Embryonic development of the liver


  • Roong Zhao,

    1. Department of Cell Biology Neurobiology and Anatomy, Medical College of Wisconsin, Milwaukee, WI
    Search for more papers by this author
  • Stephen A. Duncan

    Corresponding author
    1. Department of Cell Biology Neurobiology and Anatomy, Medical College of Wisconsin, Milwaukee, WI
    • Department of Cell Biology, Neurobiology and Anatomy, Medical College of Wisconsin, 8701 Watertown Plank Road, Milwaukee, WI, 53226
    Search for more papers by this author
    • fax: 414-456-6517

  • Potential conflict of interest: Nothing to report.

The combined use of genetic manipulation in mice and advanced culturing techniques, along with the availability of model genetic systems, underlies the recent expansion in our understanding of the fundamental processes that govern the development of the liver. Although gaps in our knowledge still exist, many of the molecules and pathways that act during the earliest stages of hepatic development have been identified. This knowledge has been applied to control the differentiation of embryonic stem and hepatic progenitor cells into hepatocytes and cholangiocytes, raising the prospect of generating quantities of hepatic cells in culture that are suitable for therapeutic uses.


mRNA, messenger RNA; E, embryonic day; FGF, fibroblast growth factor; BMP, bone morphogenetic protein.

From Where Do Hepatic Cells Arise?

The principal functional cell type of the liver is the hepatocyte, which accounts for 78% of parenchymal volume in the rat liver.1 The hepatocyte is a polarized epithelial cell that exhibits both endocrine and exocrine properties. Biliary epithelial cells (cholangiocytes), sinusoidal endothelial cells, hepatic stellate cells (Ito cells), Kupffer cells, and pit cells (liver-specific natural killer cells) represent the majority of non-hepatocyte cell types in the liver. Although most research has traditionally focused on the hepatocyte, it is now clear that non-hepatocyte cell types make crucial contributions to liver function and pathology, and efforts to understand their embryonic origins and the mechanisms underlying their development have intensified.2 In addition, there is mounting evidence to support the proposal that some nonparenchymal cells have a direct impact on the development of hepatocytes.3–5 For example, hepatoblasts purified from fetal mouse liver have been shown to require nonparenchymal cells for proliferation and expression of hepatocyte messenger RNA (mRNA).4

To understand how the liver is generated during embryogenesis, it is first necessary to define the lineage relationships and the morphogenic movements of the cells that contribute to the liver's function. Several decades of work in numerous vertebrate model systems have firmly established that hepatocytes and bile duct cells originate from a common precursor, the hepatoblast, that derives from the definitive endoderm (Fig. 1).6–13 At around embryonic day (E) 7.0 in the mouse, definitive endoderm emerges from the primitive streak to displace the extra-embryonic endoderm of the yolk sac. Shortly after this, the endoderm invaginates to form a portal at the anterior region of the developing embryo that will ultimately define the foregut of the mouse. By approximately E8.0, in mouse embryos containing seven somite pairs, the ventral wall of the foregut endoderm is positioned adjacent to the developing heart, and signals from the heart induce the underlying endoderm to initiate its development toward a hepatic fate.6, 9, 14, 15 The endoderm responds to this induction by generating the primary liver bud that can be identified as an anatomical outgrowth from the ventral wall of the foregut by E8.5 to E9.0 in mouse embryos containing 10 to 12 somite pairs. Recently, lineage tracing experiments in cultured mouse embryos have accurately defined the position of endodermal cells that give rise to the hepatic progenitors prior to their interaction with the heart.16 Two domains of endodermal cells were identified when E8.5 embryos containing one to three somite pairs were labeled: a lateral endodermal cluster that migrates toward the midline as the foregut is generated, and ventral midline endodermal lip cells that grow caudally to generate the midline of the ventral aspect of the gut. By E9.5, in mouse embryos containing approximately 15 to 20 somite pairs, the basement membrane surrounding the liver bud is lost, and cells delaminate from the bud and invade the surrounding septum transversum mesenchyme as cords of hepatoblasts (see Fig. 1; Fig. 2A-B). Expression of a number of proteins has been used to identify cells undergoing this phase of hepatic development, including expression of the transcription factors HNF4α and GATA4, which identify the migrating hepatic cells and the septum transversum mesenchyme, respectively (see Fig. 2). Although the molecular mechanisms that control the onset of delamination and migration of the hepatoblasts are not well defined, the process does require the transcription factor prospero-related homeobox 1 (Prox1).17 Examination of Prox1−/− embryos revealed that hepatoblasts failed to migrate and invade the septum transversum mesenchyme, and the mutant cells remained clustered and surrounded by basement membrane. The Prox1−/− hepatoblasts also expressed abnormally high levels of the cell adhesion molecule E-cadherin, which may contribute to the inability of these cells to fully delaminate.17 Surprisingly, the Prox1−/− embryos did generate liver lobes that were devoid of hepatocytes.17 This finding implies that the mesenchymal component of the liver contributes a significant portion of the morphogenetic cues required for at least some aspects of hepatic architecture.

Figure 1.

Overview of early stages of liver development in mouse embryos. At E8.0, in embryos containing fewer than six to seven somites, a ventral portion of the definitive endoderm (blue area) is positioned close to the developing heart (red area) and in the vicinity of mesenchyme that will contribute to the septum transversum (green area). In response to inductive cues from these tissues, including BMPs and FGFs, the ventral endoderm is specified to adopt a hepatic fate. By E8.5, in embryos containing 8 to 10 somites, the specified hepatic endoderm forms the liver bud (blue area), which expresses several liver mRNAs, including Albumin. Endothelial cells (white area) surround the bud and are necessary for its expansion. By E9.5, the nascent hepatoblasts delaminate from the endoderm and cords of hepatoblasts invade the septum transversum mesenchyme, which is the source of stellate cells as well as sinusoidal endothelial cells that begin to form vessels (white area). Transcription factors with roles in specific aspects of early hepatic development discussed in the text are shown below. He, heart; S, septum transversum; VE, ventral endoderm; FGF, fibroblast growth factor; BMP, bone morphogenetic protein; E, endothelial cells; Lb, liver bud; H, hepatoblasts.

Figure 2.

Expression of HNF4α and GATA4 identifies migrating hepatoblasts and septum transversum mesenchyme, respectively. (A,C) HNF4α and (B,D) GATA4 were identified in transverse sections through the developing liver of (A,B) E9.5 and (C,D) E10.5 mouse embryos by immunohistochemistry. HNF4α is detected within the hepatoblasts of the liver bud (brown nuclear staining) (solid arrow), but not the septum transversum mesenchyme (brown nuclear staining) (open arrow). In contrast, at E9.5 GATA4 is detected within the mesenchyme but not the heptoblasts. At E10.5, GATA4 is identified within the endothelial cells (arrowhead) lining newly formed blood vessels and sinusoids, while HNF4α is restricted to the hepatoblasts (arrow). Boxed areas are shown in a higher resolution within the insets.

Generation of the liver bud is also accompanied by development of the hepatic vasculature, which forms through a combination of angiogenesis and vasculogenesis.5, 18–20 As soon as the primary liver bud emerges from the endoderm, angioblasts or endothelial cells surround it and separate it from the septum transversum mesenchyme (see Fig. 1).5, 21 Studies of Flk1−/− mouse embryos, which lack mature endothelial cells and blood vessels, revealed that although initiation of hepatic development had occurred in these embryos, the liver bud failed to expand, and there was no evidence of hepatoblasts invading the septum transversum mesenchyme.5 These results suggest not only that the hepatic vasculature develops in concert with the hepatoblasts, but that endothelial cells have an integral role in controlling growth of the hepatic primordium. In the adult mammal, the afferent blood vessels of the liver consist of branches of the hepatic artery and portal veins, and the efferent vessels consist of centrilobular veins. These two systems are connected by a network of small capillaries called sinusoids, which are separated from the basal surface of the hepatocytes by the space of Disse, which also contains retinoid-storing hepatic stellate cells. The sinusoidal capillaries consist of phagocytic Kupffer cells, which scavenge spent cell debris from the circulation, and fenestrated endothelial cells that are highly specialized to facilitate selective transport between hepatocytes and the blood.22 Although Kupffer cells are believed to develop from bone marrow–derived monocytes in adults, their presence in the fetal liver precedes bone marrow development, and they may originate from the yolk sac.23 The sinusoidal capillaries and portal veins are among the first hepatic vessels to develop, with centrilobular veins and portal arteries forming later.19 Descriptive studies of both human and mouse development are consistent with the proposal that sinusoids develop through vasculogenesis of vessels that originate within the mesenchyme of the septum transversum.5, 19, 20, 23, 24 In chick embryos, the septum transversum develops later than seen in mammals; however, lineage-tracing experiments have shown that the liver mesothelium may supplant the septum transversum as the origin of the hepatic sinusoids during avian development.18 Hepatic stellate cells also appear to originate from the septum transversum mesenchymal cells, and many transcription factors with roles in stellate cell activation (e.g., the forkhead box factor FoxM1 and the LIM homeobox protein Lhx2) are expressed in both cell types.24–28

Which Pathways Are Required to Initiate Hepatogenesis?

Signaling Molecules.

A concerted effort has been made over the last decade to identify the factors that regulate the onset of hepatogenesis. Although experiments involving the culture of explanted chick/quail tissues had identified the heart and septum transversum mesenchyme as important sources of signals for development of the liver, nothing was known about the identity of the signaling molecules.6, 15 Advances were made when an assay was developed that reconstituted hepatic specification in vitro.9 In this assay, expression of hepatic mRNA within cultured ventral endoderm, isolated from four to six somite stage (E8.0) mouse embryos, was shown to be dependent upon the presence of cardiogenic mesoderm.9 Using this assay, Jung et al. demonstrated that fibroblast growth factors (FGF) 1 and 2, but not FGF8, could substitute for cardiogenic mesoderm to induce the onset of hepatic development in primary ventral endoderm cultures.14 In addition, inclusion of a soluble inhibitor of FGF signaling in the culture medium could repress cardiogenic mesoderm-induced hepatic gene expression within the ventral endoderm explants. Not only is the presence of FGF essential for induction of hepatic fate, but so is the local concentration of FGF.10 Recent work by Serls et al. demonstrated that when ventral endoderm explants were cultured in the absence of FGF, they expressed pancreatic mRNA; when FGF2 was included at 5 ng/mL, they expressed hepatic mRNA; and at 50 ng/mL FGF2, they expressed lung markers.29 These data imply that thresholds of FGF concentration emanating from the heart are used to pattern ventral endoderm fate during endoderm development. Although FGF8 was incapable of inducing hepatic gene expression in endodermal explant assays, it was found to contribute to outgrowth of the cultured hepatic bud.14 In addition to acting during the onset of hepatic development, FGFs also function during the later stages (E10–E12) of hepatogenesis; FGF8, in particular, has been shown to contribute to changes in cell morphology, proliferation, and survival in fetal organ cultures.30 It therefore appears that specific FGFs act at multiple and distinct stages of hepatic development.14 While it is clear that FGF1 and FGF2 have the ability to induce hepatogenesis, and both are expressed in the cardiogenic mesoderm,14 it is worth noting that no liver defect has been described in Fgf1−/−Fgf2−/− mice, which are viable, fertile, and display only mild neurological defects.31 Because there are 22 members of the FGF protein family, it seems likely that other FGFs will be involved in inducing hepatic specification and that the regulation of liver development by this family of proteins is complex.

Initially it was believed that FGF signaling alone was sufficient to induce the onset of hepatic development. However, a role for bone morphogenetic protein (BMP) signaling in hepatic induction was implied when Rossi et al. included a BMP antagonist, Xnoggin, in the in vitro hepatic induction assay.32 In their study, when cardiac mesoderm, septum transversum mesenchyme and ventral endoderm were cultured together in the presence of Xnoggin, the endoderm failed to express Albumin mRNA. Moreover, addition of exogenous BMPs, with BMP2 being the most potent, could circumvent blocking Albumin expression with Xnoggin. In addition, the requirement for BMP during hepatic induction was independent from FGF, because inclusion of FGF2 in explants containing Xnoggin did not facilitate Albumin expression. These data suggest that FGFs and BMPs act cooperatively to induce hepatic development within the ventral endoderm. The sources of BMP signals that are required for hepatic specification are likely to be diverse. Analyses of transgenic mouse embryos expressing LacZ from the endogenous Bmp4 locus are consistent with abundant levels of BMP4 being expressed from the septum transversum mesenchyme. If the septum transversum mesenchyme is the source of BMP-inductive cues, then the wealth of data supporting the view that cardiogenic mesoderm alone is sufficient to induce hepatic development must be reinterpreted.6, 9, 33, 34 However, BMPs are members of the transforming growth factor β superfamily, and several have been shown to be competent to overcome Xnoggin-mediated suppression of hepatic induction.32 Although the cardiogenic mesoderm expresses relatively low levels of BMP4, it does express several additional BMPs, suggesting it could still act as a source of BMP-inductive cues.35, 36 To definitively establish whether the septum transversum mesenchyme is essential for hepatic specification by providing BMP signals will therefore require defining which specific transforming growth factor β family proteins are neccessary for hepatic specification and whether ablation of the septum transversum mesenchyme prevents the onset of liver development.

Transcription Factors.

The response of the endoderm to inductive cues is to initiate a program of hepatic gene expression. Because gene expression is primarily controlled through transcription, it seems likely that specific transcription factors exist that contribute to the onset of hepatic differentiation. Studies in several model organisms have identified a number of factors that are required for formation of the definitive endoderm, but few have been identified that specifically affect the onset of hepatogenesis.37 An exception to the paucity of transcription factors identified to be involved in the onset of hepatogenesis was the discovery that the livers were absent from midgestation (E13.5) stage mouse embryos lacking the homeobox transcription factor Hex.38, 39 Close examination of Hex−/− embryos revealed the presence of a presumptive hepatic bud at E9.0, but no indication that this diverticulum expressed hepatic mRNA, including those encoding albumin and α-fetoprotein. The requirement for Hex during hepatic development appears to be evolutionarily conserved, because knockdown of Hex in zebrafish also prevents development of the liver.40 Although the absence of hepatic markers in Hex-null embryos at E9.5 would be consistent with a failure in hepatic specification, recent analysis of Hex−/− ventral endoderm from E8.5 embryos containing 10 somite pairs found that Alb, Ttr, and Prox1 could in fact be detected by reverse-transcriptase polymerase chain reaction.41 Moreover, cardiogeneic mesoderm was capable of inducing hepatic markers within the Hex−/− endoderm in cultured tissue explants. Although hepatic markers were expressed, the proliferation of the specified endodermal cells was markedly reduced in Hex−/− embryos.41 These studies imply that although Hex is dispensable for induction of hepatic development, it is essential for expansion of the nascent hepatoblast population. In addition to being essential for development of the early liver bud, studies in chick embryos have shown that expression of Hex in the ventral endoderm requires both FGF and BMP signaling.13 Moreover, Hex appears to be a direct target of BMP activity, because analyses of transcriptional regulatory elements within the Hex promoter have identified the presence of a BMP-responsive element that can be activated through SMAD1 and SMAD4.42 It would therefore appear that early development of the liver is regulated at least in part by direct FGF/BMP-mediated induction of Hex within the ventral endoderm.

In addition to genetic analyses, molecular and biochemical studies have also identified transcription factors that may act during the onset of liver development. Such studies have focused on the Albumin gene, which encodes one of the earliest liver markers to be expressed in the hepatic endoderm following specification. Furthermore, Albumin mRNA expression is restricted to the developing liver and is one of the few hepatic mRNAs not detectable in the gut or other endodermal organs within the embryo.43 Studies of Albumin's transcriptional regulatory elements revealed that FoxA forkhead box (previously HNF3) and GATA zinc finger transcription factors interact with the Albumin enhancer before the onset of ventral endoderm expression of Albumin mRNA.9, 44–46 Biochemical studies indicated that these transcription factors exhibit an unusual capacity to identify their binding sites within compacted chromatin, which is normally inaccessible to transcription factor binding.44, 46–49 The cooperative binding of FoxA and GATA was also shown to decompact local chromatin structure and reposition nucleosomes within the Albumin enhancer. This reorganization of chromatin structure is believed to facilitate the interaction of transcriptional activators with their binding sites, resulting in the onset of Albumin expression.44, 47, 48 This model, shown in Fig. 3, is appealing from a developmental perspective because it proposes that specific transcription factors known as “pioneer” factors could mark sets of genes as competent to be expressed when presented with the correct developmental cues.45 Moreover, it offers one explanation for why a limited set of signaling molecules has the capacity to generate such diverse responses at the level of gene expression and cell fate. Although appealing, a number of questions are raised by this model. The extent to which the proposed mechanism is applicable is unknown, because the effect of FoxA and GATA on enhancers and decompaction of chromatin within a developmental context has only been studied using the Albumin gene. Still, FoxA has been shown to relieve p53-mediated transcriptional repression of the α-fetoprotein gene in vitro.50–52 Also, it is yet to be determined whether, prior to formation of the ventral endoderm, the regulatory elements of hepatic genes are indeed in a compacted state and whether specific mechanisms exist to generate such a state during embryogenesis.

Figure 3.

Model illustrating how FoxA and GATA factors may act to control competency of the endoderm to adopt a hepatic fate. It has been proposed that in tissues such as ectoderm—which are not competent to adopt a hepatic developmental program—chromatin containing hepatic transcriptional regulatory elements is compacted and unable to bind transcriptional activators. GATA4 and FoxA, however, are able to stably interact with their binding sites in the context of compacted chromatin. As a consequence of binding, these “pioneer” factors remodel chromatin around the transcriptional regulatory elements, making the elements accessible to transcriptional activators that can control the onset of expression of hepatic mRNA. Fgf, fibroblast growth factor; Bmp, bone morphogenetic protein.

Genetic studies of nonmammalian organisms including nematodes, zebrafish, frogs, and fruitflies support a role for GATA and FoxA factors in development of the endoderm.53–60 In mammals, genetic evidence that definitively demonstrates a requirement for FoxA and GATA factors in controlling competency of the endoderm to adopt a hepatic fate has been difficult to produce. This is due either to functional redundancy between individual FoxA and GATA factor proteins or an essential role for some of these factors prior to the onset of hepatic development.61–70 For example, development of Gata6−/− mouse embryos is blocked before hepatic specification due to an indispensable requirement for GATA6 in extraembryonic endoderm differentiation.63, 65 This early embryonic lethality has, however, recently been circumvented by providing Gata6−/− embryos with a wild-type extra-embryonic endoderm via tetraploid embryo complementation.71 In contrast to expectations, examination of Gata6−/− “rescued” embryos revealed that loss of GATA6 did not affect the competency of endoderm to express hepatic markers following specification. However, the absence of GATA6 did prevent expansion of the liver bud and commitment of the nascent hepatoblasts to a liver cell fate.71 The observation that GATA4 was still expressed in the ventral endoderm of GATA6 null embryos suggested that GATA4 could compensate for the absence of GATA6 in controlling hepatic competency. Moreover, although GATA4 was identified within the ventral endoderm before and during hepatic specification, GATA4 protein was dramatically lost from the nascent hepatoblasts as they expanded into the septum transversum mesenchyme (see Fig. 2).71 These data support the proposal that GATA4 and GATA6 have overlapping functions in regulating competency of the endoderm and hepatic specification but that only GATA6 is available to facilitate early hepatoblast differentiation and expansion of the liver diverticulum (see Fig. 1). Further support for the idea that GATA6 may be the most important player in regulating differentiation of hepatoblasts was provided by recent studies in Xenopus. In these experiments, morpholino knockdown of GATA factors demonstrated that GATA6 but not GATA4 was required for expression of HNF1β in the developing frog liver.60 To definitively determine whether the GATA factors—or for that matter FoxA proteins—are essential for hepatic specification within the ventral endoderm will require the generation of embryos lacking multiple members of a family specifically in the endoderm using either the Cre/loxP system or via expression of dominant–negative alleles.

The list of transcription factors and signaling molecules that contribute to the onset of liver development has expanded significantly over the last decade.72, 73 Nevertheless, it is likely that many more factors will be involved in this complex process. Mammalian systems such as those of the mouse and rat have traditionally been favored for studying liver development because the hepatic physiology and anatomy resembles that of humans. However, identification of novel factors in these mammalian models is inefficient and expensive compared with other vertebrate model systems such as frogs (Xenopous laevis and Xenopous tropicalis), zebrafish (Danio rerio), and the Japanese medaka (Oryzias latipes), which have recently been subjected to high throughput screens.37, 74–78 These screens have already begun to provide a wealth of new information, and it is likely that many novel genes will soon be identified as making an essential contribution toward early formation of the liver.

What Governs the Establishment of Hepatic Architecture and Maturation of the Parenchyma?

Once the liver bud is generated and the hepatoblasts delaminate from the foregut and invade the septum transversum mesenchyme, the cells within the embryonic liver environment must organize to generate the complex hepatic architecture that is so crucial for normal liver function.79 This requires extensive differentiation of parenchymal and nonparenchymal cell types, organization of extracellular matrix, development of the biliary tract, maturation of sinusoidal capillaries and hepatic vasculature, and the conversion of hepatocytes to polarized epithelial cells. The molecular mechanisms underlying liver organogenesis are only now beginning to be deciphered. Our understanding of the molecular processes governing the generation of the liver as an organ, although still rudimentary, has been facilitated by the availability of mice harboring conditional alleles of genes whose expression can be disrupted specifically in liver cells. Advances in organ culture systems and the availability of cultured hepatic progenitor cells have also been invaluable.72, 80–84

During the mid-gestation stages of development, the hepatoblast population must expand to define the total volume of the liver. Studies in rat embryos have shown that between E13.5 and E20.5, the volume of the liver expands 84-fold, during which time the hepatoblasts undergo 8-doublings.85–87 Several mechanisms have been identified through gene knockout studies in mouse embryos that are required for growth of the liver. These include protection of the fetal liver from tumor necrosis factor α–mediated apoptotic cell death through the nuclear factor κB signaling pathway, regulation of cell proliferation by the transcription factors Xbp1 and Foxm1b, and signaling mechanisms that are mediated through the AP-1 transcription factor cJun.88–99 In addition, when expression of β-catenin, a key component of the Wnt signaling pathway, was blocked in fetal mouse livers cultured ex vivo, cell proliferation was reduced.100 Similarly, when Wnt signaling was inhibited in chick embryos, liver size was reduced, and when β-catenin was overexpressed in chick embryos, liver size increased, which was attributed to an expansion of the liver progenitor cell population.100, 101 Studies in chicks and mice have also shown that the mesenchymal component of the liver, which derives from the septum transversum mesenchyme, is essential for proliferation of hepatoblasts.6, 15, 102 Disruption in mouse embryos of the homeodomain transcription factor Hlx, whose expression is restricted to the hepatic mesenchyme, results in severe hepatic hypoplasia.103, 104 While the targets of Hlx action have yet to be defined, mouse embryos lacking either hepatocyte growth factor, which is expressed in the hepatic mesenchyme, or the hepatocyte growth factor receptor cMet, which is expressed in the hepatoblasts, also have severe liver hypoplasia.105–108 These findings imply that the mesenchyme is a critical source of mitogenic activity that could potentially be controlled by Hlx. A challenge for the future is to integrate signaling mechanisms that originate from the mesenchyme with transcription factors such as Foxm1b, Xbp1, and cJun that control hepatoblast proliferation.

The exact mechanisms through which bipotential hepatoblasts decide to become hepatocytes or biliary epithelial cells are still unclear, although several factors that contribute to this cell fate decision have recently been identified (as reviewed by Lemaigre and Zaret72). In mouse embryos, differentiation toward a biliary epithelial cell phenotype appears to be promoted by Notch signaling pathways and antagonized by hepatocyte growth factor, which may in turn promote hepatocyte differentiation.109 The requirement for Notch signaling in biliary development is evolutionarily conserved; in humans, haploinsufficiency of the Notch ligand Jagged1 results in Alagille syndrome, which is characterized by a reduction in intrahepatic bile ducts. Remarkably, given distinct differences in biliary anatomy between mammalian and teleost livers, Notch signaling is also required for biliary development in zebrafish.110–112 In addition to Notch, Wnt signaling may also be involved in regulating biliary epithelial cell fate; in ex vivo fetal liver cultures, the addition of Wnt3A supported biliary epithelial cell differentiation, while inhibition of β-catenin activity prevented hepatoblast expression of biliary epithelial cell markers.100, 113

The onset of biliary epithelial cell differentiation occurs at around E13.5 in the mouse when primitive biliary cells can be detected in proximity to the portal mesenchyme.114 Studies of mouse embryos lacking the transcription factor HNF6 showed that differentiation of biliary epithelial cells initiated prematurely, resulting in an increase in the allocation of biliary epithelial cells from the hepatoblast precursors.115 Biliary epithelial cells were also seen to abnormally extend as chords into the parenchyma of Hnf6−/− embryos instead of being restricted to the neighborhood of the portal mesenchyme. In addition, the gallbladder was absent from Hnf6−/− mice, and development of the extrahepatic bile ducts was disrupted. Analyses of mice in which HNF1β (a homeodomain transcription factor) was conditionally ablated in hepatoblasts revealed a similar phenotype.116 Moreover, HNF1β expression was reduced in the intrahepatic bile ducts of Hnf6−/− embryos and molecular analyses revealed that HNF6 could activate the Hnf1β promoter.115 These results suggest that HNF6 controls the timing of the onset of biliary epithelial cell differentiation and the positioning of these cells close to the portal mesenchyme, at least in part by regulating expression of Hnf1β. Direct support for the relationship between HNF6, HNF1β, and development of the biliary tract comes from studies using zebrafish, in which biliary phenotypes associated with knockdown of HNF6 can be rescued by forced expression of HNF1β.117

In contrast to the parenchyma, the portal mesenchyme is a rich source of extracellular matrix that consists of laminin, nidogen, collagens I and IV, and fibronectin, raising the possibility that the extracellular matrix could influence the fate of hepatoblasts.21, 118, 119 Careful immunohistochemical analyses at defined stages of mouse development have revealed that deposition of extracellular matrix components changes quite dramatically during hepatogenesis, which coincides with the onset of biliary epithelial cell differentiation.21 Consistent with the proposal that extracellular matrix is a significant determinant of hepatoblast cell fate, it has been demonstrated that the profile of gene expression within primary hepatic cultures is significantly affected by the composition of extracellular matrix.79, 80, 118, 120–123 For example, culture of bipotential mouse embryonic liver cells in matrigel, a soluble basement membrane preparation that includes laminin, collagen IV, heparan sulfate proteoglycans, and entactin, induced them to express biliary epithelial cell markers and form ductile like structures.83, 124 In addition, cells lacking β1-integrin, a subunit of a heterodimeric protein complex that contributes to cell–matrix interactions, are unable to colonize the liver, and inhibition of the α3-integrin subunit expression blocks extracellular matrix–induced differentiation of a hepatic cell line.125, 126 One fruitful direction for future research is to clearly define the mechanisms through which extracellular matrix signaling controls biliary epithelial cell gene expression.

A classic series of experiments from the Darnell laboratory demonstrated that hepatocyte gene expression is primarily regulated through transcription.127 Since then, several transcription factors have been identified that are essential for expression of the complete repertoire of proteins that define hepatocyte function.73, 81, 128, 129 These include the hepatocyte nuclear factors HNF1α and β, c/EBPα, HNF4α, and FoxA, which act in combination to control aspects of hepatocyte differentiation and liver function.67, 73, 130–138 Of these transcriptional regulators, the nuclear hormone receptor HNF4α appears to be particularly potent in controlling hepatocyte differentiation.133, 134, 139Hnf4 knockout embryos die before the onset of liver development due to a crucial role for this factor in regulating extra-embryonic endoderm function.140, 141 However, HNF4α was also found to be essential for hepatocyte differentiation when the extraembryonic endoderm deficiency was circumvented either by generating mice from Hnf4−/− embryonic stem cells using tetraploid embryo complementation or by ablation of Hnf4 specifically in hepatoblasts using the Cre/loxP system.133, 134, 142 Analyses of livers from E18.5 Hnf4loxP/loxPAlfpCre embryos revealed that hepatic morphology, including the distribution of sinusoids, was severely perturbed and hepatocytes lacking HNF4α failed to express a plethora of genes associated with hepatocyte activity.133 HNF4α is not, however, expressed in hepatic endothelial cells, suggesting that the requirement for this transcription factor in organization of the hepatic sinusoids is non–cell-autonomous.133 HNF4α-null hepatocytes were also unable to generate cell-to-cell contacts or form bile canaliculi and expression of several proteins that are integral components of cell junctions was undetectable, including E-cadherin. These findings suggest that HNF4α plays a fundamental role in transforming the fetal liver into an epithelial parenchyma during embryogenesis (Fig. 4).133 This proposal was supported by the observation that forced expression of HNF4α in either NIH3T3 fibroblasts or in F9 embryonal carcinoma cells induces a mesenchymal to epithelial transformation that includes expression of cell junction proteins.133, 143 HNF4α's role in hepatic epithelial formation is also likely to have biomedical significance. Expression of HNF4α is lost in a mouse model of hepatocellular carcinoma progression in which slow-growing tumors lose their epithelial characteristics and transform to a fast-growing invasive form.144 Moreover, when expression of HNF4α is forced in the fast-growing invasive hepatocellular carcinoma cells, proliferation is reduced and tumor formation is suppressed.144

Figure 4.

HNF4α regulates hepatocyte differentiation, controls conversion of the hepatic parenchyma to an epithelium and is essential for maintaining sinusoidal (red area) architecture. During midgestation stages (≈E14.5) of mouse embryonic development, HNF4α is required for expression of cell adhesion and junction proteins. Expression of adhesion molecules (blue area) within the developing hepatoblasts (green area) facilitates the formation of bile canaliculi and converts the immature cells into a polarized hepatocyte.

In addition to controlling hepatocyte differentiation during embryogenesis, HNF4α is also required to maintain a differentiated hepatocyte phenotype, since disruption of Hnf4 in adult livers also dramatically affects hepatic gene expression and severely disrupts liver function.135 Although the data still need to be validated by genetic analyses, when chromatin immunoprecipitation was used to probe human promoter microarrays for HNF4α DNA elements within HepG2 cells, HNF4α was shown to occupy a surprising 12% (1,575 genes) of genes represented on the Hu13K microarray. Prior to identifying HNF4α-bound elements, no transcription factor had been seen to interact with more than 2.5% of promoters using the same approach. Moreover, the occupancy of promoters by HNF4α is significantly higher than that found for either HNF1α (1.6%; 222 genes) or HNF6 (1.7%; 227 genes).137 Combined with available genetic, biochemical, and gene array data, these results suggest that HNF4α is a central determinant of hepatocyte function and differentiation and is crucial for establishing normal liver architecture during development.139, 145, 146

In mammals, maturation of the liver is further complicated by the fact that the fetal hepatic environment is also the site of hematopoiesis between E11.5 and E16.5. In this regard, there appears to be a close relationship between hematopoietic cells and the developing hepatic parenchyma, which is governed by cytokine signaling.147, 148 Studies using a hepatoblast culture system have shown that oncostatin M, which is secreted from hematopoietic cells within the fetal liver, contributes to control late stages of hepatocyte differentiation, possibly by increasing HNF4α expression.149–151 Furthermore, mice lacking the gp130 subunit of the oncostatin M receptor fail to accumulate normal levels of glycogen, which is consistent with a requirement for oncostatin M in completion of hepatocyte maturation.152 Not only do hematopoietic cells contribute to development of the liver parenchyma but, conversely, differentiation of the hematopoietic cells is also influenced by the hepatocytes. Cultures of fetal liver cells were found to be capable of supporting growth and expansion of hematopoietic cells. However, if the fetal cultures were induced to differentiate, or if differentiated liver cells were isolated from E18.5 embryos, they no longer supported hematopoietic cell growth.149 This finding implies that differentiation of the parenchyma contributes to the attenuation of the fetal liver as a hematopoietic organ.

Summary and Future Directions

The last two decades of research, driven by remarkable technical advances, have been focused on identifying the players that control development of the liver. While this is still ongoing, it has led to a reasonably detailed picture of how a combination of signaling molecules, transcription factors, and cell movements drives the formation of the liver rudiment, as well as differentiation of the different hepatic cell types. How the differentiated cells ultimately generate the complex architecture that defines hepatic functions may be less well understood. However, this is likely to improve in the near future with the availability of mouse strains in which factors can be disrupted in a cell-type specific manner and ongoing genetic screens in additional model organisms. The information that has been generated by the study of hepatic development is also now being used in the production of hepatocytes and other liver cells from progenitor cell populations and embryonic stem cells.124, 153–155 For example, Teratani and colleagues demonstrated that embryonic stem cells could be induced to generate cells with hepatic gene expression profiles and functions using a differentiation procedure that included culture with FGF, hepatocyte growth factor, and oncostatin M.156 The ability to efficiently produce hepatocytes from progenitor or embryonic stem cells is exciting, because such cells could potentially be used for hepatocyte transplantation, gene therapy, the generation of artificial liver devices, or even tissue engineering.157–159 By building on our understanding of the basic mechanisms that lead to development of the liver, it is highly likely that such therapies could soon become a reality.