The liver has an unusual capacity to regenerate after a loss of mass and function caused by surgical resection or toxic liver injury. Over the last 10 years there have been major advances in our understanding of the molecular and cellular mechanisms underlying liver development and regeneration. The numerous factors crucial to these phenomena have been identified mainly by using knockout mice. Forward-genetics studies using zebrafish and medaka have also generated many mutants with liver disorders or defects in liver formation. Our goal is to translate knowledge gained from laboratory work and animal models into novel therapies for human liver diseases. Exciting progress has been achieved using human partial liver transplantation and autologous cell therapy.
The liver plays a central role in metabolic homeostasis because this organ is responsible for the metabolism, synthesis, storage and redistribution of nutrients, carbohydrates, fats and vitamins. The liver is also the main detoxifying organ of the body, removing waste and xenobiotics by metabolic conversion and biliary excretion. Most of these functions are carried out by the hepatocyte, a parenchymal cell type that comprises about 80% of hepatic cells. The other 20% is made up of non-parenchymal cells, including Kupffer cells, stellate cells, endothelial cells and lymphocytes. Kupffer cells also reside in intrahepatic circulatory vessels (sinusoids) and are essential for the phagocytosis of foreign particles and infecting organisms as well as the production of cytokines. Stellate cells have various functions, including the storage of vitamin A and the production of the extracellular matrix. These cells also synthesize most of the factors that lead to hepatic fibrosis (Taub 2004). Endothelial cells line the liver sinusoids and provide a large surface area for nutrient absorption. Lymphocytes resident within the liver mediate adaptive immune responses that combat infection. All of these cell types are activated in response to hepatic injury and all play a role in liver regeneration.
Embryonic liver development occurs in multiple stages that are governed by hormonal factors as well as by intercellular and matrix–cellular interactions. In mice, liver ontogeny initiates around embryonic day 9 (E9) when epithelial cells of the foregut endoderm interact with the cardiogenic mesoderm and commit to becoming the liver primordium (Fig. 1A). The liver primordium proliferates and invades the mesenchyme of the septum transversum to give rise to the hepatic codes and bud at E9.5. At E10.5–12.5, the fetal liver takes over hematopoiesis from the yolk sac and aorta-gonad-mesonephros region. This switch in hematopoietic organ represents a critical checkpoint in murine embryogenesis. The next major stage occurs around E14.5 when both hepatocytes and bile-duct epithelial cells develop from a common founder cell, the bipotential hepatoblast (LeDouarin 1975; Houssaint 1980). These cells become the postnatal liver and the organ's main function switches from hematopoiesis to metabolism. This capacity also dominates in the adult liver.
Each stage of hepatic maturation has been characterized by its expression pattern of liver- and stage-specific genes (Fig. 1B). For example, alpha-fetoprotein (AFP) is an early fetal hepatic marker whose expression commences at E9 but decreases as liver development progresses (Shiojiri et al. 1991). In contrast, the expression of albumin, the most abundant protein synthesized by hepatocytes, starts in early fetal hepatocytes (E12) and reaches its maximal level in adult hepatocytes (Tilghman & Belayew 1982). Using ventral foregut endoderm isolated from E8.25 embryos, it has been shown that close proximity of cardiac mesoderm causes the foregut endoderm to develop into the liver (Jung et al. 1999; Zaret 2000). This induction depends on the expression of fibroblast growth factors (FGFs) 1, 2 and 8 by the cardiac mesoderm. At late stages of hepatogenesis (E14.5), blood cells in the fetal liver use a paracrine mechanism to produce an interleukin (IL)-6 family cytokine called oncostatin M (OSM) that promotes hepatocyte development (Kamiya et al. 1999). We are only beginning to define the complex network of transcription factors, cytokines and growth factors that drive the development and differentiation of liver cells.
Tumor necrosis factor-α (TNFα) is an inflammatory cytokine that elicits a wide range of biological responses, including inflammation, tumor necrosis, and cellular differentiation, proliferation and apoptosis. In the liver, TNFα engagement of TNF receptor 1 (TNFR1) activates three separate signaling pathways: the extrinsic pathway of apoptosis induction; NF-κB activation; and the activation of stress-activated protein kinase SAPK (also called c-Jun N-terminal kinase, JNK) (Fig. 2). Embryos of knockout mice lacking the function of genes involved in NF-κB signaling exhibit massive liver degeneration and apoptosis during mid-gestation at E12.5–16. Examples include mice deficient in the RelA subunit of transcription factor NF-κB (die at E15–16; Beg et al. 1995), IκB kinase β/IKK2 (die at E12.5–14; Li et al. 1999; Li et al. 1999; Tanaka et al. 1999), NEMO/IKKγ (die at E12.5–13.0; Rudolph et al. 2000), or TRAF2-associated kinase (T2K; die at E12.5–14.5; Bonnard et al. 2000). Importantly, the embryonic lethality and liver apoptosis observed in RelA-, IKK2- or T2K-deficient mice can be rescued by the simultaneous inactivation of TNFR1, suggesting that the apoptosis is induced by TNFα present in the embryonic circulation. Thus, the NF-κB pathway triggers essential antiapoptotic signaling in hepatoblasts that allows complete liver development.
The stress-activated protein kinase/extracellular signal-regulated kinase kinases, SEK1/MKK4 and SEK2/MKK7, are direct activators of SAPK/JNK. Both SEK1 and MKK7 are activated in response to a variety of cellular stresses, such as metabolic poisons, DNA damage, changes in osmolarity, heat shock, or the presence of the inflammatory cytokines IL-1 and TNFα. SEK1 and/or MKK7-mediated activation of SAPK/JNK phosphorylates c-Jun and activates c-Jun/Fos heterodimeric AP-1 transcriptional complexes. We and several other groups have used gene targeting to disrupt the sek1 gene in mice and have found that SEK1-deficient embryos display severe anemia and die at E10.5–12.5 (Nishina et al. 1997a, 1997b, 1999; Yang et al. 1997; Ganiatsas et al. 1998). Interestingly, although vasculogenesis and hematopoiesis from yolk sac precursors are normal in SEK1-deficient embryos, hepatogenesis and liver formation are severely impaired such that the number of hepatocytes is greatly reduced in E11.5–12.5 SEK1-deficient embryos. Formation of the primordial liver and hepatic bud appeared to be normal in SEK1-deficient embryos but the hepatocytes underwent massive apoptosis at E12.5. Embryos with a disrupted c-jun gene also display defective liver organization and die at E11.5–15.5 (Hilberg et al. 1993; Johnson et al. 1993), and MKK7-deficient mice at E11.5–12.5 show impaired liver formation (Wada et al. 2004). These results indicate that SEK1, MKK7 and c-Jun deliver a crucial and specific survival signal for fetal hepatogenesis. In an effort to understand the mechanisms underlying the defective liver formation in these mutants, we prepared monoclonal antibodies (mAbs) specific for murine fetal livers and characterized their binding to paraffin sections of mouse embryos at various stages (Watanabe et al. 2002). For example, the anti-Liv2 mAb bound specifically to hepatoblasts at E9.5–12.5. When we used bromodeoxyuridine (BrdU) incorporation to assess hepatoblast proliferation in sek1–/– embryos at E10.5 (before the onset of liver formation defect and massive apoptosis), we found that the mutant hepatoblasts displayed impaired growth. Similar BrdU incorporation defects were revealed in the hepatoblasts of mkk7–/– embryos. Thus, SEK1 and MKK7-mediated activation of SAPK/JNK and c-Jun are crucial for the hepatoblast proliferation needed to build the liver.
Over the last decade, studies in rats and mice have greatly expanded the list of molecules known to contribute to liver development. However, it is likely that many more factors are involved in this complex process. Identification of novel factors using mammalian model organisms is inefficient and expensive compared with studies of other vertebrate model organisms such as frog (Xenopus laevis), zebrafish (Danio rerio) and medaka (Oryzia latipes). High-throughput screens based on these models are already yielding a wealth of new information on liver gene functions (Zhao & Duncan 2005). Our group has used systematic mutagenesis in medaka to generate numerous mutations affecting various aspects of liver development and function (Watanabe et al. 2004). The nature of each mutation was established by morphological and metabolite marker screens. The 22 recessive mutations identified were assigned to five phenotypic groups (Fig. 3). Phenotypic group 1 contains mutations in six genes that affect the formation of endoderm, endodermal rods and the hepatic bud from which the liver develops. As a result of these mutations, the development of hepatoblasts is impaired. Phenotypic group 2 includes mutations that affect hepatoblast proliferation and result in defective liver morphogenesis. At present, group 2 contains mutations in four genes thought to be involved in the regulation of growth or patterning of the gut endoderm. Phenotypic group 3 comprises mutations in three genes that affect the laterality of the liver. In kendama (ken) mutants of this group, the laterality of the heart and liver is uncoupled and randomized. Phenotypic group 4 includes mutations in three genes altering bile color, indicative of defects in hemoglobin-bilirubin metabolism and globin synthesis. Phenotypic group 5 consists of mutations in three genes that result in decreased accumulation in the gallbladder of a fluorescent metabolite of the phospholipase A2 substrate, PED6. Lipid metabolism or the transport of lipid metabolites may be affected by these mutations. We believe that mutations in groups 4 and 5 may provide useful animal models for the corresponding human liver diseases. Studies of these models could aid hepatologists and surgeons in learning more about the mechanisms of liver formation, function and disease.
The concept of liver regeneration dates back to ancient Greek mythology and has become a widely studied experimental model for understanding mechanisms that regulate hepatocyte proliferation and survival. Mammals maintain a constant liver-to-body mass ratio. For example, livers from smaller animals that are transplanted into larger animals increase to the appropriate size; the converse is also seen (Francavilla et al. 1994). In a normal adult murine liver, mature differentiated hepatocytes are quiescent (in the G0 stage of the cell cycle) and exhibit minimal turnover (Fausto & Webber 1994; Overturf et al. 1997). However, upon receiving a regenerative stimulus (such as occurs during toxic injury or resection), 95% of hepatocytes undergo cell division while maintaining their metabolic function. The process of regeneration is tightly regulated through controlled delivery of ‘start and stop’ signals, including cytokines and growth factors that control the proliferation of hepatic cells (Mohammed & Khokha 2005).
Several toxic agents that injure the liver have been well-studied, in particular carbon tetrachloride (CCl4) and galactosamine. Acute CCl4 injury is the model most widely used in rodents to mimic acute human toxic liver injury and regeneration from hepatocytes. However, the cell-cycle response of rodent hepatocytes is less synchronized in CCl4–treated livers (Neubauer et al. 1998; Koniaris et al. 2003). Galactosamine also inflicts liver damage, but in this case the replicative capacity of most normal hepatocytes is diminished and liver progenitor cells known as oval cells must proliferate to replace the hepatic parenchyma (Dabeva & Shafritz 1993). The derivation of oval cells with the potential to differentiate into hepatocytes and biliary epithelial cells is still controversial.
As well as in response to injury, the liver is capable of amazing regeneration after resection. The most commonly used model is the 70% partial hepatectomy (PH) model in rodents, which was originally described by Higgins in 1931 (Higgins & Anderson 1931) and remains essentially unchanged today. However, the enlargement of the hepatic lobes that remain after resection is not true regeneration, unlike the restoration of severed limbs in amphibian models. Rather, a resected rat or mouse liver undergoes compensatory hyperplasia in which the initial liver mass (and its functions) is restored within approximately 1 week after surgery (Fig. 4). This hyperplasia associated with standard PH is due to the synchronized entry of approximately 95% of hepatocytes into the S phase of the cell cycle, followed by mitosis. In rat and mouse hepatocytes, a first peak of DNA synthesis occurs at around 24 h and 40 h post-PH, respectively. However, DNA replication in non-parenchymal cells is delayed. Subsequent levels of DNA synthesis in hepatocytes are lower, as complete restoration of the initial liver mass requires an average of ∼1.6 cycles of replication per cell for all cells (Taub 2004). Within minutes of liver resection, hepatocytes undergo a coordinated cellular activation termed the ‘acute phase response after PH’. This highly regulated process is mediated by multiple growth factors and cytokines that coordinate to transduce the response signal into kinase and transcription factor activation. As a result of the acute response, a surviving hepatocyte initiates the transcription of more than 100 early genes and accumulates triacylglycerol and cholesterol esters in intracellular lipid droplets. These droplets supply the energy and building materials needed to support rapid cell division and tissue regrowth and so are essential for normal liver regeneration (Farrell 2004; Shleyer et al. 2004).
Cytokines are also important for the initiation of liver regeneration. These intercellular messengers, particularly the inflammatory cytokines TNFα and IL-6 produced by activated Kupffer cells, prime hepatocytes to enter the cell cycle and respond to the mitogenic effects of growth factors (Fig. 5). Indeed, experiments with knockout mice suggest that TNFR1 and IL-6 may be essential for complete liver regeneration. After PH of normal mice, levels of TNFα and IL-6 increase very rapidly. Both these cytokines are components of the TNFα→TNFR1→NF-κB→IL-6→STAT3 pathway that drives hepatocyte gene expression. Each component of this pathway appears sequentially in the regenerating rodent liver during the first 12 h after PH (Fausto & Riehle 2005). However, it is not known how PH triggers the initial increase in TNFα levels. Interestingly, PH-induced production of TNFα and IL-6 is impaired in mice lacking MyD88, a key intracellular signal transducer in the pathway triggered by bacterial endotoxin. Mice lacking the complement components C3a and C5a also show impaired production of TNFα and IL-6 after PH. It is tempting to speculate that PH stimulates the binding of a presently unknown ligand to MyD88 and thus triggers TNFα and IL-6 production, but this issue remains to be clarified (Fausto et al. 2006).
While cytokines are responsible for the entry of quiescent (G0) hepatocytes into the G1 phase of the cell cycle, growth factors drive cell cycle progression from the G1 phase through to the S phase. Hepatocyte growth factor (HGF), transforming growth factor-α (TGFα) and heparin-binding epidermal growth factor-like growth factor (HB-EGF) are growth factors of major importance for liver regeneration. HGF is produced by non-parenchymal cells in the liver and other tissues and may act on hepatocytes by a paracrine or endocrine mechanism. HGF binding to the c-met receptor leads to the activation of the MAP kinases ERK1 and 2. TGFα and HB-EGF are members of the epidermal growth factor (EGF) family of ligands, and both bind to EGF receptors that are known to activate a phosphorylation cascade leading to DNA replication.
Other molecules crucial for liver regeneration are extracellular proteases such as MMPs and ADAMs. A recent study has shown that these proteases act on cytokines and growth factors that require processing for their bioactivity or which are anchored to the hepatic extracellular matrix (Mohammed & Khokha 2005). Another recent report has demonstrated the importance of caveolae to liver regeneration. Caveolae are distinct domains within the plasma membranes of most cells and are thought to regulate intracellular signaling and membrane sorting. Significantly, mice lacking caveolin-1, which is an essential component of caveolae, exhibit impaired liver regeneration and low survival after PH. Hepatocytes of these animals show a dramatic reduction in lipid droplet accumulation and do not advance through the cell cycle (Fernandez et al. 2006).
Liver cirrhosis (LC) is the end stage of chronic liver disease in humans (Fig. 6A) and is very difficult to treat. Currently, liver transplantation is the only truly effective therapy available to these patients. However, liver transplantation has serious limitations: a lack of donors, the risk of surgical complications, possible organ rejection and high cost. Therapies designed to encourage the patient's own liver to regenerate may offer alternatives that are minimally invasive and have few side-effects. Stem cells in bone marrow can be induced in vitro to differentiate into mature hepatocytes or intestinal cells. Bone marrow cell (BMC) transplantation is commonly performed to treat hematological diseases, and several clinical studies have used BMC injection to induce regeneration of myocardium and blood vessels in patients with limb ischemia (Tateishi-Yuyama et al. 2002). Furthermore, it has been shown that reaggregates of BMCs can form cartilage as do limb bud cells using mice (Masaki & Ide 2007). Taken together, these findings suggest that BMC might be an effective source for regenerative liver therapy (Fig. 6B). To confirm the potential of BMC therapy, Terai et al. developed an in vivo mouse model (GFP/CCl4) in which BMC differentiation into hepatocytes was monitored (Terai et al. 2003). Chronic liver injury in mice was induced by continuous administration of carbon tetrachloride (CCl4) and GFP-positive BMC were transplanted into the damaged liver to repopulate it. The GFP-BMC successfully differentiated into hepatoblast intermediates and then into albumin-producing hepatocytes. The BMC-infused mice showed elevated serum albumin levels, reduced liver fibrosis and an improved survival rate. In humans, am Esch et al. reported that portal administration of autologous CD133+ BMC accelerated liver regeneration (am Esch et al. 2005). These results suggest that BMC infusion is a potentially effective treatment for patients in liver failure. More recently, Terai et al. reported on a clinical trial in which some of the nine LC patients treated with autologous BMC infusion therapy showed improved liver function (Terai et al. 2006).
Studies of rodents and lower organisms such as medaka have revealed much new knowledge on the molecular and cellar mechanisms of liver development and regeneration. This knowledge should prove useful in the future in resolving liver problems in human patients. Inflammatory cytokines such as TNFα and IL-6, and growth factors such as HGF, TGFα and HB-EGF, may be useful for clinical therapy. Current efforts to apply BMC transplantation to liver regeneration are already bearing fruit. Additional work by the many researchers in this field will likely soon generate effective new therapeutic modes and drugs to treat liver disease.
We thank Drs Toshiaki Katada, Takashi Nakamura and Mary Saunders for their critical reading of the manuscript.