AMERICAN ASSOCIATION FOR THE STUDY OFLIVERD I S E ASES

RAPID COMMUNICATION | HEPATOLOGY, VOL. 66, NO. 5, 2017 A Single-Cell Transcriptomic Analysis Reveals Precise Pathways and Regulatory Mechanisms Underlying Hepatoblast Differentiation

Li Yang,1,2* Wei-Hua Wang,1,2* Wei-Lin Qiu,1,3* Zhen Guo,1 Erfei Bi,4 and Cheng-Ran Xu1

How bipotential hepatoblasts differentiate into hepatocytes and cholangiocytes remains unclear. Here, using single-cell transcriptomic analysis of hepatoblasts, hepatocytes, and cholangiocytes sorted from embryonic day 10.5 (E10.5) to E17.5 mouse embryos, we found that hepatoblast-to-hepatocyte differentiation occurred gradually and followed a linear default pathway. As more cells became fully differentiated hepatocytes, the number of proliferating cells decreased. Surprisingly, proliferating and quiescent hepatoblasts exhibited homogeneous differentiation states at a given developmental stage. This unique feature enabled us to combine single-cell and bulk-cell analyses to define the precise timing of the hepatoblast-to- hepatocyte transition, which occurs between E13.5 and E15.5. In contrast to hepatocyte development at almost all levels, hepatoblast-to-cholangiocyte differentiation underwent a sharp detour from the default pathway. New cholangiocyte gener- ation occurred continuously between E11.5 and E14.5, but their maturation states at a given developmental stage were heterogeneous. Even more surprising, the number of proliferating cells increased as more progenitor cells differentiated into mature cholangiocytes. Based on an observation from the single-cell analysis, we also discovered that the protein kinase C/mitogen-activated protein kinase signaling pathway promoted cholangiocyte maturation. Conclusion: Our studies have defined distinct pathways for hepatocyte and cholangiocyte development in vivo, which are critically important for understanding basic liver biology and developing effective strategies to induce stem cells to differentiate toward specific hepatic cell fates in vitro.(HEPATOLOGY 2017;66:1387-1401).

he liver comprises multiple cell types, among Once the hepatoblasts have been specified from the which hepatocytes and cholangiocytes (biliary foregut endoderm and form the hepatic diverticulum T duct cells) are the major epithelial cell types; at approximately embryonic day 9.5 (E9.5) in the and the hepatic lineages are of endodermal origin. mouse, they delaminate from the epithelium and Both hepatocytes and cholangiocytes are derived from migrate into the septum transversum, where they con- the bipotential progenitors, known as hepatoblasts at tinue proliferating and form the liver bud.(2) Hepato- the early stage of hepatogenesis.(1) blast proliferation is regulated by several cell signaling

Abbreviations: CK19, cytokeratin 19; DLK, delta-like; EpCAM, epithelial cell adhesion molecule; ERK, extracellular signal–regulated kinase; Ex, embryonic day x; FACS, fluorescence-activated cell sorting; FGF, fibroblast growth factor; GO, ontology; HNF, hepatocyte nuclear factor; MAPK, mitogen-activated protein kinase; MEK, MAPK/ERK kinase; PC, principal component; PCA, principal component analysis; PKC, protein kinase C; RNA-seq, RNA sequencing; SOX, sex-determining region Y-box; TBX3, T-box 3; TF, ; TGF, transforming growth factor; TPPB, (2S,5S)-(E,E)-8-(5-(4-(trifluoromethyl)phenyl)-2,4-pentadienoylamino) benzolactam. Received December 1, 2016; accepted June 30, 2017. Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.29353/suppinfo. *These authors contributed equally to this work. The accession number for the raw data files of the RNA sequencing analyses reported in this paper is GEO: GSE90047. Supported by the Ministry of Science and Technology of the People’s Republic of China (2015CB942800, to C.-R.X.), the National Natural Science Foundation of China (31521004, 31471358, and 31522036, to C.-R.X.), and National Institutes of Health (GM115420, to E.B.). Copyright VC 2017 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI 10.1002/hep.29353 Potential conflict of interest: Nothing to report.

1387 YANG, WANG, ET AL. HEPATOLOGY, November 2017 pathways, including the WNT, hepatocyte growth fac- maturation of hepatic lineages by performing single- tor, fibroblast growth factor (FGF), mitogen-activated cell RNA-seq analyses on isolated cells of the mouse protein kinase (MAPK) 9, and transforming growth hepatic lineages at time points from E10.5 to E17.5. factor beta (TGFb) pathways.(3) The cell fate decision In addition, by combining single-cell and bulk-cell occurs during liver development when the hepatoblasts analyses, we have precisely defined the developmental start specifying into hepatocytes and cholangiocytes. timing of the hepatoblast-to-hepatocyte transition. The transcription factors (TFs) T-box 3 (TBX3), CCAAT/enhancer binding protein alpha, prospero 1, and hepatocyte nuclear factor 4 alpha Materials and Methods (HNF4a) have been shown to promote hepatocyte dif- ferentiation.(4-7) The segregation of cholangiocytes ANIMALS occurs before E15.5, and these cells undergo morpho- F1 progenies of C57BL/6 and C3H (B6C3F1) genesis to form the biliary duct.(8) The TFs HNF6, mice were used (Supporting Information). All experi- sex-determining region Y-box 4 (SOX4), SOX9, and HNF1b, as well as several signaling pathways, includ- mental animal procedures were approved by the Insti- ing the Notch, WNT, FGF, and TGFb pathways, tutional Animal Care and Use Committee of Peking have been shown to promote cholangiocyte differentia- University. All mice were housed and maintained tion.(9-15) Despite these findings, the specific pathways under specific pathogen-free conditions on a 12-hour 6 8 and mechanisms underlying the differentiation of hep- day/night cycle at 23 2 C and received an autoclaved atoblasts into hepatocytes and cholangiocytes remain standard diet and water ad libitum. unclear. Cell fate decisions during lineage differentiation are GENERATION OF SINGLE-CELL coordinately regulated at the transcriptional level by the RNA-SEQ LIBRARIES activation and repression of specific gene sets.(16) How- ever, due to the lack of cellular markers that can be used After fluorescence-activated cell sorting (FACS) to efficiently distinguish and purify hepatoblasts, hepa- sorting, single cells were manually picked up under the tocytes, and cholangiocytes from different stages of the microscope with a mouth pipette and transferred into developing fetal liver, the conventional population- 4 lL of cell lysis buffer containing 0.05 lLofa based RNA sequencing (RNA-seq) method is inappro- 1:300,000 dilution of ERCC Spike-in RNA (Life priate for investigating the transcriptomic regulation of Technologies; 4456740). The complementary DNA hepatocyte versus cholangiocyte segregation. This limi- was synthesized, amplified (18 cycles) using the Smart- tation can be circumvented using single-cell RNA-seq, seq2 protocol,(18) and purified using VAHTS DNA a powerful approach that has been used to identify cell Clean Beads XP beads (Vazyme; N411-03). Two subpopulations, trace the developmental map, and deci- nanograms of complementary DNA were used to pre- pher the regulatory networks involved in the cell fate pare the sequencing libraries using the TruePrep DNA decision during organogenesis.(17) Library Prep Kit (Vazyme; TD502) with 0.43 the In this study, we have discovered precise pathways standard reaction volume. PCR amplifications were and potential mechanisms for the specification and conducted in eight cycles. Purification and size-based

ARTICLE INFORMATION:

From the 1Ministry of Education Key Laboratory of Cell Proliferation and Differentiation, College of Life Sciences, Peking-Tsinghua Center for Life Sciences, Peking University, Beijing, China; 2Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China; 3PKU-Tsinghua- NIBS Graduate Program, Peking University, Beijing, China; 4Department of Cell and Developmental Biology, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA. ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO:

Cheng-Ran Xu, Ph.D. Peking-Tsinghua Center for Life Sciences, Peking Ministry of Education Key Laboratory of Cell University Proliferation and Differentiation Beijing 100871, China College of Life Sciences E-mail: [email protected]

1388 HEPATOLOGY, Vol. 66, No. 5, 2017 YANG, WANG, ET AL. selection of PCR products were performed to obtain EpCAM signal in P2 cells became stronger than the 1 libraries with a peak size of 350 bp. signal observed at E13.5. Alb was expressed in DLK cells (P1) at an extremely high level, whereas Sox9 and Krt7 were exclusively detected in P2 cells. Thus, the Results P1 population must have contained hepatoblasts/hepa- SINGLE-CELL ANALYSIS REVEALS tocytes, whereas the P2 population contained cholan- giocytes (Supporting Fig. S1A,B). Similarly, we LINEAR AND BRANCHED identified the gating strategies at E11.5, E12.5, E14.5, PATHWAYS FOR THE and E15.5 for FACS of the hepatic lineages (Support- DIFFERENTIATION OF ing Fig. S1C). HEPATOBLASTS INTO Next, we performed single-cell RNA-seq on these HEPATOCYTES AND sorted hepatic cells (Fig. 1A). After quality control and CHOLANGIOCYTES ERCC spike-in normalization, 447 cells remained for further analyses; and on average, 7,000-9,000 To establish a road map for the differentiation of were detected with >1 million mapped reads in each hepatoblasts into hepatocytes and cholangiocytes at the cell (Supporting Information and Fig. S2A-E). To single-cell level, we decided to isolate distinct cell characterize the developmental process and define the populations from mouse fetal livers at different devel- populations of cells involved, we performed a principal opmental stages and perform single-cell RNA-seq component analysis (PCA) of all 447 cells. As shown analysis. We used well-characterized cell surface in the PCA plot, most hepatic cells obtained from the markers to isolate distinct cell populations through early to late developmental stages were located on the FACS. Delta-like (DLK) protein is an authentic hepa- main line along principal component (PC) 1, whereas toblast cell surface marker that is continually expressed some formed a distinct branch along PC2 (Fig. 1B,C). in hepatocytes until the neonatal stages,(19,20) whereas Hierarchical clustering of the PC2 higher loading epithelial cell adhesion molecule (EpCAM) is genes divided the cells into two groups (I and II). The expressed in early hepatoblasts and differentiated chol- cells in the P1 FACS-gated population were exclu- 1 angiocytes.(20) To determine the conditions for FACS sively group I cells. EpCAM cells (P2 gating) were of the hepatobiliary lineages during liver development, mainly group I cells prior to E12.5 but switched to 1 we stained dissociated single fetal liver cells obtained at group II thereafter. Beginning at E15.5, EpCAM E10.5, E13.5, and E17.5 with antibodies against cells were mainly observed in group II (Supporting DLK and EpCAM (Supporting Information). At Fig. S3A,B). Moreover, four distinct clusters were 1 E10.5, the DLK cells in the P1 population primarily identified among the heterogeneously expressed genes coexpressed with EpCAM, whereas the cells in the during hepatobiliary development. Cluster “a” genes DLK-low group were EpCAM– (Supporting Fig. were highly expressed in most of the group I cells from S1A). Quantitative RT-PCR analysis confirmed that E10.5 to E13.5 but less frequently expressed in group the P1 cells were hepatoblasts because they expressed II cells, as well as in group I cells from E15.5 and high levels of Alb, Afp, Foxa2, and Hnf4a but not the E17.5 (Fig. 1D). According to (GO) cholangiocyte markers Sox9 and Krt7 (Supporting Fig. analysis, cluster “a” was enriched with genes controlling S1B). At E13.5, the cells in the P1 population were DNA replication and cell cycle progression, including 1 still DLK , but the EpCAM level was close to the genes regulating the G1/S and G2/M transitions (Fig. background. However, we detected a few cells in the 1E; Supporting Table S1), such as Foxm1, Ccna2, 1 P2 population that were EpCAM . Compared to the Ccnb1, Ccne2, and Cdk1 (Fig. 2A). Based on these cells in the P3 (DLK–/EpCAM–) and P4 (DLK–/ results, cells expressing cluster “a” genes were prolifer- EpCAM-low) populations, the cells in the P1 popula- ating. Cluster “b” genes were highly expressed in the tion expressed high levels of Alb, Afp, Foxa2, and earlier stages of hepatobiliary development and gradu- Hnf4a, whereas cells in the P2 population also ally decreased along a pseudo–chronological order expressed Sox9 and Krt7, indicating that the P1 popu- (pseudotime) (Fig. 1D). This cluster was enriched lation mainly contains hepatoblasts/hepatocytes, the with genes involved in cell proliferation and the WNT P2 population contains cholangiocytes, and the P3 and and bone morphogenetic protein pathways (Fig. 1E; P4 populations include other non-endoderm-derived Supporting Table S1), including key genes known to cell types (Supporting Fig. S1A,B). At E17.5, the control stem cell proliferation, such as Id3, Lin28b,

1389 YANG, WANG, ET AL. HEPATOLOGY, November 2017



FIG. 1. Single-cell RNA-seq uncovers the road map of hepatobiliary lineage development. (A) Schematic workflow of hepatobiliary single-cell RNA-seq. (B) PCA plot of 447 hepatobiliary single-cell transcriptomes across seven developmental stages identifying hepa- toblasts/hepatocytes (circle) and cholangiocytes (triangle). Different stages are color-coded. Arrowhead curves indicate hepatoblast/ hepatocyte and cholangiocyte paths. (C) Schematic summary of hepatobiliary lineage development based on the PCA plot in (B). Each oval represents the distribution of hepatoblasts/hepatocytes or cholangiocytes of each developmental stage. (D) Hierarchical clus- tering of 1,761 heterogeneously expressed genes, correlated with the first two PCs (P < 5 3 10–15) identifying clusters “a”-“d.” Fea- tured TFs of each cluster are listed on the right. Pseudotime ordering of groups I and II is based on the projection of hepatoblasts/ hepatocytes on PC1 and cholangiocytes on PC2, respectively. Genes in clusters “a”-“c” and “d” are reordered by correlations with PC1 and PC2, respectively. (E) Selected GO terms enriched in clusters “a”-“d.” Abbreviations: BMP, bone morphogenetic protein; GTPase, guanosine triphosphatase; RTK, tyrosine kinase.



Etv5, and Lgr5(21-24) as well as the hepatoblast expan- (Fig. 1D). Cluster “c” genes were divided into two sub- sion enhancing factor Tbx3 (Fig. 2B).(4) Given the clusters, “c1” and “c2.” Cluster “c1” genes were critical roles of WNT signaling and TBX3 in control- expressed at an earlier stage of hepatoblast develop- ling hepatoblast proliferation,(4,13) we designated the ment and gradually increased with hepatoblast/hepato- cells expressing high levels of cluster “b” genes as hepa- cyte development, whereas “c2” genes showed delayed toblasts. Expression of the cluster “c” genes in group I expression. However, as shown in the GO analysis, was increased with the developmental pseudotime genes from these two subclusters exhibited enrichment

1390 HEPATOLOGY, Vol. 66, No. 5, 2017 YANG, WANG, ET AL.



FIG. 2. Expression of cell type–representative genes during hepatobiliary development. (A-D) levels from the RNA-seq data are projected onto the PCA plots (left); dot size represents the expression level (transcripts per million). Cells from different stages are color-coded. Gene expression levels in E11.5 hepatoblast, E17.5 hepatocyte, and cholangiocyte validated by single-cell quantitative RT-PCR are projected on box plots (right). The y axis represents the relative expression values normalized to Actb expression. “n” indicates number of single cells. *P < 0.05, **P < 0.01, and ***P < 0.001, Wilcoxon rank-sum test. Abbreviation: TPM, transcripts per million.

 in similar GO terms related to hepatocyte functions, Moreover, cluster “c” contained hepatocyte-related such as various metabolic and transmembrane trans- metabolic regulators such as Cps1, Ppara, Apoh, port processes (Fig. 1E; Supporting Table S1). Cyp2d10, and Cyp2d26 (Fig. 2C).(25,26) Thus, cells

1391 YANG, WANG, ET AL. HEPATOLOGY, November 2017 expressing high levels of cluster “c” genes were differ- expression patterns of cluster I and cluster II genes, we entiated hepatocytes. Notably, the cells at the same classified the proliferating cells into G1/S (mainly developmental time along PC1 tightly clustered expressing cluster I genes) and S/G2/M (expressing together, indicating relative homogeneity at the tran- both cluster I and II genes). The cells that did not scriptomic level (Fig. 1C). express cell cycle genes were considered quiescent (G0/ Strikingly, group II cells exclusively expressed cluster G1 phase) (Fig. 3A-C). As shown in the PCA plots, “d” genes that were enriched in cell adhesion, cell hepatoblast differentiation occurred along PC1, migration, epithelial tube morphogenesis, and Notch whereas hepatoblasts and their progenies in various and WNT signaling, which is known to regulate chol- phases of the cell cycle are distributed along PC2 (Fig. (12,13) angiocyte differentiation (Fig. 1E; Supporting 3B,C). To determine how hepatoblasts coordinate Table S1). For example, this cluster includes the TFs proliferation and differentiation during lineage specifi- Sox9, Sox4, and Hnf1b that regulate biliary cell differ- cation, we aligned all the cells in the pseudotime order entiation and morphogenesis,(10,11) as well as the duc- (8,27) along the direction of hepatoblast development and tal markers Spp1 and Krt7 (Fig. 2D). Therefore, found that proliferating cells and their corresponding we designated the cells in group II as differentiated quiescent cells kept pace without an obvious delay cholangiocytes. Interestingly, a few cholangiocytes along the developmental process (Fig. 3D). Thus, our were specified as early as E11.5, and new cholangio- single-cell analyses revealed that the differentiation cytes were continuously branched from the main line states of proliferating and quiescent hepatoblasts are along PC1 until E14.5. At E15.5 and E17.5, the chol- synchronized during early liver development. angiocyte population was completely segregated from the PC1 line (Fig. 1B,C), suggesting that none of cholangiocytes at these stages is newly generated. HEPATOBLAST-TO-HEPATOCYTE Taken together, our study has defined a road map for DIFFERENTIATION FOLLOWS hepatobiliary development at the single-cell level. A “DEFAULT” PROGRAM More importantly, our analysis has revealed that chol- angiocyte specification follows a path that sharply To further explore the mechanism underlying the branches from the linear pathway of hepatoblast to hepatoblast-to-hepatocyte transition, we analyzed the hepatocyte differentiation. overall profile of genes up-regulated during hepatoblast development to further explore the mechanism under- THE DIFFERENTIATION STATES lying the hepatoblast-to-hepatocyte transition. We OF PROLIFERATING AND performed a hierarchical cluster analysis and identified 383 genes whose expression levels increased gradually QUIESCENT HEPATOBLASTS from E10.5 to E13.5 (Fig. 4A). Strikingly, GO analy- ARE SYNCHRONIZED sis revealed that the enriched categories were mainly Differentiation and proliferation are generally associated with hepatocyte functions, such as metabolic inversely correlated during lineage specification.(28) processes, detoxification, and transport (Fig. 4B). Researchers have not clearly determined when hepato- Based on these data, hepatoblasts specify into the blasts begin to differentiate and how they coordinate hepatocyte fate immediately after their differentiation this process with proliferation. We focused our analy- from endodermal progenitors and gradually move ses on hepatoblasts and their progenies from E10.5 to toward that fate. We also analyzed the tendency of E13.5, excluding the few identified cholangiocytes, to TFs to be up-regulated during hepatoblast/hepatocyte address these questions. According to the PCA plot development and found that their expression started to (Fig. 1B), hepatoblast clusters at different developmen- increase at the beginning of the hepatoblast/hepatocyte tal stages move along PC1, suggesting that hepatoblast developmental process (Fig. 4C,D). In contrast, TFs differentiation starts at a very early stage. Hierarchical that were up-regulated during the hepatoblast/cholan- clustering analysis enabled the separation of cell cycle giocyte transition showed little or no expression in regulatory genes into two clusters (Fig. 3A). Cluster I hepatoblasts until the latter started to develop into included the G1/S markers Ccne1 and Ccne2, as well as cholangiocytes (Fig. 4C,D). Thus, the default pathway DNA replication–related genes (Fig. 3A), whereas for hepatoblasts is to differentiate into hepatocytes, but cluster II contained the G2/M markers Cdk1, Foxm1, along the way, some hepatoblasts are regulated to dif- Ccna2, Ccnb1, and Ccnb2 (Fig. 3A). Based on the ferentiate toward the cholangiocyte fate.

1392 HEPATOLOGY, Vol. 66, No. 5, 2017 YANG, WANG, ET AL.



FIG. 3. Proliferating hepatoblasts develop synchronously with quiescent hepatoblasts. (A) Hierarchical clustering identifying two clus- ters (I and II) of cell cycle–related genes and dividing hepatoblasts into three groups with different cell cycle phases. Cluster featured genes are listed on the right. (B,C) The PCA plot of 215 hepatoblasts. Different cell cycle phases (B) or developmental stages (C) are color-coded. Solid and empty circles represent quiescent (G0/G1) and proliferating (G1/S and S/G2/M) cells in (B), respectively. (D) The distribution of quiescent (solid circles) and proliferating (empty circles) hepatoblasts from different developmental stages. The x axis represents the pseudotime ordering of maturation. Abbreviation: TPM, transcripts per million.



HEPATOBLAST-TO-CHOLANGIOCYTE than one group. According to GO analyses, cluster DIFFERENTIATION FOLLOWS PC1-a genes were enriched in terms related to hepato- A REGULATED BRANCH FROM cyte functions, such as metabolic processes, transport, and detoxification (Fig. 5C). As these enriched terms THE DEFAULT PROGRAM were similar to the terms attributed to the up-regulated To obtain insights into cholangiocyte specification genes during hepatoblast development (Fig. 4A,B), we and maturation, we analyzed the dynamics of gene compared cluster PC1-a genes to those up-regulated expression in cells of the cholangiocyte lineage during genes and found that they exhibited 58% overlap (Fig. development. The PCA of all 102 identified cholan- 5D). Because E11.5 and E12.5 cholangiocytes belong giocytes revealed cholangiocyte heterogeneity starting to group I, we conclude that the cells in this group are at E13.5 (Fig. 5A). Hierarchical clustering of PC1 newly specified cholangiocytes that retain significant genes divided the cells into three groups (I-III) (Fig. features of their progenitors. The GO terms enriched 5B). Group I and group III cells expressed high levels in cluster PC1-b included cell adhesion, migration, of cluster PC1-a and PC1-b genes, respectively, while and tube morphogenesis, which are associated with group II cells expressed both cluster PC1-a and PC1-b functional cholangiocytes (Fig. 5C). Therefore, we genes at relatively low levels (Fig. 5B). The cholangio- conclude that the cells in group III are more differ- cytes identified at E11.5 and E12.5 belonged to group entiated cholangiocytes. Moreover, group I-III cells I, whereas E13.5-E17.5 cells were distributed in more represent three developmental stages (I-III) of

1393 YANG, WANG, ET AL. HEPATOLOGY, November 2017



FIG. 4. Hepatoblast adopts hepatocyte fate by default. (A) Hierarchical clustering of 383 up-regulated genes correlated with the first 2 two PCs in Fig. 3B (P < 1310 7) during hepatoblast development. Genes are reordered by the correlation with PC1. TFs are listed on the right. (B) Selected GO terms enriched in genes identified in (A). (C) The expression dynamics of up-regulated TFs across hepatobiliary lineage development. The red line represents the average tendency curve of relative expression levels of 21 TFs up- regulated during the hepatoblast-to-hepatocyte transition. The blue line represents 44 TFs up-regulated during the hepatoblast-to- cholangiocyte transition. The gray line represents the relative expression curve of an individual TF. (D) The lists of up-regulated genes during hepatoblast-to-hepatocyte or hepatoblast-to-cholangiocyte development. Genes bolded are known to be necessary for hepatic development. Abbreviation: TPM, transcripts per million.

 cholangiocyte early differentiation/maturation. Based differentiation. Early-stage cholangiocyte specification on these analyses, the genes contributing to PC1 can involves a gradual silencing of the expression of cluster be used to distinguish the states of cholangiocyte PC1-a genes that are normally expressed at high levels

1394 HEPATOLOGY, Vol. 66, No. 5, 2017 YANG, WANG, ET AL.



FIG. 5. Characterization of cholangiocyte lineage develop- ment. (A) PCA plot of all 102 cholangiocytes. Cells are color- coded by the developmental stage. (B) Hierarchical cluster- ing of 261 genes correlated with PC1 in (A) (P < 1 3 10–6) identifying two genes clusters (“a” and “b”) and dividing chol- angiocytes into three develop- mental stages (I-III). Cluster featured TFs are listed on the right. Genes are ordered by cor- relation with PC1. (C) Selected GO terms enriched in clusters “a” and “b” identified in (B) and genes related to PC2 in (A). (D) Venn diagram showing the overlap between cluster “a” genes in (B) and the genes in Fig. 4A. (E) PCA plot showing the distribution of hepatobiliary cells with different cell prolifer- ation states. Different stages of cholangiocytes and hepatoblasts/ hepatocytes are color-coded. Individual cells are marked by empty (proliferating) or solid (quiescent) circles. (F) Percent- age of proliferating cholangio- cytes from stages I to III (top) and hepatoblasts/hepatocytes from E10.5 to E17.5 (bottom). Abbreviations: RTK, receptor tyrosine kinase; TPM, tran- scripts per million.

 in hepatoblasts and hepatocytes, whereas cholangiocyte PC2 axis (Fig. 5E). Surprisingly, the percentage of maturation involves inducing the expression of genes proliferating cells increased as cholangiocytes devel- required for the structure and function of this cell type. oped from stage I to stage III, in sharp contrast to GO analysis also showed that genes contributing to hepatoblast-to-hepatocyte development, during which PC2 were enriched in terms related to cell cycle regula- the proliferation rate steadily decreased (Fig. 5F; Sup- tion (Fig. 5C). Based on the expression of the cell cycle porting Fig. S4A,B). To verify the proliferation rate genes identified in Fig. 1D, we defined proliferating determined by single-cell transcriptomic analyses, we 1 and quiescent cells in the three groups of cholangio- collected DLK cells at E11.5, E13.5, and E17.5 and 1 cytes (Supporting Fig. S4A). We projected these cells EpCAM cells at E13.5 and E17.5 for immunostain- on the PCA plot shown in Fig. 1B and found that ing using antibody against Ki67, a marker for prolifer- 1 cholangiocyte maturation mainly occurred along the ating cells.(29) Consistently, the percentage of Ki67

1395 YANG, WANG, ET AL. HEPATOLOGY, November 2017 cells gradually decreased during hepatoblast-to- (5-(4-(trifluoromethyl)phenyl)-2,4-pentadienoylamino) hepatocyte development (Supporting Fig. S4C,D). benzolactam (TPPB).(34) After 48 hours of culture, the Because the transcript of cytokeratin 19 (CK19), which treated liver explants did not show a significant difference is expressed in differentiated cholangiocytes, was rarely in size compared to the controls (Fig. 6A), and the per- 1 detected in the single cholangiocyte at stage I (Sup- centages of DLK hepatoblasts/hepatocytes between porting Fig. S4E), we considered that E13.5 groups were similar (Supporting Fig. S6A). Based on 1 1 EpCAM cells and E17.5 EpCAM /CK19– repre- bromodeoxyuridine incorporation experiments, the pro- 1 sented less mature cholangiocytes. The percentage of liferation rates of DLK cells were indistinguishable 1 Ki67 cells significantly increased in more mature between the controls and the TPPB-treated samples 1 1 E17.5 EpCAM /CK19 cells compared with less (Supporting Fig. S6B). However, we observed an mature cells (Supporting Fig. S4C,D). Based on these approximately 3-fold increase in the number of 1 analyses, hepatoblast-to-cholangiocyte differentiation EpCAM cells after the TPPB treatment (Fig. 6B,C), defines a regulated branch of the hepatoblast-to- suggesting that PKC activation stimulated the produc- 1 hepatocyte pathway, and these distinct fate choices tion of EpCAM cells. PKC is known to stimulate involve distinct mechanisms coordinating proliferation MAPK/ERK kinase 1/2 (MEK1/2), which, in turn, and differentiation. activates MAPK.(35) To determine whether the TPPB effect was mediated by the MAPK pathway, we treated MAPK SIGNALING PROMOTES E12.5 fetal livers with both TPPB and U0126,(36) a MEK1/2 inhibitor. The U0126 treatment abolished the CHOLANGIOCYTE 1 MATURATION TPPB-induced increase in the number of EpCAM cells (Fig. 6C). Thus, the PKC–MEK–MAPK pathway 1 The single-cell analysis not only provides a high- is responsible for the expansion of EpCAM cells. 1 resolution map of hepatoblast differentiation but also To determine whether the EpCAM cells from the offers potential mechanisms that can be experimentally TPPB-treated samples were more mature cholangio- tested. For example, our GO analysis indicated that the cytes, we performed single-cell RNA-seq on the 1 1 MAPK pathway was enriched in more mature (stage EpCAM and DLK cells from both TPPB-treated III) cholangiocytes (Fig. 5C). Differential gene expres- and control samples. The PCA of all previously sion analyses between stage I and stage III ductal cells described hepatic cells (Fig. 1B) and cells from this 1 identified high levels of expression of 21 MAPK path- experiment showed that the DLK cells from the con- way–related genes in stage III cholangiocytes (Support- trols and the TPPB-treated samples were located along ing Fig. S5A,B). Many of the genes involved in the the hepatoblast/hepatocyte line, whereas the 1 canonical RAS–protein kinase C (PKC)–extracellular EpCAM cells were distributed in the cholangiocyte 1 signal–regulated kinase (ERK)/MAPK pathway were direction (Fig. 6D). Remarkably, the EpCAM cells up-regulated during cholangiocyte maturation, such as from the TPPB-treated samples and the control sam- the ERK1/2 activators Erbb2/4 and Fgfr2/3,(30,31) as ples displayed a significant separation along the PC1 well as the ERK downstream factors Sorl1 and axis (Fig. 6D), suggesting that TPPB treatment pro- Wwc1.(32,33) To confirm that the MAPK/ERK pathway motes cholangiocyte maturation. This hypothesis was was activated in the more mature cholangiocytes, we supported by the observation that the cholangiocyte performed immunostaining of the active phosphorylated markers Epcam and Sox9 were expressed at high levels 1 ERK1/2 on FACS-sorted E13.5 and E17.5 EpCAM in TPPB-treated cells, whereas expression of the hepa- 1 cells. E13.5 EpCAM cells, most of which belong to toblast/hepatocyte marker Alb was totally repressed. stage I, were phosphorylated ERK– cells, whereas Foxa3, which is down-regulated in more mature chol- 1 1 >60% of the more mature E17.5 EpCAM /CK19 angiocytes (Supporting Fig. S6C), was expressed at a 1 cells were phosphorylated ERK (Supporting Fig. S5C, lower level in TPPB-treated cells (Fig. 6E). Finally, D). Thus, the MAPK/ERK pathway is activated in GO enrichment analysis indicated that TPPB-treated vivo during cholangiocyte maturation. cells exhibited enrichment in the similar cellular func- To functionally determine whether the canonical tion terms as cholangiocytes compared to controls MAPK pathway stimulates cholangiocyte maturation, we (Figs. 1E and 5C; Supporting Fig. S6D). Expression treated fetal livers at E12.5 (Supporting Information), of genes regulating cell proliferation was up-regulated thetimebeforethefateofhepatoblastsisspecifiedinto in TPPB-treated cells (Supporting Fig. S6D,E), con- cholangiocytes, with a PKC agonist, (2S,5S)-(E,E)-8- sistent with the finding described above showing that

1396 HEPATOLOGY, Vol. 66, No. 5, 2017 YANG, WANG, ET AL.



FIG. 6. PKC/ERK/MAPK signaling promotes cholangiocyte maturation. (A) Schematic workflow of the E12.5 liver lobe cultures (top). Liver lobes were treated with dimethyl sulfoxide (control), TPPB, U0126, or TPPB1U0126. Images of explants after treatments (bottom) do not show overt differences in morphogenesis between each type of treatment. Scale bar, 200 lm. (B) FACS gating strategy used to sort and analyze 1 1 EpCAM cells. (C) Statistical analysis of the percentage of EpCAM cells observed after treatment. Data are presented as means 6 SEM; “n” indicates number of biological replicates; “sample #” indicates the number of cultured liver lobes; ***P < 0.001, t test. (D) PCA plot showing that the TPPB treatment promotes cholangiocyte maturation compared to the control (left). The number of single cells is presented in parenthe- ses. The PC1 values of individual cells in the PCA plot are projected in a box plot (right). ***P < 0.001, Wilcoxon rank-sum test. (E) Expression levels (transcripts per million) of marker genes are projected onto PCA plots (top) and box plots (bottom). **P < 0.01, ***P < 0.001, Wilcoxon rank-sum test. Abbreviations: Gapdh, glyceraldehyde 3-phosphate dehydrogenase; TPM, transcripts per million.



1397 YANG, WANG, ET AL. HEPATOLOGY, November 2017 the more mature cholangiocytes exhibit higher prolif- whole–fetal liver study but rarely detected in our eration rates (Fig. 5F). Based on our results and analy- RNA-seq results (Supporting Fig. S8A). Based on ses, PKC-triggered MAPK activation promotes these findings, the transcriptomic data generated from cholangiocyte maturation. the whole fetal liver are not suitable for defining the timing of the hepatoblast-to-hepatocyte transition. THE HEPATOBLAST-TO- According to hierarchical clustering analysis, all HEPATOCYTE TRANSITION time points retained their developmental orders, and OCCURS BETWEEN the differentially expressed genes were divided into clusters “a” and “b” (Fig. 7A). Cluster “a” genes were E13.5 AND E15.5 mainly down-regulated during development, whereas Because the differentiation of hepatoblasts into hep- cluster “b” genes displayed an opposite trend. The atocytes occurs progressively, defining the time point expression patterns of these two clusters of genes sug- of the lineage transition is challenging. However, this gest that hepatoblast-to-hepatocyte differentiation definition might be achieved using an unbiased might occur between E14.5 and E15.5. This interpre- approach, i.e., comparing the similarity between the tation is supported by the results of the GO analysis transcriptomes of hepatoblasts and their progenies at (Fig. 7B). The terms enriched in cluster “a” included different developmental stages. Although single-cell the cell cycle, DNA replication, and many cell signal- RNA-seq is an ideal method to profile cell-to-cell vari- ing pathways, such as mothers against decapentaplegic, ability and reveal the interplay between intrinsic cellu- WNT, TGFb, bone morphogenetic protein, vascular lar processes, it has limited sensitivity in detecting endothelial growth factor, and MAPK pathways, as transcripts, particularly low-abundance transcripts.(37) well as histone modification and DNA methylation Considering the relative homogeneity of the cells at (Fig. 7B). Quantitative RT-PCR was performed to the same developmental stages during hepatoblast-to- validate the expression patterns of some of these genes, hepatocyte differentiation at the transcriptomic level including two DNA methyltransferases, Dnmt1 and (Fig. 1B,C), we adopted bulk-cell RNA-seq to define Dnmt3a, and two histone modifiers, Kat2a (Gcn5) and the timing of the hepatoblast-to-hepatocyte transition. Kdm5b (Jarid1b). The expression levels of all these Similar transcriptomic dynamics studies have been per- genes decreased during hepatoblast/hepatocyte devel- formed using the whole fetal liver.(38,39) However, the opment (Supporting Fig. S8B). Thus, at the hepato- fetal liver is also the organ in which hematopoiesis blast stage, cells receive many environmental stimuli to occurs. Hepatoblasts/hepatocytes are only a small frac- promote cell proliferation and are en route to differen- tion of the total fetal liver cells (Supporting Fig. S1A, tiation. In contrast, cluster “b” mainly included the C), and most of the other cells are blood cells and lym- terms related to hepatocyte functions but lacked the phocytes. Therefore, we performed RNA-seq analyses cell signaling pathways required to drive proliferation 1 at the bulk-cell level in 200,000 sorted DLK cells on (Fig. 7B), suggesting that cells expressing this cluster each day from E10.5 to E18.5 (Supporting Informa- of genes have specified into the hepatocyte fate or will tion, Fig. S7A-C). In this study, we identified 4,077 reach that stage soon. variably expressed genes during hepatoblast/hepatocyte Cell fate changes during a developmental process development, a number that is 3 times higher than the are generally triggered by TFs, and changes in TF number identified using single-cell analyses (Figs. 1D expression precede changes in the expression of their and 7A). The gene expression profiles of the bulk-cell target genes, including genes required for cell struc- and pooled single-cell RNA-seq showed strong corre- ture and function. Of the differentially expressed lations (Supporting Fig. S7B), although the single-cell genes shown in Fig. 7A, 222 genes encoded TFs. transcriptomic data sets exhibited increased variability We then performed hierarchical clustering and dif- compared with the bulk-cell RNA-seq data sets ferential expression analyses of this pool of genes and 0 0 (Supporting Fig. S7C). We compared our data to the identified two clusters, a and b (Fig. 7C,D). These transcriptomic data obtained from the whole liver(39) clusters distinguished hepatoblasts from hepatocytes and found that the hepatic marker Alb was expressed at at a time point between E13.5 and E14.5. Taken 1 high levels in DLK cells, whereas Gata1, Cd41, and together, hierarchical clustering analyses using differ- Cd34, which are specifically expressed in erythrocytes, ent gene pools strongly suggest that hepatoblast-to- myeloerythroid progenitors, and hematopoietic stem hepatocyte differentiation begins at E13.5 and is cells, respectively,(40,41) were robustly detected in the complete by E15.5.

1398 HEPATOLOGY, Vol. 66, No. 5, 2017 YANG, WANG, ET AL.



FIG. 7. Bulk-cell RNA-seq defines the hepatoblast-to-hepatocyte turning point. (A) Hierarchical clustering analysis of 4,077 variably expressed genes (coefficient of variation >0.4) in bulk hepatoblasts/hepatocytes obtained at nine developmental stages. Cluster- featured TFs are listed on the right. Arrow indicates the timing of the hepatoblast-to-hepatocyte transition. (B) Selected GO terms enriched in clusters “a” and “b” shown in (A). (C) Hierarchical clustering of 222 TFs extracted from (A). Arrow indicates the timing of the hepatoblast-to-hepatocyte transition based on TF expression. (D) Volcano plot showing TFs that are highly expressed in hepa- toblasts (E11.5) and hepatocytes (E17.5) with an adjusted P < 0.05. The listed TFs have been reported to regulate hepatoblast/hepa- tocyte development. Abbreviations: BMP, bone morphogenetic protein; CV, coefficient of variation; padj, adjusted P; SMAD, mothers against decapentaplegic; TPM, transcripts per million; VEGF, vascular endothelial growth factor.

 lineages and systematically mapped the paths for hepato- Discussion biliary development at the single-cell level from E10.5 to Although hepatoblasts are known to be bipotential E17.5. To our surprise, nascent hepatoblasts have already hepatic progenitors that give rise to hepatocytes and initiated their differentiation in the hepatocyte direction. cholangiocytes, the precise fate map for hepatic lineage This phenomenon suggests that hepatoblast-to- specification has not been constructed. In this study, we hepatocyte lineage specification is the default process. identified the conditions required to enrich the hepatic Hepatoblasts/hepatocytes show homogeneity at the

1399 YANG, WANG, ET AL. HEPATOLOGY, November 2017 same developmental stages and are “pushed” forward promotes cholangiocyte maturation using an explant sys- progressively along a linear pathway by differentiation/ tem and single-cell analysis. Thus, we have defined a maturation stimuli. Strikingly, the differentiation states mechanism for cholangiocyte maturation. of proliferating and quiescent hepatoblast progenies In summary, our single-cell studies have established a from the same developmental stages are synchronized. molecular road map for the development of cells in the In contrast, hepatoblast-to-cholangiocyte development hepatobiliary lineage in vivo and provided significant defines a sharp branch from the main hepatoblast/hepa- insights into the mechanisms of hepatocyte and cholan- tocyte pathway. Cholangiocyte specification from hepa- giocyte segregation, cellular heterogeneity, and matura- toblast starts as early as E11.5, becomes more prevalent tion. Recently, cholangiocytes were induced to (42-44) at E13.5, and is completed at E14.5. Developing chol- differentiate from embryonic stem cells in vitro ; angiocytes differentiate around the portal vein along the however, these cells are generally immature. Thus, our hilum-to-periphery axis; more mature cholangiocytes findings are important not only for understanding basic are located near the hilum, and less mature cholangio- cellular mechanisms but also for optimizing the condi- cytes are located toward the periphery of the lobes.(8) tions for the induction of embryonic stem cells to differ- Hence, newly generated cholangiocytes apparently enter entiate into hepatocytes and cholangiocytes in vitro. the differentiation/maturation path nonsynchronously, resulting in the observed heterogeneity in maturation Acknowledgment: We thank Drs. K. Zaret and X. states at the same developmental stage. Thus, our Cheng for critical advice on the manuscript, members single-cell analysis has established a high-resolution map of the Xu laboratory for helpful advice and discus- sion, and the Peking-Tsinghua Center for the Life for hepatobiliary development with surprising features. Proliferation and differentiation are mutually exclu- Science Computing Platform. sive processes occurring during development. However, researchers have not determined how different hepatic REFERENCES lineages coordinate these processes during their specifi- 1) Zorn AM. Liver development. In: StemBook. Cambridge, MA: cation. In this study, we showed that early-stage hepa- Harvard Stem Cell Institute; 2008. toblasts maintain a higher proliferation rate but that an 2) Si-Tayeb K, Lemaigre FP, Duncan SA. Organogenesis and increasing number of cells become quiescent during development of the liver. Dev Cell 2010;18:175-189. 3) Lemaigre FP. Mechanisms of liver development: concepts for hepatoblast/hepatocyte development. At E17.5, most of understanding liver disorders and design of novel therapies. Gas- the hepatocytes are quiescent. Although most E10.5 to troenterology 2009;137:62-79. E13.5 hepatoblasts exhibit a higher proliferation rate, a 4) Ludtke TH, Christoffels VM, Petry M, Kispert A. Tbx3 promotes small fraction of cells are quiescent at these stages. This liver bud expansion during mouse development by suppression of cholangiocyte differentiation. HEPATOLOGY 2009;49:969-978. quiescent state is presumably important for hepatoblast- 5) Suzuki A, Iwama A, Miyashita H, Nakauchi H, Taniguchi H. to-cholangiocyte fate commitment because reduced Role for growth factors and extracellular matrix in controlling hepatoblast proliferation is associated with cholangio- differentiation of prospectively isolated hepatic stem cells. Devel- cyte differentiation in the Tbx3–/– liver.(4) As cholangio- opment 2003;130:2513-2524. 6) Seth A, Ye J, Yu N, Guez F, Bedford DC, Neale GA, et al. cytes become more mature, the number of proliferating Prox1 ablation in hepatic progenitors causes defective hepatocyte cells increases, suggesting that the immature cholangio- specification and increases biliary cell commitment. Development cytes are transiently noncycling cells that maintain their 2014;141:538-547. ability to reenter the cell cycle during the maturation 7) Li J, Ning G, Duncan SA. Mammalian hepatocyte differentia- tion requires the transcription factor HNF-4alpha. Genes Dev process. Thus, the hepatocyte and cholangiocyte line- 2000;14:464-474. ages must use distinct mechanisms to control prolifera- 8) Antoniou A, Raynaud P, Cordi S, Zong Y, Tronche F, Stanger tion and differentiation during their development. BZ, et al. Intrahepatic bile ducts develop according to a new Cholangiocyte specification is coordinately regulated mode of tubulogenesis regulated by the transcription factor SOX9. Gastroenterology 2009;136:2325-2333. by TGFb, WNT, Notch, and FGF signaling. However, 9) Clotman F, Lannoy VJ, Reber M, Cereghini S, Cassiman D, the signals promoting cholangiocyte maturation remain Jacquemin P, et al. The onecut transcription factor HNF6 is unknown. In this study, we classified cholangiocytes into required for normal development of the biliary tract. Develop- ment 2002;129:1819-1828. different stages based on their maturation states and 10) Poncy A, Antoniou A, Cordi S, Pierreux CE, Jacquemin P, foundthatmorematurecholangiocytesexpressedhigh Lemaigre FP. Transcription factors SOX4 and SOX9 cooperatively levels of genes involved in the PKC/MEK/MAPK sig- control development of bile ducts. Dev Biol 2015;404:136-148. naling pathway. We then showed that this pathway

1400 HEPATOLOGY, Vol. 66, No. 5, 2017 YANG, WANG, ET AL.

11) Coffinier C, Gresh L, Fiette L, Tronche F, Schutz G, Babinet C, 29) Gerdes J, Lemke H, Baisch H, Wacker HH, Schwab U, Stein et al. Bile system morphogenesis defects and liver dysfunction upon H. Cell cycle analysis of a cell proliferation–associated human targeted deletion of HNF1beta. Development 2002;129:1829-1838. nuclear antigen defined by the monoclonal antibody Ki-67. 12) Tchorz JS, Kinter J, Muller M, Tornillo L, Heim MH, Bettler J Immunol 1984;133:1710-1715. B. Notch2 signaling promotes biliary epithelial cell fate specifica- 30) Bourguignon LY, Gilad E, Peyrollier K. Heregulin-mediated tion and tubulogenesis during bile duct development in mice. ErbB2-ERK signaling activates hyaluronan synthases leading to HEPATOLOGY 2009;50:871-879. CD44-dependent ovarian tumor cell growth and migration. 13) Decaens T, Godard C, de Reynies A, Rickman DS, Tronche F, J Biol Chem 2007;282:19426-19441. Couty JP, et al. Stabilization of beta-catenin affects mouse embryonic 31) Aoyama E, Kubota S, Takigawa M. CCN2/CTGF binds to liver growth and hepatoblast fate. HEPATOLOGY 2008;47:247-258. fibroblast growth factor receptor 2 and modulates its signaling. 14) Yanai M, Tatsumi N, Hasunuma N, Katsu K, Endo F, FEBS Lett 2012;586:4270-4275. Yokouchi Y. FGF signaling segregates biliary cell-lineage from 32) Rohe M, Synowitz M, Glass R, Paul SM, Nykjaer A, Willnow chick hepatoblasts cooperatively with BMP4 and ECM compo- TE. Brain-derived neurotrophic factor reduces amyloidogenic nents in vitro. Dev Dyn 2008;237:1268-1283. processing through control of SORLA gene expression. 15) Clotman F, Jacquemin P, Plumb-Rudewiez N, Pierreux CE, J Neurosci 2009;29:15472-15478. Van der Smissen P, Dietz HC, et al. Control of liver cell fate 33) Yang S, Ji M, Zhang L, Chen Y, Wennmann DO, Kremerskothen decision by a gradient of TGF beta signaling modulated by One- J, et al. Phosphorylation of KIBRA by the extracellular signal–regu- cut transcription factors. Genes Dev 2005;19:1849-1854. lated kinase (ERK)–ribosomal S6 kinase (RSK) cascade modulates 16) Davidson EH. Emerging properties of animal gene regulatory cell proliferation and migration. Cell Signal 2014;26:343-351. networks. Nature 2010;468:911-920. 34) Chen S, Borowiak M, Fox JL, Maehr R, Osafune K, Davidow L, 17) Treutlein B, Brownfield DG, Wu AR, Neff NF, Mantalas GL, et al. A small molecule that directs differentiation of human ESCs Espinoza FH, et al. Reconstructing lineage hierarchies of the distal into the pancreatic lineage. Nat Chem Biol 2009;5:258-265. lung epithelium using single-cell RNA-seq. Nature 2014;509:371-375. 35) Mackay HJ, Twelves CJ. Targeting the protein kinase C family: 18) Picelli S, Faridani OR, Bjorklund AK, Winberg G, Sagasser S, are we there yet? Nat Rev Cancer 2007;7:554-562. Sandberg R. Full-length RNA-seq from single cells using Smart- 36) Favata MF, Horiuchi KY, Manos EJ, Daulerio AJ, Stradley DA, seq2. Nat Protoc 2014;9:171-181. Feeser WS, et al. Identification of a novel inhibitor of mitogen- 19) Tanimizu N, Nishikawa M, Saito H, Tsujimura T, Miyajima A. activated protein kinase kinase. J Biol Chem 1998;273:18623-18632. Isolation of hepatoblasts based on the expression of Dlk/Pref-1. 37) Saliba AE, Westermann AJ, Gorski SA, Vogel J. Single-cell J Cell Sci 2003;116:1775-1786. RNA-seq: advances and future challenges. Nucleic Acids Res 20) Tanaka M, Okabe M, Suzuki K, Kamiya Y, Tsukahara Y, 2014;42:8845-8860. Saito S, et al. Mouse hepatoblasts at distinct developmental 38) Jochheim A, Cieslak A, Hillemann T, Cantz T, Scharf J, Manns stages are characterized by expression of EpCAM and DLK1: MP, et al. Multi-stage analysis of differential gene expression in drastic change of EpCAM expression during liver development. BALB/C mouse liver development by high-density microarrays. Mech Dev 2009;126:665-676. Differentiation 2003;71:62-72. 21) Nakayama M, Matsumoto K, Tatsumi N, Yanai M, Yokouchi Y. Id3 39) Li T, Huang J, Jiang Y,ZengY,HeF,ZhangMQ,etal. is important for proliferation and differentiation of the hepatoblasts Multi-stage analysis of gene expression and transcription regulation during the chick liver development. Mech Dev 2006;123:580-590. in C57/B6 mouse liver development. Genomics 2009;93:235-242. 22) Takashima Y, Terada M, Udono M, Miura S, Yamamoto J, 40) Welch JJ, Watts JA, Vakoc CR, Yao Y, Wang H, Hardison RC, Suzuki A. Suppression of lethal-7b and miR-125a/b maturation et al. Global regulation of erythroid gene expression by transcrip- by Lin28b enables maintenance of stem cell properties in hepato- tion factor GATA-1. Blood 2004;104:3136-3147. blasts. HEPATOLOGY 2016;64:245-260. 41) Mikkola HK, Orkin SH. The journey of developing hematopoi- 23) Akagi T, Kuure S, Uranishi K, Koide H, Costantini F, Yokota etic stem cells. Development 2006;133:3733-3744. T. ETS-related transcription factors ETV4 and ETV5 are 42) Dianat N, Dubois-Pot-Schneider H, Steichen C, Desterke C, involved in proliferation and induction of differentiation- Leclerc P, Raveux A, et al. Generation of functional associated genes in embryonic stem (ES) cells. J Biol Chem cholangiocyte-like cells from human pluripotent stem cells and 2015;290:22460-22473. HepaRG cells. HEPATOLOGY 2014;60:700-714. 24) Huch M, Dorrell C, Boj SF, van Es JH, Li VS, van de Wetering 43) Ogawa M, Ogawa S, Bear CE, Ahmadi S, Chin S, Li B, et al. 1 M, et al. In vitro expansion of single Lgr5 liver stem cells induced Directed differentiation of cholangiocytes from human pluripo- by Wnt-driven regeneration. Nature 2013;494:247-250. tent stem cells. Nat Biotechnol 2015;33:853-861. 25) Willer CJ, Schmidt EM, Sengupta S, Peloso GM, Gustafsson S, 44) Sampaziotis F, Cardoso de Brito M, Madrigal P, Bertero A, Kanoni S, et al.; Global Lipids Genetics Consortium. Discovery Saeb-Parsy K, Soares FA, et al. Cholangiocytes derived from and refinement of loci associated with lipid levels. Nat Genet human induced pluripotent stem cells for disease modeling and 2013;45:1274-1283. drug validation. Nat Biotechnol 2015;33:845-852. 26) Cui JY, Renaud HJ, Klaassen CD. Ontogeny of novel cyto- chrome P450 gene isoforms during postnatal liver maturation in Author names in bold designate shared co-first mice. Drug Metab Dispos 2012;40:1226-1237. 27) Rodrigo-Torres D, Affo S, Coll M, Morales-Ibanez O, Millan authorship. C, Blaya D, et al. The biliary epithelium gives rise to liver pro- genitor cells. HEPATOLOGY 2014;60:1367-1377. 28) Ruijtenberg S, van den Heuvel S. Coordinating cell proliferation Supporting Information and differentiation: antagonism between cell cycle regulators and cell type–specific gene expression. Cell Cycle 2016;15:196-212. Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.29353/suppinfo.

1401