Human extrahepatic and intrahepatic cholangiocyte organoids show region-specific differentiation potential and model cystic fibrosis-related bile duct disease

The development, homeostasis, and repair of intrahepatic and extrahepatic bile ducts are thought to involve distinct mechanisms including proliferation and maturation of cholangiocyte and progenitor cells. This study aimed to characterize human extrahepatic cholangiocyte organoids (ECO) using canonical Wnt-stimulated culture medium previously developed for intrahepatic cholangiocyte organoids (ICO). Paired ECO and ICO were derived from common bile duct and liver tissue, respectively. Characterization showed both organoid types were highly similar, though some differences in size and gene expression were observed. Both ECO and ICO have cholangiocyte fate differentiation capacity. However, unlike ICO, ECO lack the potential for differentiation towards a hepatocyte-like fate. Importantly, ECO derived from a cystic fibrosis patient showed no CFTR channel activity but normal chloride channel and MDR1 transporter activity. In conclusion, this study shows that ECO and ICO have distinct lineage fate and that ECO provide a competent model to study extrahepatic bile duct diseases like cystic fibrosis.


Results
Efficient organoid initiation from the human common bile duct. Samples from donor liver and the common bile duct (eBD) were dissociated and used for culture initiation (Fig. 1A). EpCAM and LGR5 frequencies were determined in the cells that were isolated from eBD and liver biopsies. Flow cytometric analysis show that EpCAM levels were low 1.2 ± 1.6% in eBD and 0.9 ± 0.7% (average ± SD) in liver starting populations as anticipated and as previously shown 18 . Cells isolated from ICO and ECO expressed ~ 100% EpCAM as shown before 14 (Supplementary Fig. S1).
LGR5 expression was too low to be detected by flow cytometry (not shown). Organoids could be initiated from human eBD tissue with > 95% success rate from over 40 biopsies from both male and female organ donors or patients with a wide age range (16-70 years). On average, the first ECO were visible in cultures at 7 days (6.8 days ± 1.0 SD). Organoids were grown 3-dimentional in hydrogels (Matrigel) and after initiation organoid cultures were passaged one to two times per week (split 1:3). Organoid initiation efficiency and expansion potential was similar to paired ICO cultured from liver biopsies from the same donor (data not shown). In accordance with previous results 14 , ECO could be viably frozen and thawed and in general, behaved similarly to ICO. Several organoid lines were biobanked early but many ECO cultures (n = 15) were passaged over 30 times for a period of 8 months or longer without clear signs of exhaustion. The ICO and ECO cultures were morphologically similar and no apparent histological differences were found (Fig. 1B-I). Both cultures remained proliferative at late passages (> 6 months), with 11.0 ± 2.1% and 17.3 ± 2.4% (mean ± SEM) Figure 1. Culture initiation and characteristics of liver-and extrahepatic bile duct-derived organoids. Schematic representation of organoid culture initiation from liver biopsies (intrahepatic bile duct; iDO) and extrahepatic bile duct (eDO) (A). Organoids were cultured and passaged weekly (1:3 for over 30 passages/ > 8 mo) and harvested at different passages for further analysis. ECO were cultured as efficiently as liver-derived ICO. Both look similar at the microscopic level (B, C), when stained with phalloidin/DAPI (D,E), and in FFPE sections stained by immunofluorescent KRT7 (F,G-10×) and KRT19 (H,I-10×). EdU-incorporation showed a similar percentage of proliferating cells (11-14% in S-phase) in both organoid types at passage 10. Shown is one representative result of three paired donor lines tested (J). The individual percentages are shown in Supplemental Fig. S2. Despite no significant difference in proliferation (p = 0.122), a significant difference in organoid size was found when analyzing individual organoids for 30 h (K) Shown is mean ± SEM of 100 organoids from paired ECO and ICO of three different donors. TROP2 protein expression was found in almost 100% of the organoid cells, both for ECO and ICO (L,M) (three paired organoid lines). Expression heatmap of selected genes described by Aizarani et al., as markers for liver progenitors 15 . The expression levels per gene were collected from the normalized RNA sequencing analysis performed on four ICO and three ECO lines, and visualized as Z-score. Though the level of expression varied per gene, no significant difference in expression was found between ICO and ECO (N). ◂ of the cells in S-phase at P10, respectively. Additional Ki67 stainings of ECO and ICO organoid sections confirmed this ( Fig. 1J and Supplementary Fig. S2), and were similar to previously described percentages for ICO 14 . Despite similar proliferation rates, the increase of relative organoids size as measured by time-lapse bright-field microscopy was significant differences for paired organoid types (Fig. 1K). Organoid size was measured starting 3.5 days after splitting and followed for 30 h. After 12 h of measuring the size of ICO were significantly larger than ECO from the same donor (n = 100 organoids per type, p < 0.001). As shown in Fig. 1L, both organoid types express the trophoblast cell surface protein 2 (TROP2) protein which is associated with organoid-initiating cells) and no significant difference was found in TROP2 protein expression (Fig. 1M, 99.1% ± 1.2% in ECO and 98.5% ± 1.0% in ICO; p = 0.67). Together, these results show that organoids are initiated efficiently from human eBD and iBD and have similar characteristics with only a small but significant differences in organoid size expansion.
Gene expression analyses of ECO and ICO. When looking at the expression of specific genes known to be markers for intrahepatic organoid-initiation cells identified by single cell RNA sequencing and cell sorting 15 , all genes were present and no significant difference in expression was observed between ECO and ICO (Fig. 1N). These genes include epithelial cell adhesion molecule (EpCAM), leucine-rich repeat-containing G-protein coupled receptor 5 (LGR5), the low-density lipoprotein receptor-related protein (LRP5/6) co-receptor group 12 , tumor-associated calcium signal inducer 2 (TACSTD2 or TROP2), SRY-Box 9 (SOX9), Cytokeratin 19 (KRT19), fibroblast growth factor receptor 2 (FGFR2), transmembrane 4 L Six Family member 4 (TM4SF4) and others 15 . Though all these markers were found positive in organoids, some genes were highly expressed (TM4SF4, KRT19) and some genes relatively low (LGR5 and OCT4).
Extending the gene expression analysis of specific genes associated with bi-potential progenitors, or progenitors with hepatocyte-fate or cholangiocyte-fate, again as described in the human liver atlas study 15 , did not show any significant different expressed between ICO or ECO ( Fig. 2A). Looking at the global gene expression of three paired organoid sets and one additional non-paired ICO, revealed a high degree of similarity and only small differences. As shown in Fig. 2B, this RNA sequencing analyses showed that only 29 genes were significantly differentially expressed (p adjusted < 0.05), of which 13 were expressed at higher levels in ECO and 16 were expressed at higher levels in ICO. Among the genes that were more abundantly expressed in ECO were Aquaporin5 (AQP5), a classic water transporter-channel known to be expressed on cholangiocytes and pancreatic epithelium and to a lesser degree by hepatocytes, and Insulin Like Growth Factor Binding Protein 1 (IGFBP1) known to be involved in ductular reactions of the biliary epithelium in response to liver damage 19 . Genes associated with the alcohol dehydrogenase pathway (ALDH1A3, AADAC) and lipid metabolism (STS), typically expressed in hepatocytes, are differentially expressed and upregulated in ICO. In particular ALDH1A3, highly expressed in ICO, is known to be involved in the biosynthesis of retinoic acid (RA) and RA signaling transduction. Interestingly, a previous study with lung organoids showed that inhibition of RA pathway causes an increased ALDH expression and an increase the size of organoids 20 . Thus, higher ALDH expression in ICO may explain their larger organoid size as compared to ECO (Fig. 1K), however, this requires further experimental proof. Other hepatocyte-related genes as CYP3A4 and HNF4a were not differently expressed in ICO and ECO (Fig. 2B). The regional-specific gene for common bile duct tissue which was retained in common bile duct organoids 17 , HOXB2, was not significantly different between ECO and ICO (not shown). This same study reported that PROM1, SOX9 and Albumin were significantly higher expressed in intrahepatic versus extrahepatic bile duct organoids. These difference was not confirmed in our ICO and ECO (data not shown) and may be related to the different culture conditions. Taken together, although these results show that gene expression in ICO and ECO was generally quite similar, several genes were found to be differentially expressed.
Cholangiocyte-related ion transport activity in ECO and ICO. To execute their function to modify hepatocyte-derived bile, cholangiocytes harbor a number of transporter-channels on their apical membranes. This include the cAMP-activated CFTR channel and the Ca 2+ -activated Cl − channel CaCCs. As shown in Fig. 3, basolateral forskolin stimulation of ECO or ICO cells grown as a 2D monolayer, induced a short-circuit current (Isc). This Isc signal could be completely inhibited by the CFTR-inhibitor GlyH-101 (N-(2-naphthalenyl)-((3,5dibromo-2,4-dihydroxyphenyl)methylene)glycine hydrazide) 21 . Furthermore, apical UTP, which elicits Ca 2+ mobilization through activation of purinergic (P2Y) receptors, induced an lsc that was inhibited by the CaCC inhibitor T16inh-A01 (Fig. 3A). No differences were found in activation of the CaCC by UTP and subsequent inhibition by T16inh-A01. Of note, although in both organoids types clear CFTR activity was measured, the most pronounced response to forskolin was observed in ECO. This observation was further supported with another CFTR measure, the 3D forskolin-induced swelling (FIS) assay. In this FIS assay, due to changed osmolarity upon forskolin stimulation, water is transported inside the lumen, causing the organoids to swell. Cholangiocyte-fate differentiation potential is similar for ECO and ICO. Different protocols have been developed for differentiation towards a cholangiocyte-like phenotype, both for iPSC and ICOs. As the starting cell population prior to differentiation already have many cholangiocyte-like properties (i.e. ion and water channel activities), this cholangiocyte-fate differentiation can also be considered maturation. As shown in Fig. 4A, after 14 days culture in differentiation medium, organoids became more dense, had thicker outer walls and stopped proliferating as shown in Fig. 4B, at gene expression level, stem cell and proliferation markers LGR5  www.nature.com/scientificreports/ and Ki76 were significantly downregulated in DM-chol both for ECO and ICO, confirming loss of stemness and proliferation. To benchmark the cholangiocyte-differentiated ICO and ECO, RNA expression profiles from bile duct tissue, published by Rimland et al. 17 , were used. For this, 20 genes known to be highly expressed in cholangiocytes were used. Gene expression clustering of ICO, ECO and the bile duct tissue showed that in both ICO and ECO in DM-cholangiocyte conditions, these cholangiocyte genes were upregulated (Fig. 4C). The P-glycoprotein (Pgp) membrane transporter family member MDR1, is an apical transporter known to be expressed by cholangiocytes. Differentiated ECO readily took up the tracer dye and actively transported this to the luminal side via MDR1, resulting in a fluorescent organoid lumen (Fig. 4D). MDR1 dependency was confirmed by blocking luminal transport of Rh123 with the MDR1 antagonist Verapamil, resulting in accumulation of dye in the cell cytoplasm. Similar results were observed for ICO in DM-chol (not shown). Finally we confirmed that cholangiocyte-lineage differentiation is highly dependent on Notch signaling (Fig. 4E). For this, a specific small molecular inhibitor of the Notch pathway, DAPT, was added during the differentiation culture in DM-chol. Clearly the cholangiocyte-specific genes AQP1, AE2 and HNF1b which were upregulated during DM-chol, were completely downregulated by treatment with DAPT. Also the Notch-related genes Notch2 and Jag1 were downregulated by DAPT (Fig. 4E). These results indicate that organoids initiated from both eBD and iBD can be driven towards a similar adult cholangiocyte-like phenotype.  22 were included as a benchmark for hepatocyte-fate differentiation. Expression of the hepatocyte-specific genes were clearly different between ICO and ECO. In the DM-hep condition ECO showed less upregulation of hepatocyte-specific genes as compared to ICO, whereas progenitor-and cholangiocyte-related genes were not significantly different. Of note, one of the ICO lines showed less prominent hepatocyte-fate differentiation but it is well known that differentiation potential varies between individual donors.
Assessing general differences in ICO and ECO gene expression showed that undifferentiated organoids in EM have only limited number of significantly different genes (16 genes significantly higher in ICO and 13 in ECO, shown in Fig. 2B). As shown in Fig. 5C, in the DM-chol condition the difference in gene expression between ICO and ECO completely disappeared (1 significantly different gene). However, in the DM-hep condition clearly heterogeneity in gene expression profiles increased. In ECO 136 genes were significantly higher and 89 genes were found significantly higher in ICO (Fig. 5C), indicating a difference in response to the DM-hep condition between these organoid types. The expression of hepatocyte-specific genes ALB (Fig. 5D) and CYP3A4 (Fig. 5E) was further confirmed in 3 independent differentiation experiments and quantified by qRT-PCR. Again, though progenitor-related gene KRT19 was not significantly different regulated between ICO and ECO (Fig. 5F), again only ICO showed significant upregulation of ALB and CYP3A4 in the DM-hep condition whereas ECO clearly did not upregulate ALB and CYP3A4. This was also confirmed on a functional level testing CYP3A4 activity in an enzymatic assay (Fig. 5G). Also here, only ICO and not ECO showed a significant increase in CYP3A4 activity in DM-hep condition. Together these results indicate that though both ECO and ICO have similar cholangiocyte lineage potential (Fig. 4), only ICO and not ECO show hepatocyte lineage potential. This indicate both organoid types retain some regional specific differentiation potential, with the extrahepatic bile duct region lacking clear hepatocyte-fate differentiation.

Modelling cystic fibrosis biliary disease in ECO.
To determine the feasibility to use ECO as a model for bile duct diseases, organoids were initiated from eBD tissue collected from a CF patient at time of liver transplantation. This patient was known to carry a compound heterozygous mutation in the CFTR gene, resulting in a non-functional CFTR protein. Western blot analysis of protein lysate from organoids confirmed the absence of mature CFTR protein (band at 170 kDa) in the CF-ECO (Fig. 6A) which was clearly visible in the healthy ECO. Immature CFTR (band at 130 kDa) was detected in both CF and healthy organoids as expected. Initiation and expansion of the CF-ECO was similar efficient compared to ECO from healthy donors. As shown in Fig. 6B, EdU incorporation assays showed similar levels level of cell proliferation, with 18% of cells S-phase for CV-ECO, as compared to healthy ECO (14%, Fig. 1J). Functional transporter assays in Ussing chambers showed a clear lack of forskolin-activated CFTR-mediated chloride currents (Fig. 6C). To demonstrate that this non-responsiveness was specific for the CFTR channel, the Ca 2+ activated Clchannel was activated with UTP at the apical side of the cells. Upon UTP addition, the channel was clearly activated in the CF-ECO cells and could be specifically inhibited by CaCC channel inhibitor T16inh-A01. As previously showed in Fig. 3, ECO of healthy donors show clear activity of both CFTR and the Ca 2+ activated Clchannel.
The CFTR channel is known to co-localized with other transporter channels, like MDR1 23 . To test whether the CFTR deficiency in organoids effects the function of other channels we needed to differentiate CF-ECO in DM-chol condition in order to measure MDR1 activity. As shown in Fig. 6D, the CF-ECO showed comparable cholangiocyte-fate differentiation as the healthy counterparts. The morphological changes in CF-ECO appeared smaller. In DM-chol conditions less dense organoids were observed as compared to healthy donor ECO (Fig. 4). This difference could be attributed to changes in water export due to dysfunctional CFTR channel activity. As shown in Fig. 6E www.nature.com/scientificreports/ (Fig. 4). Stem cell and proliferation markers, LGR5 and Ki67 were down regulated upon differentiation. Conversely, expression of mature cholangiocyte markers KRT19, NOTCH2, and HNF1B were upregulated (Fig. 6E).
Although the MDR1 pump and the CFTR channel are co-localized 23 , MDR1 functionality was not diminished in ECO of CF patient (Fig. 6F). Efficient transport of the Rhodamine123 was seen in the CF-ECO which again could be blocked by specific inhibitor, Verapamil. Combined these results provide evidence that ECO can be used as a patient-specific model for bile duct diseases like CF and encourage further research on CF physiology and personalized drug testing.

Discussion
The development, homeostasis and repair of iBD and eBD involves distinct mechanisms, either involving cholangiocyte or progenitor cell proliferation or maturation. In this study, organoids were successfully initiated from human eBD and pair-wise compared to the well-established iBD-organoids 14 . Overall these two types of organoids, ECO and ICO, behaved highly similar in terms of gene expression, proliferation capacity and function capacities. Despite these similarities, we observed regional-specific differences in their differentiation potential. As reported earlier, ICO have clear bipotent differentiation capacity. However, here we found that ECO are committed to differentiation towards cholangiocyte-fate and lack the ability to differentiate towards a hepatocyte-like phenotype. A similar observation was recently reported by Rimland et al. 17 . They also demonstrate that differences exist between ICO and ECO which is mostly reflected in the exclusive capacity of ICO to differentiate in the direction of hepatocyte lineage. Considering the differences in embryonic origin ECO may also be bipotent, not in respect to hepatocyte-fate, but with the potential for pancreatic-lineage and cholangiocyte-lineage differentiation potential. Although this has also been suggested by others 6,7 , it requires further research to confirm this. As shown in Fig. 2, only few genes were differentially expressed between ECO and ICO lines. Currently, a follow-up study has started to find markers that distinguish between the organoid-initiating cells of each type at a single-cell resolution and show their localization and distribution in primary liver and bile duct tissue. Recent advances in stem cell biology and novel culture technology enabled the large-scale expansion of primary biliary epithelium by generating complex 3D stem cell-derived constructs or organoids from liver biopsies 14 . Whether these organoids arise from LGR5-positive or negative cells, remains to be determined. Lineage tracking studies in Lgr5-creERT2 reporter mice suggests that LGR5-negative cells could be activated upon liver injury induction and at that point start to express LGR5 24 . This implies plasticity of liver ductal cells that respond to the need of parenchymal cells.
These liver-organoids, or more specifically, ICO enabled detailed analysis of intrahepatic biliary epithelium. In parallel, mature cholangiocytes were derived from human induced pluripotent cells (iPSC) to act as model for biliary disease and drug screening 25 . However, both liver-derived organoids and iPSC-based cholangiocytes could not provide insight into the (stem) cell populations present in eBD. With the availability of eBD biopsies collected during liver transplantation, we embarked on characterizing organoids from the extrahepatic biliary epithelium (eBD) and compared them to paired organoids initiated from iBD of the same donor. The noncanonical Wnt-stimulated extrahepatic cholangiocyte organoids as described by Sampaziotis 16 , isolated from eBD are potent cells for regenerative medicine applications 26,27 . However, these cholangiocyte organoids are assumedly derived from mature primary cholangiocytes (not stem/progenitor cells) and are not dependent on canonical Wnt signaling like the LGR5-positive ICO 14 and ECO. Although cholangiocyte organoid cultures are interesting in their own right they are not representative of extrahepatic LGR5-positive stem cell populations which reside in the peribiliary glands and which are dependent on canonical Wnt signaling 28,29 . As the ECO resemble biliary progenitor cells that are present in the bile ducts, they can specifically be employed in studies focusing on stem cell defects leading to biliary diseases and developmental biology research.
The results described in this study may reflect intrinsic differences between intra-and extrahepatic resident (stem/progenitor) cells. It is known that differences in phenotype and functionality between mature cholangiocyte populations depend on their localization along the biliary tree 17,19 . With the development of novel technology as single cell RNA sequencing, now the cellular composition of the liver can be reconstructed based on distinct gene expression profiles of individual cell types 15,30 . Sequencing single EpCAM-positive liver cells showed they are a heterogeneous set of cells including mature cholangiocytes as well as potential progenitor cell subpopulations which both express cell surface markers TROP2 (or TACSTD2), FGFR2, TM4SF4, and CLDN1. Further analyses indicated that the cell population with intermediate expression of TROP2 (EpCAM + TROP2 int ) to indicate the true progenitor population with both hepatocyte and cholangiocyte differentiation potential 15 . Interestingly, an www.nature.com/scientificreports/ earlier study on TROP2 in eBD reported that though TROP2 is non expressed in peribiliary glands, upon damage of the bile ducts TROP2 is clearly upregulated in these glands epithelial cells 31 . Although novel culture techniques provide evidence for long-term expansion of primary hepatocytes 32,33 and primary cholangiocytes 26,27 , both cell types remain challenging to maintain in culture while keeping their mature functions 13,14,28 . Expansion of ECO could provide in this shortcoming. In addition to cell expansion for use in tissue engineering, as was recently demonstrated by recellularization of decellularized human eBD 34 , we demonstrate that ECO, have the potential to be used as a model for bile duct-related diseases as CF. To show proof of concept, ECO were cultured from eBD tissue from a CF patient and assessed for disease-typical features. Transporter channels were functional in ECO derived from healthy donor bile duct as demonstrated by activating and blocking specific transporter function (CFTR and MDR1 23,35 ), whereas CFTR channel activity was impaired in the CF patient-derived ECO. Of note, we also showed that healthy ECO respond significantly higher to forskolin-activation as ICO, indicating that ECO better mimic biliary epithelium that is affected in CF patients. Modeling diseases as CF using ECO will help to study not only pathways involved in fluid transport and epithelial function, but also provide the opportunity to test (toxic) effects of bile acids and beneficial effects of CFTR targeted drugs.  23 ), served as positive controls. Assessing general differences in ICO and ECO gene expression showed that in DM-hep condition the difference between ICO and ECO is biggest whereas small differences in gene expression was observed in DM-chol (C). Gene expression levels of hepatocytespecific genes albumin (D) and Cyp3A4 (E) were further confirmed by qRT-PCR analyses in independent experiments. Both albumin (D) and Cyp3A4 (E) were again found increased in ICO and not in ECO. (C). KRT19 (F) expression levels did not differ significantly between these organoid types. Significance is indicated as ***p < 0.0005. As a functional measure of hepatocyte metabolic activity, Cyp3A4 activity was measured using a commercially available kit (G). The relative Cyp3A4 activity was significantly higher in DM-hep conditions for ICO but not for ECO. Significance is indicated as *p < 0.01. www.nature.com/scientificreports/

Scientific Reports
In conclusion, this study shows the feasibility to culture Wnt-driven organoids from small tissue biopsies of human eBD and that these ECO can be differentiated towards functional cholangiocyte-like cells. This opens the avenue for use in pre-clinical toxicology studies, development and testing of novel treatments in a personalized setting, and tissue engineering for patient-specific clinical applications.

Material and methods
Human organoid culture initiation and expansion. Tissue samples (≤ 0.5 cm 3 ) of donor (extrahepatic) common bile duct (eBD) and donor liver biopsies (n > 40) were collected during liver transplantation at the Erasmus Medical Center Rotterdam. Use of both tissues for research purposes was approved by the Medical Ethical Council of the Erasmus MC and informed consent was given (MEC-2014-060). Liver and eBD biopsies were obtained from a 17-year old male Cystic Fibrosis (CF) patient undergoing liver transplantation at the University Medical Center Groningen. Informed consent was obtained to use these biopsies for research (STEM Study-protocol number 1-402/K, informed consent was obtained from a parent). This patient was compound heterozygous for F508del and R1162X, resulting in a non-functional CFTR transporter. Biopsies were stored in organ preservation fluid (University of Wisconsin Belzer UW Cold Storage Solution, Bridge of Life Ltd. London, UK) and transported at 4 °C. The biopsies were used to initiate ECO and ICO cultures which were initiated as described for human liver tissue 14,36 . In short, the biopsies (liver and eBD) were washed in Advanced DMEM/F-12 medium supplemented with penicillin/streptomycin (with 10,000 U/ml, Life Technologies), HEPES (1 M, Life Technologies), Ultraglutamine (200 mM, Life Technologies). Both liver and eBD biopsies were minced and digested in a collagenase solution (2.5 mg/ml collagenase Type A, Sigma Aldrich in Advanced DMEM/F-12 medium), rocking at 37 °C for 20 min. Single cell solutions were gained by passaging over a 70 μm nylon mesh cell strainer and washed in Advanced DMEM/F-12 medium (5 min, 1500 rpm at 4 °C) after which the pellet was diluted in ice cold Matrigel (Corning) and seeded in 25 μl droplets in 48-wells plates. After solidifying of the Matrigel, 250 μl culture initiating medium was added and cells were incubated at 37˚C, 5% CO2 for three days. Culture initiating medium consisted of Advanced DMEM/F12 medium (Invitrogen) supplemented with 1% N2, 1% B27 (both Gibco), 1.25 mM N-Acetylcystein (Sigma), 10 nM gastrin (Sigma), 50 ng/ml EGF (Peprotech), 100 ng/ml FGF10 Peprotech), 25 ng/ml HGF (Peprotech), 10% R-spondin (conditioned medium), 10 nM nicotinamide (Sigma), 5 µM A83.01 (Tocris), 10 µM forskolin (Tocris), 25 ng/ml Noggin (conditioned medium), 30% Wnt (conditioned medium), 10 µM Y27632 (Sigma), and hES cell cloning recovery solution (Stemgent). After 3 days, the initiation medium was changed to expansion medium (EM), deprived of noggin, Wnt, Y27632 and hES cell cloning recovery solution. Organoid formation was seen within 2 or 3 days after culture initiation and medium was refreshed every 3 days. Organoids were split (1:3) every 7 days. To prepare viably frozen stocks (from different passages), organoids were mechanically dissociated and mixed with recovery cell culture freezing medium (Gibco) according to the manufacturers protocol and stored at − 196 °C.
DNA-synthesis and cell cycle analysis. Cell proliferation was analyzed by direct measurement of DNA synthesis using the Click-iT EdU Alexa Fluor 488 Flow Cytometry kit (Thermo Fisher). For this, organoids of different passage numbers were incubated with the thymidine analogue 5-ethynyl-2′-deoxyuridine (EdU) during 4 h at 37 °C, 5% CO 2 after which the organoids were harvested and made single cell using trypsin-EDTA incubation (15 min at 37 °C). The cells were further processed according to the manufacturer's protocol. The percentage of S-phase cells in the organoids was determined using standard flow cytometry methods (FACS-Calibur, BD Biosciences).
Organoid growth rate analysis. Organoid growth rates were determined using a custom-made pipeline for bright field-based image segmentation (Hof et al., in submission) in three pairs of ICO and ECO (passage 7-16). In brief, organoids were split, seeded in 5 µl of Matrigel into 96-well plates at similar densities, overlaid with 100 µl of expansion medium, and were cultured for 84 h before imaging. Organoids were imaged for 30 h at 1 h intervals using bright field microscopy (microscope: Zeiss AxioObserver.Z1; objective lens: Zeiss Plan-Apochromat 5×/0.16; tiling: 2 × 2; z-planes: 10; z-spacing: 65 µm). The recorded time-lapse image stacks were pre-processed with Fiji [ImageJ version 1.51n, Java version 1.8.0_6 (64-bit)] by reducing the dimensionality of the raw data set from 4 (2 × 2) tiles with 10 z-planes each to 1 stitched image with 1 z-plane per time frame using the plugin Grid/Collection stitching 37 . The resulting image stacks were subsequently segmented based on projected luminal areas of the organoids by using the Fiji plugin Morphological Segmentation (MorphoLibJ) 38 . Segmented luminal areas were measured with the Fiji plugin Region Morphometry (MorphoLibJ) 38 . The luminal areas of individual organoids were normalized to the average luminal area over the first 5 time-points. Organoids which were not detected in all 31 time frames were manually excluded from the analysis. Data is displayed as mean ± SEM in the corresponding graph.
Hepatocyte differentiation. Organoids were expanded for 5-7 days before initiation of hepatocyte differentiation. Differentiation towards hepatocyte fate was initiated with the addition of 25 ng/ml BMP7 (Peprotech) to the expansion medium and lasted 5 days. Subsequently, organoids were passaged with a 1:1.5 split ratio and medium was changed to human hepatocyte differentiation medium based on the original differentiation protocol as published by Huch et al. 14  Cholangiocyte differentiation. The protocol used to differentiate organoids towards cholangiocytes was adapted from the previously published protocol for iPSC differentiation towards cholangiocytes 25  of organoids was performed essentially as described by Dekkers et al. 39,40 . In brief, ICO and ECO were seeded in 96-well cell culture plates (circa 50/well), in a 5 µL droplet of Matrigel. After 2 days, organoids were loaded with calcein-green (Invitrogen; acetoxymethyl ester; 5 µmol/l; 1 h) suspended in modified Meyler solution. Calceinfluorescent, viable organoids were maintained at 37 °C in 5% CO 2 , and visualized on a confocal microscope equipped with a computer-controlled moveable stage (5× objective; Leica TCS SP5). After CFTR phosphorylation/activation was triggered by addition of forskolin (5 µM), images were acquired at 6 min intervals for 2 h. CFTR-mediated anion secretion stimulates osmotic fluid transport into the enclosed organoid lumen, leading to a volume increase. To assess organoid swelling, the area enclosed by the fluorescent cells lining the organoids was quantified on the ImageJ platform (NIH, USA), using a software module developed by the Optical Imaging Center of the Erasmus MC, which fully automates image analysis. Data depict the cumulative increase in volume over 1 h. In the graphic display of the data, each data point represents the average of triplicate wells.
Ussing chamber assay. Ussing chamber assays were performed essentially as described elsewhere 41  Functional transporter assays were done using CF-ECO (blue line) and healthy ECO (pink line) in an Ussing chamber assay (C). As expected, CF-ECO cells do not respond to CFTR stimulation with forskolin, whereas healthy ECO stimulated with forskolin showed a clear current which was inhibited by CFTR inhibitors, GlyH-101 and pyrimido-pyrrolo-quinoxalinedione (PPQ). These inhibitors had no effect on the CF-ECO. Stimulation with UTP shows functionality of other Ca 2+ -dependent Clchannels in both ECO (CF and healthy) and both are responsive to the specific inhibitor of this channel, T16A-inh-A01. www.nature.com/scientificreports/ ated software (Acquire and Analyze 2.3; Physiologic Instruments). Anion secretion was stimulated by addition of the adenylyl cyclase activator forskolin (10 µmol/l, Sigma-Aldrich) or the purinergic receptor agonist UTP (50 µmol/l, Sigma-Aldrich) to the luminal bathing solution. Forskolin-induced and UTP-induced secretion was inhibited by Glyh-101 (20 µmol/l, Sigma-Aldrich) and T16inh-A01 (50 µmol/l, Sigma-Aldrich), respectively.

Rhodamine 123 assay.
To determine the presence of calcium channels, including multidrug resistance protein 1B (MDR1b-important in the biliary excretion of large hydrophobic components-member of the P-glycoprotein membrane transporter family), Rhodamine 123 (Rh123) was added to the cultures. Rh123 is a fluorescent chemical compound that can be transported by these transporter channels 27. For this, cold Advanced DMEM/F12 medium was used to remove Matrigel and collagen from the cultures that were subsequently pretreated with DMSO or 10 μM Verapamil (Sigma-Aldrich-MDR1b inhibitor) for 30 min, followed by 5 min of incubation with 100 μM Rh123 (Sigma-Aldrich). The organoids were washed 3 times with expansion medium. Fluorescence (excitation wavelength: 511 nm; emission wavelength: 534 nm) was visualized with a Leica SPE-II confocal system, 30 min after washing and analyzed with ImageJ.
Flow cytometry. To assess TROP2 expression, paired ICO and ECO (passage 5) from three donors were dissociated using Trypsin/EDTA (15 min at 37 °C), washed in 8 ml in Advanced DMEM/F-12 medium (1500 rpm, 5 min, 4 °C) and cells were re-suspended in EBSS to make a single cell suspension. TROP2 antibody (Invitrogen; rabbit monoclonal conjugated to Alexa Fluor-488, clone MR54, used 1:100) was added (30 min, on ice) and cells were subsequently measured on a FACS Canto flow cytometer (BD Biosciences). Flow cytometric analysis of the starting cell populations, liver (n = 3) and extrahepatic bile duct biopsies (n = 3), collected during liver transplantation were treated similar to organoid initiation as described above. Instead of addingMatrigel to initiate the organoid cultures, the cells were stained for flow cytometry using monoclonal antibodies against human LGR5 (1:100, Clone 8F2, BD Biosciences) and human EpCAM (1:100, Clone 9C4. Biolegend), conjugated directly with PerCP and PerCP/Cy5.5, respectively. Cells were measured on a FACS CantoII flow cytometer (BD Biosciences) and analysed using FACS Diva software. Cells from ICO and ECO were harvested by trypsin-treatment (10 min, 37 °C), subsequently washed in Advanded DMEM/F-12 and stained with EpCAM and LGR5 to be analysed by flow cytometry.