In the beginning, there was the hepatocyte! If we limit the definition of a stem cell to its ability to self renew and to reconstitute a given tissue in vivo, hepatocytes fulfill both criteria. Serial transplantation experiments performed in the fumarylacetoacetate hydrolase (FAH)–deficient mice have underlined not only the extraordinary replication capacity of these differentiated cells but also the absence of a specific stem cell requirement to repopulate a rodent liver.1 Unfortunately, hepatocyte transplantation has very rarely produced therapeutic effects in human clinical trials mostly because hepatocytes are difficult to expand ex vivo and because their numbers are too low to achieve a biological effect.2, 3 A somatic human stem cell that could be propagated in large quantities while retaining its ability to differentiate into different cell types could serve as a highly valuable resource for the development of cellular therapy in liver diseases. But does such a cell type exist? In certain pathological conditions, mainly when hepatocyte replication is blocked, bipotent oval cells proliferate and participate in liver regeneration by giving rise to hepatocytes and bile duct cells. However, the fact that oval cells have been shown to generate hepatocellular carcinoma or cholangiocarcinoma in rodents and that they are difficult to characterize to homogeneity jeopardizes their use in therapeutic approaches. (For review, see Fausto and Campbell.4) FAH, fumarylacetoacetate hydrolase; HSC, hematopoietic stem cells; MSC, mesenchymal stem cells; MAPC, multipotent adult progenitor cell; USSC, Unrestricted Somatic Stem Cells; bFGF, basic fibroblast growth factor; HGF hepatocyte growth factor. Then came the hematopoietic stem cells. Numerous recent studies including indirect a posteriori studies in humans suggested that hematopoietic stem cells (HSCs) have a broader differentiation potential than previously expected and could generate hepatocytes. However, except when a very strong selective pressure occurs to expand this phenomenon, such as in FAH-deficient mice,5 it seems that bone marrow cells have a very low capacity to generate hepatocytes in vivo even in injured livers, and that differentiation is probably dependent on the type of injury.6-11 In contrast, hematopoietic stem cells generate numerous nonparenchymal cells, such as Küpffer cells or endothelial cells that could play a role in liver regeneration or repair. Altogether, these observations prompted N. Fausto to conclude in a recent review: “as paradoxical as it may sound to stem cell biologists, the hepatocyte seemed, until now, the most highly efficient stem cell for the liver”12 Where have we gone from there? One very recent and intriguing paper has demonstrated the repopulation of preimmune sheep fetal liver with human hematopoietic stem cell-derived hepatocytes.13 The fetal sheep model was developed more than 10 years ago by Zanjani's group.14 It takes advantage of recipient sheep fetuses' immune permissivity, allowing to obtain a long-term hematopoietic chimerism after in utero injection of human HSCs into the fetal peritoneal cavity at 48–54 days of gestation. In the recent study, a direct correlation between hepatocyte formation and hematopoietic cell engraftment was found. One of the sheep exhibited up to 17% of donor human hepatocytes at 11 months posttransplant, and bone marrow of a first recipient maintained the ability to produce human hepatocytes in secondary recipients. However, these results do not rule out the hypothesis of the presence of liver-specific progenitor cells within the sorted hematopoietic stem cell population as it has been recently proposed.15 Reexamination of some bone marrow transplantation experiments has led to the conclusion that donor cells had simply fused with resident hepatocytes. Such a fusion event has been demonstrated not only in the very specific FAH-deficient mouse model, where it has been shown to occur between myelomonocytic cells and resident hepatocytes,16, 17 but also in normal mice, using an elegant Cre-lox strategy.18 However, using the same approach, Harris et al.19 have recently brought some controversy to the subject, by demonstrating that epithelial cells can develop from bone marrow without cell fusion. Most of these experiments attempting to demonstrate the plasticity or the fusion of bone marrow derived cells have been performed in vivo. Only rare publications have shown the hepatocyte differentiation potentiality of human peripheral blood monocyte-derived cells20 or hematopoietic stem cells21 directly in vitro. Finally, the differences in the types of donor cell subpopulations studied, in the methods of detection of donor-derived cells and in the recipient models used hampered conclusions about the real plasticity of these cells. Moreover, in the light of recent discoveries, one could also think that in some of these experiments, bone marrow-derived hepatocytes could have originated from the mesenchymal compartment and not from the hematopoietic one. Mesenchymal stem cells (MSCs) meet two criteria: they grow in culture as adherent cells and they can differentiate in vitro into osteoblasts, chondroblasts, and adipocytes in response to appropriate stimuli. When injected in vivo MSCs differentiate into this same array of cell types.22 Several recent reports have shown that marrow stromal cells might also differentiate into other cell types such as neurons.23 With the demonstration of their pluripotentiality also comes a certain degree of confusion in the terminology of these cells, with some authors calling them stromal cells, multipotent progenitor cells or mesenchymal stem cells. For two years, multipotent adult progenitor cells (MAPCs) were the only documented postnatal bone marrow cell population capable at a clonal level of in vitro differentiation into mesenchymal cell lineage and hepatocytes. More recently, a human cord blood somatic stem cell population has been isolated and shown to be capable of differentiating into neural cells and hematopoietic cells in vitro and into neuron-like, hematopoietic cells, cardiomyocytes, and liver cells in fetal sheep.24 Finally, in this issue, Lee et al. describe a mesenchymal stem cell population with hepatocytic potential.25 So now, how can we reconcile all these stem cells? Human or rodent postnatal bone marrow contains primitive progenitors termed multipotent adult progenitor cells, or MAPCs, that copurify with mesenchymal stem cells and appear after a certain period in culture.26 These cells have multilineage differentiation potential27 — including differentiation into hepatocytes28 — and can be expanded ex vivo for more than 80 population doublings without signs of senescence. Moreover, they contributed to most somatic tissues, including liver, when injected into an early murine blastocyst, and engrafted and differentiated into cytokeratin 18 (CK18) and albumin positive cells when infused intravenously into NOD/SCID animals.27 In addition, no tumor formation was seen in these immunodeficient mice after transplantation. All these characteristics made MAPC an extremely promising candidate for liver cell therapy. However, two points need to be answered: (1) it is not yet clear whether MAPCs really exist as such in vivo, as a compartment of mesenchymal cells, or whether they only appear under specific in vitro conditions as a result of dedifferentiation of a mesenchymal stem cell; and (2) their therapeutic potential has never been demonstrated, and particularly their participation in liver regeneration or repopulation still awaits confirmation. More recently, a new intrinsically pluripotent CD45- population called USSC, for Unrestricted Somatic Stem Cells, has been isolated from human placental cord blood.23 These cells grow adherently and can be expanded to at least 1015 cells while maintaining a normal karyotype and a pluripotency including hematopoietic, neural, and hepatic differentiation in the noninjured fetal sheep model. More than 20% albumin-producing human parenchymal hepatic cells have been obtained with absence of cell fusion in this model. One major biological difference between USSCs and human MAPCs is the ease of generation of USSCs in cytokine-free cultures and the potential to generate hematopoietic cells in vitro. Besides their differentiation potential, USSCs can also be distinguished from mesenchymal stem cells and MAPCs by their phenotype (Table 1). But similarly to adult and fetal MSCs, they are also nonimmunogeneic and could even be immunosuppressive, a characteristic that could be particularly interesting for stem cell therapy. Previous studies have shown that umbilical cord blood may contain some hepatic progenitors.29 However, confirmation of pluripotentiality of these cells still awaited clonal analyses. Lee et al. have recently isolated a mesenchymal stem cell from umbilical cord blood that harbors a broader potential than expected.30 In this issue of HEPATOLOGY, Lee et al.25 demonstrate that a single clonally expanded cell, isolated from either cord blood or bone marrow, can differentiate not only into osteoblasts, adipocytes, and chondrocyte-like cells (as expected for MSCs), but also into different cell types including functional hepatocytes in vitro. After 4 weeks of induction with bFGF (basic fibroblast growth factor) and HGF (hepatocyte growth factor), these cells acquire a cuboidal morphology and are indeed able to express liver marker genes, present functional characteristics of hepatocytes such as glycogen storage, urea secretion, uptake of low-density lipoprotein, cytochrome P450 activity in vitro, and also express an antigen normally expressed on the bile canaliculi formed between adjacent hepatocytes. All these criteria have been listed as required to define a true functional hepatocyte. Although sorted with a two-step procedure with immunodepletion of hematopoietic and red blood cells (CD3, CD14, CD19, CD38, CD66b, Glycophorin A) that is more similar to the isolation of MAPCs than classical MSCs, these cells seem to differ from MAPCs by their phenotype (Table 1) and their doubling time. Finding a universal multipotent postnatal stem cell that could help to bridge patients to liver transplantation, provide metabolic support during liver failure or replace hepatocyte mass in metabolic liver-diseases cell sounds like the Holy Grail quest. Any claim for isolation of a new stem cell requires the demonstration of clonogenicity and differentiation into the suitable cell type. The three categories of mesenchymal cells presented here meet this requirement. However, it is not yet known whether the transplantation of MSCs decribed by Lee et al. will indeed generate hepatocytes and bile ducts in vivo. Although a demonstration of in vivo hepatocyte differentiation has been obtained after MAPC transplantation in mice and after human USSC transplantation in a fetal sheep model, all studies, including the one described in this issue, lack an in vivo approach demonstrating the liver repopulation associated with a therapeutic functional effect. Therefore, it will now be important to compare side by side these three categories of cells for their availability and their ability to participate in liver regeneration and function without inducing tumorigenicity.