Molecular mechanisms regulating the establishment of hepatocyte polarity during human hepatic progenitor cell differentiation into a functional hepatocyte-like phenotype

As for all epithelial cells, hepatocytes must be polarized to be functional in the adult liver. That is, the correct function of the liver is ensured by the establishment and maintenance of hepatocyte polarity. The polarization of hepatocytes involves the formation of functionally distinct apical and basolateral plasma membrane domains. The apical poles of front-facing and adjacent hepatocytes form a continuous network of bile canaliculi (BC), which is in contact with the external environment, into which bile is secreted. The basal membrane domain (sinusoidal), which is in contact with the blood, secretes various components into the circulation and is responsible for the uptake of recycled biliary salts. In polarized hepatocytes, the tight junctions (TJs) create the border between apical and lateral poles (Wang and Boyer, 2004; Decaens et al., 2008). Perturbation or loss of polarity is a hallmark of many hepatic diseases, e.g. cholestasis. A complete understanding of the molecular mechanisms involved in hepatocyte polarization is therefore of considerable significance to both liver cell biology and the pathogenesis of liver diseases. However, hepatocyte polarization is still poorly understood.

The establishment of hepatocyte polarity begins during liver embryogenesis. During hepatocyte differentiation, specific routes and mechanisms are defined for the delivery of proteins to the plasma membrane (Wang and Boyer, 2004; Lemaigre and Zaret, 2004; Lemaigre, 2009). Using microarray analysis, several studies have examined the key genes and pathways of potential importance for generating hepatocyte-like cells (HLCs). In these reports, the liver-specific gene expression was identified within the total, heterogeneous population of cells that differentiated from human embryonic stem cells (Chiao et al., 2008; Synnergren et al., 2010; Jozefczuk et al., 2011) and induced pluripotent stem cells (Jozefczuk et al., 2011), HepaRG liver progenitor cells (Parent and Beretta, 2008) or human adipose tissue-derived stromal cells (Bonora-Centelles et al., 2009; Saulnier et al., 2010). Although these systems have limitations inherent to their respective origins, they represent human models of hepatocyte differentiation. To gain the most specific insight possible into the molecular events driving the establishment of hepatocyte polarity, analyses should be performed on purified cells with specific lineage markers. Human fetal hepatic progenitor cells (hFHPCs) were isolated based on alpha-fetoprotein (AFP) promoter expression from fetus that were aborted during the first trimester (Wang et al., 2008). They express AFP, albumin (ALB), cytokeratin 19 (CK19), are able to proliferate long-term in vitro (Wang et al., 2008) and can differentiate into functional HLCs in vitro (Zhang et al., 2012). Thus, hFHPCs are a potential source and a useful model system to study the mechanisms that regulate the process of hepatocyte differentiation.

In this study, we used a two-stage differentiation protocol to generate functional and polarized HLCs from hFHPCs and performed global gene expression profiling on undifferentiated hFHPCs, immature-HLCs, and mature-HLCs. By comparing differential gene expression profiles of different stages of differentiation, a common pool of genes that serve as regulators of hFHPCs differentiation were identified. Subsequently, the genes involved in cell morphogenesis were further investigated for their molecular functions and their roles in the interactive network of genes associated with hepatocyte polarity. We identified PTK7, PARD3, PRKCI and CDC42, known to play key roles in the establishment or maintenance of cell polarity, as genes that were regulated during hFHPCs differentiation. CDC42 may be a core factor in regulating hepatocyte polarization. The expression of several of these genes was confirmed using real-time RT-PCR. The function of two genes (PTK7 and PARD3) in the formation of hepatic polarity was further explored by inactivating their expression using small interfering RNA (siRNA) technology.

We previously established a protocol to isolate fetal hepatic progenitor cells based on AFP promoter expression from human aborted fetus, and produced hFHPCs using this protocol (Wang et al., 2008). hFHPCs express the hepatic stem cells/progenitor cells marker AFP, ALB, CK19, epithelial cell adhesion molecule (EpCAM) (Schmelzer et al., 2006; Inada et al., 2008), Delta-like protein (Dlk) (Yanai et al., 2010), and sal-like protein 4 (Sall4) (Oikawa et al., 2009) (supplementary material Fig. S1A). In this study, we used a two-stage differentiation protocol to differentiate of hFHPCs into functional HLCs. In the first stage, hFHPCs were induced to become immature hepatic cells by 5 days of HGF treatment. In the second stage, the immature hepatic cells were further matured by the combination of HGF, DEX and OSM treatment for another 5 days. The majority of the differentiated cells at day 5 showed an epithelial morphology with binucleated centers, expressed lower levels of AFP, and expressed higher levels of ALB and CK8 than at day 0. The differentiated cells at day 10 exhibited typical hepatocyte morphology with a polygonal shape, containing distinctly round nuclei with one or two prominent nucleoli and intercellular structures resembling hepatic canaliculi. They did not express AFP, but the levels of ALB and CK8 significantly increased over day 5 (Fig. 1A; supplementary material Fig. S1B).

To further characterize the cells, glycogen storage, ALB secretion and CYP450 enzyme activity were detected in cells at different time points. As shown in Fig. 1B, nearly all the cells at day 10 stained positive for periodic acid-Schiff (PAS), whereas no positive staining was visible in the progenitor cells and the HGF-induced cells. Similar to PAS staining, the ALB secretion rate from cells at day 5 and day 10 were significantly upregulated compared to that at day 0 (Fig. 1C). As expected, ∼90±2.5% of the differentiated cells at day 10 exhibited CYP2B1/2 activity, compared to 10±4.3% of the cells at day 5 (Fig. 1D). The CYP1A1 and CYP1A2 activities in the cells at day 10 were greater than the equivalent activities in the cells at day 5 (Fig. 1E). These results suggest that this differentiation protocol gives rise to functional HLCs.

In parallel to functional differentiation, hFHPCs undergo cell morphogenesis and may acquire polarity. To determine whether hepatocyte polarity is established in this differentiation process, the assembly of TJs was analyzed at different time points. The localization of TJ-associated protein ZO-1 and F-actin were detected using anti-ZO-1 and FITC-conjugated phalloidin and visualized with confocal microscopy. As shown in Fig. 2A, we examined sections from day 0, day 5 and day 10 cultures in both the x-z and x-y planes. In x-y confocal images of hFHPCs (day 0), F-actin drew a network through the cells. In x-z cross-sections through the cells, small spots of ZO-1 were observed along an individual cell-cell border (Fig. 2A, left panel, arrow). After HGF stimulation for five days, F-actin was strongly concentrated at cell-cell boundaries, and in x-z confocal images, ZO-1 appeared as punctate spots at sites of cell-cell contact throughout the cell layer (Fig. 2A, middle panel, arrow). At day 10, thick F-actin bundles around the cells were increased in abundance and size; ZO-1 protein was concentrated along cell-cell borders in a linear pattern. At the same time, in x-z confocal images, ZO-1/F-actin double-staining spots increased at sites of cell-cell contact (Fig. 2A, right panel, arrow). The localization of another TJ-associated protein, junctional adhesion molecule A (JAM-A) (Braiterman et al., 2007; Paris et al., 2008), was determined later. Fig. 2B shows that JAM-A was predominantly expressed at the border of two adjacent cells at day 10. Moreover, ultra-thin sections displayed the TJ belt in hFHPC-derived HLCs (Fig. 2C). These results raise the possibility that TJs form between the hFHPC-derived HLCs during differentiation.

To test whether these morphogenetic features represent the acquisition of functional TJs, we evaluated the barrier function of TJs, the transepithelial electrical resistance (TER) and the paracellular fluxes. The values of TER in cells at day 5 and day 10 were significantly increased compared to the undifferentiated cells (day 0) (Fig. 2D). To determine the dynamic function of TJs, apical-to-basolateral FITC-dextran leakage across cultures was examined upon differentiation. In contrast to TER, which is an instantaneous measure of TJ functional integrity, FITC-dextran diffusion is measured over a period of 30 min and therefore may be a more sensitive measure of apical-to-basolateral leakage. The results reveal that permeability declined from day 6 after onward. Paracellular flux rates of FITC-dextran in differentiated cells at day 5, day 6, and day 8 were 101%, 81% and 69%, respectively, relative to that of the cells at day 0 (Fig. 2E).

To examine whether the hFHPC-derived HLCs formed BC-like structures, their ultra-structures were studied in cultures at day 10. Ultrathin sections displayed BC with microvilli (Fig. 3A, arrow) among the hFHPC-derived HLCs. To evaluate the functional activity of drug transporters in HLCs, 5(6)-carboxy-2,7-dichlorofluorescein diacetate (CDFDA) was added into the cultures at day 10 in the absence or presence of probenecid, an inhibitor of the multidrug resistance-associated protein (MRP) 2-mediated transport of 5 (and 6)-carboxy-2,7-dichlorofluorescein (CDF) in hepatocytes (Zamek-Gliszczynski et al., 2003). Fig. 3B shows that the fluorescent CDF was presented in the BC lumen among hFHPC-derived HLCs in the absence of probenecid. A significant decrease in CDF was observed in the BC lumen in the presence of probenecid, suggesting that the excretion of CDF in hFHPC-derived HLCs was mediated by MRP2, a transporter in the apical pore of hepatocytes that functions canalicular secretion (Fig. 3C). Quantitative comparisons of gene expression reveal that the mRNA level of phase I/II enzymes as well as that of drug transporters in hFHPC-derived HLCs were significantly higher than in hFHPCs (Fig. 3D). Fig. 3E demonstrates the basal surface staining of the organic anion transporting polypeptide (OATP) (König et al., 2000) in hFHPC-derived HLCs. These results suggest that this differentiation protocol gives rise to functional polarized HLCs in a way that resembles natural hepatocyte development.

To obtain an initial perspective on global gene expression changes, we performed a pairwise comparison of gene expression microarray data based on the stage of hFHPCs differentiation. A Volcano Plot filtering identified two major transitions in the gene expression patterns (supplementary material Fig. S2): undifferentiated hFHPCs to immature HLCs (stage I) and immature HLCs to mature HLCs (stage II). The microarray revealed that 1780 genes were significantly affected (814 upregulated and 966 downregulated) in stage I; 1835 genes were significantly affected (970 upregulated and 865 downregulated) in stage II. The regulated genes among the lists derived from the two stages were compared by a Venn diagram, which led to the classification of three gene sets: 514 significantly regulated genes were shared by the two stages; 1266 genes were unique to stage I; 1321 regulated genes were unique to stage II (supplementary material Fig. S3A). The 514 overlapping genes were classified into four gene subsets by the hierarchical clustering (supplementary material Fig. S3B; Table S2).

Functional gene annotation analysis of differentially expressed genes was performed according to the gene subsets to explore the functions of the regulated genes in hFHPCs differentiation. Analysis of the significantly upregulated genes found in stage I led to the identification of functional groups, such as ‘cell adhesion’, ‘extracellular matrix organization’, and ‘anatomical structure morphogenesis’ (Fig. 4A; supplementary material Table S3-1). The downregulated gene functional groups in stage I included ‘cell cycle’, ‘organelle organization’, and ‘cellular metabolic process’ (Fig. 4B; supplementary material Table S3-2). A similar analysis was performed for regulated genes found in stage II, and the upregulated gene functional groups included: ‘alcohol metabolic process’, ‘glycolysis’, and ‘cellular amino acid metabolic process’ (Fig. 4C; supplementary material Table S4-1). The functional classification of the downregulated included groups: ‘cell-cell signalling’, ‘anatomical structure morphogenesis’, and ‘cellular developmental process’ (Fig. 4D; supplementary material Table S4-2). The ten most significantly overrepresented annotations within the functional classification of the regulated genes shared by the two stages are listed in supplementary material Table S5.

Taken together, these results suggest that the gene expression patterns of undifferentiated hFHPCs, immature HLCs and mature HLCs are in keeping with the changes in cell phenotypes, and reflecting the specific effects of regulators on hepatocyte differentiation.

To gain a better understanding of the molecular mechanisms that regulate hepatocyte polarity generation during hFHPCs differentiation, the differentially expressed genes within the defined GO categories (GO 0000902: cell morphogenesis; GO 0022604: regulation of cell morphogenesis; GO 0010769: regulation of cell morphogenesis involved in differentiation) were selected for further examination. We found 63 genes (40 upregulated, 23 downregulated) in stage I and 69 genes (29 upregulated, 40 downregulated) in stage II, and 19 genes were common between the two stages (supplementary material Table S6). Among the common genes, six genes (ADM, APOE, CDC42EP3, EFNB1, FBLIM1 and THY1) were upregulated, whereas IL6, NOX4 and RAC2 were downregulated, in both stages; eight genes (ANTXR1, DFNB31, GDF7, ITGB4, LIMA1, MBP, MYH10 and TGFB3) were upregulated in stage I and downregulated in stage II. In contrast, SPG3A and VEGF were downregulated in stage I and upregulated in stage II (supplementary material Fig. S4).

Establishment of epithelial cell polarity is part of cell morphogenesis, which is driven by the forces that cells produce both internally and exert on neighboring cells. To examine the correlations between the candidate genes and hepatocyte polarization, the candidate genes were further categorized using GO analysis. The candidate genes were involved in ‘establishment or maintenance of cell polarity’, ‘membrane organization’, ‘cell junction organization’, and ‘cytoskeleton organization’ (Table 1). Interestingly, three gene subsets involved in the establishment or maintenance of cell polarity were identified. PTK7, PARD3 and PRKCI were upregulated in stage I, with PTK7 and PRKCI responsible for establishing or maintaining epithelial cell apical/basal polarity. KNTC2, CDC42, CENPA and CCDC99 were downregulated in stage I, with KNTC2, CENPA and CCDC99 responsible for establishing mitotic spindle orientation; CAP2, DOCK2 and NDE1 were downregulated in stage II, all of which are involved in cytoskeleton organization. One gene subset involved in membrane organization (DAB2, APOE, MYH10, PRKCI and ULK1) was upregulated in stage I, all of which are involved in ‘vesicle-mediated transport’. Among the genes involved in cell junction organization, THY1, GJA1, PRKCI and TGFB3 were upregulated in stage I; SMAD3, SMAD7, CDC42 and DLC1 were downregulated in stage I; and, SHROOM2 and TGFB3 were downregulated in stage II. CDC42 and PRKCI are involved in the formation of epithelial TJs. Thus, the GO analysis showed that the candidate genes were assigned to the key biological processes required for cell polarization.

To better visualize the functional interactions at the network level, the candidate genes were submitted to functional network reconstruction using STRING. As shown in Fig. 5, 16 genes (FYN, RHOJ, PLXNB1, DLC1, CDC42, NTN1, CDC42EP3, PARD3, EPHB3, EFNB1, PVRL1, PRKCI, TPM1, RAC2, LIMA1, and MYH10) (Fig. 5A) were clustered together. Within this functional cluster, CDC42, MYH10, PARD3 and PRKCI are involved in TJs, and FYN, CDC42, PARD3 and PVRL1 are involved in adherens junctions. CLASP1, KNTC2 (NDC80), CENPA, CCDC99 and NDE1 participate in maintenance of kinetochore integrity and kinetochore-microtubule attachments (Fig. 5B).

To validate the accuracy of the gene indexes calculated from the microarray, six differentially expressed genes selected based on their roles in hepatocyte function (ABCA1, CPS1 and PCK2) or cell morphogenesis (APOE, CDC42 and PTK7) were confirmed by real-time RT-PCR. Comparison of the microarray and real-time RT-PCR results with their correlation coefficients are presented in Fig. 6. The data reveal that our gene index calculations from the microarray data accurately represented the expression of genes across a wide range of expression levels.

As a proof of principle, PTK7 and PARD3 were selected from the differentially expressed genes involved in hepatocyte polarity based on their unique cellular function. By reducing their expression level using siRNA, their effects on the formation of hepatocyte polarity during hFHPC differentiation were assessed. As shown in Fig. 7A, we successfully reduced the PTK7 and PARD3 expression level by 80–90% compared to the transfection control. The reduction of PTK7 and PARD3 significantly decreased the TER and increased the paracellular fluxes over a period of 30–120 min at 6 days post-transfection (Fig. 7B,C).

To identify morphology changes, hFHPCs with reduced expression of PTK7 and PARD3 were challenged to undergo differentiation into hepatocytes using our two-stage differentiation protocol. As shown in Fig. 7D, cells with inactivated PTK7 and PARD3 were unable to acquire the typical epithelial morphology, which exhibited the effects on F-actin assembly and apical actin concentration. Furthermore, the Golgi apparatus was randomly distributed in the cytoplasm in PTK7 and PARD3 knockdown cells, while the Golgi apparatus was localized between the nucleus and apical pole in control cells. These data demonstrate that reduction of either PTK7 or PARD3 may inhibit hepatocyte polarity formation during hFHPC differentiation.

Generation of membrane polarity in hepatocytes is key developmental program and a multistage process that requires cell-cell interactions and the specific organization of proteins and lipids on the cell surface and in cell interior, which leads to modeling the complex architecture during hepatocyte differentiation (Wang and Boyer, 2004; Takashi et al., 2007). Although information has recently been accumulated suggesting specific routes and mechanisms for this process (Matsui et al., 2002; Michalopoulos et al., 2003; Imamura et al., 2007; Chiao et al., 2008; Bonora-Centelles et al., 2009; Kidambi et al., 2009; Saulnier et al., 2010; Synnergren et al., 2010; Jozefczuk et al., 2011; Fu et al., 2011), the exact mechanism for hepatocyte polarity generation is unclear partly because only a few in vitro models are available for developing complex human hepatocyte polarity (Decaens et al., 2008). The goal of this study was to generate functional polarized hepatocytes from the hFHPCs in vitro and to explore the molecular mechanisms involved in hepatocyte polarization.

First, we found that sequential treatment of hFHPCs with HGF, OSM, and DEX is efficient to generate functional hepatocyte polarity in hFHPC-derived hepatocytes, which mimics hepatocyte differentiation both morphologically and functionally and resembles the natural hepatocyte development (Kinoshita et al., 1999; Kamiya et al., 2001; Zorn, 2008). We observed that hFHPCs were unpolarised with spindle-shaped morphology and possessed higher AFP and lower ALB expression under the initial culture conditions. After HGF treatment, the cells developed a simple polar phenotype with the establishment of functional TJs at the points of cell-cell contact along the cell perimeters, as evidenced by the zonular assembly of F-actin, ZO-1 and JAM-A through the function of TJs (TER and paracellular flux). These results suggest that HGF stimulates a morphogenesis of tight and functional integrity in the early stage of hFHPC differentiation, which is in line with previous reports (Pollack et al., 2004). In stage II, the immature cells were induced with HGF, OSM and DEX, and the multipolar hepatic phenotype gradually manifested itself during this period. Parallel to the observed morphological changes, ALB secretion, bile acid synthesis, glycogen storage and CYP450 enzyme activity accelerated, suggesting that the hepatocytes polarized rapidly. The formation of the BC architecture and the establishment of the hepatobiliary transport system in hFHPC-derived HLCs was further confirmed by electron microscopy and hepatocyte uptake assays. Taken together, our data suggest that a human hepatocyte polarity was gradually generated during hFHPC differentiation in vitro. Therefore, this system may be a useful platform to explore the mechanisms of hepatocyte differentiation and hepatic polarity establishment.

Second, using high-density oligonucleotide microarrays, the global transcriptional profile of hFHPCs and their derivatives was investigated based on the stage of differentiation. We found 1780 genes in stage I and 1835 genes in stage II that were significantly regulated. From comparison of the transcriptional profiles between the two differentiation regimes, three main subsets of regulated genes were identified. Within each subset, the functional profiling of the upregulated and downregulated genes was further investigated using GO analysis. The upregulated genes in stage I were mainly involved in ‘cell adhesion’, ‘extracellular matrix organization’, and ‘anatomical structure morphogenesis’. The upregulated genes in stage II were mainly involved in ‘alcohol metabolic process’, ‘glycolysis’, and ‘cellular amino acid metabolic process’. Results revealed a dramatic switch in functional profile as proliferating hFHPCs differentiated into HLCs, which was in accordance with the cell phenotypes and previous reports on liver development (Lemaigre, 2009; Duncan, 2003; Zorn, 2008). These results present a detailed characterization of a unique set of genes, which can be used to explore the mechanisms of hepatocyte differentiation.

To examine which molecules govern hepatocyte morphogenesis, including polarization during differentiation, a subset of regulated genes (which included 113 genes) annotated as being associated with ‘cell morphogenesis’ in the two stages was selected (supplementary material Table S6) to explore the core regulated genes that are potentially involved in hepatocyte polarization. As expected, our data (Table 1) indicated that these core genes were involved in the key process of polarization, such as in apical/basal cell polarity and junction organization. Among them, PTK7 (Lhoumeau et al., 2011; Puppo et al., 2011), PARD3 (Ooshio et al., 2007; Achilleos et al., 2010; Ishiuchi and Takeichi, 2011), PRKCI (Eder et al., 2005), CDC42 (Harris and Tepass, 2010), and CLASP1 (Miller et al., 2009) have recently been identified to have roles in cell polarity. Our results demonstrated that PTK7 and PARD3 promote the formation of hepatic polarity and affect F-actin organization. The network analysis of the candidate genes, based on known and predicted protein-protein interactions, further illustrated the interactions of the molecules that regulate hepatocyte polarization and clustered them into functional categories. Sixteen genes (FYN, RHOJ, PLXNB1, DLC1, CDC42, NTN1, CDC42EP3, PARD3, EPHB3, EFNB1, PVRL1, PRKCI, TPM1, RAC2, LIMA1 and MYH10) may be the functionally regulated modules for hepatocyte polarity generation.

In conclusion, our results provide insights into understanding hepatocytes polarization at the transcriptional level and reveal novel key genes and a functional network of potential importance for future efforts in exploring hepatic differentiation.

hFHPCs were prepared as described previously (Wang et al., 2008) with the informed patient consent and under the approval of the Ethics Committee of Capital Medical University (Beijing, China) for this project. hFHPCs were maintained in MEM/NEAA (Hyclone, Logan, UT) containing 5% fetal bovine serum (FBS, Hyclone), 100 U/mL penicillin, 100 µg/mL streptomycin, 20 ng/mL epidermal growth factor (EGF, Peprotech, Rocky Hill, NJ) and 10 µg/mL insulin (Sigma-Aldrich, St Louis, MO). Cells from passages nine to eleven were used in the study.

To induce of hepatocyte maturation, the cells were plated at 2×10⁴ cells/cm² in culture plates and dishes, in expansion medium. Once cells reached 95% confluence, they were washed twice with PBS and differentiated in basal medium [MEM/NEAA, supplemented with 0.5 mg/mL albumin fraction V (Sigma-Aldrich), 100 U/mL penicillin, 100 µg/mL streptomycin, and 1% insulin-transferrin selenium (ITS, Sigma-Aldrich)] containing 20 ng/mL hepatocyte growth factor (HGF, Peprotech) for 5 days. Then, the differentiated cells were further matured in basal medium containing 20 ng/mL HGF, 10 ng/mL oncostatin M (OSM, Peprotech) and 10⁻⁷ M dexamethasone (DEX, Sigma-Aldrich) for another 5 days. Differentiation media were changed every 2 days.

The cells were fixed with 4% paraformaldehyde for 20 min at room temperature, followed by permeabilize with 0.3% Triton X-100 in PBS for 5 min and then, rinsed and blocked with 20% goat serum (Zsgb-Bio, China) for 60 min at room temperature. The cells were then incubated with rabbit anti-ZO-1 at 1∶200 (Invitrogen, Carlsbad, CA), mouse anti-JAM-A at 1∶100 (F11R, BD Biosciences, San Jose, CA), or goat anti-OATP at 1∶100 (Santa Cruz Biotechnology, Santa Cruz, CA); rabbit anti-AFP at 1:200 (Product Number A0008, Dako); mouse anti-ALB at 1:500 (clone HAS-11, Product Number A 6684, Sigma-Aldrich); mouse anti-CK8 at 1:200 (Product Number A, C5301, Sigma-Aldrich); mouse anti-EpCAM at 1:400 (Cell Signaling, MA); rabbit anti-Dlk at 1:500 (Abcam, Hong Kong) and rabbit anti-Sall4 at 1:50 (Abgent, San Diego, CA) at 4°C overnight. Following three 5-min washes in PBS with gentle agitation, an Alexa Fluor-conjugated secondary antibody (1∶500; Invitrogen) was added, and the samples were incubated for 60 min at 37°C. Afterwards, the cells were incubated with a mixture of FITC-labeled phalloidin (Sigma-Aldrich) and the nuclear stain DAPI (Sigma-Aldrich) at room temperature for 20 min. The cells were then washed and examined under a Leica TCS SP5 Confocal Microscope (Leica, Wetzlar, Germany). Photomultiplier settings were adjusted to the brightest signal and maintained during data acquisition. Samples were imaged either in the x-y plane or in the x-z plane.

To mark the Golgi apparatus, a plasmid containing the sequence of the Golgi-resident enzyme N-acetylgalactosaminyltransferase 2 (GalNAc-T2) (Röttger et al., 1998) fused to red fluorescent protein (RFP) was transduced into cells using CellLight® Golgi-RFP *BacMam 2.0* (C10593, Invitrogen) (Kost et al., 2005) according to the manufacturers' instructions. Then, the labeled cells were fixed with 4% formaldehyde in PBS for 20 min at room temperature, followed by permeabilization with 0.3% Triton X-100 in PBS for 5 min at room temperature.

For ultra-structural analysis, the differentiated cells were fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer, pH 7.2, for 60 min at 4°C. They were postfixed in 1% osmium tetroxide in 0.1 M phosphate buffer and embedded with the Spurr embedding kit. The sections were examined using an H-7650 electron microscope (Hitachi, Japan).

Intracellular glycogen was analyzed by PAS staining. Cultures were fixed in carnoy fixative at room temperature for 10 min and oxidised in 1% periodic acid for 15 min. After oxidation, the cultures were rinsed three times with distilled water and then treated with Schiff's reagent (Sigma-Aldrich) for 30 min. After the cells were rinsed with distilled water for 5 min, the cell nuclei were counterstained with Mayer's hematoxylin. The PAS-positive cells and whole cells were examined under a Leica light microscope (Leica).

The concentration of ALB secreted into the culture media was analyzed as previously described (Zhang et al., 2012).

To evaluate the activities of CYP1A1 and CYP1A2, ethoxyresorufin-O-deethylase (EROD) and methoxyresorufin-O-deethylase (MROD) assays were performed, respectively, as previously described (Donato et al., 1993; Kidambi et al., 2009). To evaluate the activity of CYP2B1/2, the penthoxyresorufin-O-deethylase (PROD) assay was performed. The differentiated cells were treated with 1 µM pentoxyresorufin and assessed under a Leica TCS SP5 Confocal Microscope (Leica). The PROD activity-positive cells were counted with Image-Pro Plus software (Media Cybernetics Inc., MD).

To determine TER, cells were cultured to confluence in 12-mm Transwells with 0.4-µm pore size filters (Millipore, Billerica, MA). TER was measured in culture medium using a MILLICELL-ERS (Millipore) with ‘chopstick’ electrodes at different time points. For calculation, the resistance of blank filters was subtracted from that of filters covered with cells to give the net resistance, which was multiplied by the membrane area to give the resistance in area-corrected units (Ω·cm²). The data presented as the mean ± s.d.

To determine the paracellular flux, cells were cultured to confluence on 12-mm Transwells with 0.4-µm pore size filters (Millipore) and 2.5 mg/mL FITC-dextran (FD-70s, Sigma-Aldrich) was added to the inner chamber. The apparatus was then placed in a CO₂ incubator at 37°C for 30, 60 and 120 min, respectively. Samples were collected from the outer chamber, and the fluorescent signals were measured at an excitation wavelength of 485 nm and emission wavelength of 535 nm using a spectrophotometer (Thermo Fisher Scientific Inc.). The results are expressed as the concentration of FITC-dextran in the outer chamber (mg/mL). The paracellular FITC-dextran flux rate was calculated as the measured value of the treated group relative to that of the control group.

At least three independent determinations of each parameter were compared among the treatment groups by one-way ANOVA using the statistical software SPSS 11.5. Differences were considered significant if P<0.05.

To determine the BC function, cells were incubated with 10 µM CDFDA (Sigma-Aldrich) in the absence or presence of 4 mM probenecid (Sigma-Aldrich) at 37°C for 10 min to allow its internalization and subsequent translocation into the BC lumen by MRP 2 (Zamek-Gliszczynski et al., 2003; Turncliff et al., 2006). After extensive washes, the capacity of BC to contain the fluorescent CDF was analyzed under a Leica TCS SP5 Confocal Microscope (Leica).

Total RNA was isolated from undifferentiated hFHPCs and their derivatives in triplicate for each time point using TRIzol (Invitrogen) and the RNeasy kit (Qiagen, Hilden, Germany) according to the manufacturers' instructions, including a DNase digestion step. RNA quantity and purity were determined using the ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and denaturing gel electrophoresis. Approximately 5 µg total RNA of each sample was used for labeling with a NimbleGen one-color DNA labeling kit, array hybridization using the NimbleGen Hybridization System (NimbleGen 12×135K microarrays, Roche NimbleGen) and array scanning using the Axon GenePix 4000B microarray scanner (Molecular Devices Corporation, Sunnyvale, CA).

The raw data were extracted and normalized using NimbleScan v2.5 Software. The gene summary files were imported into Agilent GeneSpring Software (version 11.0) for further analysis. Genes that had values greater than or equal to the lower cutoff of 50.0 in all samples were chosen for further analysis. Genes that were differentially expressed with statistical significance were identified through Volcano Plot filtering (fold change ≥2.0 and P-value ≤0.05).

Similarities and differences among the differentially expressed genes at different stages were analyzed by Gene Venn (Hulsen et al., 2008). Differentially expressed genes in these lists were analyzed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) tool (Huang et al., 2008) according to Gene Ontology (GO) roles. To examine potential interactions of genes involved in cell morphogenesis during the hFHPCs differentiation, we applied the Search Tool for the Retrieval of Interacting Genes (STRING) to predict the protein-protein interaction networks (Szklarczyk et al., 2011).

Total cellular RNA was extracted from 1×10⁶ cells with the RNeasy Mini Kit (QIAGEN) according to the manufacturer's instructions. For PCR analysis, 2 µg RNA was reverse-transcribed to cDNA using Superscript II reverse transcriptase and random hexamer primers (Invitrogen). Real-time PCR analysis was performed on an ABI Prism 7300 Sequence Detection System using the SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA). The reaction consisted of 10 µL of SYBR Green PCR Master Mix, 1 µL of a 5 µM mix of forward and reverse primers, 8 µL water, and 1 µL template cDNA in a total volume of 20 µL. Cycling was performed using the default conditions of the ABI 7300 SDS Software 1.3.1. The relative expression of each gene was normalized against 18S rRNA. The primers used are shown in supplementary material Table S1. The data are presented as the mean ± s.d.

hFHPCs were plated at 2×10⁴ cells/cm² in antibiotic-free basal medium 24 hours prior to transfection. siRNA transfection was performed following the manufacturer's protocol. Briefly, ON-TARGET SMARTpool siRNAs directed against PTK7 (L-003167-00-0005, Dharmacon, Lafayette) or PARD3 (L-015602-00-0005) or non-targeting siRNAs (D-001810-10-05) were mixed with Transfection DharmaFECT 4 (Dharmacon), respectively. After a 20-min incubation at room temperature, the complexes were added to the cells at a final siRNA concentration of 50 nM. The medium was replenished with medium containing antibiotic 24 hours post-transfection. Culture medium was then changed every 3 days for the duration of the experiment. hFHPCs transfected with non-targeting siRNAs were used as experimental control.

Experiments were independently repeated three times with replicate samples. Statistical analysis was performed using one-way ANOVA tests using SPSS 11.5 statistics software. Differences were considered significant if P<0.05.

Microarray experiments were performed by KangChen Bio-tech, Shanghai, China.