Purification and characterization of mouse fetal liver epithelial cells with high in vivo repopulation capacity


  • Potential conflict of interest: Nothing to report.


Epithelial cells in embryonic day (ED) 12.5 murine fetal liver were separated from hematopoietic cell populations using fluorescence-activated cell sorting (FACS) and were characterized by immunocytochemistry using a broad set of antibodies specific for epithelial cells (α-fetoprotein [AFP], albumin [ALB], pancytokeratin [PanCK], Liv2, E-cadherin, Dlk), hematopoietic/endothelial cells (Ter119, CD45, CD31), and stem/progenitor cells (c-Kit, CD34, Sca-1). AFP+/ALB+ cells represented approximately 2.5% of total cells and were positive for the epithelial-specific surface markers Liv2, E-cadherin, and Dlk, but were clearly separated and distinct from hematopoietic cells (Ter119+/CD45+). Fetal liver epithelial cells (AFP+/E-cadherin+) were Sca-1+ but showed no expression of hematopoietic stem cell markers c-Kit and CD34. These cells were enriched by FACS sorting for E-cadherin to a purity of 95% as defined by co-expression of AFP and PanCK. Purified fetal liver epithelial cells formed clusters in cell culture and differentiated along the hepatocytic lineage in the presence of dexamethasone, expressing glucose-6-phosphatase (G6P) and tyrosine amino transferase. Wild-type ED12.5 murine fetal liver cells were transplanted into adult dipeptidyl peptidase IV knockout mice and differentiated into mature hepatocytes expressing ALB, G6P, and glycogen, indicating normal biochemical function. Transplanted cells became fully incorporated into the hepatic parenchymal cords and showed up to 80% liver repopulation at 2 to 6 months after cell transplantation. In conclusion, we isolated and highly purified a population of epithelial cells from the ED12.5 mouse fetal liver that are clearly separate from hematopoietic cells and differentiate into mature, functional hepatocytes in vivo with the capacity for efficient liver repopulation. Supplementary material for this article can be found on the HEPATOLOGY website (http://www.interscience.wiley.com/jpages/0270-9139/suppmat/index.html). (HEPATOLOGY 2005;.)

The presence of both hematopoietic and epithelial cells in the fetal liver provides an ideal system to identify their common and distinct features and to determine whether these 2 cell populations are completely separate or partially overlapping. During embryonic development, the fetal liver is a major site of hematopoiesis, with hematopoietic stem cells (HSC) invading the early mouse fetal liver at embryonic day (ED) 10.5 and rapidly expanding between ED12 and ED16.1 Beginning at ED16, the site of hematopoiesis gradually shifts from the fetal liver to the spleen and eventually settles in the bone marrow.2 Different studies have reported common markers for HSC and fetal liver epithelial cells (FLEC), such as CD34,3, 4 Thy-1,5 and c-Kit,3, 6 and these results are consistent with studies in the adult liver in which oval cells express these markers.7–14 However, the precise lineage relationship between epithelial cells in the fetal liver and oval cells in the adult liver remains to be established.

For many years, monoclonal antibodies against specific surface antigens in combination with fluorescence-activated cell sorting (FACS) have been used as the major tool for isolating and characterizing HSC.15 In 1994, FACS sorting was introduced for the isolation of fetal liver cells by Sigal et al.,16 using the light scattering properties of the cells after removal of hematopoietic cells by “Panning”. More recent studies have extended this approach by using various combinations of antibodies.17–21 Although the antibodies used to isolate the cells were not specific for FLEC, these investigators were able to establish the liver epithelial progenitor phenotype and show the bipotency, clonality, and proliferative capacity of their enriched cell populations in cell culture.17–21 More recently, 3 epithelial-specific cell surface markers for the murine fetal liver have been identified: the unique developmental marker Liv2,22 the epithelial-specific marker E-cadherin,23 and Dlk, a type I membrane protein, also known as Pref-1.24 E-cadherin and Dlk have been used successfully to isolate FLEC, but very little is known about the ability of these cells to repopulate the liver after their transplantation.

The current study reports the separation of FLEC from HSC in the murine fetal liver at ED12.5 and their subsequent characterization, using a combination of cytoplasmic (α-fetoprotein [AFP], albumin [ALB], pancytokeratin [PanCK]), cell surface (Ter119, CD45, CD31, Liv2, E-cadherin, Dlk), and stem cell markers (c-Kit, CD34, Sca-1). By FACS analysis and immunocytochemistry, we showed cytoplasmic staining for AFP and ALB and confirm by double-labeling that the recently used surface markers Liv2, E-cadherin, and Dlk are all specific for epithelial cells in the fetal liver. The only overlapping marker for epithelial and hematopoietic progenitor cells was Sca-1. Using FACS sorting with E-cadherin, we isolated and enriched epithelial cells from the ED12.5 murine fetal liver to 95% purity. Crude, as well as purified, FLEC were transplanted into the liver of mice deficient in dipeptidyl peptidase IV (DPPIV) expression and showed up to 80% in vivo liver repopulation in retrorsine-treated animals with reformation of normal liver lobules containing fully integrated hepatocytes that exhibit normal differentiated function. The in vivo properties of ED12.5 FLEC suggest that they will be particularly advantageous for liver repopulation studies in a wide variety of pathophysiologic states and in mouse models of human liver diseases.


HSC, hematopoietic stem cells; ED, embryonic day; FLEC, fetal liver epithelial cells; FACS, fluorescence-activated cell sorting; AFP, α-fetoprotein; ALB, albumin; PanCK, pancytokeratin; DPPIV, dipeptidyl peptidase IV; PE, phyco-erythrin; PECAM, platelet endothelial cell adhesion molecule; SSC, side scatter characteristics; CCl4, carbon tetrachloride; G6P, glucose-6-phosphatase.

Material and Methods


Chemicals were from Sigma-Aldrich, St. Louis, MO; cell culture media and supplements from GIBCO BRL, Grand Island, NY; collagenase type 1 from Worthington Biochemical Corporation, Lakewood, NJ.


Pregnant C57Bl/6 mice were purchased from Jackson Laboratory, Bar Harbor, ME. DPPIV gene knockout (DPPIV−/−) recipients were bred in our Special Animal Core. All animal experimental procedures were conducted under protocols approved by the Animal Care Use Committee of the Albert Einstein College of Medicine and were in accordance with National Institutes of Health guidelines.

Isolation of Cells.

On ED12.5, fetal livers were microdissected from timed pregnant mice and cells isolated by a modification of the procedure of Sigal et al.16 In brief, fetal livers were triturated in modified Hank's buffered saline solution (HBSS) containing 1 mol/L ethyleneglycoltetraacetic acid, followed by digestion in modified HBSS containing 0.2% collagenase, 0.07% DNAse, and 1 mmol/L CaCl2. We routinely obtained 2.5 × 106 cells per fetal liver, with a viability of >98%, as estimated by trypan blue dye exclusion.

Immunocytochemistry on Cytospins.

Immunocytochemical detection of AFP, PanCK or CK19 was performed after fixation in acetone/methanol. For ALB detection, cells were fixed in 4% paraformaldehyde, followed by permeabilization with 0.1% Triton X-100. Primary antibodies for cytoplasmic antigens were polyclonal rabbit anti-human AFP (Neomarkers, Fremont, CA), monoclonal mouse anti-human CK19 (Novocastra, Newcastle, UK), monoclonal mouse anti-human PanCK, and polyclonal rabbit anti-human ALB (Sigma-Aldrich). Monoclonal unconjugated rat anti-mouse surface antibodies were against Liv2 (gift from H. Nishina), E-cadherin (TaKaRa, Otsu, Japan) and Dlk (gift from A. Miyajima). Secondary antibodies included Cy™2- or Cy™3-conjugated donkey anti-rabbit, Cy™3-conjugated donkey anti-rat, and Cy™3-conjugated donkey anti-mouse immunoglobulin Gs (Jackson ImmunoResearch Laboratories, West Grove, PA). Fluorescence was detected by using a fluorescence microscope equipped with a mercury lamp. Detection ranges were set to eliminate cross-talk between fluorophores.

FACS Analysis and Sorting.

Immunostaining procedures for cytoplasmic antigens (AFP and ALB) were as described previously. Surface markers included monoclonal fluorescein isothiocyanate–conjugated Ter119, CD45 (eBioscience, San Diego, CA) and CD34 (Serotec, Oxford, UK), monoclonal phyco-erythrin (PE)-conjugated c-Kit (CD117), and PE-Cy5-conjugated c-Kit and Sca-1 (eBioscience). Additionally, we used monoclonal, unconjugated rat anti-mouse antibodies for CD31 (platelet endothelial cell adhesion molecule [PECAM-1]) (Pharmingen, San Jose, CA) and for Liv2, E-cadherin, and Dlk, as described above. Secondary antibodies were PE-conjugated donkey anti-rat and Cy™2- or PE-conjugated donkey anti-rabbit immunoglobulin Gs (Jackson ImmunoResearch Laboratories). Immunofluorescence analysis was performed on a FACS Scan or FACS Calibur (Becton Dickinson, San Jose, CA), both equipped with the CellQuest System. Cell debris and dead cells were excluded from analysis by light parameters and To-Pro 3 (Molecular Probes, Eugene, OR) was added as a viability indicator if only surface markers were analyzed on the FACS Calibur. Positive staining was defined against the background obtained with isotype-matched irrelevant (control) antibodies. FACS sorting and reanalysis was performed using a MoFlo High Speed Sorter (DakoCytomation, Carpinteria, CA). For negative selection, cells were stained for Ter119, CD45, and c-Kit, and gates were set to exclude all positively stained cells and cells with low side scatter (SSC). For positive selection, cells were stained and sorted for E-cadherin and high SSC. Viability after sorting was >90%, as assessed by trypan blue dye exclusion. All samples were analyzed using FlowJo software (Tree Star, Ashland, OR).

Cell Culture.

For in vitro studies, unfractionated cells were plated on gelatin-coated 6-well dishes at 2 × 105 cells/cm2. After sorting, cells were plated on type-I collagen-coated (Vitrogen, Cohesion, Palo Alto, CA) 24-well dishes at 2 × 104 cells/cm2. Plated cells were incubated in 10% fetal bovine serum (Gemini Bio-Products, Woodland, CA) in Dulbecco's modified essential medium for 24 hours, after which the medium was changed to modified Block's medium25 containing the following growth factors: 0.1 μmol/L dexamethasone, 10 ng/mL epidermal growth factor (Upstate Biotechnology, Lake Placid, NY), 10 ng/mL basic fibroblast growth factor, and 20 ng/mL transforming growth factor alpha (Pepro Tech Incorporation, Rocky Hill, NJ).

Reverse Transcription-Polymerase Chain Reaction.

Total RNA was extracted using Trizol reagent (Invitrogen, Carlsbad, CA), following the manufacturer's instructions. Complementary DNA was synthesized using Superscript™II RNase (Invitrogen) and was amplified using specific primers (see Table 1) and recombinant Ampli Taq Gold DNA Polymerase (Applied Biosystems, Foster City, CA). The amplification protocol was 10 minutes at 95°C, followed by 30 cycles of 94°C for 30 seconds, primer-specific annealing temperature for 1 minute, and 72°C for 1 minute, and a final extension step of 7 minutes at 72°C.

Table 1. Primer Sequences Used for Reverse Transcription–Polymerase Chain Reaction
Gene NameStrandPrimer Sequence
α-Fetoprotein (AFP)Sense5′-TCGTATTCCAACAGGAGG-3′
Cytokeratin 19 (CK19)Sense5′-GTGCCACCATTGACAACTCC-3′
Glucose-6-phosphatase (G6P)Sense5′-AACCCATTGTGAGGCGAGAGG-3′
Tyrosine aminotransferase (TAT)Sense5′-TTA AGT CCA ATG CGG ACC TC-3′
Tryptophan oxygenase (TO)Sense5′-CATGGCTGGAAAGAACACCT-3′

Transplantation Studies.

Twenty-one DPPIV−/− recipients were given retrorsine (70 mg/kg body weight, intraperitoneally) at 6 and 8 weeks of age to block proliferation of endogeneous hepatocytes, as originally described in rats by Laconi et al.26 and adapted to mice by Guo et al.27 At 10 weeks of age, all animals were administered intraperitoneally 1.2 mL/kg body weight carbon tetrachloride (CCl4) to induce acute liver injury. Eight to 24 hours after the injection, 13 animals were subjected to transplantation into the spleen of up to 8 × 106 unfractionated cells, 8 animals received up to 1.4 × 105 sorted cells. Cell transplantation was followed by weekly treatment with 0.5 mL/kg body weight CCl4 for up to 8 weeks.

Enzyme- and Immuno-histochemistry.

Mice were killed at various times from 1 to 6 months after cell transplantation, and livers were snap-frozen in 2-methylbutane or fixed in 10% formalin in phosphate-buffered saline. DPPIV expression was determined on 5μm frozen sections by enzyme histochemistry, as previously reported.28 Paraffin-embedded sections were used for hematoxylin-eosin staining and for ALB detection using a polyclonal, unconjugated rabbit anti-human ALB antibody (Sigma-Aldrich). Frozen liver sections were fixed with acetone and stained with a monoclonal rat anti-mouse fluorescein isothiocyanate–conjugated DPPIV and a monoclonal, unconjugated rat anti-mouse CD31 antibody (PECAM-1) (Pharmingen). Unconjugated antibodies were detected by Cy™3-conjugated donkey anti-rabbit or anti-rat secondary antibody, respectively (Jackson ImmunoResearch Laboratories).

G6P Activity.

Unfixed cryosections were incubated for 20 minutes in substrate reagent: 10 mmol/L D-glucose-6-phosphate, 270 mmol/L sucrose, 2.4 mmol/L Pb(NO3)2, and 40 mmol/L Tris-maleate buffer, pH 6.5. After fixation, cryosections were immersed in 0.22% (NH4)2S diluted in 0.1 mol/L Tris-maleate/0.1 mol/L NaCl and counterstained with hematoxylin.

Glycogen Detection (Periodic Acid-Schiff Reaction).

Cryosections were incubated for 5 minutes in 0.5% periodic acid solution, followed by Schiff's reagent for 25 minutes. Sections were counterstained with hematoxylin.


FACS Analysis and Sorting.

To fully characterize epithelial cells and distinguish these from other cell fractions in the murine fetal liver, distribution profiles were established for a broad set of antibodies specific for epithelial cells (AFP, ALB, Liv2, E-cadherin, Dlk), hematopoietic/endothelial cells (Ter119, CD45, CD31), and stem/progenitor cells (c-Kit, CD34, Sca-1), using 2- and 3-color immunostaining. AFP+/ALB+ cells represented up to 2.5% of total cells in unfractionated single-cell suspensions and showed higher SSC, forward scatter, and autofluorescence relative to other cell populations (Supplementary Fig. 1; Available at: http://www.interscience.wiley.com/jpages/0270-9139/suppmat/index.html). All AFP+ cells stained positive for surface markers E-cadherin, Liv2, and Dlk, and vice versa (Fig. 1). AFP+/E-cadherin+ cells were Sca-1+ but showed no overlap with the hematopoietic stem/progenitor markers c-Kit or CD34 (Fig. 2). A very small number of E-cadherin+ cells were found in the positive range for CD34 (Fig. 2E) or c-Kit (Fig. 2F), but these were interpreted as doublets between hematopoietic and epithelial cells. However, we cannot exclude the possibility that a very small percentage of the epithelial cell fraction is CD34+ or c-Kit+. Using a combination of non–epithelial cell marking antibodies (Ter119, CD45, CD31, and c-Kit), we accounted for more than 95% of total fetal liver cells, and by negative selection for Ter119/CD45/c-Kit, epithelial cells were enriched to 25% of total cells (Fig. 3A-C). Positive selection, based on staining for E-cadherin, showed enrichment of FLEC from 2% in the initial cell preparation to 95% purity as determined by reanalysis on MoFlo (Fig. 3D-F). Epithelial cells were found to be clearly distinct from non-epithelial (hematopoietic/endothelial) cell populations in the fetal liver by staining for ALB, AFP, Liv2, E-cadherin, and Dlk, whereas Sca-1 was the only overlapping marker.

Figure 1.

FACS analysis of AFP+ cells for cell surface markers. Double-immunostaining for AFP and rat-isotype control (A) versus double-immunostaining for AFP and E-cadherin (B), Liv2 (C), or Dlk (D). All AFP+ cells stain positive for the surface markers and vice versa.

Figure 2.

FACS analysis for E-cadherin and HSC markers. Analysis of total cells (A-C) versus E-cadherin+ cells (D-F) after additional labeling for Sca-1, CD34, and c-Kit. Histograms for Sca-1 (A, D), CD34 (B, E), and c-Kit (C, F) (red) are presented in direct comparison to their isotype controls (blue). Crude fetal liver contains cell populations staining positive for Sca-1 (∼4%) (A), CD34 (∼1%) (B), or c-Kit (∼25%) (C), whereas cells gated for E-cadherin stain homogeneously positive only for Sca-1 (D). A very small number of E-cadherin+ cells were found in the positive range for CD34 (E) or c-Kit (F), but these were interpreted as doublets between hematopoietic and epithelial cells.

Figure 3.

FACS sorting of FLEC. (A) CD45+/c-Kit+ HSC in gate R1 are clearly separate from Ter119/CD45/c-Kit epithelial cells in gate R2 (negative selection). Immunostaining for E-cadherin shows <0.5% positive cells in gate R1 (B), but up to 25% positive cells in gate R2 (C). Immunostaining for rat-isotype control (D) versus E-cadherin (E) and reanalysis after positive selection (F) showed that E-cadherin+ FLEC were enriched from 1.7% to 95% by FACS sorting.


Double-label immunocytochemistry was performed on cytospins to confirm results obtained by FACS analysis. Crude fetal liver cell preparations contained approximately 4% AFP+ cells in relation to total nucleated cells. In double-immunocytochemistry, all AFP+ cells stained positive for PanCK and vice versa (Fig. 4A-C). Sorted cells were confirmed positive by immunocytochemistry for both AFP and PanCK (Fig. 4D-F). Immunostaining for ALB detected approximately 4% of total cells, and double-stainings revealed that all ALB+ cells expressed E-cadherin, Liv2, and Dlk and vice versa (Supplementary Fig. 2; Available at: http://www.interscience.wiley.com/jpages/0270-9139/suppmat/index.html). This was consistent with our findings using double-labeling for AFP together with these surface markers in FACS analysis and also confirmed previously published results showing the co-expression of ALB and E-cadherin or Dlk.23, 24 CK19, a marker for biliary differentiation, was not detected by immunocytochemistry at ED12.5.

Figure 4.

Immunocytochemistry for AFP and PanCK. Crude (A-C) and sorted cells (D-F) were immunostained for AFP (green) (A, D), PanCK (red) (B, E) and DAPI (blue). Merged pictures (C, F) show that all AFP+ cells are PanCK+ and vice versa. Few cells are stained in crude preparations (A-C), but >90% of cells are stained after sorting for E-cadherin (D-F). Original magnification: 400×.

Cell Culture.

Studies were performed to show that our cells exhibit the expected properties of fetal liver epithelial progenitor cells in culture. After isolation, crude ED12.5 fetal liver cells were plated on gelatin-coated dishes with up to 5% of total cells adherent within 24 hours. Up to 60% of these cells stained positive for PanCK. Using modified Block's medium, adherent cells formed epithelial clusters of >100 cells after 5 days (Fig. 5A). Isolated E-cadherin+ cells also grew into epithelial clusters within 5 days and showed the same characteristics as crude fetal liver cells (Fig. 5B). Messenger RNA (mRNA) from freshly isolated and cultured cells was amplified and compared with mRNA from unfractionated liver cells at different stages in liver development and from adult liver tissue (Fig. 5). ALB mRNA was strongly expressed in all samples, whereas AFP mRNA expression was weak at ED12.5, but readily detectable in later stages of fetal liver development (ED14.5 and ED18.5). CK19 mRNA was weakly expressed at ED12.5 but was strongly expressed after cell culture. G6P and tyrosine amino transferase mRNAs, although negative at ED12.5, were expressed at day 5 in cell culture in the presence of dexamethasone. Tryptophan oxygenase, a marker for terminal hepatocyte differentiation, was expressed in adult liver but not in isolated or cultured fetal liver cells. These results are consistent with the previous findings by Kamiya et al.29, 30 and demonstrate the expected properties of our isolated FLEC in vitro.

Figure 5.

Cell culture and reverse transcription–polymerase chain reaction. Epithelial cell clusters after 5 days in cell culture of (A) unfractionated fetal liver cells (original magnification: 200×) and (B) purified FLEC (original magnification: 200×). Reverse transcription–polymerase chain reaction for mRNA from ED12.5 fetal liver was compared with mRNA after cell culture, from different stages in liver development and from adult liver, using β-actin as internal control (C). In all samples, albumin (ALB) was strongly expressed. Alpha-fetoprotein (AFP) expression was weak at ED12.5, but was readily detectable after cell culture and at ED14.5 and ED18.5. Cytokeratin 19 (CK19) was weakly expressed at ED12.5, but was readily detectable after cell culture and after ED14.5. Glucose-6-phosphatase (G6P) and tyrosine amino transferase (TAT), although negative at ED12.5, were expressed at day 5 in cell culture in the presence of dexamethasone. Tryptophan oxygenase (TO) was expressed only in adult liver.

Transplantation Studies.

To show their in vivo differentiation, proliferation, and repopulation capacity, unfractionated and purified FLEC were transplanted into retrorsine-treated recipients. Our transplantation model used DPPIV−/− recipients, in which DPPIV+ donor cells can be readily detected by enzyme- or immunohistochemistry.28, 31, 32 The DPPIV enzyme is a differentiation marker that is not expressed by FLEC at ED12.5 but is expressed uniquely on the bile canalicular domain of differentiated hepatocytes, whereas biliary epithelial cells show diffuse cytoplasmic staining for DPPIV. DPPIV+ cells were detected in 19 of 21 retrorsine-treated recipients, 12 of 13 after transplantation of unfractionated fetal liver cells, and 7 of 8 after transplantation of sorted cells. At 1 month after transplantation of unfractionated fetal liver cells, DPPIV+ hepatocytes were detected in small clusters of up to 10 cells in 2-dimensional sections (Fig. 6A), whereas at 2 months after transplantation, DPPIV+ cells comprised hepatocytic clusters ranging in size from 20 to 100 cells per cluster (Figs. 6B, 7D). Markedly increased repopulation by DPPIV+ hepatocytes was observed at 4 months after cell transplantation with confluent clusters repopulating complete liver lobes and representing up to 80% of total liver mass (Figs. 6C, 8B, D). This time course reflects a high proliferative capacity of transplanted FLEC. These results were confirmed with purified FLEC, in which we saw small hepatocytic clusters of up to 10 donor-derived hepatocytes in 2-dimensional sections at 2 months (Fig. 6D) and large hepatocytic clusters with up to 100 hepatocytes per cluster at 4 months after transplantation (Fig. 6E-F). DPPIV+ hepatocytes were integrated into the liver parenchymal structure, as demonstrated in paraffin-embedded sections with hematoxylin-eosin staining (Fig. 8A). Immunostaining for ALB confirmed the hepatocytic phenotype of DPPIV+ donor cells (Fig. 8C). Normal adult hepatocytic function by the progeny of transplanted FLEC was further established by positive staining of DPPIV+ cells for G6P and glycogen in normal lobular distribution at 4 months after transplantation (Fig. 8D-F). As reported previously in the rat,33 clusters of donor-derived endothelial cells, positive for both DPPIV (cytoplasmic staining) and CD31 (PECAM-1), were also observed in the recipient's liver sinusoids (Fig. 7A-C) but were readily distinguished from donor-derived hepatocytes, which show canalicular staining for DPPIV and are CD31 (Fig. 7D-F). Both cell types integrated normally and line up side by side with the host's hepatocytes and endothelial cells, reconstituting the normal lobular structure. This observation also emphasizes the need for purification of epithelial cell fractions for transplantation studies in mouse models without hepatocyte-specific donor identification.

Figure 6.

Time course of liver repopulation by transplanted FLEC. DPPIV+ hepatocytes in clusters of progressively increasing size are shown at 1 month (A), 2 months (B), and 4 months (C) after transplantation of unfractionated fetal liver cells. DPPIV+ hepatocytic clusters derived from transplanted E-cadherin+–sorted cells are also shown at 2 months (D) and 4 months (E) after cell transplantation. (F) Large hepatocytic cluster at 4 months after transplantation of negatively sorted cells (Ter119/CD45/c-Kit). Original magnification: 400×.

Figure 7.

Immunohistochemistry for CD31 (PECAM-1) in DPPIV+ cell clusters after transplantation. Double-immunostaining for DPPIV (green) (A, D), CD31 (red) (B, E), and DAPI (blue). Merged picture (C) shows, at 2 months after transplantation of unfractionated fetal liver cells, 2 clusters of DPPIV+/CD31+ donor-derived sinusoidal endothelial cells, in which green and red fluorescence overlap (yellow), surrounded by DPPIV- host tissue. In the same animal, canalicular DPPIV staining shows a hepatocytic cluster of donor origin (D), and CD31 staining shows the recipient's sinusoidal endothelial cells (E). In the merged picture (F), DPPIV staining (donor-derived hepatocytes) remains totally separate from the recipient's DPPIV sinusoidal endothelial cells, indicating that these 2 cell types line up side by side in the host liver tissue. Original magnification: 200×.

Figure 8.

Normal integration and differentiation of transplanted FLEC. (A) Hematoxylin-eosin staining of the liver of retrorsine-treated mice at 4 months after transplantation of crude FLEC shows normal liver architecture. (B) In the same animal, DPPIV+ donor cells (red) represented 80% of total liver mass. (C) Immunohistochemistry shows normal albumin expression (red). Enzyme histochemistry for DPPIV (red) (D), G6P (brown) (E), and glycogen (purple) (F) in serial sections demonstrates that donor-derived DPPIV+ hepatocytes (D) express G6P (E) and show glycogen storage (F) in normal lobular distribution at 4 months after transplantation. Original magnification: 100×.


Distinction Between Hematopoietic and Hepatopoietic Progenitor Cells.

The fetal liver at ED12.5 provides an ideal opportunity to study common and distinct markers of HSC versus FLEC, because both of these cell types are present at this time during liver development. We studied antigens known to be expressed by HSC that have also been reported to be expressed on FLEC, such as CD343, 4 and c-Kit,3, 6 or on adult liver oval cells, such as CD45, CD34, Sca-1, and c-Kit.7–10, 12–14 Of these markers, only Sca-1 was clearly detected in both the hematopoietic and epithelial cell fractions. Sca-1 is a frequently used marker for enrichment and characterization of murine HSC.34 It has been shown to be required for the regulation of HSC self-renewal and for development of committed progenitor cells.35 In previous studies, Sca-1 has also been described in nonhematopoietic progenitor cell compartments, such as mammary gland epithelial progenitor cells,36 adult cardiac progenitors,37 and lung epithelial–specific progenitors,38 so that its expression in FLEC is not surprising.

C-Kit, the classical marker for HSC, was not expressed by ED12.5 FLEC, although a low percentage of c-Kit+ epithelial cells cannot be excluded. Our results are consistent with those of Suzuki et al., who reported that endodermal stem cells in the mouse fetal liver are c-Kit- at ED13.5,18, 19 and Tanimizu et al.,24 who reported that Dlk+ FLEC isolated from ED14.5 murine fetal liver are c-Kit. However, hematopoietic markers CD34 and CD45 were not expressed on mouse ED12.5 FLEC, in contrast to their expression on mouse oval cells.14

The percentage of epithelial cells, as defined by FACS analysis after staining for AFP or ALB, was approximately 2.5% of total cells, a percentage similar to that originally described by Sigal et al.16 in ED15 rat fetal liver. On cytospins, all AFP+ cells were also PanCK+ and vice versa. Using FACS analysis and double-immunostaining for the hepatic epithelial-specific cytoplasmic markers AFP and ALB and surface markers Liv2, E-cadherin and Dlk provided a definitive approach to characterize epithelial cell populations in the fetal liver and proved to be useful in separating these cells from HSC.

Liv2 has been identified as a specific marker for epithelial cells in the ED9.5 to ED12.5 murine fetal liver by Watanabe et al.22 and is not expressed in other tissues. Terai et al.39 used Liv2 to show transdifferentiation of transplanted bone marrow cells into hepatocytes, but the physiological role of Liv2 has not yet been determined. In our study, we demonstrated by double-labeling immunocytochemistry on cytospins, as well as in FACS analysis, that Liv2 is a highly specific surface marker for AFP+/ALB+ FLEC.

E-cadherin has been used previously by Nitou et al.23 as a selection marker for epithelial cells in the murine fetal liver, obtaining 95% purity by microbead technology. Unfortunately however, these investigators did not perform liver repopulation studies to demonstrate the in vivo potential of their enriched FLEC. In the fetal liver, staining for E-cadherin detected all AFP+ cells, and using E-cadherin in FACS sorting, we obtained up to 95% purity and a viability >90% by trypan blue dye exclusion. These cells were viable in cell culture and yielded high levels of repopulation after their transplantation.

Dlk is expressed from ED10.5 to ED18.5 in the murine fetal liver and disappears in neonatal and adult liver. Tanimizu et al.24 used Dlk to isolate epithelial cells from ED14.5 murine fetal liver by Automacs and FACS sorting, and Dlk was recently described as a marker for a subset of oval cells in hepatocytic differentiation in the 2-AAF/PH and retrorsine/PH rat models.40, 41 However, Dlk is not specific for epithelial lineages, as it is found at ED13.5 also in pituitary, lung, vertebra, and tongue.42 Dlk-null mice show no apparent defects in liver formation or hematopoiesis.43 We detected Dlk by double-label immunocytochemistry on cytospins and in FACS analysis, showing that ALB+/AFP+ cells are Dlk+ and vice versa, and demonstrating that Dlk is specific for epithelial cells in the mouse fetal liver.

Transplantation Studies.

The hallmark feature of a progenitor or stem cell is its ability to repopulate an organ or tissue with functionally differentiated progeny. Up to now, the in vivo differentiation potential of immuno-enriched murine fetal liver cells has been difficult to show. Suzuki et al.18 demonstrated small hepatocytic clusters after transplantation of sorted Ter119/CD45/c-Kit/CD49f+/CD29+ cells in retrorsine/CCl4-treated recipients,18 and Tanimizu et al.24 reported scattered GFP+/ALB+ cells after transplantation of Dlk+ cells into the Jo1/Fas model.24 Crude EGFP-marked murine fetal liver cells also showed clusters after transplantation into uPA transgenic mice, but repopulation was reported to be only 1% to 2%.44 In the current study, using the DPPIV cell transplantation model, we obtained up to 80% repopulation 4 months after transplantation of crude mouse fetal liver cells, proving engraftment, differentiated hepatocyte function, and replacement of whole liver lobules. In addition, we show large hepatocytic clusters after transplantation of purified ALB+/AFP+ murine FLEC, demonstrating their high proliferative potential and repopulation capacity in vivo. We also have demonstrated the hepatocytic phenotype of the repopulating cells by 4 different methods: enzyme histochemistry for DPPIV, immunohistochemistry for ALB, and enzyme histochemistry for G6P and glycogen. Suzuki et al.19 detected bile ducts after transplantation of cultured cells in the DAPM model, and Strick-Merchand et al.45 showed donor-derived bile ducts after transplantation of fetal liver epithelial cell lines into uPA transgenic mice. However, the retrorsine/CCl4 liver injury model favors hepatocytic proliferation and differentiation, and in the present study, we had only limited evidence for donor-derived bile ducts.

We also observed extensive repopulation of the recipient's liver sinusoids by transplanted endothelial cells, as determined by double-staining with DPPIV and CD31 (PECAM-1). The striking relation between the repopulation by endothelial and hepatocytic cells suggests a possible synergy, cooperation, or at least common mechanisms during engraftment between epithelial and endothelial cells. Hoppo et al.21 identified a mesenchymal cell fraction in the murine fetal liver that facilitated the in vitro culture of isolated epithelial cells, and this or a related cell fraction might also facilitate the engraftment of isolated epithelial cells after transplantation.

In conclusion, we characterized a population of ED12.5 murine FLEC that is essentially homogeneous for all analyzed markers (AFP, ALB, PanCK, Liv2, E-cadherin, Dlk, Sca-1), can be sorted with specific surface markers to high purity (95%), and can differentiate in vivo into mature, functional hepatocytes with high liver repopulation capacity. This cell population is clearly distinct from the abundant hematopoietic populations present at this stage of liver development. The in vivo properties of ED12.5 FLEC suggest that they will be particularly advantageous for liver repopulation studies in a wide variety of pathophysiologic states and in mouse models of human liver diseases.


The authors thank E. Hurston for technical assistance, Drs. H. Nishina and A. Miyajima for kindly providing Liv2 and Dlk antibodies, respectively, and A. Caponigro for secretarial assistance.