AFP+, ESC-Derived Cells Engraft and Differentiate into Hepatocytes in Vivo



A major problem in gene therapy and tissue replacement is accessibility of tissue-specific stem cells. One solution is to isolate tissue-specific stem cells from differentiating embryonic stem (ES) cells. Here, we show that liver progenitor cells can be purified from differentiated ES cells using alpha-fetoprotein (AFP) as a marker. By knocking the green fluorescent protein (GFP) gene into the AFP locus of ES cells and differentiating the modified ES cells in vitro, a subpopulation of GFP+ and AFP-expressing cells was generated. When transplanted into partially hepatectomized lacZ-positive ROSA26 mice, GFP+ cells engrafted and differentiated into lacZ-negative and albumin-positive hepatocytes. Differentiation into hepatocytes also occurred after transplantation of GFP+ cells in apolipoprotein-E- (ApoE) or haptoglobin-deficient mice as demonstrated by the presence of ApoE-positive hepatocytes and ApoE mRNA in the liver of ApoE-deficient mice or by haptoglobin in the serum and haptoglobin mRNA in the liver of haptoglobin-deficient mice. This study describes the first isolation of ES-cell-derived liver progenitor cells that are viable mediators of liver-specific functions in vivo.


A major obstacle in the use of embryonic stem (ES) cells for tissue replacement, gene therapy, and tissue engineering has been the ability to regulate the complex spontaneous differentiation of these totipotent cells into specific cell types or tissues. One solution is to isolate tissue-specific stem cells from differentiating ES cells, if we can identify reliable specific markers for specific tissue stem cells. Here, we describe the use of a fetal liver-specific protein, alpha-fetoprotein (AFP), to identify liver progenitor cells from differentiated ES cells. AFP is expressed in a tissue-specific manner during mammalian development. In early mouse embryos, AFP expression is specific to the visceral endoderm of the yolk sac and the gut endoderm before being restricted to the fetal liver and fetal gut later in development [14]. AFP is one of the earliest proteins known to be expressed in the hepatic lineage during embryonic development. AFP is first detected in the gut endoderm at the four-somite stage of the mouse embryo and precedes albumin expression at the seven- to eight-somite stage [3, 4]. In addition, AFP expression declines rapidly after birth, and the level of its mRNA in adult liver is less than 0.01% of that in fetal liver. During rapid hepatocyte proliferation, such as liver regeneration or tumorigenesis, AFP expression is reactivated [5]. Together, these results suggest that AFP may be a marker for liver progenitor cells or proliferating liver cells. AFP is also expressed in differentiating ES cells. ES cells differentiate in vitro to form embryoid bodies (EBs), which are complex, three-dimensional aggregates consisting of endodermal, mesodermal, and ectodermal stem cells [6, 7]. During differentiation of ES cells to EBs, AFP is first expressed in the EB outer layer, which is the equivalent of embryonic primitive endoderm that develops into the visceral endoderm [810]. The early embryonic expression pattern of AFP, other mesodermal proteins, and some hepatic-specific genes is essentially recapitulated during the development of ES cells into EBs [812]. Based on this background information, we rationalized that AFP may be a good marker for identifying endodermal progenitor cells that constitute a committed hepatic lineage from differentiated ES cells.

Materials and Methods

Gene Targeting

The targeting construct for knocking the green fluorescent protein (GFP) gene into the AFP locus was a replacement type of vector designed to replace AFP coding sequences with those of GFP. The vector was generated by ligating the ATG start codon of AFP in-frame to that of the GFP gene. Prior to ligation, an NcoI site was created at the AFP start codon gene by converting the second translation codon from AAG to GAG. A plasmid vector with a GFP insert (Clontech Laboratories; Palo Alto, CA; was partially cleaved by NcoI at the GFP translation start codon and ligated to NcoI-cleaved AFP DNA. The resulting hybrid gene was essentially a GFP gene flanked at the 5′ end by about 2 kb of AFP genomic sequences. At the 3′ end, it was flanked by a loxP/PGKTK/ PGKNeo/loxP (LTNL) fragment followed by 5 kb of AFP genomic sequences comprising exon 3 to exon 5. The construct was linearized with NotI and electroporated into CS-1 ES cells. CS-1 is a 129Sv-derived cell line (a gift of C.S. Lin) as previously reported [13]. Transfected cells were selected in the presence of G418. Targeted ES cell clones were identified by Southern blot hybridization using a single copy probe that was derived from exon 6 and intron 6 of the AFP gene.


B6.129S7-Gtrosa26 mice and apolipoprotein-E- (ApoE) deficient mice (B6.129P2-ApoEtm1Unc) were purchased from the Jackson Laboratory (Bar Harbor, ME; 129/Sv mice were purchased from the Animal Resource Centre (Australia). For transplantation, lacZ-positive F1 hybrid mice from a cross between B6.129S7-Gtrosa26 and 129/Sv strains were used. Mice were anesthetized with 0.1 ml of a cocktail consisting of 1 part Hypnorm, 1 part Midazolom, and 2 parts distilled water per 10 gm of body weight. Two-thirds partial hepatectomy was performed before cells were injected intrasplenically in a 20-μl volume of saline. In sham-transplanted mice, saline was injected in the place of cells. For cell transplantation using haptoglobin-deficient or ApoE-deficient mice, the mice were injected intraperitoneally with 15 μg/g of cyclosporin (Novartis; Basle, Switzerland; daily. Before removing the liver for analysis, mice were perfused with saline followed by 0.5% volume by volume (v/v) glutaraldehyde. The livers were removed and soaked in 30% weight by volume (w/v) sucrose in phosphate-buffered saline (PBS) and 0.5% glutaraldehyde overnight at 4°C. For analysis of haptoglobin-deficient mice, the mice were treated with 0.1 mg lipopolysaccharide (LPS) per 10 g body weight 24 hours before the mice were sacrificed. All animal experimentation protocols were approved by the National University of Singapore Animal Ethics Research Committee.

β-Galactosidase Assay

To analyze livers for the production of β-galactosidase, glutaraldehyde-fixed livers were cryosectioned at a 12-μm thickness, and the sections were placed in a tissue culture dish. The sections were stained for β-galactosidase as previously described [14] with some modifications. Briefly, tissue sections were immersed in X-gal staining solution and incubated at 37°C for 2-3 hours. After staining, the sections were washed with PBS, mounted on slides, photographed, and then stained with hematoxylin and eosin (H&E).

ES Cell Differentiation

ES cells were differentiated in vitro using either the suspension culture method [15] or a modified two-step method [16]. In the former, exponentially growing ES cells were trypsinized gently into cellular aggregates and plated on bacteriological Petri dishes to form simple EBs in ES medium. After 4 days, the simple EBs were diluted and transferred to fresh bacteriological Petri dishes. At intervals of 1, 4, and 8 days, EBs were removed and placed on tissue culture slides for 2 hours and then observed using a confocal microscope. In this short incubation, EBs adhered lightly to the slide and there was minimal disruption to the three-dimensional structure. In the modified two-step method [16], ES cells were first cultured in a methylcellulose-based medium consisting of 3.9 ml methylcellulose (MethoCult M3134; StemCell Technologies, Inc; Vancouver, Canada;, 4.2 ml Iscove's modified Dulbecco's medium (Life Technologies; Rockville, MD;, 1.5 ml serum, 100 μl monothioglycerol stock solution (37.8 μl in 10 ml PBS) (Sigma; St Louis, MO;, and 100 μl 100× L-glutamine/penicillin/streptomycin stock solution (Life Technologies). There were 2 × 104 ES cells in 100 μl ES medium. Six days later, the EBs were then dissociated into single-cell suspensions by collagenase and replated in the same methylcellulose-based medium supplemented with Hepa-cell-conditioned medium (100 μl per 3 ml culture medium) for another 6 days. Cells were harvested, and GFP+ cells were isolated using a fluorescence-activated cell sorter (FACStarplus; Becton Dickinson; San Jose, CA; Hepa-cell-conditioned medium was prepared from the media of Hepa cells, a mouse hepatoma cell line. Confluent Hepa cells were grown in Dulbecco's modified Eagle's medium with 10% fetal calf serum for 2 days; the media were then harvested and centrifuged twice at 800 g for 5 minutes to remove cellular debris. The media was filtered through a 0.2-μm and then a 0.1-μm filter. This conditioned media was stored in aliquots at -80°C.

RNA and DNA Analyses

RNA and DNA were quantified using, respectively, the RiboGreen RNA Quantification kit and the PicoGreen dsDNA Quantification kit (Molecular Probes; Eugene, OR; Reverse-transcriptase polymerase chain reaction (RT-PCR) was performed as previously described [17]. The types of mRNA, the size (bp) of the amplified PCR product, and the RT-PCR primers were : A) AFP mRNA (471 bp) 5′-TCC ACG TTA GAT TCC TCC CAG-3′ and 5′-TTG CAG CAT GCC AGA ACG ACC-3′; B) albumin mRNA (866 bp) 5′-TTG CTG CTG ATT TTG TTG AGG-3′ and 5′-GAG AAG GTT GTG GTT GTG ATG-3′; C) triose phosphate isomerase (TPI) mRNA (523 bp) 5′-CCC TGG CAT GAT CAA AGA CTT-3′ and 5′-GAT GGG CAG TGC TCA TTG TTT-3′; D) haptoglobin mRNA (895 bp) 5′-AAA CGA CGA GAA GCA ATG GGT-3′ and 5′-GAA GGC AGG CAG ATA GGC ATG-3′, and E) ApoE mRNA (451 bp) 5′-CTG AAC CGA TTC TGG GAT TAC-3′ and 5′-GCT CAC GGA TGG CAC TCA CAC-3′. All RT-PCR primers spanned at least one intron. PCR amplification of genomic DNA for the AFP/GFP hybrid marker gene by PCR was performed using conditions similar to those for RT-PCR. The 5′ primer was specific for AFP exon 1 (5′-AAA GGA CTT CAG CAG GAC TGC-3′), and the 3′ primer was specific for GFP (5′-TCG TCC TTG AAG AAG ATG GTG-3′). The expected PCR fragment size was 363 bp. All RT-PCR and PCR primers were designed using genomic sequences retrieved from Genbank.


For the detection of albumin and ApoE in tissue sections, standard immunohistochemistry was performed using 1:100 diluted biotinylated rabbit anti-albumin (Accurate Chemical and Scientific Corp; Westbury, NY; and 1:100 diluted goat anti-mouse ApoE antibody (Santa Cruz Biotechnology, Inc.; Santa Cruz, CA;, respectively. The primary antibodies were detected using streptavidin-conjugated horseradish peroxidase or horseradish peroxidase conjugated rabbit anti-goat IgG, respectively, and DAB (Sigma). The sections were counterstained with Mayer's hematoxylin.

Immunoprecipitation and Western Blot Hybridization

Serum ApoE and haptoglobin were assayed by Western blot hybridization using standard procedures. For serum ApoE, 2 μl of serum were separated on 10% SDS-PAGE, blotted onto nitrocellulose, and probed with goat anti-mouse ApoE antibody (Santa Cruz Biotechnology, Inc.) and then a biotinylated rabbit anti-goat antibody. To assay for the presence of serum haptoglobin, 100 μl of serum were first immunoprecipitated with 60 μg goat anti-human haptoglobin (Sigma) that was bound to Protein G-Agarose (Roche Molecular Biochemicals; Indianapolis, IN; in a 2:1 weight ratio according to the manufacturer's protocol. The immunoprecipitate was then separated on 10% SDS-PAGE, blotted onto nitrocellulose, and probed with a goat anti-human haptoglobin (Sigma) and then a biotinylated rabbit anti-goat antibody. For detection, the blots were incubated with streptavidin-conjugated horseradish peroxidase and then developed using a chemiluminescent substrate (SuperSignal; Pierce; Rockford, IL;


Genomic Modification of ES Cells to Tag AFP-Expressing ES Cell Derivatives

To facilitate the selection and purification of AFP-positive cells during ES cell differentiation, the GFP gene was knocked into the AFP locus in CS-1 ES cells by homologous recombination (Figs. 1A, 1B) [13]. The ATG translation start codon of the GFP gene was inserted in-frame with the translation start codon of the AFP gene (Fig. 1A). Consistent with the AFP expression pattern in ES cells, recombinant ES cells did not express GFP in the undifferentiated state (Fig. 1C). However, when induced to differentiate into EBs by suspension culture [15], we observed a different spatial distribution of green fluorescence in 4-day- and 8-day-old EBs (Fig. 1C), suggesting that GFP expression was regulated in differentiating ES cells.

Figure Figure 1..

GFP recombination at the AFP locus and its expression in EBs. A) AFP locus, targeting vector for GFP knock-in at the AFP locus, and homologously recombined locus. The vector was a replacement targeting vector consisting of the GFP gene and a floxed (PGK Neo-PGK TK) selectable marker gene flanked by 2 and 5 kb of AFP homologous sequences at the 5′ and 3′ ends, respectively. The ATG translation start codon of the GFP gene was cloned in-frame with that of the AFP gene. The vector was designed to replace AFP coding sequences in exon 1 and 2 with those of GFP. A HindIII/BamHI fragment derived from exon 6 and intron 6 of the AFP gene was used as probe for Southern blot hybridization. The probe detects a 12-kb HindIII WT fragment and a 5-kb HindIII recombined fragment. B) Southern blot hybridization of genomic DNA from WT and recombinant ES clones. C) Fluorescence microscopy of recombinant ES cell during in vitro differentiation using the suspension culture method [15]. STrypsinized ES cells were resuspended at 106/ml to form aggregates of EBs, and after 4 days, the simple EBs were transferred to a bacteriological Petri dish. At intervals of 1, 4, and 8 days, EBs were harvested and allowed to adhere to tissue culture slide before viewing using confocal microscopy (ii-iv).

In Vitro Differentiation of Modified ES Cells to GFP+ Cells

Although GFP+ cells could be easily isolated from EBs grown in suspension culture, it was difficult to obtain evenly sized EBs from suspension culture. These different sizes of EBs suggested that the EBs prepared from suspension cultures might be in different stages of differentiation. To obtain better synchronization of the differentiation process, the recombinant ES cells were differentiated in vitro using a modified two-step differentiation protocol [16] that was originally developed for the generation of hematopoietic cells from ES cells (Fig. 2A). We rationalized that if hepatocytes can be generated from bone marrow cells, then it is possible that the hepatic progenitor cells may share some similar developmental and differentiation pathways. In this protocol, ES cells were first grown in methylcellulose-based medium to form EBs for the production of hematopoietic stem cells and then induced to terminal differentiation with hematopoietic growth factors in the second step. EBs prepared in methylcellulose-based medium were fairly uniform in size and morphology. At 6 days, less than 10% of the cells in these EBs were GFP+. To increase the yield of these cells and to replenish the medium, the EBs were disaggregated into single cells and replated in methylcellulose-based medium as described above. No hematopoietic growth factors were added. After another 6 days, the cells were harvested. At the time of harvest, GFP+ cells constituted 30%-50% of the cells by fluorescent-activated cell sorting (FACS) analysis (Fig. 2B). If the medium was supplemented with 3% (v/v) Hepa-cell-conditioned medium, we observed a small but consistent ∼10% increase in viable cells (Fig. 2A). However, we did not observe an increase in the percentage of GFP+ cells (Fig. 2A). As expected, AFP mRNA was detectable in GFP+ cells but not in ES cells, EBs, or GFP-negative (GFP) cells. In our differentiation protocol, albumin mRNA was detected earlier than AFP mRNA in EBs (Fig. 2B). AFP and albumin mRNAs were not detectable in the GFP cells.

Figure Figure 2..

In vitro differentiation of recombinant ES cells. A) Schematic diagram of the two-step in vitro differentiation of ES cells and the subsequent purification of GFP+cells for transplantation. B) Upper panels show FACS analysis of AFP/GFP recombinant ES cells before and after the two-step in vitro differentiation. Lower panels show ES cells and differentiated ES cells as viewed using fluorescence microscopy. Green flourescent or GFP+cells are in the upper right quadrant. Lower panels show ES cells and differentiated ES cells as viewed using fluorescence microscopy. The cells were stained with propidium iodide. Green fluorescent cells represent GFP+cells. C) RT-PCR analysis for AFP, albumin, and TPI mRNAs in differentiated ES cells. The recombinant ES cells were differentiated in vitro using a two-step differentiation protocol. Cells from each stage of differentiation were harvested for total RNA preparation and RT-PCR analysis. Six sets of cells were harvested. They are: undifferentiated ES cells (ES); primary EBs harvested from the first step of the differentiation protocol; differentiated ES cells from the second step of differentiation and cultured in the absence of Hepa-cell-conditioned medium (Diff. cell – HCM) or in the presence of Hepa-cell-conditioned medium (Diff. cell + HCM); and differentiated ES cells from the second step of differentiation, cultured in the presence of Hepa-cell-conditioned medium and sorted by FACS for GFPand GFP+cells.

Engraftment and Differentiation of GFP+ Cells to Hepatocytes

When undifferentiated ES cells were injected intrasplenically, teratomas were observed in all four mice that had undergone partial hepatectomy (Fig. 3A). In mice that were sham-operated and injected intrasplenically with ES cells, none of the mice (n = 4) developed teratomas (data not shown). The teratomas were all ES cell derived as evidenced by the lack of β-galactosidase staining using X-gal (Fig. 3A). Therefore, partial hepatectomy is important for the engraftment and formation of ES-cell-derived teratomas in the liver. By extrapolation, we rationalized that partial hepatectomy may also be essential for the engraftment and differentiation of putative liver stem cells.

Figure Figure 3..

Histology of liver after transplantation of AFP/GFP-targeted ES cells and their differentiated, GFP+cell derivatives. A) Recombinant ES cell transplantation into two-thirds partially hepatectomized, lacZ-positive F1 hybrid mice (B6.129S7-Gtrosa26 × 129/Sv). ES-cell-derived teratomas were paraffin embedded, sectioned at a 6-μm thickness, and stained with H&E (i and ii-vii viewed under 20× and 100× magnification, respectively). Some of the transplanted livers were cryosectioned at a 12-μm thickness and stained for E. coli β-galactosidase using X-gal. The same section was then stained using H&E (viii and ix viewed under 20× magnification). (B) Transplantation of differentiated, GFP+ES cells into two-thirds partially hepatectomized lacZ-positive F1 hybrid mice. Livers were cryosectioned at a 12-μm thickness. Sham-operated and transplanted liver sections were stained together for β-galactosidase using X-gal (i and ii viewed under 10× and 20× magnification, respectively). Liver section from sham-operated mice (iii) as viewed under 40× magnification. (iv-vi) Liver sections from transplanted mice (viewed under 100×, 100×, 200×, and 400× magnifications, respectively). After X-gal staining (iv), the section was stained with H&E (v). (C) Albumin synthesis in lacZ-negative cells. Liver section from transplanted B6.129S7-Gtrosa26 mouse was stained with X-gal and then stained for albumin by immunohistochemistry. Asterisks indicate lacZ-negative and albumin-positive cells (i). The same section was then counterstained with Mayer's hematoxylin (ii). (D) PCR analysis of genomic DNA from transplanted liver. Genomic DNA was extracted from livers of transplanted animals (test liver), targeted ES cells, and F1 hybrid mice from crosses between B6.129S7-Gtrosa26 and 129/Sv strains of mice (host liver). DNA (2 ng) was assayed for the presence of the AFP/GFP hybrid marker gene by PCR.

To determine if GFP+ cells can engraft and differentiate into hepatocytes in vivo, 6 days after the second step of differentiation, these cells were isolated by FACS and transplanted into lacZ-positive F1 hybrid mice (B6.129S7-Gtrosa26 × 129/Sv). B6.129S7-Gtrosa26 is a transgenic mouse line that expresses E. coli lacZ in all tissues [18]. Since the parental ES cell line, CS-1, is derived from the 129/Sv strain of mice, cells derived from this ES line can be transplanted into F1(B6.129S7-Gtrosa26 × 129/Sv) hybrid mice without using immunosuppressive drugs. ES-cell-derived hepatocytes can be easily distinguished from the lacZ-positive host hepatocytes by staining for β-galactosidase. When varying numbers of differentiated GFP+ ES cells (2 × 103 – 2 × 105) were injected into partially hepatectomized mice, no teratomas were observed in >100 mice. Mice injected with 2 × 104 GFP+ cells were sacrificed 3 weeks after transplantation (four series of experiments; n = 10 for test and n = 10 for sham in each series). Livers were harvested, frozen, cryosectioned into 12- to 15-μm sections, and stained for the presence of β-galactosidase. To ensure complete staining, test sections were always stained in the same dish with liver sections from sham-transplanted mice.

β-galactosidase-negative hepatocytes were observed in liver transplanted with GFP+ cells. These cells of ∼15-20 μm in diameter were integrated into the architecture of the recipient liver, forming continuous hepatic cords with recipient liver cells (Fig. 3B). They usually formed clusters of 10-15 cells, and these clusters were not uniformly distributed throughout the livers. Therefore, random sections from the different lobes had to be first prepared and stained to identify engraftment sites. More sections from the engraftment sites were then prepared for staining. The β-galactosidase-negative hepatocytes stained positive with anti-albumin antibody, suggesting that the engrafted GFP+ cells had differentiated into mature functional hepatocytes in vivo (Fig. 3C). To further confirm the engraftment of GFP+ cells into the livers of recipient mice, the AFP/GFP hybrid marker gene was easily detected by PCR amplification of genomic DNA from whole transplanted liver (Fig. 3D). By comparing the ratio of this hybrid gene to the TPI gene in the transplanted liver and in the recombinant ES cell, we estimated that about 0.01% of the cells in the liver were derived from the transplanted GFP+ cells.

In Vivo Differentiation of GFP+ Cells into Hepatocytes

To further verify if GFP+ differentiated ES cells can differentiate into hepatocytes, 2 × 104 GFP+ cells were transplanted into partially hepatectomized ApoE- [19] or haptoglobin-[17] deficient mice (three series each; n = 5 for test and n = 5 for sham) (Fig. 4A). Three weeks after transplantation, ApoE and haptoglobin mRNA were detected in the livers of ApoE- or haptoglobin-deficient mice, respectively (Fig. 4B). The levels of both mRNAs were low but consistent with our estimated ∼0.01% engraftment efficiency of transplanted GFP+ cells in the host liver as described above. Clusters of four to five ApoE-expressing hepatocytes were clearly evident in the transplanted ApoE-deficient mice (Fig. 4C). To detect ApoE in the serum of transplanted animals by Western blot analysis, an additional series of transplantation into partially hepatectomized ApoE mice was also carried out using 2 × 105 GFP+ cells per mouse (n = 5 for test and n = 5 for sham) (Fig. 4D). The level of serum ApoE in the transplanted animals was ≤0.5% of that in the wild-type (WT) animals. Unlike humans, serum haptoglobin levels in mice are very low [17]. To facilitate detection, the mice were treated with LPS to induce upregulation of haptoglobin expression. In LPS-treated WT mice, serum haptoglobin was readily detected using 1 μl of serum. Consistent with an ∼0.01% engraftment efficiency of GFP+ cells in transplanted animals, 100 μl of serum were required to detect haptoglobin in the serum of LPS-treated transplanted animals (Fig. 4D).

Figure Figure 4..

Generation of ApoE- and haptoglobin-synthesizing hepatocytes in ApoE- and haptoglobin-deficient mice, respectively. A) Schematic diagram for detection of reconstituted ApoE and haptoglobin (Hp) in ApoE- and haptoglobin-deficient mice by the transplantation of differentiated GFP+ES cells. B) RT-PCR analysis of liver RNA. Total liver RNA was analyzed for the presence of ApoE (i) or haptoglobin (ii) mRNA in transplanted ApoE- or haptoglobin-deficient mice, respectively. TPI mRNA was used as internal control. C) Detection of ApoE+hepatocytes in transplanted ApoE-deficient mice by immunohistochemistry. Transplanted ApoE-deficient mice were perfused fixed with glutaraldehyde as described in Materials and Methods. Liver cryosections at a 5-μm thickness were prepared and stained with anti-ApoE polyclonal antibody, and using an HRP detection system, anti-ApoE antibody was detected as a brown precipitate. The sections were then counterstained with Mayer's hematoxylin that stained the nuclei blue. Hepatocytes that expressed ApoE had brown cytoplasm and blue nuclei. D) Western blot analysis: i) Two μl of serum from transplanted ApoE mice were analyzed with positive control serum from C57BL6/J (or ApoE+/+) mice and negative control serum from ApoE−/–) mice; ii) Mice have low levels of serum haptoglobin. LPS challenge increases serum haptoglobin level by 30 times. To increase the detection sensitivity, mice were challenged with LPS 24 hours before bleeding. Serum was immunoprecipitated with anti-haptoglobin antibody before analysis by Western blot hybridization.


ES cells have been described as the ultimate stem cell for tissue replacement and engineering. However, the complex and spontaneous differentiation of these totipotent stem cells into different tissues in vitro or into teratomas in vivo has severely limited practical applications of these cells. We rationalized that isolating committed tissue-specific stem cells from differentiating ES cells may eliminate the potential danger of forming aggressive ES-cell-derived teratomas in vivo, and ensure that the appropriate tissue was regenerated. Consistent with our rationale, all mice transplanted with ES cells developed teratomas, but none of 100 or so mice transplanted with differentiated ES cells developed teratomas.

Our study demonstrated that, using AFP as a marker, liver progenitor cells could be isolated from in vitro-differentiated ES cells. As was evident by the cellular morphology of β-galactosidase-negative, albumin-positive hepatocytes and their integration into hepatic cords in Rosa26 host mice after transplantation, GFP+ cells differentiated into hepatocytes in vivo. This was further confirmed by the presence of ApoE and haptoglobin mRNA in the livers of ApoE- and haptoglobin-deficient mice, respectively, after transplantation of these GFP+ cells. In addition, ApoE protein was detected in hepatocytes and serum of transplanted ApoE-deficient mice, and haptoglobin protein was detected in sera of transplanted haptoglobin-deficient mice. Together, these results confirmed that ApoE- and haptoglobin-producing hepatocytes were generated.

The levels of haptoglobin and ApoE detected in transplanted animals were not robust and clearly further experimental manipulation is needed to improve the extent of repopulation in the liver by transplanted progenitors. Nevertheless, this study demonstrates that in vitro differentiation of ES cells is potentially a viable and practical source of liver progenitor cells for both genetic and tissue engineering. Furthermore, several important issues will have to be addressed before practical applications can be considered. For instance, it remains to be determined if GFP+ cells are capable of homing specifically to the liver. The consequences of GFP+-cell engraftment into nonhepatic tissues will also have to be determined. The AFP/GFP-modified ES cells are not adequately equipped to resolve these issues because GFP expression in the ES cells is driven by AFP expression, and any further differentiation of GFP+ cells would invariably shut down the expression of GFP. Therefore, the use of a second specific marker will be necessary for assessing the homing specificity of the transplanted cells and for tracking the small number of transplanted cells. By this criterion, y chromosome detection by fluorescence in situ hybridization will not be an adequate marker.

During in vitro differentiation of ES cells to AFP+/GFP+ cells, increasingly lineage-restricted stem cells are generated. It is likely that the progenitor cells preceding AFP+/GFP+ cells are also capable of differentiating into liver cells in vivo. Consistent with this hypothesis, GFP cells from the first and second step of ES cell differentiation also gave rise to β-galactosidase-negative hepatocytes (data not shown). This observation suggested the presence of at least a more primitive liver progenitor cell than the AFP+/GFP+ cells. As the first liver progenitor cells to be purified from ES cells, the AFP+/GFP+ cells will provide a good reference for the identification of more primitive liver progenitor cells.


We thank Dr. Bing Lim (Harvard Medical School) for his critique of the manuscript and advice. S-KL thanks Dr. Lynne E. Maquat (Rochester Medical School) for advice and encouragement. This work was supported by a National Medical Research Council grant, RP6600011 to S-KL.