The iPSC cell lines of iNFB3 and IBMS-iPSC-01-02 (IB0102) were obtained from Professor Hong-Nerng Ho’s laboratory at National Taiwan University Hospital (Huang et al., 2019). The cells were periodically frozen and thawed while being continuously maintained in ReproCELL serum-free medium (ReproCELL, Shinyokohama, Yokohama, Japan). The cells were passaged weekly by using 30-gauge insulin needles (Terumo, Shinjuku, Tokyo, Japan) (Chen et al., 2015). iNFB3 is an iPSC cell line derived from human dermal fibroblasts that were episomally reprogrammed into human iPSCs by using electroporation to introduce OCT4, SOX2, KLF4, and c-MYC, as described previously (Chen et al., 2015). IBMS-iPSC-01-02 (IB0102) is an iPSC cell line derived from human peripheral blood mononuclear cells reprogrammed using Sendai virus vectors to transiently drive the expression of OCT4, SOX2, KLF4, and c-MYC. A human ESC line (H9; karyotype 46, XX; WiCell, WI, United States) (Thomson et al., 1998) was also obtained from Professor Hong-Nerng Ho’s laboratory at National Taiwan University Hospital. These cells were periodically frozen and thawed while being continuously maintained using the same protocol as that for human iPSCs (Chen et al., 2015).

The protocol and detailed materials information on cholangiocyte differentiation from the human pluripotent stem cells in this study were provided in Supplementary Tables S1, S2. In brief, the cholangiocyte differentiation protocol developed in this study was optimized on the basis of previous studies (Dianat et al., 2014; Sampaziotis et al., 2015a; Ogawa et al., 2015; De Assuncao et al., 2015; Takayama et al., 2016; Sampaziotis et al., 2017a; Matsui et al., 2019; Luce and Dubart-Kupperschmitt, 2020; Ogawa et al., 2021; Jalan-Sakrikar et al., 2022). In this protocol, 28 days are required to differentiate mature cholangiocytes from human PSCs. Induction of human PSC–derived definitive endoderm (DE) cells was performed by adding a high concentration of Activin A in stage 1 (days 1–4). Subsequently, DE cells were induced to differentiate into ventral foregut endoderm (VFE) in stage 2 (days 5–8). In stage 3 (days 9–13), VFE was differentiated into hepatoblasts (HBs), which are bipotent progenitors of hepatocytes and cholangiocytes (i.e., capable of developing into either hepatocytes or cholangiocytes). In stage 4 (days 14–17), HBs committed to the biliary lineage, and cholangiocyte progenitors (CPs) were generated as early cholangiocytes. In the final stage, 3D culture system was utilized to generate mature cholangiocytes from CPs until day 28 of the differentiation.

Total RNA was isolated using RNeasy Mini Kit (QIAGEN, Valencia, CA, United States) following the manufacturer’s guidelines. The concentration of total mRNA was assessed using a microvolume spectrophotometer (NanoPhotometer N60, Implen, Munich, Germany). mRNA was then reverse transcribed into cDNA by using Moloney Murine Leukemia Virus (MMLV) reverse transcriptase (Invitrogen, Carlsbad, CA, United States), followed by polymerase chain reaction (PCR) amplification by using the Fast SYBR Green Master Mix (Thermo Fisher Scientific, Waltham, MA, United States) on a real-time (RT) PCR system (LightCycler 96, Roche Diagnostics, Basel, Switzerland). The amplification process comprised initial denaturation at 95°C for 30 s, followed by 35 cycles of denaturation at 95°C for 3 s and annealing/extension at 60°C for 30 s. The GAPDH gene served as an internal control. The primer sequences and experimental conditions employed in this study are detailed in Supplementary Table S3.

Human iPSCs and ESCs were cultured for 4–5 days after passaging and were then evaluated for alkaline phosphatase activity. Subsequently, cells were fixed with 4% paraformaldehyde (Sigma-Aldrich, St. Louis, MO, United States) for 2 min. Alkaline phosphatase activity in both human iPSCs and ESCs was analyzed using an alkaline phosphatase detection kit following the manufacturer’s guidelines (Merck Millipore, Darmstadt, Germany).

Cells were treated with Triple Select Enzyme (1X) without phenol red (Gibco, Grand Island, NY, United States) for 5 min to facilitate their dissociation into single cells. Following neutralization with cell medium, cells were harvested and washed with phosphate-buffered saline containing 5% fetal bovine serum (Gibco). Cell blocking was conducted by incubation with 10% fetal bovine serum for 30 min, followed by incubation with the primary antibody at 37°C for 50 min. Cells were washed twice in phosphate-buffered saline, resuspended, exposed to the secondary antibody for an additional 20 min, and then again washed twice in phosphate-buffered saline. For intracellular markers, cells were fixed with 4% paraformaldehyde (Sigma-Aldrich) for 10 min, permeabilized with 0.5% Triton X 100 for 20 min, incubated with the primary antibody for 50 min, and then incubated with the secondary antibody for 20 min. The stained cells were assessed using a flow cytometer (FACSVerse, BD, San Jose, CA, United States) and quantified using FlowJo (version 10.5.3, Becton Dickinson). Details regarding the antibodies and their experimental conditions are summarized in Supplementary Table S4.

The 2D and 3D culture system–generated cholangiocyte organoid (CO) samples were fixed with 4% paraformaldehyde (Sigma-Aldrich). After fixation, the 2D culture system-generated cholangiocytes were permeabilized for 10 min using 0.1% Triton X 100 containing 1% bovine serum albumin (Sigma-Aldrich), and the 3D culture system–generated CO samples underwent the same permeabilization treatment for 30 min. All samples were then blocked with 3% bovine serum albumin and incubated with the primary antibody overnight at 4°C. On the following day, the samples were washed with phosphate-buffered saline (Sigma-Aldrich) and exposed to the secondary antibody for either 1 h (for 2D culture system–generated cholangiocytes) or 2 h (for 3D culture system–generated CO samples). After washing, the samples were stained with DAPI for an additional 10 min and then stored using an antifade-added cover glass or in a chambered dish to preserve the organoid structure of the cells for subsequent confocal microscopy. CO images were captured using a confocal microscope (STELLARIS 8, Leica Microsystems, Wetzlar, Germany) employing Z-stacking to generate 3D images. The antibodies and experimental conditions employed in this study are detailed in Supplementary Table S4.

Cell extracts were prepared using lysis buffer (10 mM Tris, pH 7.5; 150 mM NaCl; 1 mM EDTA; 0.5% sodium deoxycholate; 0.5% NP-40; and 0.1% SDS) supplemented with a protease inhibitor cocktail (Roche Diagnostics) and phosphatase inhibitor cocktail (Roche Diagnostics). The protein concentration was determined using an assay kit (Pierce BCA, Thermo Fisher Scientific). Equivalent amounts of total protein were separated through 10%–12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and were subsequently transferred to nitrocellulose membranes (Immobilon-P PVDF Membrane, Merck Millipore). These PVDF membranes were then blocked with 5% bovine serum albumin and incubated with specific primary antibodies. Signal detection was accomplished using a biomolecular imager (ImageQuant LAS 4000 mini, GE Healthcare, Uppsala, Sweden). Western blot analysis was performed in accordance with a previous protocol (Peng et al., 2023). The obtained Western blot image was subjected to quantification using the ImageJ software (Schneider et al., 2012). Details regarding the antibodies and their experimental conditions are summarized in Supplementary Table S4.

To investigate the presence of calcium channels, with a focus on multidrug resistance protein 1B (MDR1b) transporter on COs, the CO samples were placed into a chambered dish. Each sample was pretreated with dimethyl sulfoxide (Sigma-Aldrich) and verapamil (ChemScene, New Jersey, United States) in William’s E medium (Gibco) for 20 min and then washed twice in William’s E medium. The samples were then treated with Rhodamine 123 (ChemScene) for 30 min, then washed in William’s E medium (Gibco), and observed using a confocal microscope. Obtained images were subjected to quantification using ImageJ software (Schneider et al., 2012).

The COs were treated with forskolin (10 µM) to evaluate CFTR function in the COs. The dimethyl sulfoxide (DMSO) was used in the control group. The COs were incubated with forskolin for specific durations up to 120 min. Residual forskolin was then removed by a careful washing process, and the abundance, size, and morphology of the COs were examined under a light microscope. Images were taken for quantitative analysis.

Data are presented as mean ± standard error of the mean (SEM), except where indicated otherwise. The data were analyzed using unpaired Student’s t-test or one-way ANOVA, followed by post hoc analysis for pairwise comparisons between groups. Data were analyzed using SPSS (IBM) and visualized using GraphPad Prism 8. Statistical significance was indicated by p < 0.05. Livak (2^−ΔΔCT) method was utilized to determine the genes expression (Livak and Schmittgen, 2001).

The human PSCs used in this study were iPSCs and ESCs. Both the iPSCs and ESCs presented colony morphology and strong alkaline phosphatase activity (Figure 1A). Further analysis using RT quantitative PCR (qPCR) revealed that the expression of the pluripotent transcription factors OCT4, NANOG, and SOX2 was significantly higher in both the human PSCs than in a somatic liver cancer cell line (HepG2215; Figure 1B). Protein concentrations were analyzed through IF staining (Figure 1C) and Western blotting (Figure 1D), with quantitative analysis provided in Figure 1E.

To generate mature cholangiocytes from human PSCs, a 28-day novel induction protocol was established (Figure 2A). The human PSCs were differentiated into primitive streak, mesoderm, and DEs by using Activin A, Ly294002 (an inhibitor of phosphoinositide 3-kinase [PI3K]) and CHIR99021 (an inhibitor of glycogen synthase kinase 3β [GSK3β]). A morphological change and the epithelial-mesenchymal transition revealed the first-stage differentiation of human PSCs into DE (Figure 2B, human PSC to DE). The primary DEs were continuously induced with a high concentration of Activin A and basic FGF (bFGF) until days 3 and 4. RT-qPCR analysis revealed that by days 3 and 4, the expression of the pluripotent transcription factor OCT4 decreased and that the expression of the DE marker SOX17 increased (Figure 2C). At day 4, the expression of SOX17 protein was determined using IF staining (Figure 2D). Successful DE differentiation was further confirmed using flow cytometry, which revealed the double-positive signals of EpCAM⁺ (99.4%) and CXCR4⁺ (95.2%; Figure 2E). An additional surface marker was also employed, the c-KIT (92,7%; Supplementary Figure S1).

The differentiation of DE into VFE was induced using Activin A, a TGF-β superfamily member that determines the cell fate in respiratory and digestive tissues. Moreover, introducing FGF signaling will also enhance the differentiation of DE into VFE. By the end of day 9, cells had formed confluent epithelium, and cells also exhibited a characteristic of rhomboidal morphology (Figure 2B for hPSC iNFB3, Supplementary Figure S2 for hPSC IB0102 and H9) and significantly increased the VFE marker GATA4 (p = 0.026) as revealed by RT-qPCR (Figure 3C, Supplementary Figure S3).

For the differentiation of VFE into HB, the FGF signaling pathway was activated by adding bFGF, which synergistically acted with BMP and Wnt/β-catenin signaling. SB431542 was also added to inhibit the TGF-β pathway and promote HB differentiation (Figure 2A, stage 3: VFE–HB). Cell proliferation decreased, indicating HB differentiation (Figure 2B for hPSC iNFB3, Supplementary Figure S2 for hPSC IB0102 and H9). Concentrations of alpha-fetoprotein (AFP) in the differentiated HBs were determined using IF staining (Figure 3A) and flow cytometry at day 13, with more than 85% positive signals (Figure 3B). The expression of specific markers in the DE, VFEs, and HBs was examined using RT-qPCR. At the end of the VFE stage, we observed a significant increase in the expression of VFE-related genes, such as CLDN6 and GATA4, accompanied by a notable decrease in the expression of the DE-associated gene SOX17. HNF4A and AFP expression was significantly higher in the HBs than in the DE and VFEs (Figure 3C, Supplementary Figure S3). Together, these results highlight the successful differentiation of VFEs into HBs under our culture protocol.

To differentiate the bipotent HBs into CPs, the Notch and FGF signaling pathways were activated by the addition of Activin A, retinoic acid, and FGF10 (Figure 2A, stage 4: HB–CP). Cell proliferation was greater in the CPs than in the HBs (Figure 2B, day 18). The expression of the early cholangiocyte marker SOX9 was determined using RT-qPCR (Figure 4A, Supplementary Figure S3), and IF staining (Supplementary Figure S4A). The 2D culture system–generated CPs expressed the mature cholangiocyte markers CK19 and EpCAM at day 28 (Supplementary Figure S4B). At this time, the HBs were still present in the culture dishes, indicating that the time of inducing the differentiation of HB into CP is a crucial factor of effective biliary commitment.

Several studies have reported that activating Notch signaling can promote cholangiocyte maturation and inhibit hepatocyte differentiation (Tanimizu and Miyajima, 2004; Pasqua et al., 2021). However, we did not observe the formation of tube-like structures or functional cholangiocytes under a similar 2D culture condition (Figure 2B; Cholangiocytes stage).

To generate mature COs, CPs were dissociated into small clumps and cultivated using a 3D culture system. The size of cell clumps in the Matrigel dome and the EGF-supplemented culture medium used in the 3D culture system have critical effects on the efficiency of CO generation. The cell clumps turned into small or large cholangiocytes, forming tube-like structures within 10–15 days of cultivation (Figure 2B; Cholangiocyte stage). The expression of mature cholangiocyte markers and region-specific cholangiocyte markers was determined using RT-qPCR (Figure 4A, Supplementary Figure S5A). Importantly, the generated COs also exhibited the expression of region-specific markers of extrahepatic cholangiocytes (AQP1 and MUC1) and intrahepatic cholangiocytes (YAP1 and JAG1; Figure 4A). The expression levels of each target gene throughout the differentiation process are shown in Supplementary Figure S3.

The expression of the mature cholangiocyte markers CFTR and CK19 in small or large cholangiocytes was determined using IF staining (Figure 4B, Supplementary Figures S5B, S5C). CO generation efficiency was evaluated using the percentage of cells that had a normalized CFTR/DAPI ratio of 72.33% ± 8.7 for the 2D culture system and 77.12% ± 6.2% for the 3D culture system on day 28 (Figure 4C). The COs were further analyzed using 3D reconstruction in the Leica LAS X software package, showing the hollow tube-like organoid (Supplementary Videos S1, S2).

To compare the maturation of generated-COs under 2D and 3D cultures, we evaluated the core biliary markers and region-specific markers of cholangiocytes at day 28 of differentiation. The expression of mature cholangiocyte markers was higher in the 3D culture system–generated cholangiocytes than in the 2D culture system–generated cholangiocytes (Figure 5A). Further, IF staining analysis also revealed the tube-like structure of COs, a formation unattainable in a 2D culture system (Figure 5B). With long-term cultivation, the 3D culture system–generated organoids formed a branching tubular structure after passage (Supplementary Figure S6A), but their proliferative capacity will decrease starting from day 55–60 after passage under prolonged culture evaluated by IF staining for Ki67 (Supplementary Figure S6B).

To determine the functionality of the 3D culture system–generated human PSC–derived COs, the function of P-glycoprotein was examined using rhodamine 123 and forskolin-induced swelling assay. The accumulation of rhodamine 123 within the lumen was indicative of P-glycoprotein activity, which in turn indicated the presence of mature and functional cholangiocytes (Figures 5C, D). Verapamil was used to block the MDR1 transporter; this blocking was confirmed by the absence of accumulation of fluorescence in the cytoplasm of the cholangiocytes (Figures 5C, D, Supplementary Figure S5D).

Furthermore, the CFTR function of the cholangiocyte 3D organoids was further assessed through a rapid and quantitative water-uptake swelling assay induced by forskolin (Ramalho et al., 2022). As depicted in Figure 5E, forskolin was observed to enhance organoid size compared to the DMSO control group. These results strongly demonstrated the successful generation of functional COs.

To examine the potential underlying mechanism in cholangiocyte transdifferentiation, EGF was used in this study as EGF has been known to regulate liver development and regeneration (McLin et al., 2007). We found that under EGF-absent conditions, the expression of the core biliary markers CK19, CK17, and CFTR increased, and after EGF treatment, the expression of the extrahepatic cholangiocyte markers AQP1 and MUC1 significantly increased (Figure 6A).

Further evaluation of core biliary markers, CFTR and CK19, using IF staining confirmed that the addition of EGF enhanced the expression of CFTR, as demonstrated by the 3D cholangiocyte morphology and improved protein expressions of CFTR and CK19 (Figure 6B). Additionally, our observations revealed that EGF significantly increased the expression of Wnt signaling target genes, including both canonical (LGR5) and non-canonical (NKCC1), as well as the biliary progenitor marker NOTCH2 (Supplementary Figure S7). These preliminary findings suggest a potential correlation between EGF and Wnt signaling, supported by the upregulation of both canonical and non-canonical Wnt signaling target genes, along with the biliary progenitor gene.