Tbx3 promotes liver bud expansion during mouse development by suppression of cholangiocyte differentiation


  • Potential conflict of interest: Nothing to report.


After specification of the hepatic endoderm, mammalian liver organogenesis progresses through a series of morphological stages that culminate in the migration of hepatocytes into the underlying mesenchyme to populate the hepatic lobes. Here, we show that in the mouse the transcriptional repressor Tbx3, a member of the T-box protein family, is required for the transition from a hepatic diverticulum with a pseudo-stratified epithelium to a cell-emergent liver bud. In Tbx3-deficient embryos, proliferation in the hepatic epithelium is severely reduced, hepatoblasts fail to delaminate, and cholangiocyte rather than hepatocyte differentiation occurs. Molecular analyses suggest that the primary function of Tbx3 is to maintain expression of hepatocyte transcription factors, including hepatic nuclear factor 4a (Hnf4a) and CCAAT/enhancer binding protein (C/EBP), alpha (Cebpa), and to repress expression of cholangiocyte transcription factors such as Onecut1 (Hnf6) and Hnf1b. Conclusion: Tbx3 controls liver bud expansion by suppressing cholangiocyte and favoring hepatocyte differentiation in the liver bud. (HEPATOLOGY 2009.)

Hepatocytes and cholangiocytes constitute the liver parenchyme and the bile-transporting cells of the intrahepatic and extrahepatic bile ducts, respectively. Both cell types derive from a bipotential precursor cell, the hepatoblast, whose specification, expansion, and differentiation is intimately linked with morphogenesis of the liver.1 Liver development in the mouse begins at embryonic day (E) 8.25 after the formation of the definitive endoderm. Signals from the precardiogenic mesoderm and the underlying septum transversum region act in combination to induce and delineate the hepatic from the neighboring pancreatic and intestinal endoderm. Hepatoblasts activate an early liver gene program and form a thickened columnar epithelium that becomes pseudo-stratified at E9.0. Starting from E9.5, the basal lamina degrades, and finger-like protrusions arise from which individual cells migrate into the underlying mesenchyme and populate the hepatic lobes. Although most hepatoblasts differentiate into hepatocytes, a subset of these cells maintain their precursor character and differentiate into cholangiocytes that form the lining of the bile ducts, starting from E13.5. Thus, differentiation of hepatoblasts into hepatocytes or bile duct cells is temporally and spatially separated, suggesting the existence of localized inducers or repressing mechanisms that direct either fate.2, 3

Phenotypical analysis of mutant mice has provided substantial insight into a molecular network of transcriptional regulators that control distinct subprograms of liver organogenesis.2, 3Tbx3, a member of the T-box gene family, has recently emerged as an additional player in the genetic circuit underlying the hepatic lineage decision. Heterozygosity of TBX3 causes Ulnar-mammary syndrome in humans, an autosomal-dominant disorder characterized by upper limb skeletal malformations, severe hypoplasia of the breast, and hair and genital defects.4Tbx3-homozygous mice present ulnar-mammary syndrome–related features, including severe defects in limb and mammary gland development. Tbx3-mutant mice bred on a C57Bl6/129 mixed genetic background additionally exhibit a hypoplastic liver that was hypothesized to be secondary to impaired vascularization or hematopoiesis.5 However, a recent study has provided strong evidence for a primary requirement of Tbx3 in hepatogenesis.6, 7 The authors showed that Tbx3 expression in multipotent hepatoblasts supports proliferation and hepatic differentiation of these progenitor cells by repression of the tumor suppressor gene p19Arf (Cdkn2a).

Here, we extend the analysis of hepatic Tbx3 function and provide novel insight into the temporal and spatial requirement of the gene in orchestrating multiple aspects of early liver organogenesis. We correlate the hepatic expression of Tbx3 with onset of liver defects in Tbx3-deficient embryos and show that Tbx3 controls morphogenesis of the liver bud by coordinately regulating proliferation, migration, and differentiation of hepatoblasts. Based on molecular phenotyping and overexpression experiments in vitro, we propose that de-repression of p19Arf and proliferation defects, and impairment of hepatoblast migration is a consequence rather than a cause of aberrant cholangiocyte differentiation.


BrdU, bromodeoxyuridine; Cebpa, CCAAT/enhancer binding protein (C/EBP), alpha; Hnf, hepatic nuclear factor; mRNA, messenger RNA; PCR, polymerase chain reaction; Prox, prospero-related homeobox; qRT-PCR, quantitative reverse transcription polymerase chain reaction.

Materials and Methods

Mice and Genotyping.

Mice carrying a null allele of Tbx3 (Tbx3tm1.1(cre)Vmc, synonym: Tbx3cre)7 were maintained on an outbred NMRI (National Marine Research Institute) background. For timed pregnancies, vaginal plugs were checked in the morning after mating; noon was taken as embryonic day (E) 0.5. Embryos were harvested in phosphate-buffered saline, fixed in 4% paraformaldehyde overnight, and stored in 100% methanol at −20°C before further use. Genomic DNA prepared from yolk sacs or tail biopsy specimens was used for genotyping by polymerase chain reaction (PCR).7 All mice received humane care, and their use was approved by the Institutional Animal Care Committee of Hannover Medical School.

Histological Analysis and Immunofluorescence.

Embryos were embedded in paraffin wax and sectioned to 5 μm. For histological analyses, sections were stained with hematoxylin-eosin. For the detection of antigens, the following primary antibodies were used: rabbit anti-mouse E-cadherin (gift from Rolf Kemler), laminin (Sigma), and cytokeratin18 (Acris Antibodies).

In Situ Hybridization Analysis.

In situ hybridization analysis on 10-μm transverse sections of embryos was performed following a standard procedure with digoxigenin-labeled antisense riboprobes.8

Proliferation and Apoptosis Assays.

Cell proliferation in tissues of E9.0 and E9.5 embryos was investigated by detection of incorporated bromodeoxyuridine (BrdU) similar to published protocols.9 A total of nine sections from three individual embryos per genotype and time point were used for quantification. Statistical analysis was performed using the two-tailed Student t test. Data were expressed as mean ± standard deviation. Differences were considered significant when the P-value was below 0.05.

For detection of apoptotic cells in 5-μm paraffin sections of E9.5 embryos, the terminal deoxynucleotidyl transferase-mediated nick-end labeling assay was performed as recommended by the manufacturer (Serologicals Corp.) of the ApopTag kit used.

Cell Culture and Transfection.

Bipotential mouse embryonic liver cell line 9A1, a kind gift from M.C. Weiss, has been described previously.10 Murine hepatoma Hepa1-6 cells were obtained from the American Type Cell Culture. Cells were seeded in six-well plates and (co-) transfected after 24 hours with 10 μL lipofectamine 2000 (Invitrogen) and 1250 ng of the expression vectors pcDNA3.GL.Onecut1.Myc, pcDNA3.GL.Tbx3.Myc, pcDNA3.GL.Hnf1b.Myc or pVP16.Tbx2-DB, and 1.5 μg pMACS4.1 (Miltenyi Biotech) for enrichment of transfected cells.11 The total amount of plasmid DNA was adjusted to 4 μg by adding pcDNA3.

Semiquantitative Reverse Transcription PCR.

Total RNA was extracted from dissected E9.5 liver buds or cells with RNAPure reagent (Peqlab). RNA (500 ng) was reverse transcribed with RevertAid M-MuLV Reverse Transcriptase (Fermentas). For semiquantitative PCR, the number of cycles was adjusted to the mid-logarithmic phase. Quantification was performed with Quantity One software (Bio-Rad). Assays were performed at least twice in duplicate, and statistical analysis was done as described previously. Primers and PCR conditions are available on request.


Documentation of whole-mount specimens and sections was done as described previously.9


Liver Hypoplasia in Tbx3-Mutant Mice.

To study the hepatic requirement of Tbx3, we maintained a new Tbx3 null allele (Tbx3cre)7 on an NMRI outbred background that supported viability of homozygous mutant embryos until E14.5. At this stage, Tbx3−/− embryos showed an overall size reduction and a disproportionately small liver that was filled with blood cells (Fig. 1A, B). Tbx3-deficient livers were characterized by a dramatic reduction of the hepatoblast marker gene alpha fetoprotein (Afp) and the hepatocyte marker gene albumin (Fig. 1C), whereas expression of Ck18, an antigen confined to cholangiocytes, was increased and detected in 10% to 20% of cells in the mutant liver (Fig. 1D). This suggests that expansion and differentiation of hepatic progenitors is severely disturbed in Tbx3-deficient hepatic tissue.

Figure 1.

Tbx3-deficient mice exhibit severe liver hypoplasia at E14.5. Morphology of whole embryos and livers (A); and hematoxylin-eosin stainings (B), in situ hybridization analysis of Afp and albumin (Alb) (C) expression, and immunofluorescent detection of Ck18 protein (D) on transverse sections of wild-type (wt) and Tbx3-deficient (mt) embryos. Section planes and magnified regions are as indicated by lines and boxes, respectively. Mutant livers are still encapsulated and of normal shape but dramatically reduced in size. Blood cells and Ck18+-cholangiocytes have replaced Alb-expressing hepatocytes. Genotypes and probes are as indicated. White arrowheads point to liver. Abbreviations: pa, pancreas; st, stomach.

Early Disruption of Liver Development in Tbx3−/− Embryos.

To determine at which stage liver organogenesis becomes impaired in Tbx3-deficient embryos, we analyzed Afp expression on sections to evaluate specification of hepatic tissue and morphogenesis of the organ (Fig. 2). In Tbx3−/− embryos of E9.0 (16-somite stage), the ventral foregut endoderm expressed Afp indistinguishably from the wild-type, indicating that a hepatic diverticulum with a pseudo-stratified organization was formed. At E9.5, the mutant hepatic epithelium appeared thickened but lacked the characteristic wild-type protrusions and delaminations. At E10.5 the mesenchyme adjacent to the liver bud was densely populated with hepatoblasts in the wild-type, whereas few Afp-positive cells were found in the underlying mesenchyme of the mutant. Dramatic reduction of Afp-positive cells at E12.5 confirmed the complete failure to expand this hepatic cell population in the mutant. In summary, hepatic specification occurred normally in Tbx3−/− embryos. However, the hepatic primordium failed to expand and to delaminate hepatoblasts into the underlying mesenchyme.

Figure 2.

Hepatic development is disrupted at the emergent liver bud stage in Tbx3−/− mice. In situ hybridization analysis of Afp-expression on transverse sections at the foregut level of wild-type (wt) and Tbx3−/− (mt) embryos during liver development. Stages are as indicated in the figure, dorsal is oriented up. Afp-expression shows that hepatic fate is specified in the Tbx3-deficient foregut endoderm, but delamination of hepatoblasts does not occur. Abbreviations: fe, foregut endoderm; he, hepatic endoderm; hl, hepatic lobe; lib, liver bud; li, liver; stm, septum transversum mesenchyme.

Tbx3 Is Strongly Expressed in the Liver Bud.

We performed section in situ hybridization analysis to correlate the spatio-temporal profile of Tbx3 expression with the phenotypical changes in the mutant (Fig. 3). Tbx3 messenger RNA (mRNA) was first detected at low levels in the hepatic endoderm at the 18-somite stage (E9.0). Tbx3 expression was strongly up-regulated at the 23-somite stage (E9.5) and completely overlapped expression of Afp in the liver bud. At E10.5, Tbx3 was markedly down-regulated and confined to hepatoblasts populating the hepatic lobes. From E12.5, Tbx3 expression was barely detectable in the liver by in situ hybridization analysis. Hence, the profile of Tbx3 expression is compatible with a primary role of Tbx3 in hepatoblast expansion, migration, or differentiation during liver bud development.

Figure 3.

Tbx3 is strongly expressed in the liver bud. Analysis of Tbx3 expression during early liver development by RNA in situ hybridization on transverse sections of wild-type embryos at the foregut level. Dorsal is oriented up. Developmental stages are as indicated in the figure. Tbx3 expression is low in the pseudo-stratified epithelium of the hepatic endoderm (E9.0) but is strongly up-regulated in the hepatic endoderm during expansion of the liver bud and delamination of hepatoblasts into the stroma (E9.25-E9.5). Expression in hepatoblasts is strongly reduced at E10.5, and barely detected at later stages. Abbreviations are as in Fig. 2.

Proliferation Defects in the Tbx3−/− Liver Bud.

To evaluate whether changes of cellular proliferation rates may underlie the morphological defects in liver bud expansion in Tbx3−/− embryos, we performed a BrdU incorporation assay that detects cells in S-phase of the cell cycle (Fig. 4A). At the 18-somite stage (E9.0), when Tbx3 is only weakly expressed, labeling indices of epithelial cells of wild-type and mutant hepatic endoderm were similar. Proliferation in the lateral and dorsal foregut endoderm was higher than that in the hepatic endoderm at this stage but, as expected, was not significantly altered between the two genotypes. At the 23-somite stage, the BrdU labeling index remained high in the wild-type liver bud, exceeding the value of the lateral foregut endoderm considerably. The proliferation rate in the mutant hepatic endoderm reached only half of the wild-type level and was severely decreased in comparison with the adjacent foregut endoderm. Terminal deoxynucleotidyl transferase-mediated nick-end labeling staining showed that apoptosis was unaffected in mutant liver at this stage (Fig. 4B). Hence, severe reduction of cell proliferation in the hepatic epithelium at E9.5 is likely to cause the failure of hepatoblast liver bud expansion in Tbx3−/− mice.

Figure 4.

Proliferation of hepatic endoderm is severely reduced in Tbx3−/− embryos. (A) Analysis of cell proliferation in hepatic and lateral foregut endoderm performed on transverse sections of wild-type (wt) and Tbx3-mutant embryos (mt) at E9.0 (18-somite stage, 18s) and E9.5 (23s) by immunohistochemistry for BrdU. Quantified regions, the epithelium of the lateral foregut, and the hepatic epithelium of the liver primordium, respectively, are marked by colors. Statistical analysis of proliferation rates (% proliferation, as defined by the ratio of BrdU-positive cells to total cell number in the analyzed area) at E9.0 and E9.5 of regions and genotypes as color-coded. Proliferation rates between wild-type and mutant differ significantly in the hepatic endoderm but remain similar in the foregut endoderm that is devoid of Tbx3 expression at E9.5. (B) Terminal deoxynucleotidyl transferase-mediated nick-end labeling staining for apoptosis in the hepatic endoderm (white outline) at E9.5 (23s) does not reveal differences between wild-type and Tbx3-mutant embryos. Abbreviations are as in Fig. 2.

Expression of Cell-Cycle Regulators Is Unchanged in Tbx3-Deficient Liver Buds.

Several studies have implicated Tbx3 in the control of the cell cycle by direct repression of genes encoding inhibitors of cell-cycle–dependent kinases. To uncover the primary molecular changes that may underlie the proliferation defect of the Tbx3−/− hepatic endoderm, we analyzed expression of a number of genes encoding cell-cycle regulators by in situ hybridization analysis in the liver bud at E9.5 when morphological differences were manifested (Supporting Fig. 1A). Expression of p15INK4b, p16INK4a, p18INK4c, p19INK4d, p19Arf, and p27Kip1 was detected in the hepatic endoderm of neither wild-type nor mutant embryos, although extrahepatic expression domains confirmed the quality of probes and experimental conditions. Expression of p21Cip1 was found throughout the hepatic and foregut endoderm in either genotype. p57Kip2 was expressed in the hepatic endoderm and in the underlying mesenchyme in wild-type and indistinguishably in mutant embryos.

Targeted mutations have demonstrated the requirement for the genes encoding hepatocyte growth factor (Hgf) and its receptor c-Met for proliferation or survival of hepatocytes.12 Expression of c-Met in the hepatic endoderm, and of Hgf in the underlying mesenchyme was unaltered in the mutant. Expression of the proto-oncogene c-Myc, and of the genes encoding the cell cycle regulator cyclin D1 (Ccnd1) and the forkhead transcription factor Foxm1 were similarly unaffected (Supporting Fig. 1B).

Quantitative PCR on reverse-transcribed mRNA isolated from E9.5 liver buds (qRT-PCR) independently confirmed that expression levels of the cell cycle regulators p19Arf, p21Cip1, p27Kip1, Cdk1, and CyclinD1 are unchanged in Tbx3−/− liver buds (Supporting Fig. 1C). Therefore, it is unlikely that phenotypical defects in Tbx3−/− liver buds are an immediate consequence of transcriptional deregulation of these cell-cycle regulators, particularly p19Arf.

Hepatoblast Migration Requires Tbx3.

After establishment of the pseudo-stratified hepatic epithelium at E9.0, tissue protrusions arise from which hepatoblasts delaminate. This process is accompanied by down-regulation of the cell adhesion molecule E-cadherin and disintegration of the basal lamina (Fig. 5A). In Tbx3-mutant embryos, hepatic cords did not protrude from the liver bud, and hepatoblasts failed to invade the underlying septum transversum mesenchyme. E-cadherin remained high in the hepatic epithelium at the 30s-stage, and the basal lamina that surrounded it stayed intact (Fig. 5A). To score for molecular changes instrumental in this phenotype, we analyzed expression of genes regulating migration of hepatoblasts. Expression of hematopoietically expressed homeobox that regulates the transition from a columnar to a pseudo-stratified epithelium13 was unchanged in the Tbx3-deficient liver bud at E9.5 (Fig. 5B). Similarly, expression of the genes encoding the H2.0-like homeodomain protein (Hlx) and the Zn-finger transcription factors GATA-binding proteins 4 and 6 (Gata4 and Gata6)14, 15 was unaltered in the mesenchyme of the septum transversum region (Supporting Fig. 2). In Prospero-related homeobox 1 (Prox1) mutant mice, proliferation of hepatoblasts is decreased and delamination from the hepatic diverticulum is disturbed.16 Expression of Prox1 in the Tbx3−/− hepatic epithelium was unchanged at E9.0 but severely down-regulated at E9.5 (Fig. 5B). Thus, failure of migration of hepatoblasts into the surrounding mesenchyme may be caused by the inability to maintain Prox1 expression.

Figure 5.

Tbx3 controls hepatoblast migration. (A) Immunofluorescent detection of E-cadherin and laminin expression on transverse sections through the foregut region of wild-type (wt) and Tbx3−/− embryos (mt) at the 30-somite stage. Expression of both antigens remains high in the mutant liver bud. (B) In situ hybridization analysis for hematopoietically expressed homeobox 1 and Prox1 expression in the hepatic epithelium at the indicated stages. Prox1 expression is severely down-regulated at the 25-somite stage.

Hepatic Differentiation Defects in Tbx3−/− Embryos.

Absence of albumin-expressing hepatocytes but presence of Ck18-positive cholangiocytes at E14.5 (Fig. 1C,D) indicated that the hepatoblast lineage decision was affected in Tbx3-deficient livers. To determine the temporal onset of differentiation defects and their possibly causal relation with the observed cellular defects in proliferation and migration of hepatoblasts, we analyzed expression of a panel of genes central to hepatocyte and cholangiocyte lineage decision, respectively, at E9.0 and at E9.5, that is, before and at the onset of phenotypic changes in the Tbx3−/− liver bud (Fig. 6A).

Figure 6.

Hepatobiliary differentiation is affected in the hepatic endoderm of Tbx3−/− mice. (A) In situ hybridization analysis of expression of genes controlling hepatic differentiation on transverse sections through the foregut region of wild-type (wt) and Tbx3−/− (mt) embryos at E9.0 and E9.5. Dorsal is oriented up. Genotypes, probes, and stages are as indicated in the figure. (B) QRT-PCR analysis of marker genes on mRNA from E9.5 liver buds. Expression levels are relative to wild-type (100%). Expression of Cebpa and Hnf4a is down-regulated, Onecut1 and Hnf1b expression is up-regulated in the E9.5 Tbx3−/− liver bud, indicating that the hepatocyte fate is lost at the expense of the cholangiocyte fate. (C) Scheme for the role of Tbx3 in liver development. Tbx3 promotes the progression from the pseudo-stratified epithelium (E9.0) to a cell-emergent liver bud (E9.5). Cholangiocyte differentiation is prevented by repression of Onecut1 (and its target Hnf1b), whereas Cebpa and Hnf4a expression is maintained, leading to differentiation of hepatocytes with their high proliferation and migration potential.

CCAAT/enhancer binding protein (C/EBP), alpha (Cebpa) and hepatic nuclear factor 4a (Hnf4a) encode transcription factors that are involved in the early stages of hepatocyte differentiation,17–19) whereas the transcriptional regulators Onecut1 (Hnf6), Onecut2, and Hnf1b control cholangiocyte differentiation.20–22 At E9.0, Cebpa and Hnf4a were expressed in the hepatic endoderm, whereas expression of Onecut1, Onecut2, and Hnf1b was hardly detected. In the hepatic epithelium of E9.5 wild-type embryos, hepatocyte genes Hnf4a and Cebpa were strongly expressed, whereas expression of Onecut1, Onecut2, and Hnf1b was not detected, arguing that hepatoblasts started to differentiate into hepatocytes. Down-regulation of Cebpa and Hnf4a and up-regulation of Hnf1b and Onecut1 expression in the Tbx3−/− hepatic epithelium suggest that hepatoblast differentiation became redirected to cholangiocytes. QRT-PCR analysis on mRNA obtained from E9.5 liver buds independently confirmed the observed changes of expression (Fig. 6B). Expression of Pdx1, a marker for pancreatic fate,23 was unchanged in the Tbx3−/− embryo (Supporting Fig. 2), arguing against an expansion of pancreatic fates into the liver region. Expression of the gene encoding the signaling molecule Sonic hedgehog (Shh) remained excluded from the hepatic endoderm in the Tbx3-mutant embryo (Supporting Fig. 2), indicating that deregulation of Sonic hedgehog does not underlie this hepatic phenotype as proposed for hematopoietically expressed homeobox–mutant mice.13

Together these findings suggest that Tbx3 controls hepatic development by suppressing cholangiocyte and favoring hepatocyte differentiation in the liver bud at E9.5 (Fig. 6C).

Tbx3 and Onecut1 Antagonistically Regulate Hepatobiliary Fate Decision.

To further decipher the molecular pathways regulated by Tbx3 in hepatoblast differentiation, we employed overexpression approaches in cellular systems (Fig. 7). For loss-of-function experiments, we used a Tbx2-VP16 expression construct encoding the Tbx2 DNA-binding domain in fusion with the viral transcriptional activation domain VP16. Tbx2-VP16 competes with Tbx3 for the same conserved DNA binding sites and activates transcription, thus acting as a dominant negative version of the transcriptional repressor Tbx3. We performed overexpression experiments with Tbx3 and Onecut1 to analyze the role of either factor in cholangiocyte differentiation in both the hepatoblast cell line 9A110 and the hepatoma cell line Hepa1-6.24 Semiquantitative RT-PCR was used to judge the changes in expression of hepatocyte (albumin) and cholangiocyte (Ck7)25 differentiation markers, of transcriptional regulators for the hepatocyte (Hnf4a, Cebpa) and cholangiocyte lineage (Onecut1), and for the cell cycle inhibitor p19Arf. Overexpression of Tbx3 in 9A1 hepatoblasts led to increased levels of Hnf4a and Cebpa whereas expression of Tbx2-VP16 resulted in downregulation of Hnf4a, Cebpa and albumin (Fig. 7A). This supports the in vivo analysis and suggests that Tbx3 is required to maintain and enhance hepatocyte differentiation. Because Tbx2-VP16 represents a constitutive transcriptional activator, down-regulation of Hnf4 and Cebpa is compatible with the notion that Tbx3 functions indirectly by repressing a transcriptional repressor of these genes. Transfection of a Onecut1 expression construct similarly resulted in down-regulation of Hnf4a and Cebpa (Fig. 7A). Inhibition of Tbx3 function by Tbx2-VP16 in the hepatocyte cell line Hepa1-6 also resulted in the repression of hepatocyte marker genes, arguing that continued albeit low-level expression of Tbx3 is required to maintain the phenotype of hepatocytes (Fig. 7B). Cotransfection of Onecut1 in these cells further repressed Hnf4a expression and activated the cholangiocyte differentiation marker Ck7, robustly suggesting that inhibition of Tbx3 and activation of Onecut1 synergize in (trans-) differentiation into cholangiocytes (Fig. 7B). However, it is unlikely that endogenous Onecut1 mediates Tbx3 function on hepatocyte marker genes in these cellular systems, because Onecut1 expression was surprisingly activated by Tbx3 in all cell lines tested (Fig. 7). Levels of p19Arf were unaffected by changes of Tbx3 and Onecut1 expression, providing additional evidence that p19Arf is not involved in the hepatobiliary lineage decision.

Figure 7.

Tbx3 and Onecut1 antagonistically regulate hepatobiliary fate decision in cellular systems. Analysis of marker gene expression by qRT-PCR in the hepatoblast cell line 9A1 (A) and the hepatoma cell line Hepa1-6 (B) 48 hours posttransfection with expression constructs for Tbx3, Tbx2-VP16 (a dominant-negative version of Tbx3), Onecut1, and their combinations. Messenger RNA expression of indicated markers is shown relative to the empty vector control. Tbx3 enhances and maintains hepatocyte marker expression, whereas Onecut1 down-regulates hepatocyte markers and up-regulates in combination with Tbx2-VP16 the cholangiocyte marker gene Ck7.

In summary, our in vitro experiments further support the role of Tbx3 in the maintenance of hepatocyte differentiation by controlling an early transcription factor network. They indicate an important role of Onecut1 as an inducer of cholangiocyte differentiation.


Differentiation of hepatoblasts in hepatocytes and bile duct cells is temporally and spatially separated during liver development. Although the bulk of hepatoblasts in the liver bud differentiates into hepatocytes, cholangiocytes derive from a group of hepatoblasts located in proximity to the portal vein.26,27 In the latter case, the existence of inducing signals from the mesenchyme surrounding the portal vein has been suggested,28 whereas it was less clear how hepatocyte differentiation is favored in early liver development. Here, we have identified Tbx3 as a transcriptional regulator of hepatobiliary lineage decision in the liver bud. We propose that Tbx3 maintains the hepatocyte and suppresses the cholangiocyte lineage by antagonistically regulating expression of transcription factor genes required for either differentiation pathway. Proliferation and migration defects in the Tbx3-deficient liver bud are a consequence rather than a cause of aberrant cell differentiation.

A Primary Function of Tbx3 in Hepatobiliary Cell Fate Decision.

We found changes in the gene expression pattern of the epithelium of the Tbx3-deficient liver bud, including down-regulation of Hnf4a, Cebpa, and the hepatocyte marker albumin, and up-regulation of Onecut1, Hnf1b, and the biliary marker CK18 at later stages that are fully compatible with the notion that Tbx3 controls hepatobiliary fate decision by antagonistically regulating expression of key transcriptional mediators of the hepatocyte and cholangiocyte pathways. On the molecular level, gene expression and cell fate changes can be rationalized by two opposing models. First, Tbx3 is primarily required to maintain expression of Cebpa and Hnf4a; thus, the hepatocyte gene program in the hepatic epithelium.2, 19, 27 Down-regulation of Cebpa in Tbx3 mutants may cause premature biliary differentiation, and the increase in Hnf6 and Hnf1b expression might be secondary to Cebpa misregulation. This model is supported by the temporal profile of loss of Cebpa expression and the known role of Cebpa as a suppressor of cholangiocyte differentiation.28 However, it requires the presence of a transcriptional mediator because Tbx3 is a bona fide transcriptional repressor.29 Indeed, our cell culture experiments with a dominant-negative form of Tbx3, Tbx2-VP16, argue for the presence of a Tbx3-repressed transcriptional repressor of Hnf4a and Cebpa transcription. In a second model, Tbx3 is primarily required to suppress the cholangiocyte gene program. Up-regulation of transcriptional regulators of cholangiocyte differentiation including Onecut1 and its target Hnf1b20, 22 in Tbx3−/− embryos may result in cholangiocyte differentiation, which in turn represses Cebpa and Hnf4 expression and hepatocyte differentiation. This model gains support from a number of experimental findings. First, up-regulation of Onecut1 expression directly correlates with the expression profile of Tbx3 and the temporal onset of defects in the Tbx3-deficient embryo. Second, ectopic expression of Onecut1 in hepatoblasts and hepatocytes represses transcription of Hnf4a and Cebpa and enhances the effect of Tbx3 inhibition on cholangiocyte differentiation. Third, up-regulation of Onecut1 is compatible with the nature of Tbx3 as a repressor of transcription and does not need further intermediary steps. Although a direct regulation of Onecut1 by Tbx3 is circumstantially supported by the finding that the closely related gene Onecut2 is a direct target of the T-domain protein Tbet,30 we currently have no molecular evidence for such a mode of regulation. Indeed, the up-regulation of Onecut1 in Tbx3-overexpressing cells seems to contradict this assumption. We currently cannot resolve the discrepancy between the regulation of Onecut1 by Tbx3 in vivo and in vitro. Possibly, the cellular system is inadequate to fully reflect the endogenous regulation by Tbx3 because of lack of cofactors present in the early liver bud. Future work will analyze the possibility of combinatorial regulation of Onecut1 by Tbx3 and other transcriptional regulators.

As a third possibility, Tbx3 may simultaneously maintain hepatocyte and suppress cholangiocyte differentiation. This may be achieved by independently maintaining the transcription of regulators of hepatocyte differentiation and repressing regulatory genes for cholangiocyte differentiation.

Proliferation and Migration Defects in the Tbx3-Deficient Liver Bud Are Secondary to Cell Fate Changes.

Recent analysis of Tbx3-mutant hepatoblasts suggested that the cell-cycle inhibitors p21Cip1 and p19Arfmight be primary molecular targets of Tbx3 function in the liver.6 This implied that proliferation and migration defects of hepatoblasts precede and cause aberrant cell differentiation.6 Our molecular analyses of Tbx3 loss- and gain-of-function scenarios both in vivo and in vitro, however, clearly show that loss and gain of Tbx3 expression does not result in immediate changes of cell-cycle regulators, including p19Arf and p21Cip1. Hence, proliferation and migration defects and up-regulation of p19Arf are likely to be secondary and late consequences of the cell fate changes in the liver bud.

We assume that severe reduction of cellular proliferation in the hepatic epithelium at E9.5 is a consequence of the failure to maintain Cebpa and Hnf4a expression, and thus hepatocyte fate. However, the molecular mediators of this phenotype remain unknown.

In Tbx3−/− embryos, delamination of hepatoblasts failed, the laminin-rich membrane around the liver bud remained intact, and cells of the hepatic epithelium retained strong expression of E-cadherin. This phenotype mimics the findings in Prox1−/− animals16 and suggested an epistatic relation between the two genes in liver development. Intriguingly, Prox1 expression was established normally in Tbx3−/− embryos, but expression dramatically declined from E9.5 on. Hence, Tbx3 does not establish Prox1 expression but is indirectly required for its maintenance. Because Prox1 remains continuously expressed in hepatocytes but is lost from cholangiocytes,31 down-regulation of Prox1 in the mutant is likely to reflect the hepatoblast fate switch at E9.5. The fact that Prox1 suppresses gallbladder-specific genes may further contribute to or reinforce cholangiocyte differentiation in Tbx3 mutants.32

Intriguingly, Onecut1, whose expression is up-regulated in the Tbx3-deficient liver bud, has been implicated in both cell proliferation and migration in ways contrary to our findings. It was previously shown that forced expression of Onecut1 stimulates hepatocyte proliferation and leads to increased expression of hepatocyte growth factor-alpha, cyclin D1, and Foxm1 in mature hepatocytes.33 Hence, increased expression of Onecut1 in the Tbx3-deficient liver bud should stimulate proliferation by increased expression of these target genes. Yet, proliferation in the Tbx3-mutant liver bud is reduced and hepatocyte growth factor-alpha, cyclin D1, and Foxm1 expression was unchanged. We cannot explain the discrepancy of these findings but suggest that Onecut1 transcriptional activity may depend on cofactors as shown before for the activation of Foxa2 transcription by a Onecut1/Cebpa binary complex.34 Because Cebpa is dramatically reduced in the Tbx3-deficient liver bud, it is plausible that Onecut1 transcriptional activity may shift by changed complex formation.

In Oc1/Oc2 double mutants, the basal lamina surrounding the liver bud remains intact, and hepatoblasts fail to delaminate from the epithelium.21 Up-regulation of Onecut1 in the Tbx3-decifient liver buds makes it unlikely that the two factors are causally involved in the proliferation and migration defects. We favor instead that down-regulation of Prox1 may cause this phenotype. It is currently unclear whether Prox1 and Oc1/Oc2 act independently in liver bud expansion or whether they represent independent pathways.

Our in situ hybridization analysis has shown that high Tbx3 expression is confined to a short time window in hepatic development and tightly correlates with the onset of morphological and molecular changes in Tbx3-mutant livers at E9.5. Yet, our overexpression experiments in hepatoma cells argue that low-level expression of Tbx3 in mature hepatocytes is required to maintain the fate of these cells and prevent trans-differentiation into cholangiocytes. Conditional ablation of Tbx3 at later time points may open avenues to further analyze its role in liver development and homeostasis.


The authors thank Dr. D. Tosh for providing us with complementary DNAs for Cebpa and Hnf4a, Dr. C. J. Sherr for plasmids with various INK4 DNAs, M.C. Weiss and U. Kossatz-Böhlert for cells, Dr. R. Kemler for antibodies, H. Farin for practical advice, and H. Farin, M.-O. Trowe, and N. Malek for discussion and critical reading of the manuscript.