Direct hepatic fate specification from mouse embryonic stem cells


  • Conflict of interest: Nothing to report.


The molecules responsible for hepatic differentiation from embryonic stem (ES) cells have yet to be elucidated. Here we have identified growth factors that allow direct hepatic fate-specification from ES cells by using simple adherent monolayer culture conditions. ES cell–derived hepatocytes showed liver-specific characteristics, including several metabolic activities, suggesting that ES cells can differentiate into functional hepatocytes without the requirement for embryoid body (EB) formation, in vivo transplantation, or a coculture system. Most importantly, transplantation of ES cell–derived hepatocytes in mice with cirrhosis showed significant therapeutic effects. In conclusion, this novel system for hepatic fate specification will help elucidate the precise molecular mechanisms of hepatic differentiation in vitro and could represent an attractive approach for developing stem cell therapies for treatment of hepatic disease in humans. Supplementary material for this article can be found on the HEPATOLOGYwebsite ( (HEPATOLOGY 2005.)

Mouse embryonic stem (ES) cells are capable of differentiating into any adult animal cell type.1, 2 Although hepatocyte differentiation from mouse ES cells has been achieved3–11 the utility of ES cells as (1) a developmental model and (2) a source for pharmaceutical screening and transplantation is hampered by limited control over the differentiation process. To address this limitation, we have established a highly reproducible in vivo method to acquire abundant functional hepatocytes from ES cells through liver-regenerating animals.3 These hepatocytes express several differentiation markers of mature hepatocytes and showed sufficient functions to rescue experimental liver injury when transplanted in vivo. Our next goal was to elucidate the molecules responsible for directing this hepatic differentiation from ES cells, and to apply this information to developing a reproducible hepatic differentiation system in vitro.

To this end, we applied the results obtained from DNA-chip analysis to induce the direct differentiation of functional hepatocytes from an adherent monolayer culture of ES cells. These cells display the characteristics of mature hepatocytes apropos of liver-specific gene expression and biochemical analysis in vitro. Importantly, transplantation of monolayer-differentiated ES cell–derived hepatocytes improved liver function and prolonged the survival of mice with cirrhosis. Although the use of human ES cells for therapeutic purposes must be rigidly controlled, the procedure described in this study for inducing transplantable hepatic cells from mouse ES cells is, theoretically, transferable to human ES cells, leading to the production of a valuable source of hepatocytes to treat liver disease in humans.


ES, embryonic stem; LIF, leukemia inhibitory factor; GFP, green fluorescent protein; HIFC, hepatic induction factor cocktail; RA, retinoic acid; FGF, fibroblast growth factor; HGF, hepatocyte growth factor; PBS, phosphate-buffered saline; OsM, oncostatin M; HCM, hepatocyte culture medium; DMN, dimethylnitrosoamine; TTR, transthyretin; RT-PCR, reverse transcription-polymerase chain reaction; IGF, insulinlike growth factor, TGF, transforming growth factor; HNF, hepatocyte nuclear factor; ALB, albumin; TDO2, tryptophan 2,3-dioxygenase; AFP, alpha-fetoprotein; CK, cytokeratin; EB, embryoid body; HE, hematoxylin-eosin.

Materials and Methods

Culture of ES Cells.

pALB-EGFP/ES cells,3, 12 a J1 cell clone of 129X1/SvJ male origin, were cultured on STO feeder cells with leukemia inhibitory factor (LIF) (Funakoshi, Tokyo, Japan) (+) medium to maintain pluripotency according to Wurst and Joyner.13 After the removal of STO feeder cells, ES cells alone were plated on gelatin-coated plates (Asahi Techno Glass, Chiba, Japan). In a typical experiment, 5 to 8 gelatin-coated plates (100 mm in diameter) were seeded with 1 × 106 ES cells/plate and were cultured for further experiments with LIF (−) culture medium. Hepatic differentiation was identified by green fluorescent protein (GFP) expression, which was monitored by fluorescence microscopy (Nikon, Tokyo, Japan).

Microarray Analysis.

We followed MIAME (minimum information about a microarray experiment) guidelines for the presentation of our data. Gene expression analysis was performed between CCl4-treated and untreated mouse liver using complementary DNA microarrays (Affymetrix, Tokyo, Japan) to identify genes responsible for hepatic commitment. The procedure was conducted according to recommended protocols aspreviously described.14 Microarrays were scanned using the GeneArray scanner (Affymetrix) at 3-μm resolution, and the scanned image was quantitatively analyzed by using Microarray Suite 4.0 software (Affymetrix). For normalizing of data to compare mRNA expression levels among samples, we arbitrarily selected 1,000 as an average of AD sources corresponding to the signal intensities of all probe sets in each sample, as recommended by the manufacturer.

In the gene expression analysis of ES cell–derived GFP-positive cells, we extracted RNA from day 0 (undifferentiated ES cells), day 10− (hepatic induction factor cocktail [HIFC] untreated sample) and day 10+ (HIFC treated sample) cells and 129X1/SvJ mouse liver (8 weeks old). As a reference total RNA, we mixed equal amounts of day 0, day 10−, and day10+ RNA. Hybridization and washing were performed according to the AceGene Mouse Oligo Chip Subset A and B protocol. The fluorescence intensities were scanned using ScanArray (PerkinElmer Life Sciences) and analyzed using DNASIS Array (Hitachi Soft, Kanagawa, Japan).

In Vitro Culture.

pALB-EGFP/ES cells were cultured on gelatin-coated dishes for 3 days without feeder cells and treated for 3 days in ES cell culture medium containing LIF (100 units/mL) and 10−8 mol/L all-trans-retinoic acid (RA) (Step 1). After 3 days, the cells were passaged and cultured with LIF (−) culture medium for 5 days in the presence of fibroblast growth factor (FGF)1, 100 ng/mL; FGF4, 20 ng/mL; hepatocyte growth factor (HGF), 50 ng/mL (VERITAS, Tokyo, Japan) at 106 cells per 100-mm gelatin-coated plate (Step 2). Henceforth this mixture shall be referred to as hepatic induction factor cocktail (HIFC). Cells were then washed 3 times with cold phosphate-buffered saline (PBS) (−) and incubated for 20 minutes at 37°C in PBS containing 0.05% collagenase (GIBCO-BRL, Tokyo, Japan) and 1,000 units/mL dispase (Godoshusei, Tokyo, Japan). The dissociated cells were washed twice with serum-free Dulbecco's minimum essential medium (DMEM) and then resuspended in LIF (−) culture medium containing oncostatin M (OsM) (10 ng/mL) at 106 cells per 100 mm type I collagen-coated plate (Asahi Techno Glass) (Step 3). Two days after plating, cells were fed with hepatocyte culture medium (HCM), modified William E medium containing transferrin (5 μg/mL), hydrocortisone-21-hemisuccinate (10−6 mol/L), bovine serum albumin (0.5 mg/mL), ascorbic acid (2 mmol/L), insulin (5 μg/mL), and Gentamicin (50 μg/mL) (Sanko Junyaku, Tokyo, Japan) (Step 4). GFP gene expression was monitored by fluorescence microscopy. The GFP-positive rate was calculated as percentage of GFP-expressing cells per total cells in HIFC-treated plates. The matrices laminin, vitronectin, and fibronectin were used at 10 μg/mL, 6 μg/mL, and 1 μg/mL, respectively (Asahi Techno Glass). Mouse primary hepatocytes were cultured as described previously.15 In brief, hepatocytes were isolated from male 129X1/SvJ mice weighing 20 g by means of the collagenase S-1 (Nitta Gelatin, Tokyo, Japan) perfusion method. The dispersed cells were suspended in HCM at 1.0 × 107 cells/mL and plated onto collagen-coated plates (Asahi Techno Glass) and cultured for a week before use. Between 95% and 97% of the harvested cells were viable and positive for albumin gene expression.

Animal Treatment and Cell Transplantation.

Ten-week-old female 129X1/SvJ mice and BALB/c nude mice (SLC, Tokyo, Japan) were used in all studies. 1.0% dimethylnitrosamine (DMN) (Wako) dissolved in saline was injected intraperitoneally 3 consecutive days per week for 4 weeks into female 129X1/SvJ mice, and ES cell–derived GFP-positive hepatocytes (5 × 106 cells/0.1 ml/mouse) or control PBS(−) was given intravenously at day 28 after the initiation of DMN treatment. ES cell–derived GFP-positive hepatocytes were isolated by fluorescence-activated cell sorting. Each experimental group contained 10 mice. The mice were sacrificed periodically and dissected, and their livers were harvested and examined for the presence of GFP-positive fractions. Histological analysis of liver tissues was conducted by serial tissue section at 1 and 21 days after cell transplantation. Both plasma fibrinogen and albumin expression were measured.

The firefly luciferase-expressing pALB-EGFP/ES cell line carrying pEGFPLuc plasmid DNA (CLONTECH, Tokyo, Japan) was established and used for in vivo imaging analysis of the fate of transplanted ES cells into mice with liver injury. In vivo bioimaging was conducted in a cryogenically cooled IVIS system (Xenogen), using living image acquisition and analysis software (Xenogen, Alameda, CA). The amount of light generated was directly related to the amount of luciferase-producing cells. The IVIS system could distinguish as few as 300 luciferase-positive ES cells in vivo.15

Electron Microscopy.

To detect the hepatic cells differentiated from ES cells, GFP-positive fractions were observed by ordinary electron microscopy, as described previously.3

Immunohistochemical and Reverse Transcription Polymerase Chain Reaction Analysis.

Cultured cells were washed twice with PBS(−) and fixed in 4.0% formaldehyde at 4°C for 10 minutes. Rehydrated sections were treated with 3.0% H2O2 in methanol for 15 minutes followed by incubation with 5% fetal calf serum in PBS for 30 minutes at 4°C. The cells were then analyzed by immunohistochemistry with transthyretin (TTR) (1:100) and CK18 (1:100) antibodies overnight at 4°C. The rhodamine isothiocyanate–conjugated secondary antibody (1:5,000) was applied for 30 minutes followed by incubation with 5% fetal calf serum. All antibodies were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). An aliquot of total RNA isolated from undifferentiated ES cells and GFP-positive cells using ISOGEN solution (Nippon Gene, Tokyo, Japan) was treated with DNase I (amplification grade; TaKaRa, Kyoto, Japan) according to the manufacturer's guidelines. Reverse transcription-polymerase chain reactions (RT-PCR) reactions were performed by using a One-Step RT-PCR kit (QIAGEN, Tokyo, Japan). Primers and PCR details are available in Table 1.

Table 1. Primers and Conditions for RT-PCR
GenBank Accession No.PrimersAnnealing Temperature (°C)Cycles

Real Time-PCR Analysis.

RNA expression normalized to the housekeeping gene GAPDHwas measured by real-time PCR using the iCycler System (BIO-RAD, Tokyo, Japan) and Real-Time Detection System Software (Version 2.1). Real time-PCR monitoring was performed by adding the double-stranded DNA dye, SYBR Green I.

Biochemical Analyses.

One day after plating at 2 × 105 cells/60-mm dish, ES cell–derived GFP-positive hepatocytes or control ES cells and normal mouse hepatocytes were analyzed for glucose levels in the culture supernatant by the glucose oxidase method, as described previously.16 To examine the cellular activity of ammonia detoxification, GFP-positive cells or control ES cells and normal mouse hepatocytes were cultured at 2 × 105 cells/60-mm dish in 1.0 mL DMEM containing 2.5 mmol/L NH4Cl and further incubated for 24 hours. The culture media were tested for concentration of NH4Cl at 0, 6, 12, and 24 hours by Ammonia-Test Wako (Wako Pure Chemicals, Tokyo, Japan). To assay urea synthesis ability, cells were cultured with Hank's balanced salt solution in the presence of 5 mmol/L NH4Cl. The medium was harvested after incubation for 0, 2, 4, and 6 hours and assayed according to the manufacturer's guidelines.

Blood samples were obtained periodically and centrifuged for 20 minutes at 5,000 rpm, and the serum was collected. Serum samples were tested for albumin concentration using the Albumin II-HA Test Wako kit (Wako Pure Chemicals). Plasma fibrinogen concentration was measured using the Drihemato Fib kit (Wako Pure Chemicals).

Statistical Analysis.

The results are given as mean ± SEM. The Student t test was performed for statistical evaluation, with P < .05 considered significant. Where more than 2 samples were compared, the one-way ANOVA statistical analysis was performed, and results are given as ±SEM.


Identification of Hepatic Induction Factors.

In vivo imaging analysis of luciferase-expressing mouse ES cells showed that transplanted ES cells homed specifically to the injured liver (Fig. 1A) and subsequently differentiated into GFP-positive cells (Fig. 1B, inset). As previously reported, tumor formation was observed in the liver of all CCl4-treated mice.3 Analysis of the tumors showed that 14% to 28% of the tissue comprised mature, transplanted hepatocytes. In contrast, transplanted mice with normal liver showed no such effects. These results indicate that the injured liver produces key regulatory factors that are necessary for the homing of ES cells to the site of liver injury and, additionally, for supporting hepatic differentiation. These observations led us to a rational design of differential-gene expression analysis between CCl4-treated and untreated mouse liver. To this end, DNA microarray technology was used to identify genes responsible for hepatic differentiation. Initially, we focused specifically on induced growth factor genes, based on the pivotal role that growth factors appear to play in the differentiation of several cell types from ES cells (Table 2). Twelve species of growth factor genes, whose expression levels were undetectable in normal liver, were newly induced 24 hours after CCl4-treatment; FGF-3, -4, -5, -8, -10, -13, -18, HGF, nerve growth factor-β (NGF-β), insulin-like growth factor (IGF)-2, transforming growth factor (TGF)α, and TGFβ2. In addition, the expression level of a further 5 species of growth factor genes were significantly increased in the regenerated liver (FGF-1: 1.8-, mast cell growth factor: 2.1-, hepatoma derived growth factor: 2.2-, IGF-1: 5.2-, and TGFβ1-4047μ: 1.8-fold). Furthermore, growth factor receptors for FGF (FGFR1, 3, and 4), IGF (IGF1R and 2R), oncostatin M receptor, and TGFβ (TGFβ1R and β2R) were also significantly upregulated.

Figure 1.

Hepatic differentiation from ES cells. (A) In vivo imaging analysis of luciferase-expressing mouse ES cells transplanted into CCl4-treated (+) (n = 3) or CCl4-untreated (−) (n = 3) mice. All images shown are the visible light image superimposed on the optical CCD image with a scale in relative light units/min as shown. (B) Tumor formation was observed in the livers of CCl4-administered mice at 3 weeks after the administration of CCl4. The inset shows ES-derived tumors containing abundant GFP-positive cells (scale bar: 2.5 mm). Hepatocytes differentiated from pALB-EGFP/ES cells were recognized as GFP-positive cells.3 No tumors were observed in the livers of the control mice without CCl4-treatment. Scale bars represent 5.0 mm. ES, embryonic stem; GFP, green fluorescent protein.

Table 2. GeneChip Analysis of Change-up in Growth Factor and Receptor Gene Expression in CCl4-Treated Mouse Liver
Genbank Accession No.GeneFold Change
  1. NOTE. Microarray analysis used human U95A oligonucleotide probe arrays (12,626 transcripts). The asterisk mark (*) indicates new expression og gene in CCl4-treated mouse liver. The numbers-fold increase for gene expression was measured by analyzer software (Affymetrix).


Hepatic Differentiation of Mouse ES Cells in an Adherent Monolayer Culture.

The 12 growth factors upregulated during liver regeneration were selected to investigate their effects on the hepatocyte differentiation from ES cells in adherent monoculture conditions. Undifferentiated pALB-EGFP/ES cells3 whose GFP expression is driven by the mouse albumin promoter/enhancer were initially cultured on gelatin-coated plates without feeder cells and treated with ES cell culture medium containing LIF and RA for 3 days. At day 3, there were no GFP-positive cells detectable in the culture (data not shown). In the second stage of culturing, LIF and RA were removed from the medium, and cells were transferred to new gelatin-coated plates and exposed to various growth factors. Using the above culture system, 12 different recombinant growth factors were initially added individually to the culture medium (ESM)3 during stage 2. Hepatic differentiation was directly identifiable as the GFP-positive cell fraction appearing, in some cases, as early as day 5 of culture (Fig. 2A). Among the 12 individual growth factors assayed, those that increased the numbers of GFP-positive cells were subjected to further analysis of synergistic effects in combination with other growth factors (Fig. 2B). We discovered that cells treated with a combination of FGF1, FGF4, and HGF, each of which is upregulated in the regenerating liver, consistently produced the most dramatic increase in GFP-positive cell numbers (22.9 ± 5.8%), which were scattered throughout the monolayer culture (Fig. 3C). Thus, the application of DNA-chip technology to our system provided an effective platform for identification of factors sufficient for establishment of direct hepatic fate specification from ES cells in vitro. OSM is thought to be effective for hepatic maturation,17–19 and, interestingly, OSM receptor was upregulated in our DNA microarray analysis. Five days after growth factor treatment and hepatic induction, OsM was assayed for its effect on hepatic differentiation by addition to the culture medium. OsM increased both GFP intensity and albumin production concomitantly in the GFP-positive cell fraction. However, the actual number of GFP-positive cells was not affected (data not shown).

Figure 2.

Optimization of ES cells differentiation into hepatocytes. (A) Thirteen individual growth factors were tested for their ability to direct differentiation of hepatocytes from ES cells in adherent monoculture conditions (n = 5). (B) Among 11 different combinations, cells treated with FGF1, FGF4, and HGF produced the most significant numbers of GFP-positive cells (n = 5). (C) Expression of FGF1, FGF4, HGF, OsM, and p450 was analyzed by RT-PCR in the livers of CCl4 (+) and (−) mice. Mouse genomic DNA from liver was used as a negative control. ES, embryonic stem; FGF, fibroblast growth factor; HGF, hepatocyte growth factor; GFP, green fluorescent protein; NGF, nerve growth factor; EGF, epidermal growth factor; TGF, transforming growth factor; OSM, oncostatin M; IGF, insulinlike growth factor.

Figure 3.

Hepatic differentiation of mouse ES cells in adherent monoculture. (A) Kinetic analysis of GFP-positive cells cultured on a plate coated with several types of matrices. Data represent percentage of GFP-positive cells in total culture cells (N≥3). (B-D) Morphological changes and induction of GFP expression during the course of hepatocyte differentiation from ES cells: untreated (day 0) (B); the end of Step 2 (day 8) (C), 8 days after Step 4 (day 18) (D). Scale bars represent 50 μm. (E) An integrated schematic representation of the differentiation protocol for the induction of hepatocytes from ES cells in monolayer culture, steps 1 to 4 (see Materials and Methods) (n = 5). ES, embryonic stem; GFP, green fluorescent protein; LIF, leukemia inhibitory factor; RA, retinoic acid; FGF, fibroblast growth factor; HGF, hepatocyte growth factor; OsM, oncostatin M; HCM, hepatocyte culture medium; ESM, ES medium.

Confirmatory analysis of gene expression profiles for FGF1, FGF4, HGF, and OsM in the regenerating liver were peformed by RT-PCR (Fig. 2C). Each of the assayed growth factor genes was detected and found to be upregulated in the regenerating liver as compared with the control, normal liver, thus verifying the data obtained by DNA-chip analysis. Taking these data together, we have elucidated a subset of growth factors sufficient for hepatic cell differentiation from ES cells in vitro, namely, FGF1, FGF4, and HGF.

Next, matrix-dependent hepatic cell growth was examined. After culturing with HIFC for 5 days on gelatin-coated dishes, the cells were trypsinized and replated on various matrices. Among 5 different extracellular matrix components assayed, type I collagen produced the highest yield of GFP-positive cells; 31.5 ± 6.3% at 7 days after culturing in the presence of OsM (Fig. 3A). GFP-positive cells production on the other matrices was as follows: gelatin 13.3 ± 2.0%; laminin at 24.4 ± 4.0%; fibronectin at 7.3 ± 0.3% and vitronectin at 0.7 ± 0.3%. These data show that the growth and enrichment sensitivity of ES cell–derived hepatocytes is sensitive to extracellular matrix component type and indicate that type I collagen may be an optimal substrate for hepatocyte induction

Thus, 2 days after OsM treatment, ES cell–GFP-positive cells were cultured with a serum-free HCM on type I collagen-coated plates for 3 weeks. Within 24 hours of replating, proliferating ES cell–derived GFP-positive cells were identified. Cells in these colonies proliferated 2 to 5 times in the first week and retained their GFP expression for a further 3 weeks in culture (Fig. 3D). Furthermore, the use of a serum-free HCM eliminated serum-dependent cells, nearly all of which were nonhepatic. After 2 weeks in culture, ES cell–derived GFP-positive cells stopped proliferating, even when split to a lower density, as might be expected for terminally differentiated hepatocytes. Thus, HCM culture conditions facilitate the enrichment of ES cell–derived GFP-positive cells up to 80% in culture. In summary, the following steps outline the strategy for maximal hepatic cell induction from ES cells: Step 1, RA and LIF treatment of ES cells for 3 days on gelatin-coated plates (endoderm induction stage); Step 2, HIFC treatment for 5 days on gelatin-coated plates (hepatic induction stage); Step 3, OsM treatment for 2 days on collagen-coated plates (hepatic maturation stage); Step 4, culture in serum-free HCM on collagen-coated plates (hepatocytes selection stage) (Fig. 3E). This differentiation program is highly efficient, with approximately 1 × 106 GFP-positive cells produced from 1 × 105 undifferentiated ES cells.

Characterization of GFP-Positive Cells.

Microscopic analysis of ES cell–derived GFP-positive cells showed a hepatocyte-like morphology, with binucleate cells frequently observed (Fig. 4A-B). Electron microscopic analysis showed that large glycogen areas were often observed in epithelial cells of bile duct–like structures, which are abundant in the GFP-positive fractions (Fig. 4C). Additionally, ultrastructural characters of mitochondria (Mt) in the liver progenitor–like cells resemble those of mature hepatocytes. Immunogold particles indicating GFP molecules were detected in the cells containing bile canaliculi and peroxisomes (data not shown). We next ascertained whether ES cell–derived GFP-positive cells exhibit hepatocyte-specific gene expression profiles. ES cell–derived GFP-positive cells were collected by fluorescence-activated cell sorting, and RT-PCR analysis of sorted GFP-positive cells at Step 3 (see above) confirmed that the population expressed the liver-specific marker and enzyme genes: ALB, TDO2, TTR, TAT, G6P, CK8, Lst-1, CPS1, PEPCK, and p450 (CYP1A1), and transcriptional factor genes: hepatocyte nuclear factor (HNF)-3β (FoxA2) and -4α (see supplemental Fig. 1). This expression pattern is similar to that of primary hepatocytes. Importantly, GFP-positive cells were negative for c-kit and ERas gene expression which are both markers of undifferentiated ES cells and important for their tumor-like growth characteristics. The ES cell–derived GFP-positive cells also were negative for alkaline phosphatase activity, also a marker for undifferentiated ES cells (data not shown). In addition, CK19, an intrahepatic bile duct cell (cholangiocyte) marker, was not detected in ES cell–derived GFP-positive cells (data not shown), indicating that the ES cell–derived GFP-positive population does not contain other liver cell types. Kinetic and quantitative PCR analysis showed that albumin (ALB) gene expression was induced on the first day of Step 2 (see supplemental Fig. 1). Quantitative levels of ALB and tryptophan 2,3-dioxygenase (TDO2) mRNA expression were normalized to the internal reference gene GADPH at day 0, day 6, and day 10. ALB mRNA concentration increased gradually from day 6 to day 10, and TDO2 expression was detectable at day 10, but not at day 0 or day 6, by real-time PCR analysis. The level of ALB mRNA at day 6 was approximately 20% of levels in ES cell–derived GFP-positive cells at day 10. Before ALB expression, alpha-fetoprotein (AFP) gene expression was identified on the second day of Step 1, reached maximal expression levels at the end of Step 1, and diminished thereafter. TDO2, a specific marker of mature hepatocytes, was identified in the late stage of Step 2, reached maximal expression levels at Step 3, and subsequently retained its expression in HCM for several weeks. Immunohistochemical staining for hepatocyte-specific proteins indicated that HNF3β was positive at day 3. GFP-positive cells were double positive for cytokeratin (CK)18/ALB and TTR/ALB at day 10 (Fig. 5). We next compared liver-specific gene expression profiles between ES cell–derived hepatocytes and normal mouse liver by DNA-chip analysis, using microarrays from Mouse Genome Informatics ( Differentiated, ES cell–derived hepatocytes and normal mouse liver revealed a 98% correspondence of gene expression (number of genes assayed = 88) (see supplemental Table 1). Biochemical analyses indicated that cultured GFP-positive cells display glucose-producing ability, capacity to clear ammonia from the culture media, and urea synthesis ability, characteristics of mature hepatocytes (see supplemental Fig. 2). Moreover, levels of glucose and urea produced by ES cell–derived GFP-positive cells were similar to those produced by monolayer cultures of primary hepatocytes. Conversely, HIFC-untreated ES cells did not produce glucose and urea and lacked the capacity to clear ammonia from the culture media. Together, these biochemical data indicate that our ES cell–derived GFP-positive cells are functional hepatocytes. Karyotyping of more than 40 ES cell–derived GFP-positive cells by G-banding analysis showed that all cells had the normal chromosome number (data not shown). In addition, no teratoma formation was observed when 1 × 106 GFP-positive cells cultured for 1 week were subcutaneously injected into BALB/c nude mice (n = 6) and 129X1/SvJ mice (n = 4). Thus, hepatocytes differentiated from ES cells by our culture system display the phenotypic and biochemical characteristics and functional activity of normal mature hepatocytes.

Figure 4.

Morphological characteristic of GFP-positive cells. (A) Arrow indicates binucleate cell in phase contrast. (B) Fluorescence image. Scale bars represent 50 μm. (C) Electron microscopy of the GFP-positive fractions. Gl, glycogen areas; Mt, mitochondria; bc, bile canaliculi; N, nuclear; L, lipid granule. Scale bars, 0.5μm. GFP, green fluorescent protein.

Figure 5.

Immunohistochemical analysis of GFP-positive cells. (A, D) Phase contrast of ES cell–derived-hepatocytes. (B, E) Fluorescent images of isolated ES cell–derived GFP-positive hepatocytes in hepatocyte culture medium and analyzed at 2 weeks after step 3. (C, F) Immunofluorescent images of CK18 (C) and TTR (F) staining in ES cell–derived ALB-positive hepatocytes. ES, embryonic stem; GFP, green fluorescent protein; TTR, transthyretin.

Transplantation of GFP-Positive Cells Into Mice With Cirrhosis.

Finally, to address our ultimate goal of examining whether these hepatocytes are therapeutically applicable, we transplanted 5 × 106 GFP-positive cells per mouse into mice with DMN-induced cirrhosis. The transplanted ES cell–derived GFP-positive cells were immediately diffused through the liver and integrated adjacent to the fibrotic region, along with the expression of GFP (Fig. 6A-B, arrows). By counting the number of GFP-positive cells in 100 tissue sections, the number of GFP-positive cells that migrated into the host liver was estimated to be at least 1 × 106 cells per mouse. Histological analysis indicated that transplantation of ES cell–derived GFP-positive cells significantly suppressed the onset of fibrosis/cirrhosis in mice (Fig. 6C-E). Survival analysis showed that mice receiving GFP-positive cells survived for at least twice as long as control mice injected with PBS(−) alone (Fig. 7A). In addition, plasma fibrinogen and albumin concentrations were increased 1.6- and 1.3-fold, respectively, over those of the control mice administered with PBS(−) (Fig. 7B-C). No teratoma or liver tumor formation was observed at 80 weeks after the transplantation of GFP-positive cells into either normal 129X1/SvJ or liver-damaged mice (n = 10) (Fig. 8). Thus, our ES cell–derived hepatocytes functioned therapeutically in vivo, and their transplantation ameliorated the effects of DMN-induced cirrhosis.

Figure 6.

Transplantation of ES cell–derived GFP-positive cells into mice with DMN-induced cirrhosis. (A) Hematoxylin-eosin (HE) staining of liver sections 1 day after the transplantation of ES cell–derived GFP-positive hepatocytes. (B) Grafted cells, indicated by arrows in (A), were detected by GFP fluorescence in the same section. (C) HE staining of liver section from control DMN-treated mice administered with saline (21 days after treatment). (D) HE staining of liver section from DMN-treated mice administered with ES-derived hepatocytes (21 days after transplantation). (E) HE staining of liver sections from age-matched, untreated normal mice. Scale bars represent 10 μm (A, B) and 50 μm (C-E). ES, embryonic stem; GFP, green fluorescent protein; DMN, dimethylnitrosoamine.

Figure 7.

In vivo functions of ES cell–derived hepatocytes. (A) Survival curve of mice with DMN-induced cirrhosis. The arrow indicates time of transplantation of ES cell–derived GFP-positive hepatocytes (open circles: mice administered with ES cell–derived GFP-positive hepatocytes; closed circles: control mice administered with saline). Plasma fibrinogen (B) and albumin (C) concentrations were measured in the same mice. N = 10 in each experimental group. ES, embryonic stem; GFP, green fluorescent protein; DMN, dimethylnitrosoamine.

Figure 8.

Transplantation of ES cell–derived GFP-positive cells into mice with cirrhosis at 80 weeks. (A) No tumor was observed in the transplantation of ES cell–derived hepatocytes into mice with DMN-induced cirrhosis (n = 10). (B) Hematoxylin-eosin (HE) staining of a serial section from A. ES, embryonic stem; GFP, green fluorescent protein; DMN, dimethylnitrosoamine.


Recent reports have highlighted the differentiation of hepatocytes from ES cells in vitro5, 6, 8–11 and in vivo.3, 4, 7 To elucidate the molecular mechanisms underlying the development of the liver, previous reports have attempted the differentiation of ES cells into hepatocytes by forming embryoid bodies (EBs) in vitro. However, because at least some of the cells of EBs are not terminally differentiated, this material is not useful for transplantation. Moreover, EB differentiation is not a scalable process. Nevertheless, none of these articles have clarified the precise molecular mechanisms of hepatic induction, especially in regards to what kind of growth signals are required for hepatocyte differentiation from ES cells. To address this question, we deduced the growth factors that direct hepatic fate specification and applied them to establish a methodology for the direct differentiation of functional hepatocytes from an adherent monoculture of ES cells. These cells display the characteristics of mature hepatocytes with respects to liver-specific gene expression and functionality in vitro. More importantly, transplantation of our ES cell–derived hepatocytes improved liver function and prolonged survival in a clinically relevant model of cirrhosis. We had previously shown that hepatic cells were efficiently induced when ES cells were transplanted into mice after liver injury.3 Although these hepatocytes are chromosomally normal, for clinical application of ES cell–derived hepatocytes, in vitro differentiation of hepatocytes from ES cells is imperative from a quality control standpoint. In the current study, we have advanced to successfully induce the direct differentiation of functional mature hepatocytes without the use of animals or the formation of EB, which are, hence, more suitable for clinical use. Although the application of human ES cells for therapeutic purposes is currently subject to regional regulatory restrictions, the induction of transplantable hepatic cells from mouse ES cells described here may be instantly transferable to human ES cells, leading to the production of human hepatocytes. In support of this, differentiation of hepatocytes from Cynomolgus monkey ES cells is achievable using HIFC (T. Teratani, unpublished observation).

The immortality and rapid growth of ES cells are attractive features for their use in stem cell therapies.20–22 However, the finding that ES cells produce teratomas when transplanted in syngenic animals23 is likely to preclude their therapeutic usage.24–26 In our system, although serum-free HCM allowed enrichment of hepatocytes to almost 80% in the culture dish, the remaining 20% fraction contains non-hepatic cells, including other types of differentiated cells and, potentially, immature ES cells. However, our ES cell–derived hepatocytes neither formed skin tumors when subcutaneously injected into athymic nude mice and 129X1/SvJ mice nor displayed anchorage independence in soft agar. Additionally, ERas message, the transforming oncogene important in the teratoma-forming ability of ES cells,27 was undetectable in the ES cell–derived hepatocytes. Nevertheless, although cell sorting may increase the purity of hepatocytes up to 98%, there is a finite risk that residual undifferentiated ES cells could form teratomas when they are implanted into a recipient. Such a risk, when balanced against the therapeutic benefits of stem cell therapy, might be managed by the use of human ERas-knockout human ES cells.

It was recently reported that bone marrow–derived adult stem cells differentiated into hepatocyte-like cells in vitro.28 However, bone marrow–derived adult stem cells may also repair damaged liver by cell fusion within the host liver and not by converting directly into hepatic cells.29–31 Although possibly induced fusion of cells may be therapeutically exploited to achieve the rescue of damaged liver tissue, problems are likely to arise because the fusion mechanisms are not fully understood. In this regard, it may be preferable to use ES cells as a source of functional hepatocytes rather than adult stem cells for treating human diseases. Finally, during mouse liver development, AFP and ALB are differentially regulated. AFP synthesis starts soon after fertilization, whereas the earliest time ALB can be detected is at day 17 or 18.32, 33 In our system, HIFC-stimulated AFP mRNA expression preceded ALB mRNA, indicating that ES cell–derived hepatocytes mimic the normal liver developmental program. Additionally, during embryonic liver development, late stage, hematopoietic cells produce OsM, which induces maturation of mouse fetal hepatocytes17, 18 and induces the differentiation of cultured human fetal hepatocytes.19 In our system, OsM was critical for G6P and TDO2 expressions at late stage and an increase of GFP intensity. Thus, our system was shown the supporting of these papers. Furthermore, HNF3β (FoxA2) expression is important for the endoderm lineage,34 and HNF4α is essential for morphological and functional differentiation of hepatocytes, accumulation of hepatic glycogen stores, and generation of the hepatic epithelium.35 These important transcription factor genes, for HNF4α and HNF3β (FoxA2), were expressed in our system through to the final differentiation stage. Thus, our in vitro hepatic cell induction system appears to mimic in vivo hepatic development and will therefore be useful for studying regulatory mechanisms of hepatocyte-specific and liver-enriched transcription factor gene expression.

In conclusion, this study documents an experimentally deduced combination of growth factors and matrixes that is sufficient for reproducible and efficient hepatic differentiation leading to direct induction of mature functional hepatocytes from ES cells in adherent monoculture conditions, with therapeutic efficacy. Our system will be a valuable tool for studying the molecular basis of the developmental processes influencing hepatic cells in vitro and bring us a step closer to establishing a safe and effective stem cell therapy to treat hepatic failure in vivo.


The authors thank Yusuke Yamamoto, Ayako Inoue, Kimi Honma, Maho Kodama, Shinobu Ueda, Akemi Sugai, and Masako Hosoda for their excellent technical assistance.