grainyhead-like 2 (Grhl2) that we identified as cholangiocyte-specific transcription factors (Senga et al., 2012). Their expression was higher in adult cholangiocytes than in neonatal cells (Fig. 4A). Furthermore, Grhl2 and Hes1 were maintained at lower levels in neonatal cells than in adult cells during culture (Fig. 4B). Interestingly, Grhl2 expression was further inhibited in neonatal culture after inducing hepatocytic differentiation by sequential treatment with OSM and MG. Downregulation of Grhl2 in neonatal cholangiocytes and its continuous expression in adult cells during the culture were further confirmed by immunofluorescence analysis (supplementary material Fig. S7). Therefore, we considered the possibility that constant expression of Grhl2 in adult cholangiocytes might inhibit hepatocytic differentiation.

To test this hypothesis, we introduced Grhl2 into neonatal cholangiocytes and induced hepatocytic differentiation, and found that Grhl2 inhibited induction of hepatocyte markers (Fig. 4C). We further confirmed that Grhl2 blocked expression of albumin, CPSI, HNF4α, and C/EBPα proteins induced by OSM and MG (Fig. 4D). Moreover, the downregulation of Grhl2 by short interfering RNAs (siRNAs) in adult cholangiocytes slightly induced hepatocytic characteristics (supplementary material Fig. S8). These results suggest that maintenance of Grhl2 at a high level is a crucial factor fixing adult EpCAM⁺ cells in the cholangiocyte lineage.

Given that Grhl2 is implicated in maturation of cholangiocytes (Senga et al., 2012), we considered the possibility that neonatal and adult cholangiocytes might be different in terms of their maturation status as epithelial cells, although bile duct structures are formed in neonatal liver (Fig. 5A). To examine epithelial characteristics of cholangiocytes, we cultured them to develop monolayers, and first measured transepithelial resistance (TER). In the culture condition used here, cholangiocytes formed a monolayer during 2 days of incubation. During and after the formation of the monolayers by neonatal and adult cholangiocytes, values of TER increased and reached a plateau (supplementary material Fig. S9). After 4 days of incubation, the monolayer of adult cholangiocytes showed the higher TER value than that of neonatal cells (Fig. 5B). We also examined the efflux of 4 kDa fluoresceinisothiocyanato-dextran (FITC-dextran) and found that FITC-dextran passed through the monolayer derived from neonatal cholangiocytes more readily than through that of adult cells (Fig. 5C). These results indicated that neonatal cholangiocytes formed relatively immature tight junctions (TJs) compared with adult cells.

As we previously reported, maturation of TJs promotes epithelial morphogenesis, which could be correlated with enlargement of the apical lumen of cysts formed in three-dimensional culture of epithelial cells (Senga et al., 2012). After 10 days of three-dimensional culture, about 1% of neonatal and adult cholangiocytes formed cysts with a central lumen. However, the lumen size of neonatal cysts was significantly (P<0.0001) smaller than that of adult cysts, further suggesting that neonatal cells form relatively immature TJs compared with adult ones (Fig. 5D). These results indicate that neonatal cholangiocytes are immature epithelial cells.

In this study, we demonstrated that cholangiocytes possess the ability to convert into hepatocytes in the neonatal period but this capability is lost in the adult. Similarly, it has been demonstrated that pancreatic duct cells have the potential to differentiate into endocrine and exocrine cells in the neonatal period but their differentiation potential becomes limited in the adult (Kopp et al., 2011). Thus, tubular epithelial cells may generally lose lineage plasticity during postnatal development.

Although, as we mentioned above, it has been shown that neonatal pancreatic duct cells lose the capability to differentiate to multiple types of cell during development, it is not known how the plasticity of epithelial cells is limited. We unexpectedly found that neonatal cholangiocytes are still developing epithelial characteristics even after forming the tubular structure. It can be assumed that production of bile by neonatal hepatocytes is less than that by mature ones and, therefore, relatively immature TJs in neonatal livers are sufficient to prevent the leakage of bile to the parenchyma and/or to the blood vessels, including the portal vein and the hepatic artery. This assumption seems to be consistent with the fact that the accumulation of bile in the neonatal gallbladder is much less than in the adult one (supplementary material Fig. S10). Furthermore, we showed that Grhl2 was expressed at a higher level in adult than in neonatal cholangiocytes and could inhibit hepatocytic differentiation. As we previously demonstrated, Grhl2 promotes formation of functional TJs by establishing a molecular network among claudin 3, claudin 4 and Rab25 (Senga et al., 2012). Thus, our results suggest that the molecular machinery that establishes the epithelial integrity limits the differentiation potential of epithelial cells and thereby stabilizes the lineage of the cells.

It was recently shown that transcription factors could covert fibroblasts into pluripotent stem cells or other types of somatic cells (Yamanaka and Blau, 2010; Yang, 2011). The combination of Gata4, HNF1α and FoxA1, or that of HNF4α plus FoxA1, A2 or A3, was able to convert mouse skin fibroblasts to hepatocytes (Huang et al., 2011; Sekiya and Suzuki, 2011). Because these proteins are strongly expressed in MHs but not in cholangiocytes, we considered the possibility that their expression status is a key to determining the potential for hepatocytic differentiation. In addition to these transcription factors, we focused on C/EBPα, which is also important for the functions of MHs (Inoue et al., 2004). During the course of hepatocytic differentiation, neonatal cholangiocytes expressed FoxA1, HNF4α and C/EBPα. Adult cholangiocytes, however, expressed FoxA1 but neither HNF4α nor C/EBPα. To elucidate the difference in induction, we examined epigenetic modification of the promoters of HNF4α and C/EBPα. Compared with hepatocytes, methylation of CpG sequences increased in cholangiocytes (supplementary material Fig. S11). However, there was little difference between 1W and 6W cholangiocytes. Other epigenetic mechanisms or upstream factors may regulate the expression of HNF4α and C/EBPα in hepatic epithelial cells. Although C/EBPα expression was effective in conferring hepatocytic characters on cholangiocytes, the level of induction was limited. This indicates that other factors may block lineage conversion. The present study suggests that Grhl2 is one such inhibitory factor.

Although Grhl2 did not affect expression of C/EBPα mRNA, it did block induction of C/EBPα protein during hepatocytic differentiation (supplementary material Fig. S12; Fig. 4), suggesting that Grhl2 or its target inhibits translation of C/EBPα. However, downregulation of Grhl2 alone did not markedly induce expression of C/EBPα and hepatocytic differentiation in adult cholangiocytes. This result indicates that other molecules might be involved in regulating those processes. Nevertheless, when upregulation of C/EBPα and downregulation of Grhl2 simultaneously occurred in adult cholangiocytes, hepatocytic markers were further upregulated and some cells expressed albumin and CPSI proteins (supplementary material Fig. S13). Moreover, we demonstrated that the inhibition of the Notch pathway by DAPT was effective in inducing hepatocytic characteristics in adult cholangiocytes, although DAPT treatment only slightly upregulated Hes1. Given that the Notch pathway could regulate the lineage of hepatic epithelial cells independently of Hes1 (Jeliazkova et al., 2013), other targets of the pathway may be also involved in conferring hepatocytic characteristics in adult cholangiocytes. Taken together our results suggest that to induce hepatocytic differentiation in adult cholangiocytes, we may need to not only promote expression of hepatocytic transcription factors but also inhibit cholangiocytic factors and the Notch pathway.

In summary, we demonstrate here that cholangiocytes alter their lineage plasticity during epithelial maturation. We identified a possible molecular network augmenting epithelial structures and functions, which also contributes to stabilization of the epithelial cell lineage by blocking conversion to other lineages. Our results suggest that it is not easy to convert the mass of mature cholangiocytes to hepatocytes; however, several groups have reported that hepatocytes can be produced from pluripotent stem cells or somatic cells (Si-Tayeb et al., 2010; Huang et al., 2011; Sekiya and Suzuki, 2011). Although induced hepatocytes differentiate to functional hepatocytes in diseased mice, it is still difficult to control the process of hepatocytic differentiation of pluripotent and somatic cells and produce a mass of MHs in vitro. Neonatal cholangiocytes have a remarkably strong ability to convert into hepatocytes, so for pluripotent cells to achieve the differentiation status of these cells would be an important step in the differentiation process. We have successfully expanded human cholangiocytes isolated from adult human liver tissue in the same culture conditions as used for mouse cholangiocytes (supplementary material Fig. S3). In addition, cholangiocytes isolated from extrahepatic bile ducts and the gallbladder of adult mice could proliferate efficiently in the same culture conditions (data not shown). Therefore, if we could find a factor that reverts mature cholangiocytes to the differentiation status of neonatal ones, it may be possible to produce functional hepatocytes that can be used as a source of cell therapy and for drug screening.

Col-I (3 mg/ml) was purchased from Koken Co., Ltd (Tokyo, Japan). Growth factor-reduced Matrigel® (MG), which contains extracellular matrix proteins including type IV collagen, laminin-111 and nidogen, was purchased from BD Biosciences (Bedford, MA). Epidermal growth factor (EGF), hepatocyte growth factor (HGF) and OSM were purchased from R&D Systems (Minneapolis, MN).

One-week (1W)- and 6-week (6W)- old mice (C57BL6, Sankyo Lab Service, Japan) were used to isolate neonatal and adult cholangiocytes, respectively. All the animal experiments were approved by the Sapporo Medical University Institutional Animal Care and Use Committee and were carried out under the institutional guidelines for ethical animal use. A two-step collagenase perfusion method was performed through the portal vein of adult mice or through the left ventricle of neonatal mice to digest liver tissues. After the removal of parenchymal cells, the residual material including bile ducts was further digested with Liberase TM (Roche Applied Sciences, San Diego, CA) for neonatal tissues or with collagenase/hyaluronidase solution for adult tissues. Enzymatic digestion was terminated by adding ice-cold fresh medium containing 10% fetal bovine serum (FBS).

The cell suspension was passed sequentially through a 100-µm mesh and a 70-µm cell strainer (BD Biosciences). Nonspecific binding of antibodies was blocked by an antibody against the Fcγ receptor (anti-CD16/CD32 antibody; BD Biosciences). Cells were incubated with biotin-conjugated anti-EpCAM antibody (BioLegend, San Diego, CA) followed by streptavidin microbeads (Miltenyi Biotec, Gladbach, Germany). EpCAM⁺ cells were purified through a MACS column (Miltenyi Biotec). Twenty thousand cells were placed in each well of a 12-well plate. For culture on Col-I gel or MG, collagen type IAC (Koken) mixed with 10× reconstitution buffer containing 200 mM HEPES, 50 mM NaOH, 260 mM NaHCO₃, 10× Dulbecco's modified Eagle's medium (DMEM) and PBS or MG was added to each well. To coat wells with collagen type IAC or MG, these agents were diluted in 0.1 M CH₃COOH and DMEM/F12 medium, respectively, and 500 µl of solution was added to each well. The cells were cultured in DMEM/F12 medium supplemented with 10% FBS, 10 ng/ml EGF and HGF, 5×10⁻⁸ M dexamethasone (Dex; Sigma Chemical Co., St. Louis, MO) and 1× insulin-transferrin-selenium (ITS; Gibco, Carlsbad, CA). After 5–7 days in culture, cells were dissociated from the dishes and then used for subculture (supplementary material Fig. S1).

Human liver tissue was obtained from a patient who underwent hepatic resection at Sapporo Medical University Hospital, with informed consent and the approval of the Sapporo Medical University Ethics Committees. The liver tissue was digested by a method reported previously (Sasaki et al., 2008). Cholangiocytes were isolated from the remaining tissue using the same protocol as that used for the isolation of mouse cholangiocytes and then purified through an MACS column with FITC-conjugated anti-human EpCAM (BioLegend) and anti-FITC microbeads (Miltenyi Biotec).

After culture on Col-I gel, cholangiocytes were dissociated from the gel and 5×10⁴ cells were cultured in each well of 24-well plates coated with gelatin. After the cells became confluent, they were incubated with 20 ng/ml OSM, 1% DMSO, 10⁻⁷ M Dex, and 1× ITS for 4 days and then overlaid with 5% MG for an additional 4 days.

To examine the ability to eliminate ammonium ions from the culture medium, NH₄Cl was added to the culture medium at 2 mM. The concentration of ammonium ions was measured every 2 hours by using the Ammonia Test Wako (Wako Pure Chemical Industries, Osaka, Japan).

To enhance the formation of bile canaliculus (BC) structures in the colonies, 100 µM taurocholate (Tokyo Chemical Industry Co. Ltd, Tokyo, Japan) was added to the medium for 1 day. The formation of BC-like structures was confirmed by incubation with 10 µg/ml fluorescein diacetate (FDA; Sigma-Aldrich, St. Louis, MO) for 30 minutes. The accumulation of metabolized fluorescein into BC-like structures was examined.

cDNAs of C/EBPα, HNF 4α and Grhl2 were amplified by PCR and inserted into retroviral vectors to generate pMXsNeo-C/EBPα, pMXsPuro-HNF4α and pMXsNeo-Grhl2. Retrovirus was added to the culture 48 hours after starting the culture on Col-I gel. For the control, pMXsNeo or pMXsPuro was introduced to cholangiocytes. G418 (1 mg/ml) or puromycin (10 µg/ml) was added to the culture 24 hours after infection to select cells with pMXsNeo-C/EBPα or pMXsNeo-Grhl2, and pMXsPuro-HNF4α, respectively. After incubation in the presence of antibiotics for 24 hours, cells were incubated in medium without them for 2 or 3 days before replating onto gelatin-coated dishes.

Total RNA was isolated from purified EpCAM⁺ cells using an RNeasy Mini Kit (Qiagen, Hilden, Germany). cDNA was synthesized using an Omniscript Reverse Transcription Kit (Qiagen). Primers used for PCR are shown in supplementary material Table S1.

Cholangiocytes induced to differentiate or colonies derived from EpCAM⁺ cells were fixed in PBS containing 4% paraformaldehyde (PFA) at 4°C for 15 minutes. After permeabilization with 0.2% Triton X-100 and blocking with Blockace (DS Pharma, Biomedical Co. Ltd, Osaka, Japan), cells were incubated with primary antibodies (supplementary material Table S2). Signals were visualized with Alexa-Fluor-488, -555 or -633-conjugated secondary antibodies (Molecular Probes, Carlsbad, CA). Nuclei were counterstained with Hoechst 33258. Images were acquired with a Nikon X-81 fluorescence microscope.

Fifty thousand cholangiocytes dissociated from the Col-I gel were replated on a 12 mm Transwell with a 0.4 µm pore, polyester membrane coated with Col-I gel, which was placed in a 12-well plate (Corning Inc., Corning, NY). TER was measured directly in the culture medium using a Millicell-ERS epithelial Volt–Ohm meter (Millipore, Billerica, MA) during the culture. The TER values were calculated by subtracting the background TER of blank filters, followed by multiplying by the surface area of the filter (1.12 cm²). For the paracellular tracer flux assay, 4 kDa FITC-dextran (Sigma-Aldrich) was added to the medium inside the Transwell dish on day 4 at a concentration of 1 mg/ml. After incubation for 2 hours, an aliquot of medium was collected from the basal compartment. The paracellular tracer flux was determined as the amount of FITC-dextran in the basal medium, which was measured with an Infinite M1000 Pro multi-plate reader (Tecan Group Ltd, Mannedorf, Switzerland).

Neonatal and adult cholangiocytes were cultured in gel containing Matrigel® as previously reported (Tanimizu et al., 2007). Briefly, cholangiocytes were dissociated from Col-I gel and 5,000 cells were replated on the mixture of Matrigel® and type I collagen (1∶1 v/v) in a well of an 8-well coverglass chamber (Nunc, Roskilde, Denmark) covered with 5% Matrigel®. After 5 minutes of incubation, cells were fixed and used for immunofluorescence analysis.

We thank Ms Minako Kuwano and Ms Yumiko Tsukamoto for technical assistance.