Healthy tissue samples of the human liver from two males and two females were obtained during liver transplantation at the Erasmus Medical Center in Rotterdam under the approval of the Medical Ethical Council (MEC-2014-060). The human liver-derived ICOs were isolated and cultured, as described previously (Broutier et al., 2016). Briefly, the liver biopsies were washed in cold phosphate-buffered saline (PBS) and minced into small pieces. Minced tissues were washed in cold DMEM GlutaMAX medium (Gibco, Thermo Fisher Scientific, Waltham, MA, United States) supplemented with 1% (v/v) fetal calf serum (FCS; Gibco) and 1% (v/v) penicillin/streptomycin (P/S; Gibco). This was followed by digestion with 0.125 mg/ml type II collagenase (Gibco) and 0.125 mg/ml dispase (Gibco) in the DMEM GlutaMAX medium supplemented with 0.1 mg/ml DNase I (Roche, Basel, Switzerland), 1% (v/v) FCS, and 1% (v/v) P/S at 37°C in a shaking water bath. Every 10–15 min, the supernatant was collected, and the fresh digestion-supplemented medium was added and allowed for digestion three times. Single cells were obtained by passing through a 70-µm filter, and the filtrate was washed in cold DMEM GlutaMAX medium (supplemented with 1% (v/v) FCS and 1% (v/v) P/S) and centrifuged (400 g for 5 min at 4°C). The cells were diluted in cold Matrigel (Corning, New York, NY, United States) and seeded in 50 µL droplets in 24-well plates (Corning). After gelatinization of the Matrigel, a culture medium was added, and cells were incubated at 37°C, 5% CO₂ (v/v). The culture medium (expansion medium, EM) was based on the Advanced DMEM/F12 medium (Gibco) consisting of 1% (v/v) P/S, 1% (v/v) HEPES (10 mM; Gibco), and 1% (v/v) GlutaMax (Gibco). The EM was supplemented with 10% (v/v) R-Spondin-1–conditioned medium (the Rspon1-Fc-expressing cell line was a kind gift from Calvin J. Kuo), 2% (v/v) B27 supplement without vitamin A (Invitrogen, Carlsbad, CA, United States), 1% (v/v) N2 supplement (Invitrogen), 10 mM nicotinamide (Sigma-Aldrich, St. Louis, MO, United States), 1.25 mM N-acetylcysteine (NAC; Sigma-Aldrich), 100 ng/ml fibroblast growth factor 10 (FGF10; Peprotech, Rocky Hill, NJ, United States), 10 nM recombinant human (Leu15)-gastrin Ⅰ (GAS; Tocris Bioscience, Bristol, United Kingdom), 10 µM forskolin (Tocris Bioscience), 50 ng/ml epidermal growth factor (EGF; Peprotech), 25 ng/ml hepatocyte growth factor (HGF; Peprotech), and 5 µM A8301 (transforming growth factor ß inhibitor; Tocris Bioscience). Organoids were mechanically split weekly at 1:3–1:4 ratio, and the medium was changed every 2–3 days.

ICOs were removed from Matrigel using cold Advanced DMEM/F12, mechanically dissociated into cell clusters using a pipette tip, and transferred to fresh Matrigel mixed with 1.2 mg/ml rat-tail type Ⅰ collagen (Merck Millipore, Darmstadt, Germany) at a ratio of 2:3. Cultures were subsequently incubated at 37°C for 2 hours in a culture incubator for the hydrogel to polymerize after which the EM medium was added. After 1 day, the medium was changed to defined medium (cholangiocyte differentiation medium, CDM) consisting of Advanced DMEM/F12 medium supplemented with 1% (v/v) P/S, 1% (v/v) HEPES, 1% (v/v) GlutaMax, 2% (v/v) B27 supplement without vitamin A, 1% (v/v) ITS Premix (contains insulin, human transferrin, and selenous acid; Corning), 1.25 mM NAC, 100 ng/ml FGF10, 10 nM GAS, 50 ng/ml EGF, 25 ng/ml HGF, and 5 µM A8301. Under differentiating conditions, cells were cultured for 1 week, and the medium (CDM) was refreshed every other day.

The RNeasy Micro kit (QIAGEN, Hilden, Germany) was used to isolate RNA from liver organoids, M/C (Matrigel/collagen)- and HFM-cultured CLCs, following the manufacturer’s instructions. RNA concentration and purity were measured with the ND-1000 spectrophotometer (NanoDrop, Thermo Fisher Scientific). cDNA was prepared with the iScript cDNA synthesis kit, according to the manufacturer’s instructions (Bio-Rad, Hercules, California, United States). qPCR was performed to measure the relative gene expression using the SYBR Green method (Bio-Rad). Hypoxanthine phosphoribosyltransferase 1 (HPRT1) and ribosomal protein L19 (RPL19) were selected as stably expressed reference genes, and their average expression was used for normalization (Bustin et al., 2010). Details of all primers are described in Supplementary Table S1. Primary cells might lose some characteristics upon in vitro culturing. Therefore, we used the human gallbladder that contains mature cholangiocytes to compare the gene expression of markers in CLCs. Hence, we normalized the CLC gene expression to gallbladder samples for data analysis.

A rhodamine 123 (Sigma-Aldrich) transport assay was performed on day 4. Organoids were pretreated with DMSO or 10 µM verapamil [an inhibitor of the efflux pump P-glycoprotein also known as multidrug resistance protein 1 (MDR1); Sigma-Aldrich] and incubated at 37°C for 30 min. Subsequently, organoids were gently removed from Matrigel/collagen and resuspended in CDM containing 100 µM of rhodamine 123 and incubated at 37°C for 10 min. Then, organoids were centrifuged (400 g for 5 min at room temperature), and the supernatant was removed. Organoids were washed with warm Advanced DMEM/F12 and centrifuged again. Fluorescence (excitation wavelength: 470 nm; emission wavelength: 535 nm) was measured using the BZ-9000 microscope (Keyence, Osaka, Japan).

For the farnesoid X receptor (FXR) activity assay, CLCs were incubated in CDM consisting of 10 µM GW4064 (an agonist of FXR; Sigma-Aldrich) or DMSO (control) for 24 h. RNA was isolated and used to evaluate the gene expression of downstream signaling of FXR target genes (solute carrier family 51 subunit alpha and solute carrier family 51 subunit beta, SLC51A and SLC51B) in the CLCs.

The polyethersulfone HFM (SENUO Filtration Technology, Tianjin, China) was double-coated to provide the cells with an extracellular matrix, as described previously (Chen et al., 2018). Briefly, HFM were cut into 3-cm pieces and sterilized in 70% (v/v) ethanol for 30 min. Next, the HFM was washed in PBS and precoated with filter-sterilized l-DOPA solution (2 mg/ml of 3,4-Dihydroxy-l-phenylalanine in 10 mM Tris buffer, pH 8.5) for 4 h at 37°C in an incubator, during which the fibers were turned 90° every hour. Subsequently, fibers were washed with PBS and put in rat-tail type I collagen solution (25 μg/ml in PBS) for another 2 h at 37°C in an incubator, during which the fibers were turned 90° every 30 min. The collagen solution was removed, and the fibers were washed with PBS prior to cell seeding. Single cells were prepared from liver organoids by in-gel trypsinization into single cells using TrypLE™ Express (Gibco) in a culture plate for 30–45 min at 37°C in an incubator. The single cells were collected and washed with Advanced DMEM/F12. Next, the double-coated HFMs were incubated with the cell suspension (5 × 10⁵ cells/3-cm fiber in 1.5-ml Eppendorf tubes) in 1 ml EM for 4–5 h at 37°C to allow cell adherence to the membrane. The tubes were turned 90° every hour. Finally, the cells covered HFM were transferred to 6-well plates and cultured in EM for 1–2 weeks for expansion and subsequently in CDM for 1 week for differentiation to create the bile duct tubules.

ICO-derived CLCs were removed from M/C using ice-cold cell recovery solution (Corning) and incubated on a horizontal shaker at 4°C (60 rpm) for 30–60 min until the 3D drops were dissolved. M/C- and HFM-cultured CLCs were washed with PBS and fixed with 4% (v/v) paraformaldehyde solution for 30 min at 4°C and permeabilized with 0.3% (v/v) Triton X-100 in PBS for 30 min at 4°C. To avoid non-specific antibody binding, the samples were incubated with block solution [2% (v/v) goat serum (Sigma-Aldrich), 2% (w/v) bovine serum albumin (Sigma-Aldrich), and 0.1% (v/v) Tween-20 in PBS (0.1% PBST)] for 1 h. Primary antibodies were diluted in block solution, and samples were incubated overnight at 4°C with shaking (60 rpm on a horizontal shaker). After being washed with 0.1% PBST three times for 20 min while shaking at 4°C (60 rpm on a horizontal shaker), the samples were incubated with secondary antibodies (5 µM Alexa Fluor 488 or 568; Life Technologies, Thermo Fisher Scientific) and with DAPI to stain nuclei (1:1,000; Sigma-Aldrich) overnight at 4°C while shaking (60 rpm on a horizontal shaker). Finally, tissues were washed three times for 20 min while shaking at 4°C (60 rpm on a horizontal shaker) and mounted on slides with ProLong Diamond Antifade mounting medium (Invitrogen). Images were acquired using a Leica TCS SP8 X imaging system. Antibody details for experiments are shown in Supplementary Table S2.

To measure the permeability of the bioengineered bile duct, an inulin-fluorescein isothiocyanate (2–5 kDa; inulin-FITC, Sigma-Aldrich) leakage assay was performed. The bile duct fiber was connected to inlet and outlet blunt needles (18 g; 0.5-inch length; OctoInkjet, Hoyland, United Kingdom) assembled in a custom-made 3D-printed biocompatible polylactide chamber (Supplementary Figure S1). Bile ducts were perfused (2 ml/h) with 0.1 mg/ml inulin-FITC in HBSS buffer supplemented with 10 mM HEPES for 10 min at room temperature. Samples were taken from the outer compartment of the bile duct, and fluorescence was measured at an excitation wavelength of 475 nm and an emission wavelength of 500–550 nm by using a fluorometer (Promega; Madison, Wisconsin, United States).

Statistical graphs were determined by GraphPad Prism 9 (GraphPad Software). RT-qPCR results’ data were presented as data points in box and whiskers (Min to Max; Figures 1C, 2) and the mean ± standard deviation (mean with SD; Figures 3C,D, 4C). The statistical test used was the two-tailed Mann–Whitney U test, and p-value details are described in the figure legends.

The generation of CLCs from human ICOs was based on our previous strategy for murine liver organoids (Chen et al., 2018) with some modifications to enhance further differentiation (Figure 1A). Compared to the study by Chen et al., we modified the medium and used a serum-free culture system to drive the human ICOs to CLCs. Details of expansion and differentiation medium are described in Supplementary Table S5. ICOs were expanded from the liver tissue biopsy, as previously described (Broutier et al., 2016), and seeded in a matrix that consisted of Matrigel and type Ⅰ collagen. After 8 days of cholangiocyte differentiation, organoids formed cystic and tubular structures whereas the organoids growing in expanding conditions remained cystic (Figure 1B). The cholangiocyte organoids showed an increase in wall thickness, and the central lumen became more visible. In addition, the CLCs demonstrated more clearly visible membrane junctions than the organoids cultured under expansion conditions.

Next, the efficiency of the differentiation was evaluated through gene expression analysis performed at three different time points (day 4, 6, and 8) in expanding (EM) and differentiating (CDM) conditions. EM analysis served as a control to assure differentiation was performed toward more mature cholangiocytes. The first indicator of differentiation was the decreased expression of the stem/progenitor cell marker LGR5 in the CDM condition compared to EM (Figure 1C), which was decreased at all time points. The enhanced mRNA expression of the biliary markers hepatocyte nuclear factor-1β (HNF1β) and hairy and enhancer of split-1 (HES1) demonstrated a more efficient cholangiocytic phenotype. Under the CDM condition, HNF1β was increased on day 6, and HES1 was increased on days 4 and 8. This differentiation was further corroborated by the increased mRNA expression of keratin 7 (KRT7) at all time points (days 4, 6, and 8). G protein-coupled bile acid receptor 1 (GPBAR1) on day 8 and gamma-glutamyl transferase 1 (GGT1) on days 6 and 8 were increased under differentiation conditions. No significant difference in HNF1β (days 4 and 8), HES1 (day 6), keratin 19 (KRT19; days 4, 6, and 8), ATP binding cassette subfamily B member 1 (ABCB1; days 4, 6, and 8), GPBAR1 (days 4 and 6), solute carrier family 4 membrane 2 (SLC4A2; days 4, 6, and 8), solute carrier family 10 membrane 2 (SLC10A2; days 4, 6, and 8), cystic fibrosis transmembrane conductance regulator (CFTR; days 4, 6, and 8), GGT1 (day 4), and somatostatin receptor 2 (SSTR2; day 4) were found, although the mRNA expressions under CDM conditions were higher than EM. Only for the biliary markers, aquaporin 1 (AQP1; days 4, 6, and 8) and SSTR2 (days 6 and 8), the mRNA expression was slightly lower under CDM conditions than EM. More details on the statistical analyses are presented in Supplementary Table S3.

To further verify the organoid differentiation to CLCs, immunofluorescence staining was performed (Figure 1D). Epithelial markers, tight junction protein 1 (TJP1, also known as ZO1), and E-cadherin 1 (CDH1), as well as cholangiocyte markers K7 and K19, were present in CLCs. More importantly, the CLCs expressed cholangiocyte-specific transporters, secretin receptor (SCTR), and TGR5 (also known as GPBAR1, encoded by GPBAR1) that play an important role in sensing and signaling and appear in mature cholangiocytes (Guo et al., 2016; Sato et al., 2018; Wang et al., 2021). In addition, the CLCs showed F-actin indicated by the staining for phalloidin, as a marker for the apical domains which demonstrates the CLCs developed apicobasal epithelial polarity, with the apical surface facing the lumen.

In the liver, cholangiocyte functions including the reabsorption and secretion of bile acids, water, and small molecules are performed by a series of transmembrane channel proteins. Cholangiocytes harbor a variety of transport channels, for instance, P-glycoprotein (also known as MDR1, encoded by ABCB1), a mediator of multidrug resistance expressed at the apical surface. To determine the functionality of in vitro-generated CLCs, the function of P-glycoprotein was investigated by incubating CLC organoids with the fluorescent dye rhodamine 123. Functional P-glycoprotein activity, located at the apical membrane, was confirmed by the accumulation of rhodamine 123 inside the lumen of CLC organoids. Moreover, blocking the transport activity with verapamil resulted in the accumulation of the fluorescent dye in the cytoplasm of organoid cells, verifying specific P-glycoprotein-mediated transport activity (Figures 3A–C).

Activation of the nuclear bile acid receptor farnesoid X receptor (FXR) leads to bile excretion whereas its inhibition causes bile accumulation or cholestasis (Shin and Wang, 2019). Bile acid excretion is performed by two transporters: one of ASBT located at the apical membrane of the cholangiocyte and the organic solute transporter α/β (OST α/β) (encoded by SLC51A and SLC51B, respectively) at the basolateral membrane. ASBT is responsible for the uptake of bile acids and can be downregulated by FXR activation. On the other hand, OST α/β are upregulated by FXR and support bile acid excretion. Incubation of the CLC organoids with the well-known FXR agonist GW4064 resulted in the downregulation of SLC10A2 and in the upregulation of SLC51A and SLC51B (Figure 3D), indicating that the ICO-derived CLCs express functional FXR-mediated bile homeostasis.

The observed polarization and functionality of the organoids prompted us to create a HFM system amendable for transport studies. For this, single cells were acquired from trypsinized organoids and seeded on the HFM scaffold (Figure 4A). The cell-seeded constructs were cultured for 10–14 days under expanding conditions (EM) until a confluent monolayer was formed and subsequently switched to differentiating conditions (CDM) for seven more days. CLCs formed an epithelial barrier on the HFM, as confirmed by immunofluorescence staining of the tight junction proteins ZO1 and CDH1 (Figure 4B). Furthermore, the restricted inulin-FITC diffusion represented the bile duct barrier function (Figure 4C). The bile ducts show an inside-out polarization as staining for the tight junction protein and phalloidin demonstrated extraluminally expressed TJP1/ZO1 and F-actin, whereas the basolateral E-cadherin/CDH1 and SCTR appeared at the luminal side. In addition, staining for K7, K19, TGR5/GPBAR1, and CFTR confirmed that HFM-cultured CLCs also maintained cholangiocyte-specific markers while being cultured on HFM (Figure 4B).

The bioengineered bile ducts were characterized further by comparing gene expression profiles to organoids cultured in CDM on a culture plate (M/C). The results revealed that CLCs cultured on HFMs maintained a stable expression of key cholangiocyte markers (HES1, KRT19, SLC4A2, GPBAR1, GGT1, and SSTR2). Interestingly, the cholangiocyte markers HNF1β, KRT7, ABCB1, SLC10A2, CFTR, and AQP1 showed a trend toward upregulation in CLCs cultured on HFMs as compared to plates cultured under differentiating conditions, suggesting that some cholangiocyte markers show a more mature phenotype in these tubular structures (Figure 2). The details of the statistical analyses are presented in Supplementary Table S4.