Advertisement

Identification of an FXR-modulated liver-intestine hybrid state in iPSC-derived hepatocyte-like cells

Open AccessPublished:July 18, 2022DOI:https://doi.org/10.1016/j.jhep.2022.07.009

      Highlights

      • Human iPSC were differentiated to HLC and characterized by single-cell RNA-seq with complementary epigenetic analyses.
      • HLC co-expressed liver and intestinal genes in the same individual cells.
      • Bioinformatics identified low activity of the nuclear factor FXR as one key factor behind mixed cellular identity.
      • Activation of FXR in HLC induced liver-associated target genes and functions and decreased expression of hybrid state genes.

      Background & Aims

      Pluripotent stem cell (PSC)-derived hepatocyte-like cells (HLC) have enormous potential as a replacement for primary hepatocytes in drug screening, toxicology and cell replacement therapy, but their genome-wide expression patterns differ strongly from primary human hepatocytes (PHH).

      Methods

      We differentiated human induced pluripotent stem cells (hiPSC) via definitive endoderm to HLC and characterized the cells by single-cell and bulk RNA-seq, with complementary epigenetic analyses. We then compared HLC to PHH and publicly available data on human fetal hepatocytes (FH) ex vivo; we performed bioinformatics-guided interventions to improve HLC differentiation via lentiviral transduction of the nuclear receptor FXR and agonist exposure.

      Results

      Single-cell RNA-seq revealed that transcriptomes of individual HLC display a hybrid state, where hepatocyte-associated genes are expressed in concert with genes that are not expressed in PHH – mostly intestinal genes – within the same cell. Bulk-level overrepresentation analysis, as well as regulon analysis at the single-cell level, identified sets of regulatory factors discriminating HLC, FH, and PHH, hinting at a central role for the nuclear receptor FXR in the functional maturation of HLC. Combined FXR expression plus agonist exposure enhanced the expression of hepatocyte-associated genes and increased the ability of bile canalicular secretion as well as lipid droplet formation, thereby increasing HLCs’ similarity to PHH. The undesired non-liver gene expression was reproducibly decreased, although only by a moderate degree.

      Conclusion

      In contrast to physiological hepatocyte precursor cells and mature hepatocytes, HLC co-express liver and hybrid genes in the same cell. Targeted modification of the FXR gene regulatory network improves their differentiation by suppressing intestinal traits whilst inducing hepatocyte features.

      Lay summary

      Generation of human hepatocytes from stem cells represents an active research field but its success is hampered by the fact that the stem cell-derived ‘hepatocytes’ still show major differences to hepatocytes obtained from a liver. Here, we identified an important reason for the difference, specifically that the stem cell-derived ‘hepatocyte’ represents a hybrid cell with features of hepatocytes and intestinal cells. We show that a specific protein (FXR) suppresses intestinal and induces liver features, thus bringing the stem cell-derived cells closer to hepatocytes derived from human livers.

      Graphical abstract

      Keywords

      Introduction

      Human hepatocytes are required for clinical therapy, as well as studies in pharmacology and toxicology.
      • Godoy P.
      • Hewitt N.J.
      • Albrecht U.
      • Andersen M.E.
      • Ansari N.
      • Bhattacharya S.
      • et al.
      Recent advances in 2D and 3D in vitro systems using primary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and their use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.
      Currently, much effort is directed towards understanding hepatic cell differentiation
      • Camp J.G.
      • Sekine K.
      • Gerber T.
      • Loeffler-Wirth H.
      • Binder H.
      • Gac M.
      • et al.
      Multilineage communication regulates human liver bud development from pluripotency.
      and improving protocols for the generation of hepatocyte-like cells (HLC) from embryonic stem cells or induced pluripotent stem cells (iPSC). HLC express several adult hepatocyte markers and offer the perspective of an unlimited supply of human hepatocytes.
      • Sachinidis A.
      • Albrecht W.
      • Nell P.
      • Cherianidou A.
      • Hewitt N.J.
      • Edlund K.
      • et al.
      Road map for development of stem cell-based alternative test methods.
      ,
      • Godoy P.
      • Schmidt-Heck W.
      • Hellwig B.
      • Nell P.
      • Feuerborn D.
      • Rahnenführer J.
      • et al.
      Assessment of stem cell differentiation based on genome-wide expression profiles.
      However, HLC exhibit major differences compared to primary human hepatocytes (PHH). As shown using computational tools, such as CellNet
      • Cahan P.
      • Li H.
      • Morris S.A.
      • Lummertz Da Rocha E.
      • Daley G.Q.
      • Collins J.J.
      CellNet: network biology applied to stem cell engineering.
      • Radley A.H.
      • Schwab R.M.
      • Tan Y.
      • Kim J.
      • Lo E.K.W.
      • Cahan P.
      Assessment of engineered cells using CellNet and RNA-seq.
      • Morris S.A.
      • Cahan P.
      • Li H.
      • Zhao A.M.
      • San Roman A.K.
      • Shivdasani R.A.
      • et al.
      Dissecting engineered cell types and enhancing cell fate conversion via Cellnet.
      and liver-specific gene expression algorithms
      • Kim D.S.
      • Ryu J.W.
      • Son M.Y.
      • Oh J.H.
      • Chung K.S.
      • Lee S.
      • et al.
      A liver-specific gene expression panel predicts the differentiation status of in vitro hepatocyte models.
      that objectify the differentiation status of stem cell-derived cell types, HLC reach between 50% and 80% of the gene regulatory network (GRN) status of PHH and reference liver tissue.
      • Sachinidis A.
      • Albrecht W.
      • Nell P.
      • Cherianidou A.
      • Hewitt N.J.
      • Edlund K.
      • et al.
      Road map for development of stem cell-based alternative test methods.
      ,
      • Godoy P.
      • Schmidt-Heck W.
      • Hellwig B.
      • Nell P.
      • Feuerborn D.
      • Rahnenführer J.
      • et al.
      Assessment of stem cell differentiation based on genome-wide expression profiles.
      ,
      • Godoy P.
      • Schmidt-Heck W.
      • Natarajan K.
      • Lucendo-Villarin B.
      • Szkolnicka D.
      • Asplund A.
      • et al.
      Gene networks and transcription factor motifs defining the differentiation of stem cells into hepatocyte-like cells.
      Using these platforms, it was demonstrated that the relatively large difference between HLC and PHH is not only explained by the low expression of numerous liver-associated genes in HLC, but also by the expression of ‘undesired genes’ that are normally expressed in endodermal and intestinal cells but not in adult hepatocytes.
      • Godoy P.
      • Schmidt-Heck W.
      • Natarajan K.
      • Lucendo-Villarin B.
      • Szkolnicka D.
      • Asplund A.
      • et al.
      Gene networks and transcription factor motifs defining the differentiation of stem cells into hepatocyte-like cells.
      ,
      • Gao X.
      • Li R.
      • Cahan P.
      • Zhao Y.
      • Yourick J.J.
      • Sprando R.L.
      Hepatocyte-like cells derived from human induced pluripotent stem cells using small molecules: implications of a transcriptomic study.
      Using single cell (sc)RNA-seq, the present study demonstrates that the differentiation of iPSC to HLC leads to a hybrid state, in which hepatocyte-associated genes and genes not expressed in the adult liver are expressed in the same cell. An integrated Omics analysis identified the farnesoid X receptor (FXR) as a key factor whose activity was too low in the hybrid state. FXR transduction combined with exposure to FXR-activating ligands (Fig. 1A) increased the expression of liver-associated genes and suppressed undesired features, demonstrating that HLC differentiation can be improved through the targeted stimulation of specific GRN.
      Figure thumbnail gr1
      Fig. 1HLC express liver and intestine genes.
      (A) Schematic of the iPSC to HLC differentiation protocol. (B) Fluorescence imaging of immunocytochemical stainings of ALB, AFP and DDP-IV (green), F-actin (phalloidin-rhodamine; red) and DAPI (blue) in day 25 HLC. (C) Canalicular export of CMFDA and SNARF1. The time after addition of fluorophores is indicated. Exemplary bile canaliculi indicated by arrows. (D) Lipid droplet accumulation visualized by fluorescent imaging of day 25 HLC treated with 0.8 mM oleic acid and untreated day 25 control HLC stained with BODIPY, phalloidin-rhodamine and DAPI. (E) Time-resolved qPCR data for selected markers. The x-axis depicts the time period of HLC differentiation (d0-d25), freshly isolated hepatocytes (PHH) and PHH after cultivation periods of 1 and 3 days, as well as Caco-2 cells. Error bars represent standard deviation of n=5 replicates of the HLC differentiation time-course, n = 2 donors of primary hepatocytes and n = 1 for Caco-2 cells. ATAC-seq, assay for transposase-accessible chromatin followed by sequencing; DE, definitive endoderm; HLC, hepatocyte-like cells; HTS, high throughput screening; iPSC, induced pluripotent stem cells; M1, maturation medium; M2, maintenance medium; OA, oleic acid; P, progenitor medium; PHH, primary human hepatocytes; RRBS, reduced-representation bisulfite sequencing; TS, thawing and seeding medium. (This figure appears in color on the web.)

      Materials and methods

      Detailed protocols of all techniques are available in the supplementary methods.

      hiPSC culture and differentiation of HLCs

      A differentiation protocol (Takara Bio Europe) was used to differentiate hiPSCs (ChiPSC18, ChiPSC22, JHU106) via definitive endoderm (DE) to HLC in a 25-day procedure
      • Godoy P.
      • Schmidt-Heck W.
      • Natarajan K.
      • Lucendo-Villarin B.
      • Szkolnicka D.
      • Asplund A.
      • et al.
      Gene networks and transcription factor motifs defining the differentiation of stem cells into hepatocyte-like cells.
      with modifications as described in the supplementary methods. Time-resolved analysis of the genes ALB, HNF4A, HNF1A, NR1H4, ABCB11, POU5F1, NANOG, CXCR4, SOX17, FOXA2, CDX2, KLF5, ISX, TWIST, SNAI1 was performed by quantitative reverse-transcription PCR (primers and conditions are described in the supplementary methods). For further characterization, immunostainings for albumin (ALB), caudal type homeobox 2 protein (CDX-2) and anterior gradient 2 (AGR2) were performed (Sigma-Aldrich, HPA031024, HPA045669 and HPA007912, 1:200), which were subjected to high-throughput imaging using the ImageXpress Micro XLS (Molecular Devices). Imaging of canalicular export in cultures of HLC and PHH using the red- and green-fluorescent BSEP and MRP2 substrates SNARF-1 (Invitrogen) and 5-chloromethylfluorescein diacetate (CMFDA: Invitrogen), respectively, was performed according to the methods of Vartak et al. (2020) with modifications.

      Lentiviral transduction and FXR agonist treatment

      Cells were differentiated according to the described protocol above (or Fig. 1A). On day 13 of differentiation, cells were transduced with lentiviral particles (VectorBuilder) carrying FXR and EGFP constructs under control of separate promoters. Transduction efficiency was assessed by EGFP fluorescence. On day 22 and day 24, cells were exposed to FXR agonists chenodeoxycholic acid (CDCA, 100 μM) and GW4064 (1.5 μM). Secretion of 5-CMF and the area of lipid droplets were studied by imaging and image analysis as described in the supplementary materials and methods.

      Omics methods

      Single-cell RNA-seq (scRNA-seq) of HLC (2 independent preparations, 183 cells) and PHH (3 donors, 288 cells) was performed as described. Reduced representation bisulfite sequencing (RRBS), ATAC-seq, and RNA-seq as well as data analysis followed published protocols with details given in the supplementary materials and methods.

      Biostatistical analysis

      Differential pattern group (DPG) plots were generated by plotting the log2 fold change of HLC over PHH on the y-axis and the log2 fold change of iPSC over PHH on the x-axis for each gene. Tissue identity enrichment analysis, gene ontology enrichment analysis, overrepresentation analysis for transcriptional regulators and reactome pathway enrichment analysis for gene sets, as well as CellNet analysis, determination of overlap ratios, RRBS-, ATAC-seq-, and scRNA-seq data analyses, transcription factor motif enrichment in differentially methylated regions, and transcriptional regulator enrichment analysis were performed as described in the supplementary materials and methods.

      Results

      Stem cell-derived HLCs show hepatic and intestinal features

      Differentiation of hiPSC via DE to HLCs was performed according to a previously reported protocol
      • Godoy P.
      • Schmidt-Heck W.
      • Natarajan K.
      • Lucendo-Villarin B.
      • Szkolnicka D.
      • Asplund A.
      • et al.
      Gene networks and transcription factor motifs defining the differentiation of stem cells into hepatocyte-like cells.
      with modifications (Fig. 1A). HLCs showed positive immunostaining for ALB, alpha-fetoprotein (AFP) and the apical hepatocyte marker dipeptidyl peptidase IV, which visualized thin canaliculi between the HLC (Fig. 1B). Canalicular excretion was active, as evidenced by CMFDA and carboxy-SNARF-1
      • Vartak N.
      • Guenther G.
      • Joly F.
      • Damle-Vartak A.
      • Wibbelt G.
      • Fickel J.
      • et al.
      Intravital dynamic and correlative imaging reveals diffusion-dominated canalicular and flow-augmented ductular bile flux.
      (Fig. 1C). Upon the addition of oleic acid to the culture medium, HLC formed lipid droplets, which were labeled with BODIPY (Fig. 1D; Fig. S1).
      HLC were characterized by quantitative reverse-transcription PCR using marker genes (Fig. 1E). The pluripotency-associated transcription factors POU5F1 and NANOG showed the highest expression levels in iPSC, which decreased during differentiation to levels below those observed in PHH, while markers of DE (CXCR4, SOX17, FOXA2) showed stage-specific upregulation. In HLC, increased RNA expression was observed for the liver markers ALB, HNF4, HNF1A, NR1H4 (FXR), and ABCB11 (BSEP), although levels remained below those of freshly isolated and cryopreserved PHH. It should be noted that the expression of hepatocyte markers, particularly ALB and ABCB11, decreased rapidly upon the cultivation of PHH, demonstrating the importance of using freshly isolated cells as a reference. The expression of the intestine-associated transcription factors CDX2, KLF5 and ISX significantly increased compared to the levels observed in PHH.

      Genome-wide analysis identifies GRNs that could be targeted to improve HLC differentiation

      Having characterized the phenotype of the stem cell-derived HLC, we then performed RNA-seq on iPSC, DE, HLC and PHH bulk populations. In a principal component analysis (PCA), HLCs were shown to locate more than half-way from iPSC along PC1 towards primary hepatocytes, while DE clustered closely with iPSCs (Fig. 2A). This observation led to the analysis of the similarity of the bulk transcriptomes to liver and intestine using CellNet
      • Radley A.H.
      • Schwab R.M.
      • Tan Y.
      • Kim J.
      • Lo E.K.W.
      • Cahan P.
      Assessment of engineered cells using CellNet and RNA-seq.
      (Fig. 2B). Compared to the liver reference, iPSC, DE and HLC showed a GRN status similarity of 5%, 17% and 80%, respectively. For the intestinal phenotype, iPSC, DE and HLC showed GRN status similarity of 7%, 11% and 49%, further confirming that HLC express a substantial number of intestine-associated genes (additional tissue classification scores in Fig. S2).
      Figure thumbnail gr2
      Fig. 2OMICS characterization of HLC, iPSC, DE and PHH bulk preparations.
      (A) PCA of the 1,000 most variable genes among iPSC, DE, and HLC obtained by RNA-sequencing. (B) CellNet analysis of RNA-sequencing data for liver and intestine tissue identities (GRN) of the indicated cell populations in comparison to ESC, liver and intestine CellNet training datasets. Error bars indicate standard deviation of n=4 iPSC, DE and HLC samples and n=52 ESC-, n=64 intestine- and 33 liver-training samples, (C) Supervised logical clustering of differentiation trajectories obtained from RNA-sequencing data, visualized in the differentiation pattern plot (DiPa plot). The x-axis represents log2 fold changes of iPSC over PHH, the y-axis indicates log2 fold changes of HLC over PHH. Dotted lines represent cut-offs of the clustering approach. (D) Integrated heatmap of transcriptome, methylation and chromatin accessibility data obtained by RNA-seq, RRBS and ATAC-seq for iPSC (n=4), DE (n=4), HLC (n=4) and PHH (n=6 for expression and RRBS data; n=3 for ATAC-seq data) (columns). Data is represented as z-scores calculated for each assay. The identity of each gene (rows) was mapped to DPG as obtained from supervised clustering. c1: gene cluster with increased chromatin accessibility and gene expression in HLC compared to PHH; c2: decreased chromatin accessibility and gene expression in HLC compared to PHH. The heatmap was produced using the R pheatmap package; rows and columns were clustered by correlation. (E) Rank score (y-axis) and significance (x-axis) of FXR (NR1H4) binding site overrepresentation analysis with RegulatorTrail in promoters associated with closed chromatin in HLC for the 10 DPG of C. PCA, principal component analysis. DE, definitive endoderm; DPG, differentiation pattern group; ESC, embryonic stem cells; HLC, hepatocyte-like cells; iPSC, induced pluripotent stem cells; ORA, overrepresentation analysis; PHH, primary human hepatocytes. (This figure appears in color on the web.)
      We next applied the DEseq2 pipeline (Table S1) to establish a supervised clustering approach, visualized in the differentiation pattern (DiPa) plot (Fig. 2C). Here, genes were plotted based on their expression fold changes in iPSC (x-axis) and HLCs (y-axis) compared to the target population, PHH. Genes with similar expression changes between iPSC and HLC in relation to PHH were then categorized into 11 differentiation pattern groups (DPG0-10) (Fig. 2C; supplementary materials and methods). Genes expressed at a similar level in HLC and PHH are found in DPG0. Genes in DPG1 and 6 were expressed at a similar level in both iPSC and HLC and thus unaltered by the differentiation protocol. DPG3 and 8 genes show favorable up- and downregulation, respectively, whereby favorable means that gene expression in HLC comes close to that of PHH. DPG2 and 7 genes were insufficiently up- and downregulated; whereas DPG4 and 9 contain genes that were excessively up- and downregulated, respectively. Genes in DPG5 and 10 were adversely up- or downregulated and therefore represent misguided differentiation.
      Next, several techniques were applied to characterize the genes in the individual DPG, summarized in Tables S2-7. An overrepresentation of liver-associated genes was observed in DPG1, 2 and 3 using TissueEnrich, in agreement with an overrepresentation of gene ontology terms related to xenobiotic metabolism and Reactome pathways related to xenobiotic and bile acid metabolism in DPG2. In contrast, an enrichment of genes associated with the intestine and gallbladder was found in DPG4, and RegulatorTrail overrepresentation analysis indicated a strong enrichment of HNF1B binding sites, a transcription factor implicated in cholangiocyte differentiation.
      • Limaye P.B.
      • Alarcón G.
      • Walls A.L.
      • Nalesnik M.A.
      • Michalopoulos G.K.
      • Demetris A.J.
      • et al.
      Expression of specific hepatocyte and cholangiocyte transcription factors in human liver disease and embryonic development.
      DPG5 also showed intestinal enrichment, with overrepresentation of transcription factor binding sites for CDX2 and KLF5, as well as binding sites for the epigenetic regulator MBD4.
      From the complex information summarized in Table S7, we were particularly intrigued by the significant enrichment of target genes regulated by the metabolism-controlling transcription factor FXR (NR1H4) in DPG2, which contains genes that show upregulation in HLC compared to iPSC, but not sufficient upregulation to reach the levels observed in PHH. Interestingly, despite FXR being expressed (Fig. 1E), liver-associated target genes were not upregulated in HLC (Fig. S3). One may therefore hypothesize that a lack of FXR agonists is responsible for FXR activity being too low.

      Integrative OMICS analysis identifies potentially FXR-responsive chromatin among insufficiently upregulated genes

      The observed differences between HLC and PHH may be due to an epigenetic landscape that does not allow expression of its dependent genes despite the presence of FXR. Indeed, the enrichment of epigenetic regulators, such as MBD4 and KMT2A may indicate altered chromatin in DPG1 and 2 (Table S7) and adverse upregulation (DPG5) and insufficient downregulation (DPG7) may be a result of high chromatin accessibility.
      To gain deeper insight into the interplay between gene expression and the epigenetic background in HLC, we performed RRBS and ATAC-seq of iPSC, DE, HLC and PHH. The resulting integrated heatmap visualizes key features of differential gene expression, promoter methylation, and chromatin accessibility in relation to the DPG (Fig. 2D; Table S8; Fig. S4). A large set of genes in iPSC, DE and HLC showed hypermethylation compared to PHH (Fig. 2D). However, promoter hypermethylation did not necessarily correspond to decreased RNA expression, as observed in the region with open chromatin indicated as ‘c1’, including a cluster of adversely upregulated genes (DPG5). Thus, despite promoter hypermethylation, upregulation of DPG5 genes, including intestine-associated genes, such as CDX2, ISX, and GATA6, occurred during the differentiation of iPSC into PHH.
      Like ‘c1’, region ‘c2’ was also characterized by promoter hypermethylation, but had less open chromatin in HLC, and was consistently associated with RNA expression that was too low. This region mostly contained genes from DPG1 and 2, which were shown to be enriched in FXR-binding motifs (especially genes from DPG2) (Table S7). Among the 7,714 genes with closed chromatin, FXR-regulated genes were most notably enriched in DPG2 (Fig. 2E). Also, DNA methylation status at promoters showed significant associations with FXR-dependent genes, particularly in DPG2 (Fig. S5).

      Single-cell sequencing reveals a liver-intestine hybrid state in in vitro-derived HLCs

      In the previous paragraphs, we identified FXR as a possible key factor required for HLC maturation. However, manipulation of FXR signaling would not be a promising strategy to improve differentiation if HLC consisted of distinct cell populations, e.g. with hepatocyte and intestinal phenotypes. In this case, cell sorting or other selection strategies would instead be adequate. Alternatively, the expression of intestinal genes may be a result of incomplete cell fate decision-making and reshaping of the epigenetic landscape, leaving the HLC in a hybrid state. In such a scenario, manipulation of FXR signaling, which is involved in hepatic as well as intestinal gene regulation, may contribute to a shift in the direction to either the hepatic or intestinal cell state. To address this, we used scRNA-seq to study if HLC consisted of distinct subpopulations. Analysis of 2 biological HLC replicates and PHH from 3 donors showed that HLC and PHH populations clearly separated in the PCA (Fig. 3A). All 3 PHH donor populations clustered together, as did the biological replicates of HLC. No distinct HLC subpopulations were apparent in the PCA plot and Louvain clustering identified a single HLC cluster. RNA velocity analysis demonstrated that the HLCs of both replicates approach the PHH cluster (Fig. S6).
      Figure thumbnail gr3
      Fig. 3Differentiation of HLCs via a hybrid cell state in vitro.
      (A) PCA of the top 1,000 most variable genes obtained in single-cell sequencing of 2 replicates of day 25 HLC and cryopreserved PHH of 3 different donors. (B) Co-expression plots for selected gene-pairs representative of liver and intestine phenotypes presented as log2 CPM for each gene and cell. (C) Co-expression of AFP and CDX2 in single HLC and PHH (left panel) and co-immunostaining (right panel); AFP: red, cytoplasmic; CDX2: green, DAPI: blue. (D) Liver- (y-axis) and hybrid state gene (x-axis) expression z-scores in representative replicates of iPSC, HLC and PHH. The color scale indicates expression of FXR-dependent genes of the CellNet liver and colon GRN. (E) Single-cell protein levels of ALB, CDX2 and AGR2 determined by high-throughput screening of immunostainings of at least 10,000 cells per time point. Red dots indicate mean intenstiy; error bars indicate standard deviation. Numbers shown below the violins indicate the fraction of positive cells. GRN, gene regulatory network; HLC, hepatocyte-like cells; iPSC, induced pluripotent stem cells; PHH, primary human hepatocytes. (This figure appears in color on the web.)
      We then addressed the hypothesis that hepatocyte-associated and non-hepatocyte genes, are expressed in the same HLC. For this purpose, we selected hepatocyte- and intestine-associated genes from CellNet and plotted their pairwise expression levels in single HLC (Fig. 3B). As expected, PHH expressed high levels of hepatocyte-associated genes, while intestine-associated genes were below the limit of detection. In contrast, HLC not only expressed hepatocyte-associated genes, but variable levels of intestinal genes were also detectable in the same individual cells. This was confirmed by immunofluorescence co-staining of AFP and CDX2 (Fig. 3C), as well as AGR2 and HNF4α (Fig. S7) in HLC. Moreover, heatmaps of networks generated using the CellNet listings for liver and intestine-associated genes demonstrated the mixed identity of HLC (Fig. S8).
      We next described this ‘hybrid state’ on a genome-wide scale and additionally included a publicly available scRNA-seq dataset generated by Camp et al. (2017) that allowed us to define ‘adult liver genes’ as those showing at least a log2 fold change of 5 between iPSCs and PHH based on scRNA-seq data. Non-hepatocyte-associated genes further referred to as ‘hybrid state genes’, were derived from the present scRNA-seq data using a modified definition of DPG4 and 5 (supplementary materials and methods) which, according to the DiPa procedure, isolates genes that are excessively or adversely upregulated after differentiation. The mean scaled expression (z-scores) of all hybrid state genes was then analyzed in relation to the mean of the adult liver genes in individual HLC (Fig. 3D, Table S9). As per definition, adult liver genes displayed a strong increase in expression from iPSC to adult PHH, while this increase was less pronounced in HLC. In contrast to PHH, HLCs showed an increase in hybrid state gene expression that coincided with the upregulation of adult liver genes in the same cells. The HLC hybrid state genes identified in the present study and in Camp et al. (2017) overlapped 7.12 times more than randomly expected and showed enrichment of intestinal tissue identity in the overlap (Fig. S9), which is remarkable considering that these cells were generated by different laboratories using distinct iPSC lines. Highlighting the single cells according to the mean FXR target gene expression for the CellNet FXR liver and colon GRN, showed that PHH activate the liver-associated, and HLC mostly the colon-associated FXR GRN (Fig. 3D), once again indicating that activation of FXR signaling may improve HLC differentiation.
      Finally, further support of the HLC hybrid state concept was provided by a high-throughput screen of immunofluorescence staining of ALB and 2 intestinal markers (CDX2 and AGR2) (Fig. 3E; Fig. S10). ALB was detected in 87% of HLC on day 25, but the protein abundance of the intestinal genes CDX2 and AGR2 also increased during HLC differentiation up to day 25, leading to positive cell fractions of 94% and 78%, respectively, demonstrating that liver and intestinal protein expression occur in the same cell.

      The liver-intestine hybrid state is not a feature of FHs ex vivo

      In order to clarify whether the hybrid state of HLC observed in vitro also occurs in vivo during fetal gestation, we included fetal hepatocytes (FH) isolated at weeks 10 and 17 of gestation from a published data set
      • Camp J.G.
      • Sekine K.
      • Gerber T.
      • Loeffler-Wirth H.
      • Binder H.
      • Gac M.
      • et al.
      Multilineage communication regulates human liver bud development from pluripotency.
      in the analysis. A t-distributed stochastic neighbor embedding (t-SNE) representation of the integrated dataset used in the present study showed iPSC, HLC and adult PHH in separate clusters (Fig. 4A, Fig. S11). In line with the PCA analysis (Fig. 3A), the t-SNE plot did not indicate the presence of subpopulations of HLC. In contrast, FH consisted of 2 subpopulations at week 10, and of several at week 17.
      Figure thumbnail gr4
      Fig. 4The liver-intestine hybrid state is not induced in subpopulations in vitro.
      (A) t-SNE plot of HLC Replicates 1 and 2, PHH from donors 1, 2 and 3 together with single-cell sequencing data of FH at gestational week 10 and 17 from Camp et al. (2017). (B) Expression (log2 CPM) of marker genes representative of hepatocytes, hematopoietic and intestinal cells projected onto the t-SNE plot from A). (C) Liver- (y-axis) and hybrid state gene (x-axis) expression z-scores in FH week 10, FH week 17 and PHH. The color scale indicates expression of the hepatocyte marker ALB, the erythropoiesis marker KLF1 and FXR-dependent genes of the CellNet liver and colon GRN. (D) Correlation-based heatmap of regulon activity discriminating iPSC, day 25 HLC, PHH and FH populations created with the R pheatmap package. FH, fetal hepatocytes; HLC, hepatocyte-like cells; iPSC, induced pluripotent stem cells; PHH, primary human hepatocytes. (This figure appears in color on the web.)
      We next visualized the expression of selected hepatocyte-, and intestine-associated genes in the same t-SNE plot (Fig. 4B). Since liver isolates from fetal stages of hepatogenesis commonly include cell types other than hepatocytes or hepatic progenitors, such as blood forming cells, we also included hematopoiesis-associated genes. FH consisted of 2 types of cells: (i) a ‘hematopoietic cluster’ that contained FH from week 10 and to a minor degree from week 17 with high expression of hematopoietic genes (KLF1, TAL1, and GATA1) but low expression of liver and intestinal genes, and (ii) ‘hepatocyte precursor clusters’ that expressed hepatocyte-associated genes (ALB, HNF4Aα, and AFP) and no hematopoietic genes. Importantly, no expression of the intestine-associated genes was observed in the hepatocyte precursor clusters of FH (Fig. 4B). Further profiling of FH on a genome-wide scale (Fig. 4C) showed an increase in adult liver genes in the hepatocyte precursor cell population, while the increase in hybrid state genes was lower compared to HLC shown in Fig. 3D. The slight increase in hybrid state gene expression in week 17 FH should be interpreted with caution, since these cells were transiently taken into culture for purification.
      • Camp J.G.
      • Sekine K.
      • Gerber T.
      • Loeffler-Wirth H.
      • Binder H.
      • Gac M.
      • et al.
      Multilineage communication regulates human liver bud development from pluripotency.
      The color scale in Fig. 4C illustrates the distinct hepatocyte precursor and hematopoietic cell clusters of week 17 FH, expressing either ALB or KLF1, respectively. The specific features of the individual cell types were further characterized by regulon activity analysis (Fig. 4D, Table S10).
      • Aibar S.
      • González-Blas C.B.
      • Moerman T.
      • Huynh-Thu V.A.
      • Imrichova H.
      • Hulselmans G.
      • et al.
      SCENIC: single-cell regulatory network inference and clustering.
      HLC exhibited a liver-intestine hybrid state characterized by the simultaneous activity of liver-associated (NR1H4, CEBPG, NFIL3, etc.) and intestine-associated genes and regulators (PRDM1, KLF5, FOSL2, CDX2, etc.) in the same cell. In contrast, these intestine-associated regulators were not observed in FH ex vivo (Fig. 4D).

      Manipulation of FXR expression and activity enhances HLC maturation

      As described, FXR represents a candidate for targeted intervention designed to improve the differentiation of HLC. Gene regulation through FXR seems to depend not only on its expression, because mRNA levels of FXR (NR1H4) in HLC are relatively similar to those of PHH (Fig. 1E). Nevertheless, many liver-associated FXR target genes remained at lower expression levels in HLCs (Fig. S3). These observations prompted us to investigate if, besides exogenous FXR expression, the additional stimulation by agonists may cause an increase in liver-associated gene expression, while simultaneously decreasing intestinal FXR target gene expression. We proceeded to combine lentiviral expression of FXR and exposure to FXR agonists, CDCA (100 μM) and/or GW4064 (1.5 μM) (Fig. 5A). Lentiviral expression of FXR caused the expected increase in FXR mRNA levels and also induced expression of the FXR-dependent gene ABCB11 (BSEP) (Fig. 5B); immunostaining confirmed nuclear FXR and bile canalicular BSEP expression upon FXR activation (Fig. S12). Although a large overlap was obtained among the different interventions (Fig. 5C), treatment of FXR-transduced HLC with both agonists (CDCA and GW4064) – referred to as FXR intervention hereafter – induced a stronger response in expression compared to FXR plus each agonist alone (Fig. 5D; Figs. S13 and S14; Table S11). This was also visualized by expression changes in liver and colon FXR network genes, demonstrating that liver-associated genes were mostly upregulated upon FXR intervention, while less liver-associated genes decreased (Fig. 5E).
      Figure thumbnail gr5
      Fig. 5Experimental enhancement of FXR activity improves HLC maturation.
      (A) Intervention schedule for FXR transduction (MOI 20) and agonist treatment (1.5 μM GW4064; 100 μM CDCA). (B) Influence of the interventions on FXR (NR1H4) and BSEP (ABCB11) expression analyzed by qRT-PCR. Error bars represent standard deviation. Significance was calucalted using a two-tailed t test (∗∗p <0.01; ∗∗∗p <0.001). (C) Overlap of differentially expressed genes (log2 fold change > |1.5|; p value < 0.01) of FXR and FXR plus agonist interventions compared to vehicle control. (D) Differential genes with the highest absolute fold changes and FDR-adjusted p values lower than 0.001 for FXR-transduced and agonist (1.5 μM GW4064; 100 μM CDCA)-stimulated HLC. (E) Correlation-based heatmaps visualizing expression changes larger than a log2-fold change of |1.5| for liver- and intestine-associated genes for all interventions. Heatmaps were created with the R pheatmap package. iPSC, induced pluripotent stem cells; DE, definitive endoderm; HLC, hepatocyte-like cells; PHH, primary human hepatocytes; TS, thawing and seeding medium; P, progenitor medium; M1, maturation medium; M2, maintenance medium; CDCA, chenodeoxycholic acid; log2FC, log2 fold change. (This figure appears in color on the web.)
      To characterize the effect and study the reproducibility of the FXR intervention, the differentiation experiments and RNA-seq analyses were repeated using ChiPSC18, the cells used in all previous experiments, and 2 further hiPSC cell lines, ChiPSC22 and JHU106 (Figs. S15-18, Tables S12-21). Although the number of DEGs due to FXR intervention differed between the hiPSC lines, their overlap was more than 5,400-fold higher than randomly expected, demonstrating the high degree of reproducibility (Fig. 6A). Similar to the first set of RNA-seq experiments, the expression of selected liver-associated FXR network genes predominantly increased, while hybrid state genes predominantly decreased (Fig. 6B, Fig. S19 and 20). For an unbiased, genome-wide assessment of the FXR intervention, DiPa plot analysis was applied, where the means of the differences of gene expression with and without FXR intervention were calculated for each DPG and visualized by a background color code (Fig. 6C; Fig. S21). For the 3 hiPSC lines, DPG that should increase to reach similar levels as PHH were consistently induced (DPG 1, 2, 9, 10), while DiPa groups that should decrease were suppressed (DPG 3, 4, 5, 6). An exception was DPG7, where an up- instead of the desired downregulation was obtained. To identify the GRNs most influenced by the FXR intervention, CellNet network influence scores were calculated, demonstrating increased scores for liver (NR1H4, ONECUT1, NR0B2) and intestinal identity (CDX2, ISX, HNF4A) of HLC (Fig. S22A). This was also reflected in the CellNet GRN status describing intestinal cell identity, but not in the liver GRN status (Fig. S22B). For calculation of the liver GRN status, CellNet considers all genes from liver-associated transcription factor networks (13 networks, 429 genes total) and 8 additional genes, but does not include gene expression outside of those largely overlapping networks. However, gene expression changes outside of these networks may still contribute to improvement of HLC differentiation, as demonstrated by our independent analysis and reflected in the CellNet intestine GRN status. Using a second differentiation protocol (Wang et al. 2016) in ChiPSC18, we also observed a relatively large overlap of FXR-mediated expression changes with our standard protocol (Fig. S23), although the hybrid state analysis was based on the first protocol.
      Figure thumbnail gr6
      Fig. 6Increased liver and decreased hybrid state gene expression upon FXR activation.
      (A) Differential genes due to FXR activation (FXR transduction plus 1.5 μM GW4064 and 100 μM CDCA) of the iPSC lines ChiPSC18, ChiPSC22 and JHU106 with FDR-adjusted p values lower than 0.001. The overlap was assessed by the OR, where the randomly expected overlap is 1.0. (B) Examples of increased gene expression of liver-associated and decreased expression of intestine- and brain-associated genes due to FXR activation. Green stripes indicate the samples with the FXR intervention. Error bars indicate standard deviation of n=3 iPSC (ChiPSC18,JHU106), DE and HLC samples, n = 2 iPSC (ChiPSC22) and n = 4 Colon-, n = 4 PHH. (C) Influence of FXR activation visualized by the DiPa plot. Red and blue dots indicate genes significantly up- or downregulated by the FXR intervention. The background color indicates mean up- (red) or downregulation (blue) of genes within a DiPa group. The respective numbers are given in . DE, definitive endoderm; HLC, hepatocyte-like cells; iPSC, induced pluripotent stem cells; OR, overlap ratio; PHH, primary human hepatocytes. (This figure appears in color on the web.)
      Finally, we investigated if FXR activation has functional consequences, by quantification of the secretion of the fluorophore 5-CMF into bile canaliculi
      • Vartak N.
      • Guenther G.
      • Joly F.
      • Damle-Vartak A.
      • Wibbelt G.
      • Fickel J.
      • et al.
      Intravital dynamic and correlative imaging reveals diffusion-dominated canalicular and flow-augmented ductular bile flux.
      ) (Fig. 7A). The kinetics of fluorescence in the canalicular lumen showed an initial increase followed by a plateau and were fitted to an exponential association function to characterize the increase in canalicular fluorescence by the time constant Tau (the smaller the faster the increase) and the amplitude (as a measure of the fluorescence in the plateau phase). Analysis of the secretion kinetics demonstrated that FXR activation led to a significantly faster increase of fluorescence in the canaliculi and to a higher amplitude (Fig. 7B,C).
      Figure thumbnail gr7
      Fig. 7FXR activation increases canalicular secretion and lipid droplet formation.
      (A) Video stills showing the secretion of green fluorescent 5-CMF into bile canaliculi after exposure of HLC to CMFDA with and without FXR activation (FXR transduction, 1.5 μM GW4064 and 100 μM CDCA). Left margin: seconds after CMFDA addition. Quantification of the time constant (B) and amplitude (C) of canalicular fluorescence intensity, each dot representing the data of an individual bile canaliculus; ∗∗∗∗p <0.0001 (Wilcoxon test, unpaired, two-sided). (D) Images of HLC with and without FXR activation, BSA control and oleic acid treatment. Lipid droplets were visualized by AdipoRed. (E) Quantification of the total area of lipid droplets per cell. BSA: Control HLC with 400 μM bovine serum albumin; Control: HLC without FXR transduction and agonist treatment; OA800: HLC treated with 800 μM oleic acid; ∗∗∗∗p <0.0001; ∗p <0.1 (Wilcoxon test, unpaired, two-sided). (J) Schematic of the consequences of FXR activation. BSA, bovine serum albumin; FXRi; FXR intervention; OA800, oleic acid 800 µM. (This figure appears in color on the web.)
      As a second functional property, lipid droplet formation was induced by the addition of oleic acid to the culture medium of HLC and visualized by AdipoRed (Fig. 7D). The total area of lipid droplets was analyzed on the single cell level by employing the Cellpose machine learning algorithm. In control HLC, a statistically significant but only small increase in lipid droplet area was induced by oleic acid (Fig. 7E). After FXR activation, the area strongly increased compared to the controls. Thus, the FXR intervention increased canalicular secretion and formation of lipid droplets in HLC (Fig. 7F).

      Discussion

      Hepatocytes are an important tool in drug development and preclinical research
      • Nussler A.
      • Konig S.
      • Ott M.
      • Sokal E.
      • Christ B.
      • Thasler W.
      • et al.
      Present status and perspectives of cell-based therapies for liver diseases.
      ; yet their limited availability and high cost encourage the development of stem cell-based differentiation protocols. Despite remarkable achievements in the field, the use of stem cell-derived hepatocytes is hampered by their incomplete differentiation.
      • Sachinidis A.
      • Albrecht W.
      • Nell P.
      • Cherianidou A.
      • Hewitt N.J.
      • Edlund K.
      • et al.
      Road map for development of stem cell-based alternative test methods.
      ,
      • Godoy P.
      • Schmidt-Heck W.
      • Hellwig B.
      • Nell P.
      • Feuerborn D.
      • Rahnenführer J.
      • et al.
      Assessment of stem cell differentiation based on genome-wide expression profiles.
      Genome-wide studies have demonstrated that human embryonic- and iPSC-derived hepatocytes exhibit inadequate expression of a large cluster of metabolism-associated genes expressed in PHH but also excessive expression of genes not expressed in PHH, such as genes normally expressed in the gastrointestinal tract.
      • Godoy P.
      • Schmidt-Heck W.
      • Hellwig B.
      • Nell P.
      • Feuerborn D.
      • Rahnenführer J.
      • et al.
      Assessment of stem cell differentiation based on genome-wide expression profiles.
      ,
      • Godoy P.
      • Schmidt-Heck W.
      • Natarajan K.
      • Lucendo-Villarin B.
      • Szkolnicka D.
      • Asplund A.
      • et al.
      Gene networks and transcription factor motifs defining the differentiation of stem cells into hepatocyte-like cells.
      ,
      • Gao X.
      • Li R.
      • Cahan P.
      • Zhao Y.
      • Yourick J.J.
      • Sprando R.L.
      Hepatocyte-like cells derived from human induced pluripotent stem cells using small molecules: implications of a transcriptomic study.
      To characterize the stem cell-derived HLC and to design a strategy to improve their differentiation, we performed RNA-seq analysis of bulk preparations and single cells, characterized the epigenetic landscape, and performed targeted interventions to optimize the activity of GRN.
      Bulk sequencing of HLC followed by CellNet analysis demonstrated 80% liver and 49% intestine identity, confirming previous reports.
      • Godoy P.
      • Schmidt-Heck W.
      • Natarajan K.
      • Lucendo-Villarin B.
      • Szkolnicka D.
      • Asplund A.
      • et al.
      Gene networks and transcription factor motifs defining the differentiation of stem cells into hepatocyte-like cells.
      The expression of liver- and intestine-associated genes in HLC could theoretically be explained by the existence of 2 subpopulations. However, scRNA-seq did not provide evidence of this. In contrast, the results clearly demonstrate that HLC exist in a hybrid state, where genes representative of at least 2 adult cell types are expressed within the same cell. This hybrid state could be confirmed in an independent HLC dataset.
      • Camp J.G.
      • Sekine K.
      • Gerber T.
      • Loeffler-Wirth H.
      • Binder H.
      • Gac M.
      • et al.
      Multilineage communication regulates human liver bud development from pluripotency.
      In the future, profiling of HLC from additional laboratories will provide further evidence to determine whether the hybrid state represents a universal feature of stem cell-derived HLC.
      An important question addressed in the current study is if HLC differentiation can be improved by inducing liver gene expression and suppressing hybrid state genes. To identify promising candidates and to assess the efficacy of interventions, we established the DiPa technique that projects each gene onto a coordinate system in which the ratio of HLC/PHH is plotted against iPSC/PHH. Consequently, groups of genes with similar development during the differentiation process cluster together within DPGs. The DiPa plot in combination with enrichment analyses guided us to the nuclear receptor FXR (NR1H4). An interesting feature of this transcription factor is that FXR-dependent genes were not only enriched among the inadequately expressed liver genes of DPG2 and DPG3 but also among the excessively expressed intestinal genes of DPG4 and DPG5, in agreement with previous studies demonstrating that FXR is a key regulator of hepatic as well as intestinal gene expression.
      • Wagner M.
      • Fickert P.
      • Zollner G.
      • Fuchsbichler A.
      • Silbert D.
      • Tsybrovskyy O.
      • et al.
      Role of farnesoid X receptor in determining hepatic ABC transporter expression and liver injury in bile duct-ligated mice.
      ,
      • Pereira-Fantini P.M.
      • Lapthorne S.
      • Joyce S.A.
      • Dellios N.L.
      • Wilson G.
      • Fouhy F.
      • et al.
      Altered FXR signalling is associated with bile acid dysmetabolism in short bowel syndrome-associated liver disease.
      This led to the question of whether experimentally increasing FXR activity in HLCs would lead to an increase in DPG2 and DPG3 genes, while suppressing DPG4 and DPG5 genes, or if genes belonging to DPG4 and DPG5 would increase, thus promoting the undesirable intestinal phenotype. A further argument speaking for an intervention to activate FXR is that regions with closed chromatin were enriched in FXR-dependent genes and FXR is known to remodel chromatin to an open configuration. One interesting aspect was that FXR-dependent genes showed inadequate expression even though FXR itself was present, which prompted us to stimulate cells with FXR agonists, in addition to exogenously expressing FXR. Thus, we applied an intervention strategy with lentiviral FXR expression to guarantee that its levels are not limiting, together with stimulation by the endogenous FXR agonist CDCA, the synthetic FXR agonist GW4064, and a combination of both. The largest influence on gene expression was obtained by combined FXR expression and agonist exposure. Importantly, increased FXR activity not only enhanced FXR-dependent liver genes in DPG1 and DPG2, but also suppressed intestinal genes in DPG4 and DPG5. While FXR activation strongly enhanced the expression of individual liver-specific genes, the extent of the decrease of hybrid genes associated with the intestinal phenotype, although statistically significant, was comparatively moderate. FXR activation increased secretion of 5-CMF into bile canaliculi, which was due to increased expression of BSEP. Moreover, FXR increased the formation of lipid droplets upon exposure to oleic acid. The influence of FXR on lipid metabolism is complex, since on one hand FXR is known to supress lipogenesis via downregulation of SREBP leading to the therapeutic effect of, for example, obeticholic acid, while on the other hand FXR activity leads to triglyceride accumulation and the enlargement of lipid droplets via FXR-CREB under exposure to exogenous fatty acids (Wu et al., 2020).
      Since major differences between HLC and PHH remained after activation of FXR, particularly in DPG7, it will be important in future work to combine the targeting of FXR with further interventions to learn if there is a tipping point beyond which the entire expression profile is brought closer to PHH. For instance, reducing the activity of the transcription factors CDX2, GATA6, and KLF5 may further decrease the expression of hybrid state genes. Such a strategy is experimentally feasible, because the addition of viruses and agonists to the culture medium during differentiation does not require much additional experimental effort.
      In conclusion, we demonstrate that hiPSC-derived HLC generated by commonly used in vitro protocols co-express both liver and undesired intestinal genes within the same cell, and the nuclear factor FXR represents a critical control factor that can be targeted to shift the balance from a liver-intestine hybrid cell towards a hepatocyte.

      Abbreviations

      AFP, alpha-fetoprotein; AGR2, anterior gradient 2; ALB, albumin; CDCA chenodeoxycholic acid; CDX-2, caudal type homeobox 2 protein; CMFDA, 5-chloromethylfluorescein diacetate; DE, definitive endoderm; DiPa, differentiation pattern; DPG, differentiation pattern group; FH, fetal hepatocytes; FXR, farnesoid X receptor; GRN, gene regulatory networks; HLC, hepatocyte-like cells; (h)iPSC, (human) induced pluripotent stem cells; PCA, principal component analysis; PHH, primary human hepatocytes; RRBS, reduced representation bisulfite sequencing; t-distributed stochastic neighbor embedding.

      Financial support

      This study was funded by the German Federal Ministry of Education and Research (BMBF) under 01EK1604A-D.

      Authors’ contributions

      PN, KK, DF, JGH, JW were responsible for the design of the study. PN and DF produced HLC, PN, DF and KB performed staining and qPCR analysis. PN, AR, NV performed live-staining assays. PN, DF, KK prepared samples for next generation sequencing. KL isolated single cells. KK prepared sequencing libraries. GG performed sequencing. AS processed the raw sequencing data. PN, KK, BH, DF analyzed the data. PN, KE, DF, BH, JR designed DiPa supervised clustering. PN, KK, BH generated figures. PN and JGH wrote the manuscript. KE, RM, CC, DF, BKM, PG, NB, MM, DH, JW critically assessed the manuscript. JGH, JW and JR supervised the study.

      Data availability statement

      The datasets generated during the current study are available in the EGA repository under accession EGAS00001004201.

      Conflict of interest

      Patricio Godoy is also affiliated with F. Hoffmann-La Roche Ltd (Roche Innovation Center Basel, Basel, Switzerland), Barbara Küppers-Munther is affiliated with Takara Bio Europe AB (former Cellartis AB) (Arvid Wallgrens Backe 20, 41346 Gothenburg, Sweden).
      Please refer to the accompanying ICMJE disclosure forms for further details.

      Supplementary data

      The following are the supplementary data to this article:

      References

        • Godoy P.
        • Hewitt N.J.
        • Albrecht U.
        • Andersen M.E.
        • Ansari N.
        • Bhattacharya S.
        • et al.
        Recent advances in 2D and 3D in vitro systems using primary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and their use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.
        Arch Toxicol. 2013; 87: 1315-1530https://doi.org/10.1007/s00204-013-1078-5
        • Camp J.G.
        • Sekine K.
        • Gerber T.
        • Loeffler-Wirth H.
        • Binder H.
        • Gac M.
        • et al.
        Multilineage communication regulates human liver bud development from pluripotency.
        Nature. 2017; 546: 533-538https://doi.org/10.1038/nature22796
        • Sachinidis A.
        • Albrecht W.
        • Nell P.
        • Cherianidou A.
        • Hewitt N.J.
        • Edlund K.
        • et al.
        Road map for development of stem cell-based alternative test methods.
        Trends Mol Med. 2019; 25: 470-481https://doi.org/10.1016/j.molmed.2019.04.003
        • Godoy P.
        • Schmidt-Heck W.
        • Hellwig B.
        • Nell P.
        • Feuerborn D.
        • Rahnenführer J.
        • et al.
        Assessment of stem cell differentiation based on genome-wide expression profiles.
        Philos Trans R Soc B Biol Sci. 2018; 373https://doi.org/10.1098/rstb.2017.0221
        • Cahan P.
        • Li H.
        • Morris S.A.
        • Lummertz Da Rocha E.
        • Daley G.Q.
        • Collins J.J.
        CellNet: network biology applied to stem cell engineering.
        Cell. 2014; 158: 903-915https://doi.org/10.1016/j.cell.2014.07.020
        • Radley A.H.
        • Schwab R.M.
        • Tan Y.
        • Kim J.
        • Lo E.K.W.
        • Cahan P.
        Assessment of engineered cells using CellNet and RNA-seq.
        Nat Protoc. 2017; 12: 1089-1102https://doi.org/10.1038/nprot.2017.022
        • Morris S.A.
        • Cahan P.
        • Li H.
        • Zhao A.M.
        • San Roman A.K.
        • Shivdasani R.A.
        • et al.
        Dissecting engineered cell types and enhancing cell fate conversion via Cellnet.
        Cell. 2014; 158: 889-902https://doi.org/10.1016/j.cell.2014.07.021
        • Kim D.S.
        • Ryu J.W.
        • Son M.Y.
        • Oh J.H.
        • Chung K.S.
        • Lee S.
        • et al.
        A liver-specific gene expression panel predicts the differentiation status of in vitro hepatocyte models.
        Hepatology. 2017; 66: 1662-1674https://doi.org/10.1002/hep.29324
        • Godoy P.
        • Schmidt-Heck W.
        • Natarajan K.
        • Lucendo-Villarin B.
        • Szkolnicka D.
        • Asplund A.
        • et al.
        Gene networks and transcription factor motifs defining the differentiation of stem cells into hepatocyte-like cells.
        J Hepatol. 2015; 63: 934-942https://doi.org/10.1016/j.jhep.2015.05.013
        • Gao X.
        • Li R.
        • Cahan P.
        • Zhao Y.
        • Yourick J.J.
        • Sprando R.L.
        Hepatocyte-like cells derived from human induced pluripotent stem cells using small molecules: implications of a transcriptomic study.
        Stem Cell Res Ther. 2020; 11: 393https://doi.org/10.1186/s13287-020-01914-1
        • Vartak N.
        • Guenther G.
        • Joly F.
        • Damle-Vartak A.
        • Wibbelt G.
        • Fickel J.
        • et al.
        Intravital dynamic and correlative imaging reveals diffusion-dominated canalicular and flow-augmented ductular bile flux.
        Hepatology. 2020; (hep.31422)https://doi.org/10.1002/hep.31422
        • Limaye P.B.
        • Alarcón G.
        • Walls A.L.
        • Nalesnik M.A.
        • Michalopoulos G.K.
        • Demetris A.J.
        • et al.
        Expression of specific hepatocyte and cholangiocyte transcription factors in human liver disease and embryonic development.
        Lab Investig. 2008; 88: 865-872https://doi.org/10.1038/labinvest.2008.56
        • Aibar S.
        • González-Blas C.B.
        • Moerman T.
        • Huynh-Thu V.A.
        • Imrichova H.
        • Hulselmans G.
        • et al.
        SCENIC: single-cell regulatory network inference and clustering.
        Nat Methods. 2017; 14: 1083-1086https://doi.org/10.1038/nmeth.4463
        • Nussler A.
        • Konig S.
        • Ott M.
        • Sokal E.
        • Christ B.
        • Thasler W.
        • et al.
        Present status and perspectives of cell-based therapies for liver diseases.
        J Hepatol. 2006; 45: 144-159https://doi.org/10.1016/j.jhep.2006.04.002
        • Wagner M.
        • Fickert P.
        • Zollner G.
        • Fuchsbichler A.
        • Silbert D.
        • Tsybrovskyy O.
        • et al.
        Role of farnesoid X receptor in determining hepatic ABC transporter expression and liver injury in bile duct-ligated mice.
        Gastroenterology. 2003; 125: 825-838https://doi.org/10.1016/S0016-5085(03)01068-0
        • Pereira-Fantini P.M.
        • Lapthorne S.
        • Joyce S.A.
        • Dellios N.L.
        • Wilson G.
        • Fouhy F.
        • et al.
        Altered FXR signalling is associated with bile acid dysmetabolism in short bowel syndrome-associated liver disease.
        J Hepatol. 2014; 61: 1115-1125https://doi.org/10.1016/j.jhep.2014.06.025