Efficient Differentiation of Hepatocytes from Human Embryonic Stem Cells Exhibiting Markers Recapitulating Liver Development In Vivo



The potential to differentiate human embryonic stem cells (hESCs) in vitro to provide an unlimited source of human hepatocytes for use in biomedical research, drug discovery, and the treatment of liver diseases holds great promise. Here we describe a three-stage process for the efficient and reproducible differentiation of hESCs to hepatocytes by priming hESCs towards definitive endoderm with activin A and sodium butyrate prior to further differentiation to hepatocytes with dimethyl sulfoxide, followed by maturation with hepatocyte growth factor and oncostatin M. We have demonstrated that differentiation of hESCs in this process recapitulates liver development in vivo: following initial differentiation, hESCs transiently express characteristic markers of the primitive streak mesendoderm before turning to the markers of the definitive endoderm; with further differentiation, expression of hepatocyte progenitor cell markers and mature hepatocyte markers emerged sequentially. Furthermore, we have provided evidence that the hESC-derived hepatocytes are able to carry out a range of hepatocyte functions: storage of glycogen, and generation and secretion of plasma proteins. More importantly, the hESC-derived hepatocytes express several members of cytochrome P450 isozymes, and these P450 isozymes are capable of converting the substrates to metabolites and respond to the chemical stimulation. Our results have provided evidence that hESCs can be differentiated efficiently in vitro to functional hepatocytes, which may be useful as an in vitro system for toxicity screening in drug discovery.

Disclosure of potential conflicts of interest is found at the end of this article.


Human embryonic stem cells (hESCs) are able to replicate indefinitely and to differentiate into most, if not all, cell types of the body, thereby having the potential to provide an unlimited source of cells for a variety of applications. These include regenerative medicine for a broad spectrum of human diseases, elucidating mechanisms that underlie cell fate specification, and as in vitro models for determining the metabolic and toxicological properties of drug compounds [1, 2]. Many of these applications require efficient and regulated differentiation of hESCs to specific cell types. Hepatocytes, the primary cells of the liver, have attracted particular attention, as the liver plays a central role in multiple functions of the human body and liver failure is often only treatable by liver transplantation. Unfortunately, the number of donors is insufficient to cope with the growing demand for transplantation. Hepatocyte transplantation, to increase the number of functional hepatocytes, could be employed as an alternative therapeutic approach to whole organ transplantation for liver failure. Stem cell-derived hepatocytes could also be utilized for extra-corporeal support devices in the case of acute liver failure [3]. In addition to their therapeutic potential, human hepatocytes are valuable for assessing the toxicity of new drugs, as liver is the primary tissue involved in the metabolism of drug compounds and is among the most common tissues affected by drug toxicities. Primary human hepatocytes suffer from not only inconsistent availability but also significant phenotypic and genotypic variability. Although animal models have wide application in preclinical drug development, they are often not predictive for humans.

Several studies have reported the differentiation of hepatocyte-like cells from hESCs [4, [5]–6]. However, these papers have mainly focused on characterizing the final hESC-derived hepatocytes and lack of evidence that hepatocyte differentiation followed the liver developmental process in vivo. The liver, similar to the pancreas, develops from the primitive gut tube, which is formed by a flat sheet of cells, called definitive endoderm [5, 7]. Definitive endoderm is one of the three germ layers derived from the epiblast of the inner cell mass of the blastocyst and is generated during the gastrulation stage of embryogenesis. During gastrulation, cells from specific regions of the epiblast are recruited to form the primitive streak where they transform and give rise to both mesoderm and definitive endoderm [8, 9]. The definitive endoderm at the anterior ventral segment of the gut tube interacting with cardiogenic mesoderm becomes more proliferative and forms the liver bud where cells are referred to as hepatoblasts [10, 11]. The hepatoblasts proceed through a series of maturation steps that accompany autonomous proliferation, cellular enlargement, and functional maturation as the liver develops. During this process, cells express certain characteristic genes that represent cellular development with the primitive streak, with definitive markers at early developmental stages being replaced by hepatic markers at later stages.

After the emergence of the liver bud from the developing gut tube, the level of hepatic maturation is characterized by the expression of liver- and stage-specific genes [12]. For example, α-fetoprotein (AFP) is an early hepatic marker, expressed by hepatoblasts in the liver bud until birth, when expression is dramatically reduced [13]. In contrast, albumin, the most abundant protein synthesized by hepatocytes, is initially expressed at lower levels in early fetal hepatocytes but this increases as the hepatocytes mature, reaching a maximum in adult hepatocytes [14]. Furthermore, isoforms of cytochrome P450s (CYPs) proteins also exhibit differential expression levels according to developmental stages of the liver. CYP3A7 is mainly expressed in human fetal liver, whereas CYP3A4 is the predominant isoform in adult hepatocytes [15].

Previously we developed a method to differentiate hESCs to hepatocyte-like cells [16]. In the present study, we have improved the procedure further and have significantly increased hepatocyte production from hESCs. We show that by priming hESCs to differentiate through the primitive streak mesendoderm to definitive endoderm prior to treatment with dimethyl sulfoxide (DMSO) increases the proportion of hepatocytes in the end population to approximately 70%. Furthermore, we show that the gene expression profile during this differentiation process recapitulates that of in vivo development and the derived hepatocytes performed multiple liver functions.

Materials and Methods

Cell Culture and Differentiation

Human embryonic stem cells H1 and H7 were cultured and propagated in matrigel coated plates with mouse embryonic fibroblast conditioned medium (MEF-CM) supplemented with basic fibroblast growth factor in feeder-free, serum-free conditions as previously described [17, [18]–19].

The differentiation was initiated, when hESCs reached a confluence level of approximately 50%–>70%, by replacing the MEF-CM with priming medium A (RPMI1640 containing 1× B27 (both from Invitrogen, Paisley, U.K., http://www.invitrogen.com), 1 mM sodium butyrate (NaB) (Sigma-Aldrich, Dorset, U.K., http://www.sigmaaldrich.com) and 100 ng/ml activin A (PeproTech, London, http://www.peprotech.com). After 24–48 hours, the medium was changed to priming medium B, which is the same as priming medium A, except the concentration of sodium butyrate was reduced to 0.5 mM and cells were cultured for a further 48–72 hours. The cells were then split 1:2 to new matrigel coated plates and cultured in differentiation medium (serum replacement (SR)/DMSO medium: knockout-Dulbecco's modified Eagle's medium (DMEM) containing 20% SR, 1 mM glutamine, 1% nonessential amino acids, 0.1 mM β-mercaptoethanol; all from Invitrogen; and 1% DMSO; Sigma), for 7 days. Finally, the cells were cultured in maturation medium (CL15 medium [20]: L15 medium supplemented with 8.3% fetal bovine serum, 8.3% tryptose phosphate broth, 10 μM hydrocortisone 21-hemisuccinate, 1 μM insulin; all from Sigma, and 2 mM glutamine) containing 10 ng/ml hepatocyte growth factor (HGF) and 20 ng/ml oncostatin M (OSM) (both from R&D systems, R&D Systems Inc., Minneapolis, http://www.rndsystems.com). The medium was changed daily during differentiation.

RNA Isolation and Reverse Transcription–Polymerase Chain Reaction

Total RNA was isolated using RNeasy mini kit (Qiagen, West Sussex, U.K., http://www1.qiagen.com) following manufacturer's instruction and DNA was removed by the treatment with DNase (Qiagen). cDNA was synthesized using 2 μg total RNA with reverse transcriptase (Roche Diagnostics, Burgess Hill, U.K., http://www.roche-applied-science.com) in a 20–25 μl volume. Polymerase chain reaction (PCR) was carried out as previously described [21]. Primer sequences and PCR conditions are provided in supplemental online data Table 1.

Immunoblotting, Immunocytochemistry and Flow Cytometry

Cells were lysed and sonicated in sodium dodecyl sulfate (SDS) buffer and the lysates were separated in 8% polyacrylamide gels containing SDS, then transferred to nitrocellulose membrane. Western blotting was carried out as previously described [16]. For immunostaining, cells were fixed in 4% paraformaldehyde for 10 minutes at room temperature and washed with phosphate-buffered saline (PBS). After blocking with PBS containing 10% goat serum, cells were incubated with primary antibody at 4°C overnight, followed by 30 minutes incubation with appropriate secondary antibody at room temperature. The primary and secondary antibodies used are commercially available except CYP3A and CYP2D6, which are from Prof. R. Wolf's laboratory. For flow cytometry, cells were harvested and incubated in blocking buffer containing 40% goat serum for 30 minutes followed by incubation with CXCR4-PE antibody following manufacturer's instruction. The cells were analyzed in Beckon Dickinson flow cytometer after washing. All the antibodies and conditions are listed in supplemental online data Table 2. The controls were performed by replacing the primary antibodies with corresponding immunoglobulin G (IgG) antibodies and they were negative. The counting on albumin and Hepar one staining were done on 300–400 cells in four random chosen fields.

Measurement of Hepatocyte Export Proteins

Cells were cultured in 6-well plate in 3 ml appropriate media and after 24 hours supernatants were collected for export protein assay. The hepatic export proteins α-2-macroglobulin (A2M), haptoglobin, fibrinogen, and fibronectin were measured using sandwich enzyme-linked immunosorbent assays (ELISA) essentially as described by the manufacturer (DAKO, Ely, U.K., http://www.dako.co.uk). High binding EIA plates (Corning, Koolhovenlann, Netherlands, http://corning.com/index.aspx) were coated with rabbit anti-human antibody (DAKO) to each specific protein overnight at 4°C. Antibodies were diluted 1:1000 for A2M and fibronectin, 1:2000 for haptoglobin, and 1:10000 for fibrinogen. Sample supernatants were diluted 1:10 and then added to 96-well plate in triplicates. The plates were incubated for 2 hours at room temperature. Peroxidase-conjugated rabbit anti-human antibody (DAKO) directed against the appropriate protein was added and the plates were incubated for 1 hour at room temperature. The substrate o-phenylenediamine was added and the reaction stopped with 0.5M sulfuric acid. The plates were read at 490 nm with a reference wavelength of 630 nm using a MRX II plate reader (Dynatech, Billinghurst, U.K.) and the concentration of the appropriate protein in each sample was calculated from standard curves using the MRX II Endpoint software. Analysis of significance between variables was performed using the paired two-tailed t-test. A difference was considered significant using 95% confidence intervals (p < .05).

Measurement of Basal Metabolism

hESC-derived hepatocytes and HepG2 cells (ECACC no 850,11430) were incubated with a cocktail of midazolam, bufuralol, phenacetin, and tolbutamide, each at a final concentration of 10 μM except tolbutamide, which was at 100 μM in the culture medium. After 24 hours, medium was collected and treated with acetonitrile to prevent further enzyme activity. The concentration of the metabolites, 1′-hydroxymidazolam, 1′-hydroxybufuralol, acetaminophen, and 4′hydroxytolbutamide, were measured using reverse phase high-performance liquid chromatography with tandem mass spectrometric detection (liquid chromatography-mass spectrometry-mass spectrometry, or LC-MS/MS [Waters, Herts, U.K., http://www.aters.com]) and the amounts were calculated according to the standards, which were prepared 0–500 ng/ml in the same culture medium. All the P450 substrates and metabolites were obtained from Ultrafine Chemicals (Sigma-Aldrich). The metabolites were normalized over alanine aminotransferase (ALT) activity which was measured from cell lysates using a Cobas Integra 400 clinical analyzer (Roche) and an in vitro diagnostic reagent system for ALT.

CYP3A4 Induction

Triplicate T25 flasks of hESC-derived hepatocytes as well as HepG2 cells were treated with rifampicin at a final concentration of 25 μM or DMSO for 72 hours. The rifampicin-containing medium was replaced with medium containing CYP3A4 substrate midazolam at concentration of 10 μM. After 24 hours, medium was collected and 1′-hydroxymidazolam metabolite was measured with LC-MS/MS as described above.


Directed Hepatocyte Differentiation from hESCs

We have developed a three-step protocol by modifying our previous method [16] to maximize hepatocyte production from hESCs (Fig. 1A and Methods). In this protocol, hESCs, routinely cultured without feeder cells as described previously [18], were first treated with 100 ng/ml activin A and 1 mM NaB for 24–48 hours, followed by 2–3 days of 100 ng/ml activin A and 0.5 mM NaB. During the first 24-hour period, we observed dramatic cell death; the surviving cells proliferated well, reaching about 70% confluence by the end of the priming stage (Fig. 1A). The cells were then split at a 1:2 ratio and cultured in SR/DMSO medium for 7 days. The cells gradually exhibited morphological change from a spiky shape to a polygonal outline (Fig. 1B). Finally, the medium was changed to modified L15 medium (see Methods for details) supplemented with 10 ng/ml HGF and 20 ng/ml OSM for a further 7 days. During the differentiation process, the hESCs went through a series of profound morphological changes and hepatocyte morphology started emerging from day 9. By day 13, the cells revealed characteristic hepatocyte morphology: polygonal in shape and distinct round nuclei (Fig. 1B & Suppl. online Fig. 7A).

Figure Figure 1..

A three-step protocol to differentiate human embryonic stem cells (hESCs) to hepatocytes. (A) Schematic representation illustrating the three-stage procedure in differentiation of hESCs to hepatocytes. (B) Images showing sequential morphological changes from hESCs (H7) to hepatocytes through the definitive endoderm and hepatoblast during differentiation. Abbreviations: DMSO, dimethyl sulfoxide; FBS, fetal bovine serum; HGF: hepatocyte growth factor; KO-DMEM, knock-out Dulbecco's modified Eagle medium; NaB, sodium butyrate; OSM, oncostatin M; SR, knockout serum replacement.

Direct comparison of this protocol with the previous one showed that the priming protocol significantly increased the production of hepatocyte-like cells from about 10%–70%. The percentage of hepatocytes differentiated from hESCs using this protocol was estimated at 71% (±7.5%) of all cells by counting albumin positive cells and 65% (±7.1%) of all cells by counting hepatocyte specific marker, Hepar1-positive cells at day 17 of differentiation. The remaining 30% cells were mainly fibroblast-like cells. The hESC-derived hepatocytes could be maintained in the final stage culture conditions for up to 5 days after characteristic hepatocyte morphology appeared, after which they began to deteriorate with increasing numbers of fiber-like cells (Suppl. online Fig. 7B). The time taken for the hESC hepatocytes to deteriorate in culture is very similar to that observed for primary human hepatocytes.

Gene Expression Pattern during Differentiation from hESCs to Hepatocytes Reflects the Progress of Liver Development in vivo

Gene expression analysis showed that the expression of genes during the differentiation process of hESCs recapitulated that of the liver developmental process in vivo (Fig. 2A–2C and Suppl. online Fig. 8). In the first priming stage of differentiation, hESCs were transitioned through mesendoderm to definitive endoderm. This is represented by transient upregulation of Brachyury, the gene expressed by the primitive streak mesendoderm and downregulated in definitive endoderm; persistent expression of goosecoid (GSC), CXCR4, and hepatocyte nuclear factor (HNF) 3β (FoxA2), genes which are expressed by the primitive streak but continuously expressed by the definitive endoderm progenitors [8, 22]; and increased expression of Sox17, a definitive endoderm marker. Following transfer to SR/DMSO medium in stage II of differentiation, hepatocyte nuclear factors 4α (HNF4α), 1α (HNF1α), and 1β (HNF1β) were dramatically upregulated, followed by increased expression of transthyretin (TTR), which is controlled by HNF4α [23]. Towards the end of stage II, high expression of AFP was evident, indicating hepatoblast or early hepatocyte formation [24]. By stage III, expression of albumin, the most abundant protein in the liver, was significantly increased and maintained. Other proteins related to liver functions were also expressed at this stage, such as apolipoprotein F (ApoF), constitutive androstane receptor (CAR), and tryptophan dioxygenase (TO). Upregulation of tyrosine aminotransferase (TAT) and CYP7A1 (Suppl. online Fig. 9) in the hESC-derived hepatocytes indicates the hESCs were differentiated to hepatic cell fate rather than yolk sac cells [25]. There was no clear Pax6 expression after the differentiation, indicating no neural ectoderm differentiated.

Figure Figure 2..

Progressive alteration of the gene expression pattern during the differentiation from human embryonic stem cells (hESCs) (H7) to hepatocytes resembles that of liver development in vivo. Reverse transcription-polymerase chain reaction (RT-PCR) (A, C) and Western blotting (B) showing progressive downregulation of undifferentiated cell gene expression, transient expression of mesendoderm markers and upregulation of hepatocyte genes expression. Stages and days in differentiation are as indicated. (D) Flow cytometry analysis showing increase of CXCR4 expressing cells during stage I differentiation. *: positive control for Pax6 from hESC-derived neural progenitor. Abbreviations: A, adult liver RNA; AFP, α-fetoprotein; ALB, albumin; ApoF, apolipoprotein F; ES, human embryonic stem cells; F, fetal liver RNA; GSC, goosecoid; HNF, hepatocyte nuclear factor; hTERT, human telomerase reverse transcriptase; TAT, tyrosine aminotransferase; TO, tryptophan dioxygenase; TTR, transthyretin.

The gene expression pattern was further analyzed by Western blot. The results of Western blotting (Fig. 2B) confirmed the progression of gene expression detected by reverse transcription-PCR (RT-PCR). Oct4 protein was considerably downregulated after 1-day differentiation and became undetectable at stage II of the differentiation. FoxA2 was clearly upregulated early around day 2 of differentiation and HNF4α was present at high levels from day 5 of stage II. The early hepatocyte gene, AFP, appeared near the end of stage II and the more mature marker albumin was expressed at the beginning of stage III. By stage III, the hepatocyte-like cells also expressed increasing levels of c-met, the HGF receptor. In general, expression of mRNAs and proteins showed same pattern. The subtle differences between RT-PCR and Western blot on HNF4α and albumin mainly reflect sensitivity difference between the two techniques.

Immunostaining further confirmed the specificity of gene expression. During stage I differentiation, CXCR4 expressing cells increased significantly from less than 15% before differentiation to over 70% at end of stage I, which is similar as reported (Fig. 2D) [26]. The transcription factors FoxA2 and HNF4α were localized in the nuclei, and AFP, albumin, and hepar1 showed cytoplasmic staining (Fig. 3A). The hESC-derived hepatocytes were also stained positive for aminopeptidase N (CD13), which has been reported to be positive for canaculi [27]. When stained with cytokeratin antibodies, CK18 and CK19, most of the hESC-derived hepatocytes were positive (Fig. 3B).

Figure Figure 3..

Characterization of the hESCs-derived hepatocytes. (A) Immunocytochemistry showing expression of various markers during differentiation process. The label on the right upper corner of each image represents day (d) and stage (st) of their differentiation and the label on the right lower corner represents antibodies used. Scale bar represents 100 μm (Aa–Ad) and 50 μm (Ae, Af). (B) Immunocytochemistry showing expression of CK18, CK19, and CD13 in human embryonic stem cell (hESC)-derived hepatocytes (Ba, Bb, and Bd, respectively). Negative controls were performed with corresponding immunoglobulin G (IgG) and representative is shown (Bc). CD13 showed no staining in Hep3B cells (Be). Scale bar represents 50 μm. Abbreviations: AFP, α-fetoprotein; ALB, albumin; HNF, hepatocyte nuclear factor.

Activin A and Sodium Butyrate Cooperate to Increase Hepatocyte Production from hESCs

In our previous method, we applied the SR/DMSO directly on hESCs to induce differentiation [16]. The resulting hepatocytes exhibited typical polygonal morphology and performed a range of hepatocyte functions. However, the percentage of hepatocytes in the population was low, with approximately 10% of the cells developing hepatocyte morphology; the rest of the cells were of mixed cell types, indicating that differentiation induced by DMSO is not lineage-specific. We proposed that priming hESC differentiation to definitive endoderm prior to DMSO treatment would be crucial in leading to more efficient hepatocyte differentiation. Increasing evidence show that activin/nodal signaling pathway are important for definitive endoderm differentiation [28, [29]–30]. NaB has been reported to contribute to more homogeneous hepatocyte differentiation [4] and more recently, NaB and activin A together are found to induce definitive endoderm differentiation from hESCs [31]. In order to further investigate the role of activin A and NaB in the definitive endoderm differentiation, the hESCs (both H7 and H1) were treated with activin A or NaB alone, or with both. After the first 24 hours of treatment, the later two conditions displayed substantial levels of cell death. Therefore, the NaB was reduced to 0.5 mM for a further 2 days before replaced by DMSO-containing medium. At this point, the cells from all of these treatments displayed similar morphologies. However, after 4 days culture in DMSO containing medium, the cells exhibited different changes in morphology. A considerable number of the cells in activin/NaB combined treatment had revealed early hepatocyte morphology, while cells treated with activin alone resulted in less hepatocyte formation and no clear hepatocyte differentiation was observed in NaB treatment (Fig. 4A).

Figure Figure 4..

Comparison of the effect of activin A and sodium butyrate (NaB) in the differentiation of human embryonic stem cells (hESCs) (H7) to definitive endoderm and in the further differentiation to hepatocyte. (A) Phase-contrast images of hESCs treated with activin A or NaB alone, or with combined activin A and NaB for 3 days (upper panel), then transferred to DMSO-containing medium for further 4 days (lower panel). (B) Comparison of gene expression with these treatments during the first stage of differentiation. Abbreviation: GSC, goosecoid.

The profile of gene expression was analyzed throughout the first 3 days and was very similar between the combined and activin-alone treatments, but different from NaB treatment (Fig. 4B). The former showed expression of primitive streak and definitive endoderm markers—such as mix1′ GSC, Sox17, and FoxA2—which are very low or absent in NaB-alone treated cells. However, the undifferentiation marker gene Nanog had obvious higher expression in activin-alone treated cells than those treated with combined activin and NaB, indicating that in the current conditions treatment with activin alone for 3 days contained more undifferentiated cells, which may result in more heterogeneous population after further differentiation with DMSO.

Liver Function in hESC-derived Hepatocytes

To test the functionality of hESC-derived hepatocytes, we carried out several experiments. One of the functions of the liver is that it can produce and export plasma proteins that are important in maintaining homeostasis of the body. In addition to albumin, other proteins include fibrinogen, a zymogen of fibrin important for blood clotting function; fibronectin, an important extracellular protein capable of binding receptor proteins, and α-2 macroglobulin (A2M), a multifunctional binding protein. To examine if the hESC-derived hepatocytes were capable of synthesizing and releasing these proteins into the culture medium, we measured these proteins in the culture medium by ELISA during stage III, when early hepatocytes had formed. The data clearly show that these export proteins were significantly increased in the culture medium in comparison with hESC controls, and the production of these proteins peaked at the later stages of the differentiation protocol when the hESC-derived hepatocytes were more mature (Fig. 5A).

Figure Figure 5..

Functional test of human embryonic stem cell (hESC)-derived hepatocytes. (A) hESC-derived hepatocytes produce and secrete plasma proteins, fibrinogen, fibronectin, and α-2 macroglobulin, which are significantly higher in comparison to their original hESCs, * and ** representing p < .005 and p < .01 by Student's t test, respectively. (B) hESC-derived hepatocytes express cytochrome P450 reductase (CPR), CYP3A of P450 family, and c-met. They also exhibit evident glycogen storage. Scale bar represents 100 μm in glycogen and 50 μm in the rest.

Another important function of liver is metabolism and detoxification, in which P450 cytochrome enzymes play a critical role. Thus, we examined expression of several members of this multigene family, including CYP3A4, CYP3A7, CYP2D6, CYP2C9, and CYP2C19 in the hESC-derived hepatocytes either by RT-PCR or Western blotting. The Western blotting results showed a marked increase in the expression of CYP3A and CYP2D6 proteins after hESCs were differentiated to hepatocytes (Fig. 6A). The CYP3A expression was also detected by immunostaining (Fig. 5B). The RT-PCR showed that both CYP3A4 and CYP3A7 were expressed in hESC-derived hepatocytes and the specificity and fidelity of CYP3A4 and CYP3A7 PCR products was confirmed by sequencing (Fig. 6B). The expression levels of CYP3A4 in the hESC-derived hepatocytes was similar to that observed in fetal liver tissues but lower than in adult liver, and the CYP3A7 expression was lower than in fetal tissues. We also detected CYP2C9 and CYP2C19 expression in the hESC-derived hepatocytes though the levels were lower than in adult liver tissues. In addition to the P450 isozymes, the expression of the cytochrome P450 reductase (CPR) (Fig. 5B) an important redox partner for all CYPs, and expression of pregnane X receptor (PXR) (Fig. 6B) and constitutive androstane receptor (CAR) (Fig. 2A), nuclear receptors important for the transcriptional regulation of P450s and other drug metabolizing enzymes [32] (particularly CYP3A4 and CYP2B6) (Fig. 6) in the presence of foreign toxic substances, were clearly induced in the hESC-derived hepatocytes after differentiation.

Figure Figure 6..

Cytochrome P450 isozymes in human embryonic stem cell-derived hepatocytes (hESC-HC). (A) Western blotting showing protein expression of CYP3A (CYP3A4, CYP3A5, CYP3A7) and CYP2D6 in hESCs and hESC-HC from two experiments. (B) Reverse transcription-polymerase chain reaction (RT-PCR) showing CYP3A4, CYP3A7 as well as CYP2C9 and CYP2C19 expression in hESC, hESC-derived definitive endoderm (end of stage I), hESC-HCs, and fetal and adult liver. (C). Basal metabolism of P450 substrates. The P450 substrates as indicated were administered to the cells and their metabolites were detected by mass spectrometry and normalized over alanine aminotransferase (ALT). The p values were calculated by Student's t test between HepG2 cells and hESC-HC. (D). Induction of CYP3A4 by rifampicin in both hESC-HC and HepG2 cells. The cells were treated with rifampicin or vehicle, DMSO, for 72 hours prior to administering CYP3A4 substrate midazolam. The metabolite 1′hydroxymidazolam (1′OH-Midazolam) was measured by mass spectrometry and normalized over ALT. The p value was calculated in hESC-HC by Student's t test between control (DMSO) and rifampicin treatment. Abbreviations: DE, definitive endoderm; DMSO, dimethyl sulfoxide; PXR, pregnane X receptor.

To determine the activity of P450 isozymes, we measured the metabolism of the P450 substrates midazolam, tolbutamide, bufuralol, and phenacetin by measuring the formation of their metabolites after 24 hours. CYP3A4 metabolizes midazolam to 1′hydroxymidazolam while the metabolism of tolbutamide to 4′hydroxytolbutamide is catalyzed by CYP2C9. Phenacetin is converted to acetaminophen by CYP1A2 or CYP2E1 and bufuralol is metabolized to 1′hydroxybufuralol by CYP2D6. The metabolites of all substrates were detected in hESC-derived hepatocytes (Fig, 6C), which demonstrated that the P450 isozymes expressed in these cells are active. When compared to the most commonly used human hepatocyte cell line, HepG2, the hESC-derived hepatocytes showed considerably higher rates of metabolism of midazolam and tolbutamide. The tolbutamide metabolism was not detected in HepG2 cells. We further tested the induction of CYP3A4 upon chemical stimulation because it is the most prevalent P450 isozyme in adult liver and is involved in the metabolism of a significant proportion of currently used drugs [33]. Because the human form of CYP3A4 can be induced with rifampicin through the transcription factor PXR, we treated both hESC-derived hepatocytes and HepG2 cells with rifampicin for 72 hours, followed by treatment with CYP3A4 substrate, midazolam, and the formation of 1′hydroxymidazolam was measured by mass spectrometry. hESC-derived hepatocytes produced higher levels of metabolites in response to rifampicin treatment although the increase was less than twofold (Fig. 6D). This suggests that the hESC-derived hepatocytes do respond to chemical treatment and that PXR is active in these cells.

We also examined the glycogen storage function in the hESC-derived hepatocytes using periodic-acid Schiff (PAS) staining. In compare with fibroblast (Suppl. online Fig. 7B), the hESC-derived hepatocytes exhibited evident glycogen storage (Fig. 5B).


We have developed a new differentiation protocol based on our previous one for efficient differentiation of hepatocytes from human embryonic stem cells. Direct comparison of the two protocols showed that activin-priming protocol exhibited significant increase in the number of hepatocyte-like cells (70%) compared with the previous protocol (10%) though hepatocytes derived from both protocols showed similar gene expression pattern (Fig. 2 and [16]). However, due to the low efficiency, we were unable to carry more detailed analysis on previous differentiation, as we had to dissect hepatocyte foci for RT-PCR. With the current protocol, we were able to extract RNA and protein from a whole-cell population and performed gene expression analysis during the differentiation. The results indicated that gene expression in the current in vitro differentiation process from ES cells to hepatocytes recapitulates liver development in vivo. This in vitro model, therefore, will be useful for future studies to elucidate molecular mechanisms regulating hepatocyte differentiation/liver development. In stage I, the hESCs first converted to cells similar to the primitive streak mesendoderm cells, then to the cells resembling those in the definitive endoderm. This was characterized by the gradual downregulation of undifferentiated gene expression, such as Oct4 and Nanog; transient increase of Brachyury expression; and upregulation of GSC, Sox17; and FoxA2 expression. During stage II of differentiation, the definitive endoderm cells further differentiated to hepatocyte progenitor cells as shown by the gradual sequential upregulation of HNF4a, AFP, and albumin. In addition, the cell morphology changed from a triangular spiky shape to the characteristic polygonal shape. Finally in stage III of the differentiation, the hepatocyte progenitor cells further developed into more mature hepatocytes, as shown by the increasing expression of albumin, apolipoprotein F, CAR, TO, and c-met receptor; as well as evident glycogen storage and generation and secretion of plasma proteins. The cells exhibited characteristic morphology of the liver hepatocytes.

In addition, by the end of stage III, the hESC-derived hepatocytes also expressed, several cytochrome P450s, and PXR. Moreover, not only are the P450 enzymes expressed but also they could convert substrates to metabolites more efficiently than HepG2 cells. We are interested in CYP3A family members, particularly CYP3A4 because it is critical for the drug metabolism [33]. The hESC-derived hepatocytes expressed detectable levels of both CYP3A4 and CYP3A7 and responded to rifampicin treatment. This has important implications for the application of hESC-derived hepatocytes as an in vitro model for drug development. Because it is very difficult to obtain human liver tissues, we were unable to directly compare our hESC-derived hepatocytes to primary human hepatocytes on the functionality tests. However, from the gene expression analysis, we think that to be able to use hESC-derived hepatocytes for drug screening, further improvement may be required, particularly on maturation. The current protocol has significantly improved differentiation efficiency but the hESC-derived hepatocytes are probably still at fetal liver developmental stage as indicated by high expression of AFP and low expression of CYP3A4. It remains a challenge on how to further mature hESC-derived hepatocytes. Liver development in vivo occurs in a 3-dimensional (3D) environment but in vitro culture is 2-dimensional (2D). Although this 2D system may aid efficient differentiation of hESCs to hepatocytes as cells receive more synchronous induction, the 3D culture environment promotes cell-cell interactions that may further enhance maturation and function [34, 35]. In addition, the current culture conditions may not be ideal for further hepatocyte maturation. We have changed maturation medium from hepatocyte culture medium [16] to the current modified L15 medium, which has reduced appearance of vacuole-like structures in maturing hepatocytes and proliferation of nonhepatocytes, mainly fibroblasts. Further improvements of the culture conditions may help further maturation and enhance functionality. Moreover, although the majority of the cells in the liver are hepatocytes, the liver also contains other cell types, such as Kupffer cells, liver endothelial cells, and so forth. These cells may have an effect on hepatocyte maturation and liver function. Future work on maturation will need to take these factors into account.

In our differentiation protocol, the combination of activin A and NaB treatment primes hESCs more efficiently after further differentiation to hepatocytes than the activin A alone. However, when we checked the gene expression patterns after stage I in both H1 and H7 cells, the activin A-only treated group seemed to have higher expression of the definitive endoderm markers (e.g., Sox17, FoxA2) and NaB itself did not show any evidence of promoting hESCs to the definitive endoderm. This raises the question of the role of NaB in this process. NaB, a short-chain fatty acid, is a histone deacetylase inhibitor and has been reported to induce growth arrest, differentiation, and apoptosis in a number of cancer cells [36, 37]. Because NaB and the combined treatment resulted in more cell death than activin A alone in our experiments, we hypothesized that NaB functions as a selecting reagent to ablate those cells that do not differentiate. The expression of Nanog in these three treatments supported this postulation. After 3 days of treatment, the cells treated with activin A alone expressed the highest levels of Nanog relative to cells treated with NaB alone or a combination of these treatments. These data indicate that after 3 days of activin A treatment, a subset of undifferentiated hESCs remain, which accounts for the formation of other cell types after DMSO treatment (DMSO does not specifically direct hESCs to the definitive endoderm lineage). More recently, a three-step protocol has been developed in which a 70% hepatocyte yield was achieved [38] using activin A alone in the first stage. This difference to our results may be a consequence of the different reagents used in the second stage.


Protocols that direct hESCs down specific differentiation pathways efficiently are important for biomedical as well as regenerative medical applications. Here we report a simple and relatively economical strategy to differentiate hESCs efficiently to hepatocytes. This system will be very useful in the further development of using hESC-derived hepatocytes for biomedical and clinical research and application.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.


We thank Anish Majumdar and Jianjie Jiang from Geron Corporation for their technical help and discussion at early-stage of the project; other members of the laboratories who have provided assistance through-out this project. This work was sponsored by Geron Corporation, California USA, the Biotechnology and Biological Science Research Council and Institute of Obstetrics and Gynaecology Trust. D.C.H. and J.F. are currently affiliated with Centre for Regenerative Medicine, Chancellor's Building, University of Edinburgh, Edinburgh, U.K. Z.A.H. is currently affiliated with the Centre for Stem Cell Biology, The University of Sheffield Alfred Denny Building, Sheffield, U.K. D.C.H. and D.Z. contributed equally to this work.