Fibroblast growth factor 10 is critical for liver growth during embryogenesis and controls hepatoblast survival via β-catenin activation


  • Potential conflict of interest: Nothing to report.


Fibroblast growth factor (FGF) signaling and β-catenin activation have been shown to be crucial for early embryonic liver development. This study determined the significance of FGF10-mediated signaling in a murine embryonic liver progenitor cell population as well as its relation to β-catenin activation. We observed that Fgf10−/− and Fgfr2b−/− mouse embryonic livers are smaller than wild-type livers; Fgf10−/− livers exhibit diminished proliferation of hepatoblasts. A comparison of β-galactosidase activity as a readout of Fgf10 expression in Fgf10+/LacZ mice and of β-catenin activation in TOPGAL mice, demonstrated peak Fgf10 expression from E9 to E13.5 coinciding with peak β-catenin activation. Flow cytometric isolation and marker gene expression analysis of LacZ+ cells from E13.5 Fgf10+/LacZ and TOPGAL livers, respectively, revealed that Fgf10 expression and β-catenin signaling occur distinctly in stellate/myofibroblastic cells and hepatoblasts, respectively. Moreover, hepatoblasts express Fgfr2b, which strongly suggests they can respond to recombinant FGF10 produced by stellate cells. Fgfr2b−/−/TOPGAL+/+ embryonic livers displayed less β-galactosidase activity than livers of Fgfr2b+/+/TOPGAL+/+ littermates. In addition, cultures of whole liver explants in Matrigel or cell in suspension from E12.5 TOPGAL+/+mice displayed a marked increase in β-galactosidase activity and cell survival upon treatment with recombinant FGF10, indicating that FGFR (most likely FGFR2B) activation is upstream of β-catenin signaling and promote hepatoblast survival. Conclusion: Embryonic stellate/myofibroblastic cells promote β-catenin activation in and survival of hepatoblasts via FGF10-mediated signaling. We suggest a role for stellate/myofibroblastic FGF10 within the liver stem cell niche in supporting the proliferating hepatoblast. (HEPATOLOGY 2007.)

The liver primordium arises in mouse embryos from the ventral foregut endoderm as a diverticulum around embryonic day 8.5 (E8.5).1 The cranial portion of the diverticulum comprises hepatic progenitor cells, or hepatoblasts, that invade the septum transversum and envelop the omphalomesenteric vein (the portal vein precursor). Hepatoblasts are bipotent progenitor cells capable of differentiating into cholangiocytes or hepatocytes, thus expressing genes specific to hepatocytes (Albumin and α-Fetoprotein), cholangiocytes (Cytokeratin 17 and 19), and progenitor cells (c-kit, Hnf4α).2Desmin-expressing hepatic stellate cells, originating from the mesoderm-derived septum transversum, transdifferentiate into Collagen-expressing and α-Smooth muscle actin (α-Sma) –expressing myofibroblastic cells, which participate in liver regeneration and fibrogenesis in response to postnatal injury.3 Stellate cells also serve to regulate proliferation and differentiation of hepatocytes in part through release of growth factors and morphogens such as hepatocyte growth factor and pleiotrophin.4, 5 Relatively little is known about how hepatoblasts and stellate cells interact with each other within the progenitor cell niche.

Fibroblast growth factors (FGFs) are known to induce the expression of liver-specific genes in the adjacent foregut endoderm, thus initiating liver bud formation through the selective proliferation of hepatoblasts.6 FGF10 has been implicated in the proliferation or differentiation of various stem/progenitor cell populations.7–9 Mesenchymally expressed FGF10 regulates differentiation of the foregut epithelial cells toward hepatic or pancreatic cell lineages in zebrafish, suggesting a significant role for FGF10 in the differentiation of liver precursor cells. FGFs bind in a promiscuous manner to 7 different tyrosine kinase FGF receptors (FGFRs) encoded by 4 genes to activate the RAS–RAF–mitogen activated protein kinase (MAPK) pathway.10 Similar pleiotropic phenotypes of null mutants for either FGF10 or the FGFR2-IIIb isoform (FGFR2b) indicate that this ligand and receptor are each other's main—although not exclusive—binding partners.11Interestingly, FGFR2B may be crucial to postnatal liver regeneration, because adult mice expressing a soluble, dominant-negative form of FGFR2B display decreased cell proliferation after partial hepatectomy.12

The Wnt/β-catenin signaling pathway has also been implicated in the maintenance, survival, proliferation, and cell fate decisions of progenitor cell populations in several organs, including the liver.13 Classical activation of β-catenin entails binding of the Wnt ligand to the Frizzled receptor, leading to increased levels of activated β-catenin within the cytoplasm, and ultimately nuclear translocation of β-catenin. Alternatively, other pathways promote β-catenin activation in a Wnt-independent fashion. For instance, binding of Hepatocyte Growth Factor to c-Met enhances nuclear translocation of β-catenin.14 In all instances, β-catenin functions as a potent coactivator of transcription factors such as lymphoid enhancer factor and T cell factor.15

In the present study, we used Fgf10 and Fgfr2b null embryos to analyze the role of FGF10 in hepatoblast survival and proliferation during hepatogenesis. Using flow cytometric cell isolation techniques, we identified the phenotype of the Fgf10-expressing cells as stellate/myofibroblastic cells and cells with active β-catenin signaling as hepatoblasts, which also express FGFR2B. We demonstrate that FGF10 from the stellate cells promotes β-catenin activation in and survival of the hepatoblasts by binding to FGFRs (most likely FGFR2B) on the hepatoblasts. The embryonic stellate/myofibroblastic cells likely function within the liver stem cell niche to support the proliferating hepatoblast.


DMEM, Dulbecco's modified Eagle medium; E, embryonic day; FACS, fluorescence-activated cell sorting; FDG, fluorescein digalactoside; FGF, fibroblast growth factor; FGFR, fibroblast growth factor receptor; PI, propidium iodide; RT-PCR, reverse-transcription polymerase chain reaction.

Materials and Methods

Mutant Embryos.

Fgf10+/− and Fgfr2b+/− mice were bred on a C57Bl/6 and C57Bl/6xGK129 background, respectively, as reported previously.16 Mlc1vnLacz-v24 mice, called Fgf10+/LacZ hereafter, kindly provided by Dr. Robert Kelly, were maintained on a mixed agouti background. TOPGAL+/− mice, a kind gift from Dr. Elaine Fuchs,17 were maintained on a mixed CD1 background and back-crossed with Fgfr2b+/− mice to generate Fgfr2b+/−/TOPGAL+/− mice which were intercrossed to generate Fgfr2b−/−/TOPGAL+/− embryos. Wild-type C57Bl/6 and Rosa26R mice were purchased from Jackson Laboratories (Bar Harbor, ME). All animal experiments were performed in accordance with National Institutes of Health guidelines.

Localization of β-Galactosidase Activity.

Livers from Fgf10+/LacZ, TOPGAL+/−, and Fgfr2b−/−;TOPGAL+/− embryos were fixed with 4% paraformaldehyde. LacZ expression was characterized with X-gal staining as described by Kelly et al.18 For histology, vibratome sections (500 μm) were stained with X-gal, embedded in paraffin, and sectioned at 5-μm thickness. Immunofluorescence was performed using rabbit anti-human albumin (1:200) (DakoCytomation, CA) plus goat anti-rabbit Cy3 plus mouse monoclonal anti-human pancytokeratin (1:100) (Sigma-Aldrich, St. Louis, MO) plus goat anti-mouse–FITC antibody. Coverslips were mounted with Vectashield containing 4′,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, CA). Light and fluorescence microscopy images were obtained using an Axioplan microscope (Zeiss, Germany) connected to an inSight QE camera (Spot, Diagnostic Instruments Inc., Sterling Heights, MI). Images were analyzed with Spot Basic Program 4.0.9 (Diagnostic Instruments Inc.) and overlayed using Adobe Photoshop CS (Adobe Systems Inc., San Jose, CA).

Immunofluorescent Detection of Cell Proliferation, Cell Death, and Marker Expression.

Colocalization of proliferating cells was performed with (1) rabbit anti-mouse Phospho-Histone H3 antibody (1:100) (Cell Signaling Technology Inc., Danvers, MA) plus donkey anti-rabbit FITC antibody, (2) goat anti-mouse albumin antibody (1:100) (Abcam, Cambridge, MA) plus AlexaFluor 350 donkey anti-goat antibody (Invitrogen, Carlsbad, CA), and (3) pancytokeratin as described above. Terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling assays were performed using the In Situ Cell Death Detection Kit (Roche Diagnostics, Denmark) as described previously.7 Positive cells were counted in three 32× high-powered fields per animal (approximately 800-1,000 cells/field.) Statistical significance was determined via Student t test and Wilcoxon signed rank test (P < 0.05).

Analysis of Gene Expression via Real-Time Reverse-Transcription Polymerase Chain Reaction.

Total RNA from wild-type C57Bl/6 embryonic livers was extracted from pools of 3-8 livers per stage (E12.5-E17.5) using the RNEasy Mini Kit (Qiagen Inc., Valencia, CA). Total RNA was also collected from individual livers for analysis. Complementary DNA was synthesized using the iScript cDNA Synthesis Kit (Bio-Rad Life Science, Hercules, CA). Real-time reverse-transcriptase polymerase chain reaction (RT-PCR) was performed using the LightCycler Taqman Master (Roche Diagnostics) and probes from the Universal Probe Library (Roche Diagnostics) against Fgfr2b, Fgf10, Fgf1, Fgf3, and Fgf7 primers (Table 1). All primer pairs were designed with the Universal Probe Library (Roche Applied Science) to have melting temperatures of 59°C-60°C, to flank an intron at the genomic level, and to produce an amplicon of less than 100 base pairs. β-Actin was selected over 36B4 and GAPDH as an internal control for all analyses. Data were analyzed from pooled as well as individual liver RNA samples at selected stages. Triplicate PCRs were performed on the same complementary DNA for all samples.

Table 1. Reverse-Transcription Polymerase Chain Reaction Primer Pair Sequence
Gene NameSequence

Culture of TOPGAL Liver Explants and Liver Cells.

E12.5 TOPGAL+/− livers were transferred into Matrigel (BD Biosciences, San Jose, CA) and cultured in Dulbecco's modified Eagle medium (DMEM)/F12 (Gibco, Grand Island, NY) supplemented with 50 U/mL penicillin, 50 U/mL streptomycin, and 0.5% fetal bovine serum (US Bio-technologies Inc.) with or without 250 ng/mL of recombinant FGF10 (rFGF10) (R&D Systems, Minneapolis, MN). Cultures were maintained for 24 hours in 5% CO2 at 37°C and 100% humidity. Livers were then fixed and stained with X-gal as described above. To obtain a single-cell suspension, fetal livers from E12.5 TOPGAL embryos were digested with collagenase (1 mg/mL), pronase (1 mg/mL), and DNAse (0.1 mg/mL) in phosphate-buffered saline (Mediatech, Inc., Herndon, VA) for 20 minutes at 37°C. Samples were red blood cell–depleted using 1× ammonium chloride RBC lysis buffer (PharMLyse; BD PharMingen, San Diego, CA) and 1 × 106cells resuspended in phosphate-buffered saline and filtered to obtain single-cell suspension. Cells were plated in 6-well cell culture plates in DMEM/penicillin/streptomycin without growth factor or with either 500 ng/mL rFGF10, 25 ng/mL rWnt3A (R&D Systems), or 250 ng/mL rFGF9 (R&D Systems) for 12 hours. Cells were then harvested using trypsin.

FACS Isolation of LacZ+ Cell Populations.

Single cell suspensions of fetal livers from Fgf10+/LacZ and TOPGAL embryos (E12.5-13.5) were obtained as described above. Fc receptors were blocked using purified rat anti-mouse CD16/CD32 monoclonal antibodies (BD PharMingen) to prevent nonspecific binding and then cells incubated with either CD45 FITC or allophycocyanin (BD Pharmingen) at 4°C for 30 minutes. To identify β-galactosidase activity, cells were loaded for 1 minute with 2 mM hypotonic fluorescein digalactoside (FDG) (Invitrogen) according to the manufacturer's protocol. Cells were then suspended in phosphate-buffered saline/4% fetal bovine serum/10 mM HEPES, and 5 μM propidium iodide (PI) (Invitrogen) was added to identify and exclude nonviable cells. In lieu of FDG, the fluoroprobe 7-hydroxy-9H-(1,3-dichloro-9,9-dimethylacridin-2-one β-galactoside (Invitrogen) was also used with equivalent sorting quality as described.19 β-Galactosidase hydrolyses FDG to fluorescein, with excitation at 491 nm and emission at 514 nm; 7-hydroxy-9H-(1,3-dichloro-9,9-dimethylacridin-2-one β-galactoside is hydrolyzed to DDAO with excitation at 615-645 nm and emission at 680-720 nm. Both excitation and emission spectra were used to isolate cells with β-galactosidase activity. For fluorescence-activated cell sorting (FACS) analysis, cells were analyzed using a BD FACSCalibur (BD PharMingen). For FACS isolation, cells were isolated on a BD FACS-Vantage (BD PharMingen). Post-analysis was performed using the Flo-Jo program (TreeStar Inc., Ashland, OR). Compensation for FITC, PI, and allophycocyanin was performed with compensation beads (BD PharMingen); compensation for intracellular fluorescein, DDAO, and PI was conducted using positive- and negative-stained compensation beads. Negative gates for surface staining were determined using mouse IgG-stained and unstained cells. Isolated cells were washed twice with phosphate-buffered saline, pelleted, resuspended, and plated in 6-well culture plates in DMEM with penicillin/streptomycin with or without 250 ng/mL rFGF10 at 37°C, 5% CO2.

Gene Expression Analysis of Isolated Fetal Liver Cells.

Fluorescein+ CD45 fetal liver cells from heterozygous Fgf10+/LacZ or TOPGAL embryos as well as single DDAO+ CD45 cells from E13 heterozygous TOPGAL fetal livers and single hepatocytes from adult wild-type C57Bl/6 mice were isolated on a BD FACS-Vantage. Cells were centrifuged at 200g for 5 minutes to remove supernatant and total RNA was extracted for RT-PCR using the designed primer sets for Albumin, Cytokeratin-19, Hnf4α, Cd45, α-Sma, Desmin, Fibulin-2, Gfap, Fgf10, Fgfr2b, LacZ, and β-actin as described above.


Livers from Fgf10−/− and Fgfr2b−/− embryos were smaller than wild-type littermate livers and were misshapen. When normalized to embryo weight, both mutant embryos exhibited statistically significant reductions in liver/embryo weight ratios than wild-type littermate livers (Fig. 1). Phospho-Histone H3 staining demonstrated a statistically significant reduction in overall proliferation in Fgf10−/− livers compared with Fgf10+/+ livers at E12.5 (Fig. 2A -C). Terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling demonstrated a 4-fold increase in apoptosis in Fgf10−/− livers compared with Fgf10+/+ livers at E12.5 (Fig. 2D-F). Using triple label immunofluorescence, we showed a significant reduction in the quantity of histone H3+ albumin+ pancytokeratin+ cells in E12.5 Fgf10−/− livers compared with wild-type littermates indicating a reduction in proliferating hepatoblasts (Fig. 2G-I).

Figure 1.

Comparison of liver mass of wild-type and Fgf10−/− or Fgfr2b−/− embryos. (A, B) Whole mount P0 Fgf10+/+ and Fgf10−/− livers. (C, D) Whole mount E18.5 Fgfr2b+/+and Fgfr2b−/− livers. (E) Graphic representation of respective liver/embryo weight ratios of both null mutants and their wild-type littermates. *P < 0.05; n = 10. Scale bar = 1 mm.

Figure 2.

Comparison of cell proliferation and apoptosis in E12.5 Fgf10−/− livers. (A-C) Immunofluorescence imaging of phosphorylated histone H3 in Fgf10−/− and wild-type livers (n = 16). (D-F) Immunofluorescence imaging of terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling in Fgf10−/− and wild-type livers (n = 8). (G-I) Immunofluorescence imaging of phosphorylated histone H3, albumin, and pancytokeratin in Fgf10−/− and wild-type livers (n = 8). *P < 0.05.

Kelly et al. initially characterized a transgenic mouse line, Fgf10+/LacZ, with the β-galactosidase gene, LacZ, under the control of endogenous Fgf10 regulatory sequences such that expression of LacZ accurately reports Fgf10 expression.16, 18 Analysis of serial sections of X-gal stained Fgf10+/LacZ embryo livers demonstrated Fgf10 expression in the mesenchyme of the septum transversum surrounding the hepatic diverticulum at E9 (Fig. 3A). As the liver grows within the septum transversum, Fgf10 expression persists along the surface of the liver near the bile duct (Fig. 3B). By E12.5, Fgf10 expression is scattered within the liver itself but is also concentrated along the mesenchyme of the extrahepatic portal structures (Fig. 3C-F). By E14.5, parenchymal Fgf10 expression diminishes, persisting only along the extrahepatic portal vein and bile duct (not shown). This early peak in Fgf10 expression coincides temporally with maximal β-catenin activation as observed by β-galactosidase activity in the livers of TOPGAL embryos (Fig. 4). Analyses of the temporal expression patterns of Fgf10 and genes encoding other FGFR2B ligands was performed via real-time RT-PCR on total RNA from pooled livers or individual wild-type liver embryos from E12.5 to E17.5 (Fig. 5); the patterns of expression were similar in either case. Fgf10 is maximally expressed at E12.5 followed by a greater than 5-fold reduction by E14.5. In comparison, Fgf1 and Fgf7 message levels peak later at E17.5 and E15.5, respectively. Fgf3 expression was undetectable in the liver at all studied stages (data not shown.) Fgfr2b expression increased gradually from E12.5 to E17.5.

Figure 3.

Spatial and temporal expression pattern of β-galactosidase activity in Fgf10+/LacZ embryos early during hepatogenesis. (A) E9 (original magnification ×20). (B) E10.5 (original magnification ×20). (C, D) BABB cleared whole mount E12.5 liver. Arrow indicates staining along extrahepatic portal structures (original magnification ×2.5, ×10). (E, F) E12.5 (original magnification ×20, ×40). Abbreviations: b, bile duct; hd, hepatic diverticulum; p, portal vein; st, septum transversum.

Figure 4.

Comparison of β-galactosidase activity in Fgf10+/LacZ and TOPGAL livers at E11.5, E12.5, and E18.5. Scale bar = 0.2 mm.

Figure 5.

Analysis of relative expression levels of messenger RNAs encoding (A) Fgf10, (B) Fgf1, (C) Fgf7, and (D) Fgfr2b from E12.5 to E17.5 via real-time RT-PCR.

To characterize the phenotypes of β-galactosidase expressing cells from Fgf10+/LacZ and TOPGAL livers, the livers were dissected at the time point when maximal β-galactosidase activity was present, respectively E12.5 in Fgf10+/LacZ and E13.5 in TOPGAL livers. Livers were digested to a single cell suspension. Cells were then stained for the hematopoietic cell marker CD45 and subjected to cell-sorting flow cytometry based on fluorescein content generated by β-galactosidase–driven conversion of FDG. Rosa 26 liver cells loaded with FDG demonstrated bright fluorescein staining (positive control), whereas FDG-loaded C57Bl/6 controls did not (negative control) (Fig. 6A-C). The approximately 20% PI+ dead cells were excluded from analysis. Fifteen percent of the total cells from E12.5-13.5 livers demonstrated surface expression of CD45; these cells were removed from FACS isolation by gating techniques. 3.4 ± 2.0% of the total cells from E12.5 Fgf10+/LacZ livers were Fluorescein+ CD45 PI (Fig. 6D). 6.0 ± 2.5% of the total cells from E13.5 TOPGAL livers were Fluorescein+ CD45 PI (Fig. 6E).

Figure 6.

Flow cytometric and marker gene analyses of fluorescein+ cells from Fgf10+/LacZ and TOPGAL embryo livers. (A-C) Validation of flow cytometric cell sorting based on β-galactosidase–generated fluorescein in livers from C57Bl/6 and Rosa 26 mice. *P < 0.05. (D, E) Flow cytometic sorting of cells from Fgf10+/LacZ and TOPGAL embryo livers. (F) Gene expression analysis of fluorescein+ CD45 cells from either transgenic strain via RT-PCR.

To analyze these sorted cells for gene expression, total RNA was isolated for RT-PCR (Fig. 6F). Fluorescein+ CD45 PI cells from E12.5 Fgf10+/LacZ livers demonstrated expression of Fgf10 along with myofibroblastic markers, α-Sma and Fibulin-2, and stellate cell markers, Desmin and Gfap (Fig. 6F) (Fibulin-2 and Gfap data not shown), but display no detectable expression of the hepatoblast markers Albumin, CK19, and Hnf4α, nor of the hematopoietic marker CD45, nor Fgfr2b.20 This expression profile indicated that we had isolated either 2 distinct cell populations (embryonic myofibroblasts and stellate cells) or a single population of cells with dual embryonic stellate/myofibroblastic phenotype, all of which express Fgf10. Fluorescein+ CD45 PI cells from E13.5 TOPGAL livers expressed Albumin, CK19, Hnf4α, and Fgfr2b genes, but had no detectable expression of Fgf10, α-Sma, Desmin, or CD45, indicating absence of mesenchymal cells. This expression profile is consistent with that of a bipotent hepatoblast. In addition, Fgfr2b expression suggests that these cells may respond to FGF10.

To determine whether β-galactosidase expressing cells from TOPGAL livers truly represented hepatoblasts or a mixed population of hepatocytes and cholangiocytes, total RNA was immediately isolated from single individual DDAO+ CD45 cells isolated by FACS from E13.5 TOPGAL livers. RT-PCR was then performed for each single cell. Sixteen of 26 (61.5%) single cells demonstrated bipotent gene expression with Albumin and CK19 messenger RNA. In contrast, only 6.2% of adult hepatocytes had the same expression profile (P < 0.05). Thus, single-cell analysis confirms that at E13.5, the majority of the liver cells with active β-catenin signaling are hepatoblasts with bipotent gene expression (Fig. 7). However, approximately 30% of the cells yielded an amplicon for only 1 marker, suggesting that these cells had already committed to the hepatocyte or cholangiocyte lineage, respectively, unless an amplicon of the complementary marker was not generated due to sensitivity limitations of single-cell RT-PCR. Epifluorescence microscopy of sections of E12.5 livers from Fgf10+/LacZ embryos confirmed that β-galactosidase activity is present in distinct cells in close proximity to hepatoblasts coexpressing albumin and cytokeratin (Fig. 8).

Figure 7.

Gene expression analysis of single isolated TOPGAL-expressing cells. (A) A representative experiment of single-cell RT-PCR on individual fluorescein+ CD45 cells from TOPGAL livers. Bone marrow cells and adult liver are negative and positive control, respectively. (B) Comparison of percentage of Albumin and Ck19 coexpressed by fluorescein+ CD45 cells E13.5 TOPGAL liver and by mature murine hepatocytes.

Figure 8.

Immunofluorescence colocalization of pancytokeratin and albumin in proximity to Fgf10-expressing cells in E12.5 liver. (A) X-gal staining for β-galactosidase activity. (B, C) Immunodetection of albumin and pancytokeratin. (D) Merged image shows distinct Fgf10 expression (blue) in close proximity to putative hepatoblasts coexpressing cytokeratin and albumin (yellow).

To determine if FGF10 and FGFR2B act upstream of β-catenin signaling in the developing liver, both loss of function and gain of function experiments were performed. For loss of function, β-galactosidase activity was assessed via X-gal staining of TOPGAL livers lacking FGFR2b signaling. Fgfr2b−/−TOPGAL+/− livers were smaller and displayed less β-galactosidase activity than livers of Fgfr2b+/+TOPGAL+/− littermates at both E11.5 (Fig. 9 A,B) and E12.5 (not shown). Moreover, as a model of gain of function, E12.5 TOPGAL+/+ liver explants in Matrigel were treated with rFGF10 for 24 hours. They showed marked expansion of X-gal staining with more pronounced bridging of LacZ-positive regions compared with nontreated controls (Fig. 9C-F). Furthermore, treatment of E12.5 liver cells in culture with rFGF10 resulted in 2.9-fold increases in expression of CyclinD1, a downstream target of β-catenin.19 Importantly, there was some β-galactosidase activity—albeit less in the Fgfr2b−/−/TOPGAL+/− embryo livers—suggesting that other pathways for β-catenin activation are present.

Figure 9.

Comparison of β-galactosidase activity in E11.5 TOPGAL+/− livers. (A) Fgfr2b+/+. (B) Fgfr2b−/−. (C) DMEM alone (original magnification ×2.5). (D) DMEM + rFGF10 (original magnification ×2.5). (E) Inset of panel C (original magnification ×10). (F) Inset of panel D (original magnification ×10). (F') Additional view of liver in panel D (original magnification ×10).

To better quantify the effect of FGF10/FGFR2B signaling on the β-catenin pathway, we again used FACS analysis on liver cells from TOPGAL+/− embryos. E12.5 TOPGAL+/− liver cells in suspension were cultured overnight in DMEM, with or without rFGF10 or rWNT3A. Cells cultured with either rFGF10 or rWNT3A demonstrated a significant survival advantage compared with control cells after 24 hours as assessed via PI staining (Fig. 10C). Additionally, the mean fluorescence index, which is a measure of the relative fluorescence of individual cells, was significantly higher in both rFGF10- and rWNT3A-treated cells compared with controls (control, 0.7 ± 0.3; rFGF10, 3.3 ± 1.0; rWNT3A, 3.7 ± 0.8 [P < 0.05]) indicating greater β-catenin activation per cell as seen by a shift of the scatter plots to the right (Fig. 10A). Furthermore, the number of fluorescein+ cells increased significantly with treatment with rFGF10 (4.6-fold increase) or rWNT3A (5.1-fold increase) compared with controls (P < 0.05), indicating an increase in the number of cells manifesting β-catenin activation (Fig. 10B). Using LacZ expression measured via real-time RT-PCR as a readout of β-catenin activation, treatment of TOPGAL+/− liver cells with rFGF10 for 12 hours resulted in a 1.7-fold increase in β-catenin activation, whereas treatment with rFGF9, which binds specifically to FGFR3b, had no effect. From these experiments, we infer that FGF10 can act on TOPGAL-positive hepatoblasts to promote their survival.

Figure 10.

Treatment of E12.5 TOPGAL liver cells with rFGF10 or rWNT3A. (A) FACS scatter plots of TOPGAL E12.5 liver cells cultured overnight in DMEM (left), with presence of rFGF10 (middle) or rWNT3A (right). (B) Graphic representation showing relative number of LacZ-expressing (fluorescein+) cells upon rFGF10 treatment of TOPGAL liver cells. (C) Graphic representation of percentage of PI+ cells with either rFGF10 or rWNT3A treatment. *P < 0.05 compared with control; n = 6.


Hepatogenesis involves several FGFs secreted by adjacent mesenchymal tissues. FGF1, FGF2, and FGF8, secreted by the cardiac mesenchyme, induce formation of the initial liver bud from the foregut endoderm by stimulating hepatoblast proliferation and differentiation.6 We show that Fgf10 is expressed in the septum transversum as early as E9 just as the hepatic diverticulum is evaginating from the foregut endoderm. Although Fgf10 may be expressed earlier, it is not essential for induction of diverticulum formation, because Fgf10−/− embryos still develop livers. Null mutations of either Fgf10 or Fgfr2b both result in smaller livers, suggesting that FGF10 may be acting through FGFR2b, as in other organs.16, 21, 22 Furthermore, loss of Fgf10, which normally is maximally expressed around E12.5, leads to reduction in hepatoblast proliferation. This, however, does not preclude the possibility that other ligands may be relevant to FGFR2B activation or that other FGFRs may also undergo ligand binding during early hepatogenesis. Indeed, FGF1, which is known to be expressed early during hepatogenesis, promiscuously binds several FGFRs, including FGFR2B.10 Although these data do not exclude the roles of other FGFs and FGFRs during hepatogenesis, we can nonetheless conclude that FGF10 likely binds to FGFR2B early during hepatogenesis to enhance hepatoblast proliferation.

The nature of Fgf10-expressing cells in the liver is of particular interest. We show through localization of LacZ activity in Fgf10+/LacZ that mesenchymal cells surrounding the portal vein and bile duct as well as scattered single cells within the liver itself express Fgf10. Gene expression analysis of these LacZ+ CD45 cells from Fgf10+/LacZ embryo livers indicates that these cells are embryonic hepatic stellate cells and myofibroblasts, or the cells with both phenotypes. The possibility of a dual stellate/myofibroblastic cell phenotype may indicate that we have isolated a potential mesenchymally derived progenitor cell, which may give rise to both postnatal cell types. If so, transdifferentiation of hepatic stellate cells to myofibroblasts during certain types of liver injury may represent a recapitulation of the embryonic stellate/myofibroblastic cell transition. Such transdifferentiation from quiescent hepatic stellate cell to myofibroblast could possibly arise through reversion of the hepatic stellate cell to a cell of bipotent intermediate phenotype.

The Wnt/β-catenin pathway has been implicated in the survival and proliferation of epithelial progenitor cells in many organ systems, including the liver, during embryogenesis.13, 23 WNT3A-conditioned medium enhances proliferation and survival of hepatoblasts within in vitro culture systems of E12.5 liver cells.24 Inhibition of WNT/β-catenin activity via the overexpression of a dominant-negative lymphoid enhancer factor-1 or the overexpression of a secreted WNT inhibitor Dickkopf in chick embryos results in smaller, misshapen livers.25 In this study, we show that the vast majority of LacZ+ cells from TOPGAL E13.5 embryonic livers are CD45 and thus not of hematopoietic origin. Indeed, we show that most β-catenin–activated embryonic liver cells are hepatoblasts. Furthermore, it is possible that some of the cells manifesting β-catenin activation have, in fact, differentiated into hepatocytes or cholangiocytes.

FGF/FGFR activation has been previously linked to downstream β-catenin activation. FGF10 promotes downstream β-catenin activation in cell lineage specification in developing skin and mammary placode.22, 26 Sekhon et al. showed that maximal FGFR activation occurs from E10 to E12 and that embryonic liver explants treated with FGF1, FGF4, or FGF8 increases nuclear localization of β-catenin.24 We show through both loss of function and gain of function analyses that FGF10 and FGFR2B activity promotes the activation of β-catenin signaling during hepatogenesis. Loss of FGFR2B signaling results not only in smaller liver size, but also in decreased β-catenin activation. Addition of FGF10 to embryo livers grown in culture enhances β-catenin activation, both in the number of cells manifesting β-catenin activation and in the level of β-catenin activation per cell. In this study, we show coexpression of Fgfr2b in a population of LacZ+ CD45 cells, indicating that FGFR2b may be the receptor for FGF10 in β-catenin–activated hepatoblasts, although other receptors cannot be excluded. Although the exact mechanism by which FGFR activation leads to β-catenin activation in hepatoblasts has not previously been determined, it is conceivable that the PTEN/Akt pathway, which has recently been shown to lead to β-catenin activation in stem cells of the intestinal epithelium, might also play a role downstream of liver progenitor cell FGFR activation, which has been linked to Akt activation.27 Alternatively, FGF10, which is secreted by stellate cells, might induce expression of Wnts in an autocrine fashion, which could in turn bind to Frizzled on hepatoblasts to induce β-catenin activation.28

Given this spatial distribution of Fgf10-expressing cells, we speculate that early during hepatogenesis, FGF10 is secreted by embryonic stellate cells/myofibroblasts, which reside around the portal vein and scattered throughout the liver. FGF10 is then carried through the portal vein and the developing sinusoids of the liver to the hepatoblasts, which express FGFR2B. By binding FGFR2B on hepatoblasts, FGF10 induces activation of β-catenin signaling. Embryonic stellate/myofibroblastic cells thus function as part of the hepatoblast niche supporting hepatoblast survival and proliferation in part through FGF10/FGFR2B signaling.

In conclusion, stellate cells/myofibroblasts are an important element in the hepatoblast niche by FGF10 activation of the β-catenin signal pathway, most likely through FGFR2B. Further characterization of the role of FGF10/FGFR2B signaling in hepatoblast physiology as well as the role of embryonic stellate cells/myofibroblasts in the hepatoblast niche will allow us to understand how hepatic progenitor cells are maintained—not only during hepatogenesis, but also postnatally during liver injury and regeneration.


We thank Lora Barsky (Saban GISCT Program FACS Core) for technical support and thank David Warburton, and Henri Ford for stimulating discussions related to this project.