Liver regeneration and repair: Hepatocytes, progenitor cells, and stem cells

Many different terms are used to refer to cells originating in the terminal branches of he bile ductular system and the canals of Hering in rodents and humans, which can function as progenitors for hepatocytes and cholangiocytes, the mature forms of the two hepatic epithelial cell lineages. In rats and mice, in experimental situations in which large amounts of these cells proliferate, the term oval cells is used for single cells or for clusters of cells that form a ductule. Intermediate forms between ductular cells and hepatocytes are often referred to as ductular hepatocytes, and the term neocholangiole has been used to describe the structures that contain these cells. Small hepatocytes that are not completely differentiated, which probably originate from oval cells, often are referred to as small hepatocyte precursor cells. Cell lines obtained from normal rodent liver, which have progenitor cell capabilities, are known as liver epithelial cell lines. Editor's note: The reader is referred to page 1739: “Nomenclature of the finer branches of the biliary tree.” For human liver, the term hepatic progenitor cells frequently is used to refer to cells that are equivalent to oval cells in rodents. The ductular reaction that involves these cells is known as an atypical ductular reaction (to distinguish from typical ductular reactions that do not involve the generation of hepatic progenitor cells). The term ductular hepatocyte also is used commonly to describe intermediate forms, particularly for cells appearing in the repopulation process after massive hepatic necrosis. In rodents and humans, no special name has been given to the cells located in the terminal branches of the ductular system and in the canals of Hering. Because they are likely to be the cells that most commonly give rise to oval cells, they are often referred to as intrahepatic stem cells or ductular stem cells. In this article, the term oval cell is used interchangeably with hepatic progenitor cell. The terms atypical ductular reaction and ductular hepatocytes are used to describe, respectively, ductular reactions involving oval cells and intermediate forms between ductular cells and hepatocytes. FAH, fumarylacetoacetate hydrolase; AFP, α-fetoprotein; HCC, hepatocellular carcinoma; HSC, hematopoietic stem cell; MAPC, multipotent adult progenitor cell. This question has multiple answers, because the source of hepatocytes depends on the nature of the growth process. In every case, it is necessary to ascertain whether hepatocytes responsible for liver regeneration originated from the replication of existing hepatocytes, were generated by differentiation of oval cells, or were produced from bone marrow cells.2 Replication of mature hepatocytes in liver regeneration has been documented extensively.3-5 Differentiation of oval cells has also been established as a mechanism that can generate significant numbers of hepatocytes (Table 1).6-11 However, the contribution of bone marrow cells to the generation of hepatocytes in liver repopulation and regeneration remains uncertain, both regarding its extent and the mechanisms involved. As a general rule, replication of existing hepatocytes is the quickest and most efficient way to generate hepatocytes for liver regeneration and repair. Oval cells usually replicate and differentiate into hepatocytes only when the replication of mature hepatocytes is delayed or entirely blocked (Fig. 1). Bone marrow cells can generate hepatocytes in transplanted livers but so far, the frequency of hepatocytes produced by this route is very low, and such cells are not always detectable. Note however, that in transplanted livers, bone marrow cells are an important source of nonparenchymal cells such as Kupffer cells and endothelial cells (discussed later). Because liver regeneration after 70% hepatectomy requires no more than two rounds of hepatocyte replication, it was generally assumed that the proliferative capacity of mature hepatocytes is very limited. This view has now been drastically changed. First, experiments with cultured hepatocytes isolated from transgenic mice that expressed liver growth factors demonstrated that long-term hepatocyte replication is compatible with a differentiated phenotype.12 More striking were the results of hepatocyte transplantation experiments in urokinase-plasminogen activator transgenic mice, showing that liver repopulation by transplanted hepatocytes involved at least 12 rounds of replication.13 Subsequent serial transplantation experiments performed in fumarylacetoacetate hydrolase (FAH)-deficient mice (FAH knockout mice) demonstrated that hepatocytes could replicate 70 or more times.14 In the serial transplantation experiments, there was no evidence that the repopulation capacity was dependent on stem cells. This conclusion also is supported by data demonstrating that diploid, tetraploid, and octoploid hepatocytes have roughly the same capacity to repopulate damaged livers.15 Thus, although hepatocytes are quiescent in normal livers and replicate in a limited and regulated manner during liver regeneration after partial hepatectomy, these cells have an enormous proliferative potential that can be unleashed under certain conditions. Nevertheless, there is evidence that the replicative activity of hepatocytes diminishes in advanced cirrhosis in humans and in chronic liver injury in mice, reaching a state of “replicative senescence,” perhaps as a consequence of telomere shortening.16-19 Both hepatocytes and intrahepatic bile ducts originate from endodermal-derived hepatoblasts (Fig. 1) that express albumin and α-fetoprotein (AFP).20, 21 At day 14 (mice) or 15 (rats) of embryonic development, hepatoblasts located near vascular spaces, the site for portal spaces in later development, express dual markers of the hepatocyte (albumin and AFP) and biliary (cytokeratins 7 and 19) lineages.22 These hepatoblasts give rise to the primitive intrahepatic bile ducts, structures that connect parenchymal hepatocytes with the larger segments of the biliary system. Primitive intrahepatic bile ducts correspond to the canals of Hering and terminal bile ductules of adult livers which may constitute the niche for intrahepatic stem cells.23-25 The embryological origin of intrahepatic bile ducts explains some important features of oval cell proliferation in adult livers. Cell lineages in the liver. During embryonic development, hepatoblasts give rise to the two epithelial lineages of the liver, producing hepatocytes and cholangiocytes. Oval cells originate in association with the intrahepatic biliary system, formed by hepatoblasts located near portal spaces. In adult livers, hepatocytes and cholangiocytes can replicate. Oval cells form a bipotential reserve compartment capable of generating hepatocytes whenever hepatocyte replication is blocked (red lines). In adult rat liver, cells of the canals of Hering and terminal bile ductules may express AFP.26, 27 Oval cells thought to be generated from them may express AFP and may contain isozymes of aldolase, pyruvate kinase, and lactic dehydrogenase present in both adult and fetal liver cells, and glucose-6-phosphatase, a typical hepatocyte marker.28-34 However, the extent to which these markers are expressed in a population of proliferating oval cells depends on the agent that elicited oval cell proliferation. Analysis of marker expression suggests that populations of proliferating oval cells constitute a heterogeneous cell compartment (or oval cell compartment) containing cells that may differ in their differentiation capacity and stage of differentiation. Some of these cells may function as hepatocyte progenitors, whereas others may be indistinguishable from cholangiocytes, cells that do not express AFP, or hepatocyte markers. Oval cells and cholangiocytes share epitopes that react with, among others cytokeratins 7, 8, 18, and 19, the antibodies OV6 (an anti-cytokeratin 19 antibody), OC2 (anti-myeloperoxidase), and some other members of the OC series, γ glutamyl transferase, and the antigens A6 and G7.29, 30, 35-37 A very small number of hematopoietic stem cells present in fetal livers may remain in adult livers. These cells may be distinct from oval cells, but are induced to proliferate by the same conditions that cause oval cell proliferation. In this case, hematopoietic stem cells would be a component of the oval cell compartment but constitute a distinct population, which does not acquire markers of the hepatocyte lineage but shares general stem cell markers with oval cells originating from the canals of Hering. Hematopoietic cells located in the adult liver may be pluripotent stem cells present in the hepatic tissue, functioning as the equivalent of embryonic stem cells capable of generating multiple lineages, including hepatocytes. If this view is correct, hematopoietic stem cells located in the liver (perhaps in periductular spaces) would differentiate progressively,9 first into oval cells and ultimately into hepatocytes, under stimuli known to cause an oval cell response. The relationships between oval cells and bone marrow cells would be understood more easily if it could be demonstrated that oval cells can be derived from hematopoietic cells. This has been shown to occur in some experimental models,44 but in these experiments, the proportion of hepatocytes generated through this route was very small, representing approximately 0.15% of hepatocytes in the liver. More recent data using the FAH model of liver injury demonstrated that in this model, oval cells do not originate from bone marrow precursors but are generated intrahepatically.45 Another study, using three different models of rat liver injury, showed that bone marrow cells were not the source of oval cells that repopulated these livers.46 An extensive ductular reaction occurs after massive (or submassive) hepatic necrosis in humans.47-53 In this type of injury, ductular proliferation involves mature cholangiocytes and ductular hepatocytes. The latter, located at the periphery of portal tracts, proliferate and express cholangiocyte and hepatocyte markers. Ductular hepatocytes are considered to be an intermediate form between ductular cells and hepatocytes, resembling ductal plate cells in the developing human liver. Such cells are also present in massive hepatic necrosis in rats.54 It is not known with certainty whether the generation of hepatocytes from ductular hepatocytes leads to complete repopulation of injured human livers, but at least one well-documented case has been described in a patient who recovered from massive liver necrosis. Fujita et al.52 performed sequential biopsies on the natural liver of a patient with massive necrosis after receiving an auxiliary partial orthotopic liver transplant. Complete regeneration of the natural liver was observed 12 to 14 months after transplantation, through a process that involved an initial ductular reaction followed by hepatocyte differentiation from ductular hepatocytes. As oval cells proliferate in response to treatment with both carcinogenic and noncarcinogenic agents, the detection of oval cells in a carcinogenic process is not proof of the role of oval cells as cancer progenitors. Nevertheless, it has been demonstrated that oval cells can generate hepatocellular carcinoma (HCC), cholangiocarcinoma, and hepatoblastoma in rodents.55-57 Although oval cell proliferation is fairly common in experimental models of hepatocarcinogenesis, in some models, oval cell and ductular proliferation is not apparent (for instance, in carcinogenesis induced by overexpression of growth factors such as transforming growth factor α58). Indeed, there is no special reason to support the notion that HCC is generated exclusively from oval cells.59 Mature hepatocytes also can function as tumor precursors, as long as they are in a proliferative state. It is also important to note that in general, oval cells do not seem to produce tumors directly, but do so through the generation of hepatocytes, although these cells “merge” into hepatocytes of neoplastic nodules in the liver of rats fed the carcinogen 3′-methyl-4-dimethylaminoazobenzene.60 Whether hepatocytes generated from oval cells in adult liver are abnormal, immature, or have a high risk for transformation remains to be established. In retrorsine-induced hepatocellular injury combined with partial hepatectomy, the great majority of proliferating cells are incompletely differentiated hepatocytes called small hepatocyte precursor cells.61, 62 Such cells, which also have been detected in galactosamine-induced liver injury,63 presumably originate from oval cells. Their preferential accumulation in some types of injury probably reflects a variable transit flux between cellular compartments containing ductular stem cells, oval cells, small hepatocyte precursor cells, and mature hepatocytes.64 Oval cells, commonly referred to as hepatic progenitor cells, have been detected in human livers in small cell dysplastic foci, hepatocellular adenomas, chronic viral and alcoholic hepatitis, nonalcoholic fatty liver disease, hemochromatosis, primary biliary cirrhosis, and cirrhosis associated with primary sclerosing cholangitis, conditions associated with an increased risk of neoplastic development.18, 65-70 Oval cell markers such as AFP and cytokeratins 7 and 19 are expressed in approximately 50% of small cell dysplastic foci and in HCC, suggesting the possible origin of HCC from cells that express these markers.65 This conclusion is strengthened by the finding that such markers are not detected in foci of large cell dysplasia, lesions that are not considered to be tumor precursors. Hepatocyte generation from oval cells occurs at late stages of cirrhosis, apparently at a time at which hepatocyte replication has diminished.17 However, this process does not lead to extensive parenchymal regeneration and is essentially ineffective in restoring the normal parenchyma. Based on the experimental data discussed above, it may be suggested that generation of hepatocytes from oval cells in severely damaged cirrhotic livers produces hepatocytes that have a high risk for transformation. Answers to this question may come from studies of populations of cells isolated from small cell dysplastic foci and HCCs. In both animals and humans, chromosomal abnormalities present in tumors have been found in these foci.71 It would be important to know whether the cells that harbor such chromosomal defects in dysplastic foci express oval cell phenotypes. In both human and rodent livers, oval cells proliferate and differentiate in close proximity to stellate cells with myofibroblast morphological features.48, 72, 73 In rats, oval cells form ductules, which are extensions of the canals of Hering and are surrounded by a continuous basement membrane. Stellate cells penetrate through this basement membrane and establish direct contact with oval cells in the ductules.72 Oval cell proliferation is associated with increased expression of c-KIT, and also of hepatocyte growth factor, acidic fibroblast growth factor, and transforming growth factor α, which also function as growth factors for hepatocyte replication.74 In both human and rodent liver, expression of transcription factors of the hepatocyte nuclear factor family is detectable shortly after the start of oval cell proliferation, indicating an early commitment to hepatocyte differentiation.53, 75, 76 It is puzzling that growth factors that stimulate oval cell proliferation are similar to those that stimulate hepatocyte replication after partial hepatectomy and that both cell types require signaling through tumor necrosis factor receptor type I.77, 78 Yet, as discussed, oval cells and hepatocytes rarely proliferate simultaneously; oval cell replication generally occurs when hepatocyte proliferation is blocked. However, the interferon γ network is stimulated only in oval cell, but not in hepatocyte, proliferation.79 Studies in progress suggest that interactions between interferon γ and cytokines such as tumor necrosis factor may inhibit hepatocyte proliferation while they stimulate oval cell replication (Brooling J et al., unpublished manuscript, 2001). Other interesting features of oval cell phenotypes are the expression of proteins of the drug resistance gene families, of adenosine triphosphate binding cassette transporter genes, and of neuroendocrine peptides.80-84 The tremendous interest generated by stem cells during the last few years to a great extent is the result of to two newly discovered properties of these cells.85 Studies of hematopoietic stem cells (HSC) and bone marrow mesenchymal stem cells, revealed that they are capable of generating many different types of tissue cells (a property known as transdifferentiation) and can choose multiple differentiation pathways (a property called differentiation plasticity). Understanding the mechanisms of transdifferentiation is key to the field of stem cell biology and should provide important clues for the use of stem cells in organ repopulation and regeneration. So far, the most impressive demonstrations of hepatocyte generation from bone marrow cells are the production of hepatocytes in cultures of multipotent adult progenitor cells (MAPCs)86, 87 and the repopulation of livers of FAH knockout mice by transplanted HSCs.88, 89 Yet, neither of these conditions can be considered as examples of transdifferentiation. In culture, MAPCs can differentiate into cells of mesodermal, ectodermal, and endodermal lineages. Injected into a blastocyst, a single MAPC contributes to the formation of all somatic tissues. Thus, MAPC can be considered as equivalent to embryonic stem cells, which have persisted in adult tissues.86, 87 Human, mouse, and rat MAPCs, grown in matrigel in the presence of hepatocyte growth factor and fibroblast growth factor-4, differentiated into mature hepatocytes with apparently fully functional properties.90 If MAPCs are indeed adult embryonic stem cells, hepatocyte generation from these cells constitutes a process of differentiation of pluripotent, uncommitted cells, along a specific differentiation path. This is similar to the differentiation process of embryonic stem cells during development and is quite different from transdifferentiation, which implies a change in differentiation commitment of an already committed cell. These exciting results await confirmation from other laboratories. Much needs to be known about the properties of MAPCs, and most importantly, whether they can generate hepatocytes in vivo. The other dramatic demonstration of hepatocyte generation from bone marrow cells is the extensive repopulation of damaged livers of FAH knockout mice transplanted with HSC.88, 89 So far, this is the only example in animals or humans of extensive repopulation of damaged livers by cells derived from bone marrow. In this system, the kinetics of repopulation by HSC is slow and inefficient compared with that obtained by hepatocyte transplantation, although significant repopulation eventually occurs. The first hepatocytes generated from HSC appeared at approximately 7 weeks after transplantation. By contrast, transplanted hepatocytes reconstituted more than 50% of the liver in approximately 4 weeks, leading the authors to conclude that “hepatocyte replacement by bone marrow cells is a slow and rare event.”91 Nevertheless, by 22 weeks after transplantation, repopulation from transplanted HSC constituted approximately 30% of the entire liver. It has now been shown that hepatocytes generated from transplanted HSC in FAH knockout mice are the product of cell fusion rather than a result of transdifferentiation.92, 93 The fusion process created tetraploid hepatocytes, 6X hepatocytes, and aneuploid hepatocytes. The liver cell that fuses with HSCs has not been identified, but the hepatocytes produced by the fusion event do not express HSC genes. Recent data from studies of muscle regeneration suggest that HSC are not capable of directly generating myogenic progenitors.94 Instead, after muscle injury, circulating inflammatory cells such as macrophages and neutrophils fuse with myotubes. A similar mechanism may occur in the liver, that is, fusion may occur between bone marrow-derived macrophages and hepatic cells, triggering the proliferation of the fused liver cell. It is conceivable that fusion between bone marrow and liver cells occurs in FAH knockouts because of the high proliferative pressure imposed by this system. However, it may be argued that high levels of hepatocyte production from hematopoietic stem cells can be achieved only in a system in which cell fusion occurs. In any case, these results make it clear that any attempt to use HSC to reconstitute livers needs to demonstrate: (1) the mechanism by which hepatocytes are formed and (2) that hepatocytes generated from bone marrow cells function normally, do not have abnormal genomes, and are not prone to malignancy.95 Krause et al.96 injected single, highly purified bone marrow cells into irradiated mice and obtained engraftment in several organs, including skin, lung, and liver. The livers of five mice examined did not contain HSC-derived hepatocytes, and three of these mice also did not have bone marrow-derived cells in bile ducts. The other two mice had 0.4% and 2.2% of bile duct cells of bone marrow origin 11 months after transplantation. Wagers et al.97 used a different method of bone marrow cell isolation and marked cells with green fluorescent protein. Single-labeled HSC cells injected into irradiated mice failed to contribute to brain, kidney, gut, muscle, and liver, although they completely reconstituted the bone marrow of these animals. The experiments of Krause et al. and Wagers et al. differ in important technical aspects, including the type of cells used, but there seems to be complete agreement regarding the generation of hepatocytes from injected, purified bone marrow cells: no bone marrow-derived hepatocytes were detected in either of these experiments. Theise et al.98 transplanted unfractionated bone marrow cells or purified CD34+lin- cells from male mice into irradiated female mice and searched for cells containing the Y chromosome and expressing albumin mRNA in the liver of the recipient mice 1 week to 8 months after transplantation. Cells positive for both the Y chromosome and albumin mRNA were detected from 2 to 8 months after transplantation at frequencies of 0.39% to 1.1% of total hepatocytes in the liver. The authors used a correction factor to compensate for potential error sampling in detecting the Y chromosome, which raised the reported frequencies by a factor of 2. In all of these experiments, the livers of the mice transplanted with bone marrow cells were morphologically normal and apparently undamaged. It is of great interest to determine whether more efficient production of hepatocytes from injected bone marrow cells leading to liver repopulation can be achieved in injured livers. Mallet et al.99 reconstituted the bone marrow of lethally irradiated mice with bone marrow cells from Bcl-2 transgenic mice100 and asked if bone marrow-derived Bcl-2 expressing hepatocytes would repopulate the liver after hepatic injury elicited by repeated injections of Fas agonist antibody. Even with severe hepatic damage, the proportion of bone marrow-derived Bcl-2-expressing hepatocytes found in hepatic tissue was very small, varying from 0.008% to 0.8% of total hepatocytes, depending on the extent of damage. Kanazawa and Verma101 studied the generation of hepatocytes from bone marrow cells in three different models of liver injury and concluded that there was little or no contribution of bone marrow cells to the replacement of hepatocytes in these models. Fujii et al.102 transplanted green fluorescent protein-positive bone marrow cells into green fluorescent protein-negative irradiated recipient mice and identified green fluorescent protein-positive endothelial cells and Kupffer cells, but no hepatocytes. A similar result was obtained by Dahlke et al.103 using a model of acute liver failure. It can be concluded from these experiments that bone marrow cells have a minimal capacity to generate hepatocytes in normal livers and a very low capacity in injured livers. The exception is the production of hepatocytes from bone marrow cells in FAH knockout mice, which, as already discussed, is a consequence of cell fusion. Theise et al.104 investigated, in patients receiving bone marrow or liver transplants, whether hepatocytes could be generated from the bone marrow cells of the transplant recipient. The frequency of hepatocytes that were considered to be bone marrow derived varied from 1% to 3.6% in five patients and was 8% in one patient. The authors multiplied these values by a factor of approximately 5 to correct for sampling errors in the detection of the Y chromosome and reported that hepatocyte engraftment ranged from 4% to 43%. Alison et al.105 also examined the livers of patients who received bone marrow or liver transplants from sex-mismatched donors. The frequency of bone marrow-derived hepatocytes in the liver of the recipients was estimated to be 0.5% to 2%. Korbling et al.106 reported that the frequency of hepatocytes generated from the bone marrow of recipients of sex-mismatched liver or bone marrow transplants varied from 4% to 7% and was unrelated to liver injury and to the time after transplantation. Two other studies of liver transplant patients did not detect bone marrow-derived hepatocytes or found only occasional cells.107, 108 Several studies have addressed the question of cellular chimerism in liver transplants. Hove et al.109 examined the livers of 16 transplanted patients to identify cells originating from the recipient and reported chimerism of endothelial cells in 14 patients, bile duct epithelial cell chimerism in five patients, and hepatocyte chimerism in one patient. Ng et al.110 recently reported that the vast majority of recipient-derived cells present in transplanted livers were macrophages or Kupffer cells and that only 1.6% of the total recipient cells detected in these livers were hepatocytes (this corresponded to 0.62% of the total number of hepatocytes in the liver). Finally, Kleeberger et al.111 detected 91% chimerism frequency in liver transplant recipients, and the presence of hepatocyte chimerism in two of nine patients in samples obtained 4 weeks after transplantation and in five of nine patients at 12 months or longer (a quantitative analysis of the percentage of recipient hepatocytes in the chimeric livers was not reported). How can these results be interpreted? First, the lack of consistency of the results may be a consequence of the use of different techniques to identify recipient-derived hepatocytes in transplanted livers. For instance, factors often are used to correct for the inability to examine the complete surface of hepatocyte nuclei to detect the Y chromosome. This type of correction, which involves the multiplication of observed values by factors that vary from less than 2 to more than 5 can introduce significant errors and uncertainties in the reported data. Another difficulty is that large numbers of mesenchymal cells in transplanted livers originate from the recipient's bone marrow. Although various markers can be used to identify hepatocytes, the precision of the methods used for this identification is variable and not always optimal. Superimposition of images without confocal microscopy may lead to errors caused by the detection of reaction products or in situ hybridization signals in endothelial or Kupffer cells that are in close apposition to a hepatocyte. The uncertainties about the techniques used in some of these experiments have been discussed.112, 113 Until more definitive data are obtained, it can be concluded that the generation of bone marrow-derived hepatocytes occurs in the livers of some but not all transplant patients, that it is a highly inefficient process, and that, because of its very low frequency, its physiological relevance remains unproven. During the last few years, work with stem cells became one of the most exciting areas of biomedical research, generating much enthusiasm from scientists, clinicians, and the general public. Having as an endpoint the repopulation and regeneration of tissues and organs, the promise of this research is far reaching. After the initial phase of excitement and the publication of findings that often defied long-accepted knowledge, it is now time to assess the progress made, to point out pitfalls and misinterpretations, and, most importantly, to project a realistic look into the future. It is a sobering thought that definitions for stem cells are still being debated.114 Does the expression “Seeing is not being” apply to stem cells?115 Is a stem cell “an entity or a function,”116 a contextual category,1 the first, defined, component of a hierarchical system, or a functionally plastic entity (the “chiaroscuro stem cell”117)? Can this entity be isolated and studied in culture in a biologically meaningful way, or are we dealing with Heisenberg's uncertainty?1 Carried to their most extreme implications, and crudely interpreted, some of these ideas may lead us to an “anything goes” intellectual environment, in which all results even the most discordant, are accepted uncritically. Great progress in stem cell research has been made and will continue to be made by the careful scrutiny of experimental data, by examining the reliability and reproducibility of methods to identify stem cells, and by addressing the issue of the biological relevance of the findings. Progress in the field is inextricably linked to the understanding of mechanisms of cell differentiation, proliferation, and transdifferentiation and knowledge about the interactions between cells and extracellular matrix components in normal and injured tissues. Future work may show that umbilical cord cells, embryonic stem cells, or fetal liver cells are better sources of hepatocyte precursors for hepatic repopulation than bone marrow cells.118-130 The emphasis on bone marrow stem cells often obscures the fact that the remarkable regenerative capacity of the liver primarily is the result of to the replication of mature hepatocytes. When hepatocyte replication is slow or inhibited, intrahepatic stem cells give rise to oval cells, which replicate and differentiate into hepatocytes (Fig. 1). The proliferative capacity of normally quiescent, highly differentiated hepatocytes is unique among differentiated cells in mammalian tissues. As paradoxical as it may sound to stem cell biologists, the hepatocyte is the most highly efficient “stem cell” for the liver. A first observation is that the identification of bone marrow–derived hepatocytes needs to be carried out according to rigorous criteria. Methodological accuracy and reproducibility are critical factors that determine the reliability of the data reported. From the review of the literature presented here, I conclude that the generation of hepatocytes from bone marrow cells is a very rare event in liver transplantation and repopulation after injury and that such hepatocytes are produced by cell fusion rather than by a transdifferentiation mechanism (Fig. 2). This premise does not exclude a potential functional role for bone marrow-derived cells in hepatic homeostasis and, in fact, suggests experimental approaches to study this issue. Possible mechanisms for the generation of hepatocytes from bone marrow cells. Of the mechanisms shown in the figure, only cell fusion (#4) has been shown to occur in vivo. Generation of hepatocytes from pluripotent bone marrow stem cells (#2) can occur in culture.90 (Diagram redrawn from Wagers AJ, Weissman IL.140) Does cell fusion generate functionally intact hepatocytes that, despite their abnormal ploidies, may not carry a high risk for transformation? Can cell fusion techniques be applied to liver repopulation in a clinical setting? May this approach be more successful for liver repopulation than hepatocyte, oval cell, or hepatic embryonic stem cell transplantation? The role of the bone marrow in generating nonparenchymal cells in liver regeneration and repopulation seems to be much more significant than the generation of hepatocytes. Would blockage of the migration of HSC or bone marrow-derived leucocytes into the liver interfere with liver regeneration and repopulation? Is the hepatic seeding of bone marrow-derived endothelial and Kupffer cells essential for normal liver homeostasis? Do these cells (or perhaps even the small number of bone-marrow-derived hepatocytes) produce special cytokines and growth factors that are required for hepatocyte replication? The derivation of oval cells (albeit at very low levels) from the bone marrow has been reported, but so far not confirmed in other laboratories. This very important question must be settled definitively. If oval cells are not generated by bone marrow cells, could they (or cells in the canals of Hering) fuse with bone marrow cells to generate hepatocytes? Do MAPCs differentiate in vivo and generate lineages that populate the liver in adult humans or animals? Can cultures of MAPCs be used to produce large amounts of human hepatocytes suitable for transplantation? Great progress has been achieved in the purification of stem cells from embryonic liver and umbilical cord blood. Would transplantation of these types of cells into injured livers reconstruct both hepatocytes and bile ducts? Are these cells more efficient than fetal or adult hepatocytes for the repair of liver injury?