Identification and Characterization of Vitamin A-Storing Cells in Fetal Liver: Implications for Functional Importance of Hepatic Stellate Cells in Liver Development and Hematopoiesis


  • Hiroshi Kubota D.V.M., Ph.D.,

    Corresponding author
    1. Departments of Cell and Molecular Physiology, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, North Carolina, USA
    • Department of Animal Science, School of Veterinary Medicine, Kitasato University, 35-1, Higashi-23, Towada, Aomori 034-8628, Japan. Telephone: +81-176-24-9376; Fax: +81-176-23-8703
    Search for more papers by this author
  • Hsin-lei Yao,

    1. Department of Biomedical Engineering, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, North Carolina, USA
    Search for more papers by this author
  • Lola M. Reid Ph.D.

    Corresponding author
    1. Departments of Cell and Molecular Physiology, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, North Carolina, USA
    2. Department of Biomedical Engineering, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, North Carolina, USA
    3. Program in Molecular Biology and Biotechnology, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, North Carolina, USA
    • Campus Box #7038, University of North Carolina School of Medicine, Chapel Hill, North Carolina 27599-7038, USA. Telephone: 919-966-0347; Fax: 919-966-6112
    Search for more papers by this author


Hepatic stellate cells (HpSTCs) are major regulators of hepatic fibrogenesis in adults. However, their early development in fetal liver is largely unknown. To characterize fetal HpSTCs in the liver, in which hepatic development and hematopoiesis occur in parallel, we determined the phenotypic characteristics of HpSTCs from rat fetal livers, using a strategy focused on vitamin A. Storage of vitamin A in the cytoplasm is a unique characteristic of HpSTCs, permitting identification of them by vitamin A-specific autofluorescence (vA+) when excited with UV light using flow cytometry. A characteristic vA+ cell population was identified in liver as early as 13 days post coitum; it had a surface phenotype of RT1A intercellular adhesion molecule (ICAM)-1+ vascular cell adhesion molecule (VCAM)-1+ β3-integrin+. Although nonspecific autofluorescent cells were found with the antigenic profile of RT1A ICAM-1+ VCAM-1+, they were β3-integrin and proved to be hepatoblasts, bipotent hepatic parenchymal progenitors. In addition to expression of classic HpSTC markers, the vA+ cells were able to proliferate continuously in a serum-free hormonally defined medium containing leukemia inhibitory factor, which was found to be a key factor for their replication. These results demonstrated that the vA+ cells are fetal HpSTCs with extensive proliferative activity. Furthermore, the vA+ cells strongly express hepatocyte growth factor, stromal-derived factor-1α, and Hlx (homeobox transcription factor), indicating that they play important roles for hepatic development and hematopoiesis. The abilities to isolate and expand fetal HpSTCs enable further investigation into their roles in early liver development and facilitate identification of possibly novel signals of potential relevance for liver diseases.

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


Hepatic stellate cells (HpSTCs) are liver-specific mesenchymal cells in the space of Disse, located between the plates of parenchymal cells and the flanking endothelia. A prominent characteristic of HpSTCs is the presence of cytoplasmic lipid droplets containing vitamin A. Indeed, a major role of HpSTC is uptake, storage, and release of vitamin A compounds, which are requisite for embryonic development and, in adults, for vision and reproduction [1, [2]–3]. In mammals, approximately 50%–80% of the total body vitamin A is stored in HpSTCs [3].

HpSTCs are key effectors in hepatic fibrogenesis in adult liver [2, 4]. Although HpSTCs have been studied in adult livers, little is known about their embryonic origin and early development [5, 6]. In early liver development, endodermal cells in the foregut give rise to hepatic diverticulum around 10 days post coitum (dpc) in rats [7]. Subsequently, the hepatic diverticulum develops into the surrounding mesoderm, called the septum transversum, and forms hepatic cords. HpSTC are thought to derive from mesenchymal cells of the septum transversum, because histological analyses have suggested that mesenchymal cells in the septum transversum become trapped in the subendothelial space of the hepatic cords and are hypothesized to develop into HpSTCs [8, 9]. In rat liver at 13 dpc, the walls of sinusoids consist of two layers, endothelial cells and pericytes. The latter are dendritic cells and contain lipid droplets, and their cytoplasmic processes extend between parenchymal cells consisting of hepatoblasts, which are bipotent hepatic progenitors [8, [9]–10]. Although these dendritic cells are believed to be fetal HpSTCs, neither biological characteristics of the dendritic pericytes nor any HpSTC-specific marker expression has been studied to confirm that.

In the fetus, the liver is a primary site for hematopoiesis. Therefore, fetal HpSTCs may have unique roles in supporting fetal liver hematopoiesis, as well as hepatic development. Although the antigenic profiles of hematopoietic and hepatic cells in fetal livers have been studied [10, [11]–12], those of fetal HpSTCs have not. In this study, we determined the antigenic phenotype of fetal HpSTCs in the rat and revealed their unique phenotypic and biological characteristics.

Materials and Methods


Pregnant Fisher 344 rats were obtained from the Charles River Laboratories (Wilmington, MA, The morning on which the plug was observed was designated day 0. Male Fisher 344 rats (200–250 g) were used for isolation of adult HpSTCs. All animal experiments were conducted under the institutional guidelines. The University of North Carolina Institutional Animal Care and Use Committee approved all experimental procedures.

Cell Preparation

Fetal livers were isolated from 13–14-dpc rats and digested with 800 U/ml collagenase (Sigma-Aldrich, St. Louis, followed by further digestion with 0.5 mg/ml Trypsin-0.2 mg/ml EDTA · 4Na solution (Sigma-Aldrich). Subsequently, the cell suspension was treated with 200 U/ml DNase I (Sigma-Aldrich) [10]. Two litters of fetal rats were used to isolate cells, and average cell number obtained from 13- or 14-dpc fetal livers was 4.5 × 105 cells per liver (n = 302) or 2.4 × 106 cells per liver (n = 130), respectively. Isolation of HpSTCs from adult livers was performed using a standard protocol described elsewhere [13].

Cell Culture

Fetal liver cells were cultured on SIM mouse embryo-derived thioguanine and ouabain resistant (STO) cell feeders and in a serum-free hormonally defined medium (HDM) as described previously [10]. HDM consisted of a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F-12 medium (Invitrogen, Carlsbad, CA,, to which was added 2 mg/ml bovine serum albumin (Sigma-Aldrich), 5 μg/ml insulin (Sigma-Aldrich), 10−6 M dexamethasone (Sigma-Aldrich), 10 μg/ml iron-saturated transferrin (Sigma-Aldrich), 4.4 × 10−3 M nicotinamide (Sigma-Aldrich), 5 × 10−5 M 2-mercaptoethanol (Sigma-Aldrich), 7.6 μeq/l free fatty acid [14], 2 × 10−3 M glutamine (Invitrogen), 1 × 10−6 M CuSO4, 3 × 10−8 M Na2SeO3, and antibiotics (penicillin and streptomycin) [10]. STO feeder cells were prepared as previously described using STO5 transfected with pEF-Hlx-MC1neo [10]. For long-term culture of sorted vitamin A-specific autofluorescent cells, cells were cultured on STO feeders and in HDM supplemented with 10 ng/ml human leukemia inhibitory factor (LIF; Boehringer Mannheim, Mannheim, Germany, and 10 ng/ml epidermal growth factor (EGF; BD Biosciences, San Diego, Medium was changed every other day, and cells were subcultured to fresh STO feeders every week.


Staining procedures for cultured cells were described previously [10]. For nestin or desmin expression, cells were stained with anti-nestin antibody (Rat-401; Developmental Studies Hybridoma Bank, Iowa City, IA,∼dshbwww) or anti-desmin antibody (D33; DAKO, Glostrup, Denmark, followed by Alexa488-anti-mouse IgG (Molecular Probes Inc., Eugene, OR,

Flow Cytometry

Cells were analyzed and sorted by a FACStar Plus cell sorter (BD Biosciences) equipped with dual Coherent I-90 lasers. To detect vitamin A-specific autofluorescence, cells were excited at 351 nm, and fluorescence emission was detected with the use of 450DF20 filter (Omega Optical Inc., Brattleboro, VT, Fluorescence-conjugated antibodies were excited at 488 nm, and their fluorescence emission was detected by standard filters. All antibodies were obtained from BD Biosciences unless otherwise indicated. Monoclonal antibodies used for analysis of rat cells were fluorescein isothiocyanate (FITC)-anti-RT1Aa, b, l (B5), phycoerythrin (PE)-anti-rat intercellular adhesion molecule (ICAM)-1 (1A29), anti-rat vascular cell adhesion molecule (VCAM)-1 (5F10; Babco, Emeryville, CA,, anti-rat CD44 (OX-49), PE-anti-rat VCAM-1 (MR106), PE- or biotin-anti-rat β3-integrin (2C9.G2), biotin-anti-rat platelet/endothelial cell adhesion molecule (PECAM)-1 (TLD-3A12), and biotin-anti-rat Thy-1 (OX-7). For staining with unconjugated anti-VCAM-1 antibody (5F10), fetal liver cells were incubated with the anti-VCAM-1 antibody followed by staining with biotin-anti-mouse IgG2a monoclonal antibody (R19–15). Streptavidin-Cy-Chrome (BD Biosciences) was used to detect biotin-conjugated antibodies. For the experiments using fluorescence-activated cell sorting (FACS) to isolate long-term-cultured cells derived from sorted vitamin A-specific autofluorescent cells, all cells in the culture were harvested and stained with biotin-anti-mouse CD98 (H202-141) followed by streptavidin-Cy-Chrome to separate cultured rat cells and STO feeder cells [10]. To block nonspecific antibody binding, cells were incubated with 20% goat serum (Invitrogen), 1% teleostean gelatin (Sigma-Aldrich), and anti-rat CD32 antibody (D34–485) solution prior to antibody staining in FACS experiments.

Colony-Forming Assay

The protocol for a colony-forming assay (CFA) was described for hepatoblasts previously [10]. Briefly, sorted cells were placed on STO feeders in triplicate at 500 or 2,500 cells per well (3.8 cm2) in a 12-well plate and cultured in HDM for 14–15 days, with medium changes every other day. To determine whether the hepatoblasts are bipotent, double immunofluorescence staining of albumin and CK19 was performed. The colonies were stained with Diff-Quick (Baxter, Deerfield, IL, to count the number of colonies per well.

Cell Proliferation Assay

Vitamin A-specific autofluorescent cells isolated by FACS were placed in triplicate at 500 cells per well in a 96-well plates with HDM supplemented with laminin (BD Biosciences) at a final concentration of 8 μg/ml. LIF and EGF were added at the concentrations indicated. Five days after plating, cell cultures were rinsed twice to remove floating cells, and fresh medium was added with the tetrazolium salt WST-1 (Boehringer Mannheim) to measure the number of viable adherent cells [10]. After 4 hours, the absorbance was determined according to the manufacturer's protocol.

Reverse Transcriptase-Polymerase Chain Reaction

The primer sequences used for polymerase chain reaction (PCR) are shown in supplemental online Table 1. The procedure of reverse transcriptase (RT)-PCR for sorted cells by FACS was described previously [15]. cDNAs synthesized from total RNAs of sorted cells were normalized by cell number. The number of amplification cycles for each target gene varied; numbers of cycles are indicated in supplemental online Table 1.


Frozen sections (8 μm) were prepared from whole Fischer-344 rat fetuses obtained from Dr. M. Dabeva (Albert Einstein College of Medicine, Bronx, NY), mounted on slides, and stored at −80°C. For immunostaining, sections were air dried, fixed with cold acetone, and stained with primary antibodies. Primary antibodies used were mouse anti-rat VCAM-1 (MR106) and Armenian hamster anti-rat β3-integrin (2C9.G2). Goat Alexa488-anti-mouse IgG1 and goat Texas Red-anti-Armenian hamster IgG were used for secondary antibodies, respectively. The sections were counterstained with 4′,6-diamidino-2-phenylindole for visualization of cell nuclei and analyzed by confocal microscopy.


Identification of Vitamin A-Specific Autofluorescent Cells in Fetal Liver

Vitamin A-rich lipid droplets in the cytoplasm have been used as a unique characteristic for identification of HpSTCs. Vitamin A specifically produces a green-blue fluorescence when excited with the light of a 330–360-nm UV laser. Flow cytometric analysis (FCA), using a UV laser, is able to detect the vitamin A-specific green-blue fluorescence in the cytoplasm of HpSTCs in adult liver [16]. To identify fetal HpSTCs, we sought cells with the characteristic of vitamin A-specific green-blue fluorescence in fetal liver by flow cytometry. In a previous study, we identified hepatoblasts as RT1A (rat major histocompatibility complex class Ia) OX18low ICAM-1+ side scatter (SSC)high cells in 13-dpc liver of rat fetus [10]. Hepatoblasts showed relatively high SSC characteristics among the cell population of the fetal liver. Therefore, an autofluorescence signal produced by hepatoblasts must be distinguished from vitamin A-specific autofluorescence signal in FCA. Figure 1A shows the patterns of FCA of fetal liver cells at 13 dpc following staining with antibodies against RT1A and ICAM-1. Based on the SSC signal, two gates, R1 and R2, were created (Fig. 1A, entire population), and the expression patterns of RT1A and ICAM-1 were analyzed (Fig. 1A, R1 and R2). As expected, the majority of cells of R2 (SSChigh) were RT1A ICAM-1+ cells (Fig. 1A, R2, lower right), identified previously as hepatoblasts [10].

Figure Figure 1..

Flow cytometric analysis for autofluorescent cells in 13 days post coitum (dpc) rat fetal liver and lung. (A): The pattern of FSC and SSC of ALL. Based on the value of SSC, R1 and R2 gates were created and represented low (SSClo) and high (SSChi) SSC, respectively. Expression patterns of RT1A and ICAM-1 in the R1 and R2 are also shown. RT1A ICAM-1+ SSChi cells (R2, lower right) are hepatoblasts in the rat fetal liver [10]. The numbers in the grids indicate the percentage of each quadrant. (B): The autofluorescence pattern of ALL, R1, and R2 were analyzed with UV laser and 488-nm laser. UV laser-specific autofluorescent signal was detected with a 450-nm filter, whereas nonspecific autofluorescence (ns-autoflu) signal excited with a 488-nm laser was measured with a 530/30 band-pass filter. UV laser-specific autofluorescent cells were detected in R1 and R2 (upper left). (C): Expression of RT1A and UV laser-specific autofluorescence signal was studied. UV laser-specific autofluorescent cells were RT1A. The ns-autoflu+ RT1A cells (arrow) were identified and corresponded to the rat hepatoblast population. (D): UV laser-specific autofluorescence signal was analyzed in 13-dpc fetal lung cells. There were no UV-specific autofluorescent cells in the lung cell population. Most of the cells were RT1A, and no ns-autoflu cells (comparable to the hepatoblast population in the fetal liver) were detected. Abbreviations: ALL, entire population; FITC, fluorescein isothiocyanate; FSC, forward scatter; ICAM, intercellular adhesion molecule; PE, phycoerythrin; SSC, side scatter.

Subsequently, we analyzed autofluorescent activities in the 13-dpc fetal liver cells. The vitamin A-specific blue-green autofluorescence signal was measured by detecting the emission light with a 450-nm filter by excitation of a UV laser (351 nm) (described in Materials and Methods). To detect a non-UV laser-specific autofluorescence signal, a 488-nm laser and 530/30-nm band-pass filter was used. NADH, riboflavins, and flavin coenzymes commonly cause intrinsic cellular autofluorescence [17, 18], and the peak autofluorescence emission of these molecules after 488-nm excitation overlaps with the detection region of FITC, for which peak emission is 518 nm. Patterns of autofluorescence signals and UV laser-specific and non-UV laser-specific autofluorescence of whole fetal liver cell populations and two subpopulations (Fig. 1A, R1 and R2 gates) are shown in Figure 1B. In the pattern of the whole cell population (Fig. 1B, ALL), two distinct subpopulations with high autofluorescent characteristics were identified. One had an autofluorescence signal specific for UV light (Fig. 1B, ALL, upper left), which is referred to here as vitamin A-specific autofluorescence (vA+), whereas cells locating diagonally in the upper right quadrant indicate nonspecific autofluorescence, because the autofluorescence signals were detected with the 530- and 450-nm filters when excited by the 488-nm and UV lasers, respectively. The subpopulation with nonspecific autofluorescent characteristics (ns-autoflu+) exclusively derived from the SSChigh gate (Fig. 1A, R2; Fig. 1B, R2), whereas vA+ cells (Fig. 1B, upper left) were detected in both R1 and R2. Figure 1C shows the pattern of vitamin A-specific autofluorescence signal and RT1A expression, which was detected by an FITC-conjugated antibody against RT1A. FCA indicated that vA+ cells, as well as ns-autoflu+ cells, had no RT1A expression, because those two populations did not shift in the stained sample (Fig. 1C) compared with the control sample (Fig. 1B). In addition, FCA indicated that the RT1A ICAM-1+ SSChigh cells (Fig. 1A, R2, lower right), which represent the hepatoblast population, and RT1A ns-autoflu+ cells (Fig. 1C, R2, arrow) are overlapping populations by this FCA (data not shown).

To determine whether these autofluorescence signals were specific in fetal liver, fetal lung cells from the 13-dpc fetuses were isolated and analyzed by flow cytometry. The FCA showed there were neither ns-autoflu+ cells nor vA+ cells in the lung cells (Fig. 1D), indicating that the autofluorescence signals in particular subpopulations in the fetal liver represent unique phenotypic characteristics. These results clearly indicate that flow cytometry was able to detect characteristic vA+ cells in rat fetal liver as early as 13 dpc and that the vA+ cells were RT1A.

Vitamin A+ Cells Express VCAM-1 and β3-Integrin

VCAM-1 (CD106) has been identified as a unique surface marker of mature HpSTCs in adult liver [19]. Therefore, we analyzed VCAM-1 expression in fetal liver cells to investigate whether the vA+ cells express VCAM-1. By FCA, it appeared that approximately 15% of cells were VCAM-1+ in the 13-dpc fetal liver (Fig. 2A). We next analyzed the pattern of autofluorescence and RT1A expression of the VCAM-1+ cells. Interestingly, the VCAM-1+ cells contained essentially all vA+ cells, as well as the entire ns-autoflu+ cell population (Fig. 2A), indicating that HpSTCs and hepatoblasts express VCAM-1. FCA of two monoclonal antibodies against rat VCAM-1 (5F10 and MR106) showed an identical pattern of VCAM-1 expression (data not shown). In addition, fetal liver VCAM-1+ cells were RT1A ICAM-1+ cells because the R1 gate in Figure 2B included the VCAM-1+ cell population. These results indicate that fetal liver RT1A ICAM-1+ VCAM-1+ cells consist of vA+ cells, hepatoblasts, and some autoflu cells.

Figure Figure 2..

VCAM-1 and ICAM-1 expression on vitamin A-specific autofluorescence (vA+) cells. (A): Histogram of flow cytometry for VCAM-1 expression on 13 days post coitum (dpc) fetal liver. Approximately 15% of the cells express VCAM-1 on the cell surface. Closed and open histograms represent stained cells and unstained cells, respectively. VCAM-1+ and VCAM-1 cells were analyzed by flow cytometry for their autofluoresence signals. All vA+ cells and nonspecific autofluorescence (ns-autoflu)+ cells were VCAM-1-positive. The numbers in the grids indicate the percentage of each quadrant. (B): Two-color analysis of 13-dpc fetal liver cells for RT1A and ICAM-1. The R1 cell population (RT1A ICAM-1+) contained all vA+ cells and ns-autoflu+ cells. These results indicate that vA+ and ns-autoflu+ cells are VCAM-1+ RT1A ICAM-1+. Abbreviations: FITC, fluorescein isothiocyanate; ICAM, intercellular adhesion molecule; PE, phycoerythrin; VCAM, vascular cell adhesion molecule.

We next investigated surface antigens to distinguish the two autofluorescent populations, the vA+ cells and the hepatoblasts, both of which were RT1A ICAM-1+ VCAM-1+ cells. Because it has been reported that β3-integrin (CD61) is expressed on adult HpSTCs [20], two-color FCA of VCAM-1 versus β3-integrin was performed (Fig. 3B). The vA+ RT1A cells expressed β3-integrin, whereas ns-autoflu+ RT1A cells were clearly β3-integrin. autoflu RT1A cells contained some VCAM-1+ β3-integrin+ cells. The remaining major population (Fig. 3B, R4) was VCAM-1 and appears to correspond to the R2 cell population in Figure 2B. These cells were nonadherent when they were cultured on plastic dishes (data not shown). We also analyzed expression of PECAM-1 (CD31), which is known as an endothelial cell marker. FCA indicated that PECAM-1 expression in vA+ RT1A cells and ns-autoflu+ RT1A cells is negligible, whereas PECAM-1+ cells were detected in the autoflu RT1A and nonadherent cell populations (Fig. 3B). In addition, we analyzed expression of Thy-1 (CD90). Thy-1 is a surface marker for oval cells that appear in adult livers after oncogenic insults [21]. Oval cells share some characteristics with hepatoblasts in fetal liver [22, 23]. Nonetheless, FCA showed that ns-autoflu+ RT1A cells are Thy-1. By contrast, vA+ RT1A cells, autoflu RT1A cells, and nonadherent cells express Thy-1 heterogeneously. We found that ns-autoflu+ RT1A cells were CD44lo, whereas vA+ RT1A cells were CD44 (data not shown). CD44 is a cell-adherent molecule, and its ligand, hyaluronan, is a basic component of embryonic extracellular matrices [24]. CD44 appeared to be expressed differentially in the vA+ RT1A cells and ns-autoflu+ RT1A; however, the expression on the cell surface was rather weak. Together, these data suggest that only β3-integrin antibody staining, among all antibodies examined, facilitates distinguishing the vA+ cells and ns-autoflu+ cells, both of which were RT1A VCAM-1+ ICAM-1+ cell population in the fetal livers.

Figure Figure 3..

Antigenic profiles of vA+, ns-autoflu+, and autoflu RT1A cells in 13 days post coitum fetal liver. (A): Flow cytometric analysis for UV-autoflu and RT1A expression. In the RT1A cell population, three gates, vA+ RT1A (R1), autoflu RT1A (R2), and ns-autoflu+ RT1A (R3), were created based on the autoflu signals. The R4 gate covered the remaining cells in the fetal liver, which were nonadherent cells. (B): Two-color analysis of VCAM-1 versus β3-integrin, PECAM-1, and Thy-1 expression for each gated cell population (R1–R4). The numbers in the grids indicate the percentage of each quadrant. Primarily, R1 cells are VCAM-1+ β3-integrin+, whereas R3 cells uniformly express VCAM-1 but not β3-integrin, PECAM-1, or Thy-1. Abbreviations: autoflu, autofluorescence; ns-autoflu, nonspecific autofluorescence; PECAM, platelet/endothelial cell adhesion molecule; vA, vitamin A-specific autofluorescence; VCAM, vascular cell adhesion molecule.

To gain insight into in situ localization of vA+ cells in fetal liver, immunohistochemistry for VCAM-1 and β3-integrin was performed. Whereas VCAM-1 was detected throughout the liver parenchyma (Fig. 4A), we observed only a few β3-integrin+ cells in the same field (Fig. 4A). The β3-integrin+ coexpressed VCAM-1, and the VCAM-1+ β3-integrin+ cells showed a dendritic morphology with cytoplasmic processes extending between parenchymal cells. In addition, some of them located along sinusoids, suggesting that they are fetal HpSTCs.

Figure Figure 4..

Characterization of VCAM-1+ cells in fetal liver. (A): Confocal microscopy of a rat 14 days post coitum (dpc) fetal liver section immunostained with anti-VCAM-1 (green) and anti-β3-integrin (red) antibodies. Nuclear DNA was counterstained with 4′,6-diamidino-2-phenylindole (blue). Scale bar = 50 μm. (B): Immunocytochemistry of a bipotent hepatoblast colony. ns-autoflu+ RT1A VCAM-1+ β3-integrin cells were isolated by fluorescence-activated cell sorting (FACS) and placed on STO feeder cells in hormonally defined medium at a clonal cell density (250 cells in a well of 12-well plate; 66 cells per cm2). After 15 days in culture, the cells were fixed and stained with antibodies against albumin (red) and CK19 (green). Each colony was generated from a single sorted cell [10]. More than 95% (95.7% ± 0.4% [mean ± SEM]; n = 3) of hepatic colonies contained albumin+ CK19 and albumin CK19+ cells, which represent hepatocytic and biliary differentiation, respectively. Scale bar = 500 μm. (C): Reverse transcriptase-polymerase chain reaction analysis of 14-dpc fetal liver cells fractionated by FACS. Lane 1, ns-autoflu+ RT1A VCAM-1+ β3-integrin; lane 2, vitamin A-specific autofluorescence (vA+) RT1A VCAM-1+ β3-integrin+; lane 3, autoflu RT1A VCAM-1+; lane 4, autoflu RT1A VCAM-1; lane 5, remaining VCAM-1 cell population; lane 6, no cDNA. vA+ RT1A VCAM-1+ β3-integrin+ cells (lane 2) strongly expressed stromal cell-derived factor-1α and HGF. The vA+ cells were positive for HpSTC markers (desmin, nestin, vimentin, and SMαA) and negative for hepatoblast markers (albumin and Prox1). Abbreviations: HGF, hepatocyte growth factor; SDF-1α, stromal cell-derived factor; SMαA, smooth muscle α-actin; VCAM, vascular cell adhesion molecule.

Nonspecific Autofluorescent RT1A VCAM-1+ β3-Integrin Cells Are Hepatoblasts

In a previous study, we developed an in vitro CFA for hepatoblasts, which can differentiate to either hepatocytes or biliary epithelia, depending upon the microenvironment [10]. In the CFA, a single hepatoblast can generate a colony of cells with both hepatocytic and biliary markers. Using the assay, we proved that almost all hepatic cells at 13 dpc of the rat are indeed bipotent [10]. To examine whether vA+ cells have any potential to generate hepatoblast colonies, the CFA was performed. Four cell populations were isolated by FACS and subjected to the CFA for hepatoblasts: group 1, ns-autoflu+ RT1A VCAM-1+ β3-integrin; group 2, vA+ RT1A VCAM-1+ β3-integrin+; group 3, autoflu RT1A, and group 4, VCAM-1 nonadherent cells. The CFA clearly indicated that hepatic colonies were generated exclusively from group 1, ns-autoflu+ RT1A VCAM-1+ β3-integrin cells (Table 1). More than 95% of the hepatic colonies derived from the group 1-sorted cells contained both hepatocytic (albumin+ CK19) and biliary epithelial (albumin CK19+) cells (Fig. 4B). The colony-forming efficiency was approximately 31% (Table 1). A hepatic progenitor cell line (rhel4321) established in a previous study [10] had a colony efficiency in the CFA of 42.5% ± 1.8%. Therefore, the result of CFA in this experiment indicates that the group 1 sorted cells are nearly pure hepatoblasts, because the colony efficiency by established cell lines without sorting is assumed to be much higher than that of freshly isolated cells. These results clearly demonstrated that whereas the ns-autoflu+ RT1A VCAM-1+ β3-integrin cells are hepatoblasts, the other groups of sorted cells, including the vA+ RT1A VCAM-1+ β3-integrin+ cells, do not contain hepatoblasts.

Table Table 1.. The frequency of hepatic colonies from sorted rat fetal liver cells
original image

Gene Expression of Freshly Isolated Vitamin A+ RT1A VCAM-1+ β3-Integrin+ Cells

We next analyzed the gene expression pattern of vA+ RT1A VCAM-1+ β3-integrin+ cells to examine whether they express various markers for HpSTCs. Five populations were isolated by FACS, and RNAs were isolated from the five populations. RT-PCR for HpSTC markers was performed using cDNAs synthesized from the RNAs. The five populations were as follows: (a) ns-autoflu+ RT1A VCAM-1+ β3-integrin, (b) vA+ RT1A VCAM-1+ β3-integrin+, (c) autoflu RT1A VCAM-1+, (d) autoflu RT1A VCAM-1, and (e) VCAM-1 nonadherent cell population. HpSTCs in adult liver express the intermediate filaments desmin and nestin [25, 26], which are not expressed in other cell types in the liver. Vimentin is expressed broadly in mesenchymal cells, whereas smooth muscle α-actin (SMαA) is expressed in myogenic cells. Expressions of vimentin and SMαA increase after activation of HpSTCs [2, 4]. RT-PCR analyses showed that vA+ RT1A VCAM-1+ β3-integrin+, autoflu RT1A VCAM-1+, and autoflu RT1A VCAM-1 cells expressed all four intermediate filaments (Fig. 4C). ns-autoflu+ RT1A VCAM-1+ β3-integrin cells express albumin, as well as Prox1 (Fig. 4C), which is a transcription factor expressed specifically in hepatoblasts [27]. This result was consistent with the data obtained from the CFA, which demonstrated that ns-autoflu+ RT1A VCAM-1 β3-integrin cells were hepatoblasts (Table 1). There was no expression of nestin, SMαA, or vimentin in the hepatoblasts. A weak expression of desmin in the hepatoblasts (Fig. 4C) agrees with the results of the prior immunolocalization study for desmin with confocal microscopy [28]. The expression of HpSTC-specific intermediate filaments strongly suggests that vA+ cells are fetal HpSTCs.

Subsequently, expression of hepatocyte growth factor (HGF), stromal cell-derived factor-1α (SDF-1α), and the divergent homeobox transcription factor Hlx were investigated using RT-PCR. HGF is required for normal hepatic development, especially for proliferation and differentiation of hepatoblasts in the mouse [29]. In adult liver, HpSTCs are major producers of HGF [30]. SDF-1α is a potent chemokine for hematopoietic progenitors, and hematopoietic stem cells in fetal liver migrate in response to the chemokine [31]. Hlx is expressed in mesenchymal cells in developing fetal liver and plays an indispensable role in fetal liver hematopoiesis and hepatic development [32]. Interestingly, vA+ RT1A VCAM-1+ β3-integrin+ cells expressed HGF, SDF-1α, and Hlx transcripts most strongly among all cell fractions examined (Fig. 4C).

Ex Vivo Clonal Expansion of Vitamin A+ RT1A VCAM-1+ β3-Integrin+ Cells

Although HpSTCs isolated from adult liver can proliferate in vitro in serum-supplemented media, they usually transform to myofibroblast cells and show only limited proliferative activity [2]. Fetal HpSTCs may have extensive proliferative potential when cultured under the appropriate conditions. Therefore, we investigated the ex vivo growth capability of the vA+ cells in fetal livers. When vA+ RT1A VCAM-1+ β3-integrin+ cells isolated by FACS were cultured in HDM at a cell density of 500 cells per well of 96-well plates for 5 days in the presence of LIF, a pleiotropic growth factor for different types of cells including embryonic stem cells or myogenic cells [33, 34], the cells expanded in a dose-dependent manner (Fig. 5A). In addition, EGF potentiated the proliferation of fetal HpSTCs induced by LIF but did not support the expansion on its own (Fig. 5A).

Figure Figure 5..

In vitro culture of vitamin A-specific autofluorescence (vA+) RT1A vascular cell adhesion molecule (VCAM)-1+ β3-integrin+ cells isolated by fluorescence-activated cell sorting (FACS). (A): Effect of LIF and EGF on in vitro proliferation of vA+ RT1A VCAM-1+ β3-integrin+ cells. Five hundred vA+ RT1A VCAM-1+ β3-integrin+ cells isolated by FACS were placed per well of a 96-well plate with hormonally defined medium (HDM) plus laminin-supplemented LIF and/or EGF at the concentrations indicated. After 5 days of culture, the degree of cell proliferation was measured by the tetrazolium salt WST-1. LIF support proliferation of the vA+ cells at as low as 0.1 ng/ml. EGF slightly improved the vA+ cell proliferation. (B): Effect of LIF and EGF on vA+ RT1A VCAM-1+ β3-integrin+ cells cultured with STO feeders. Two hundred fifty vA+ RT1A VCAM-1+ β3-integrin+ cells isolated by FACS were seeded on STO feeder cells in HDM with EGF and/or LIF. Twelve-well plates were used. The cultures were stained with Diff-Quick after a 2-week culture period. Although STO cells express LIF, the amount of the production was not adequate to support clonal expansion of the cells in the absence of exogenous LIF supplementation. Exogenous LIF and addition of EGF dramatically improved clonal expansion of the vA+ cells. (C): Immunocytochemistry of colonies derived from vA+ RT1A VCAM-1+ β3-integrin+ cells isolated by FACS. Cells were placed on STO feeders in HDM supplemented with EGF and LIF. Fifteen days after in vitro culture, cultures were stained with antibodies for desmin or nestin. Colony-forming cells expressed nestin and desmin, whereas STO cells did not express either. Scale bar = 200 μm. (D): Immunocytochemistry of 2-month-cultured vA+ RT1A VCAM-1+ β3-integrin+ cells isolated by FACS. Sorted cells were placed on STO feeders in HDM supplemented with EGF and LIF. Proliferating cells were subcultured five times on fresh STO feeders. Cultured cells were stained with antibodies for desmin or nestin. Proliferating cells maintained the expression of nestin and desmin during the entire culture period. Scale bar = 200 μm. Abbreviations: EGF, epidermal growth factor; LIF, leukemia inhibitory factor.

The proliferation, however, did not persist in the culture conditions making use of plastic culture plates. We next placed the sorted vA+ RT1A VCAM-1+ β3-integrin+ cells on STO feeders [10]. Although STO cells produce LIF, exogenous LIF and supplementation of EGF dramatically supported colony formation from sorted vA+ RT1A VCAM-1+ β3-integrin+ cells (Fig. 5B). Proliferating cells in the culture expressed desmin and nestin, whereas STO feeders did not express either (Fig. 5C). Three single colonies were picked and placed on fresh STO feeders. The single colony-derived cells continued to proliferate in the cocultures with STO feeders supplemented with LIF and EGF for 2 months, indicating that they have extensive growth potential. Expressions of desmin and nestin were maintained in the proliferating cells (Fig. 5D). To further compare the characteristic phenotypes of 2-month-cultured cells with those of freshly isolated vA+ RT1A VCAM-1+ β3-integrin+ cells, RT-PCR was performed. Single colony-derived cells (A428-3) that were maintained for 2 months in culture were separated from STO feeder cells by FACS, and the RNA was extracted for RT-PCR analysis. RNA was isolated from STO feeder cells that were sorted simultaneously as a control. In addition, RNA was isolated from adult HpSTCs to compare with those from A428-3 and STO cells. The results demonstrated that A428-3 expressed desmin, nestin, SMαA, vimentin, β3-integrin, SDF-1α, HGF, and Hlx, indicating the expression pattern was similar to that of fresh vA+ RT1A VCAM-1+ β3-integrin+ cells (Figs. 5, 6A). Furthermore, VCAM-1 expression was confirmed by FCA (Fig. 6B). RT1A expression appeared to be induced by in vitro culture (Fig. 6B). The RT-PCR results of adult HpSTCs agreed with previous reports, in which the phenotype of normal adult HpSTCs is desmin+, glial fibrillary acidic protein (GFAP)+, and HGF+ but SMαAlo/−. The results also showed that adult HpSTCs express SDF-1α, β3-integrin, and Hlx. There was expression of GFAP neither in A428-3 cells (Fig. 6A) nor in any fractions tested in fetal liver (data not shown).

Figure Figure 6..

Phenotypic characteristics of 2-month-cultured vitamin A-specific autofluorescence (vA+) RT1A vascular cell adhesion molecule (VCAM)-1+ β3-integrin+ cells. (A): Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of cultured vA+ RT1A VCAM-1+ β3-integrin+ cells. Cells were isolated by fluorescence-activated cell sorting (FACS) and cultured on STO feeders in the hormonally defined medium with EGF and LIF. After 2-month culture, cells were fractionated by FACS. Proliferating rat cells and mouse STO feeder cells were fractionated by FACS following antibody staining of mouse CD98 monoclonal antibody. CD98 was expressed on mouse STO cells, and the monoclonal antibody reacted specifically to mouse CD98 but not rat CD98. RNAs were isolated from vA+-derived rat cells and STO cells. Normal rat HpSTCs were also used, and the RNA was isolated as a control. The cDNAs were synthesized from those RNAs and subjected to PCR with primers specific for various transcripts that expressed in HpSTCs. (B): Flow cytometry for cultured vA+ RT1A VCAM-1+ β3-integrin+ cells. Cells used for RT-PCR were stained with anti-VCAM-1 or anti-RT1A antibody and mouse CD98 antibody. The CD98-negative fraction was analyzed for VCAM-1 or RT1A expression. Continuously proliferating cells derived from vA+ cells in rat fetal livers expressed VCAM-1 and RT1A uniformly under the culture conditions examined. Abbreviations: GFAP, glial fibrillary acidic protein; HGF, hepatocyte growth factor; SDF-1α, stromal cell-derived factor-1α; SMαA, smooth muscle α-actin; VCAM, vascular cell adhesion molecule.


Vitamin A-storing cells in fetal liver were identified by flow cytometry using the specific autofluorescence generated by cytoplasmic vitamin A-rich lipid droplets. The surface phenotype of vA+ cells appeared to be uniform and consisted of cells that are RT1A ICAM-1+ VCAM-1+ β3-integrin+ PECAM-1. VCAM-1 has been shown to be a unique surface marker to distinguish HpSTCs from myofibroblasts in adult liver [19]. ICAM-1 and β3-integrin expressions on HpSTCs have also been demonstrated in previous studies [13, 20]. In addition to those surface markers, vA+ cells express intermediate filaments specific for HpSTCs, including desmin, vimentin, SMαA, and nestin. These molecular markers have been used to identify HpSTCs in adult liver [2, 4, 25, 26]. These results strongly support the interpretation that the vA+ cells are fetal HpSTCs.

VCAM-1, an immunoglobulin gene superfamily of adhesion molecules, plays a key role in supporting adhesion of hematopoietic progenitors to bone marrow stromal cells [35]. Although fetal liver is a major hematopoietic organ in fetus, hematopoietic cells are originally generated in yolk sac and the aorta-gonad-mesonephros region. Shortly after that, they colonize the fetal liver and produce large numbers of hematopoietic progenitors until birth [36]. In the later developmental stage in utero, hematopoietic stem cells move to bone marrow, which becomes the primary hematopoietic organ throughout the postnatal life [37]. Adhesion molecules expressed on bone marrow stromal cells or endothelial cells are important for homing process of hematopoietic stem/progenitor cells [38]. Studies using blocking antibodies for VCAM-1 or very late antigen (VLA)-4, which is expressed on hematopoietic stem/progenitors, demonstrated that hematopoietic stem/progenitor cells failed to colonize to bone marrow when the VCAM-1/VLA-4 adhesion pathway was blocked [39, 40]. Therefore, VCAM-1 in fetal liver is likely to involve establishment of fetal liver hematopoiesis.

Although FCA indicated that high VCAM-1 expression was detected on hepatoblasts and fetal HpSTCs, the latter may play more important roles for hematopoietic cells, because they express SDF-1α as well. SDF-1α is a potent chemoattractant for hematopoietic stem cells, which express CXCR4, the receptor for SDF-1α [41]. The chemokine plays a central role during the migration of hematopoietic stem/progenitor cells to bone marrow [42]. It was shown that SDF-1α upregulates VLA-4-dependent adhesion to VCAM-1 [43]. Therefore, it is possible that SDF-1α and VCAM-1 expressions on fetal HpSTCs are crucial to recruit hematopoietic stem/progenitor cells into fetal liver.

Interestingly, our study reveals that VCAM-1 is expressed by hepatoblasts as well as HpSTCs. In addition to the surface antigenic profile and the cells' mRNA expression, the culture assays clearly demonstrated that VCAM-1+ cells are hepatoblasts. This finding is unexpected because VCAM-1 is known as a surface marker for mesenchymal cells, such as endothelial cells, myogenic cells, or HpSTCs [19, 44, 45]. The expression appears to be developmentally controlled because adult hepatocytes are VCAM-1 by FCA (data not shown), which agrees with previous studies [46]. Several lines of evidence demonstrated that VCAM-1 and VLA-4 interaction occurs on myogenic cells and controls myogenesis in addition to the adhesion pathway between hematopoietic stem cells and endothelial cells; therefore, it is worthwhile to elucidate the biological functions of VCAM-1 on hepatoblasts. VCAM-1 expression has been found also on mouse hepatoblasts (H. Kubota, unpublished data), suggesting that the physiological role might be conserved across species.

Our data indicate that fetal HpSTCs are major HGF producers in the fetal liver. HGF is a crucial growth factor for hepatic development [29], and the factor is responsible for liver parenchymal cell growth during liver regeneration as well [47]. In addition, it has been shown that HpSTCs, but not parenchymal cells, endothelial cells, or Kupffer cells, express HGF in adult liver [30]. Therefore, our data and previous studies suggest that HpSTCs are main HGF producers from fetuses to adults in the liver. Interestingly, a recent study indicated that HGF potentiated SDF-1α-mediated recruitment of hematopoietic progenitors to the liver [48]. The study suggested that HGF produced from HpSTCs in injured adult liver is important for homing of hematopoietic cell into the livers. Likewise, fetal HpSTCs may play a crucial role for hepatic and hematopoietic development in the fetal liver because the cells are main producers for HGF and SDF-1α.

A divergent homeobox protein, Hlx, is expressed in septum transversum and mesenchymal cells in fetal liver [49]. A previous study of Hlx knockout mice demonstrated that the mutant mice have impaired hepatic development and fetal liver hematopoiesis [32]. A transplantation experiment indicated that the hematopoietic defect was caused by the fetal liver microenvironment but not by the hematopoietic progenitors per se. Thus, Hlx+ cells are a crucial cell population in fetal liver for supporting hepatic and hematopoietic development. Our data indicated that fetal HpSTCs in the rat strongly expressed Hlx, and we have found that a cell population in the mouse fetal liver expresses desmin, nestin, VCAM-1, HGF, SDF-1α, and Hlx as well (H. Kubota, unpublished data). Therefore, it is interesting to examine whether Hlx knockout mice have this cell population, which is a counterpart of rat fetal HpSTCs. Although the relationship between Hlx expression and HpSTC development is not clear, loss of Hlx expression in fetal HpSTCs may cause defects in the biological function of HpSTCs in the developing liver of the mutant mice. Furthermore, considering the unique phenotypic and functional characteristics of fetal HpSTCs, including expression of VCAM-1 and Hlx and production of HGF and SDF-1α, the cells might comprise a stem cell niche for hematopoietic stem cells, hepatic stem cells, or both in the liver.

GFAP is a marker used to identify HpSTCs in adult liver [50]. However, we did not detect GFAP mRNA by RT-PCR in any cell fractions examined or in the whole fetal liver sample. In addition, even after culture of isolated fetal HpSTCs, GFAP expression was not induced, whereas desmin and nestin expressions were sustained in cultured cells. This result suggests that differentiation of fetal HpSTCs will result in GFAP expression in a later developmental stage. We cannot, however, exclude another possibility, in which GFAP+ cells are derived from different precursors that do not exist in the 13-dpc fetal livers. Circulating cells in blood flow may be a source of the alternative cellular origin [51]. However, the majority of HpSTCs express GFAP in adult liver; therefore, the minor contribution of circulating cells from the blood is unlikely to become a dominant population in the liver. Thus, it seems more likely that acquisition of GFAP expression happens during maturation of HpSTCs.

Fetal HpSTCs that were purified by FACS proliferated under serum-free conditions supplemented with EGF and LIF. With the support of STO feeders, fetal HpSTCs replicated continuously for more than 2 months. The phenotypes of fresh and in vitro-cultured fetal HpSTCs were similar, indicating that the culture condition did not transform HpSTCs to myofibroblast cells, which commonly happens in serum-supplemented conditions. Until now, HpSTCs from adult livers have been cultured in medium supplemented with animal serum. HpSTCs cultured in the serum-supplemented medium give rise to myofibroblastic cells, which acquire fibroblastic characteristics and lose the original HpSTC phenotypes. Although the myofibroblastic cells can replicate and expand in culture, they eventually cease proliferation. Therefore, the serum-supplemented medium conditions are not appropriate for the culture of HpSTCs. Recently, LIM homeobox protein Lhx2 was identified as a negative regulator for myofibroblastic transformation of HsSCs [52]. In addition, Lhx2 was found as a molecule that is important for maintenance of hair follicle stem cells, in which the transcriptional factor maintains the undifferentiated status of the stem cells [53]. Thus, Lhx2 might play a similar role to maintain a fetal stage of HpSTCs. The culture system developed in this study will facilitate the investigation of molecular mechanisms of proliferation and differentiation of HpSTCs.

One of the important questions to be addressed in future studies is how to resolve the reports of shared antigens on endothelial cells, hepatoblasts, biliary epithelia, oval cells, and fetal liver-derived HpSTCs [54, [55], [56]–57]. Some of this confusion has been eliminated with the recent findings that angioblasts and HpSTCs are tightly bound to the surface of hepatic stem cells and hepatoblasts; recognition of the separate subpopulations requires rigorous flow cytometric analyses [57]. The flow cytometrically purified subpopulations comprise hepatic stem cells, hepatoblasts, biliary epithelia with antigenic profiles distinct from those of angioblasts, endothelia, and HpSTCs [58]. Thus, the overlap in antigen expression is actually among HpSTCs, angioblasts, and endothelia. The relevance of this is unknown, but it hints at possible lineage connections, a hypothesis now under study.

In this study, we identified culture conditions that are wholly serum-free and with only defined and purified factors in the medium, conditions under which the vA+ cells expanded clonogenically and yielded only the vA+-like cells and no other cell types. This strongly suggests that the vA+ characteristics are exclusive to fetal HpSTCs. However, we cannot exclude the possibility that the addition in culture of some soluble and/or matrix signal or transplanting cells in vivo might induce these cells to other fates. In vitro differentiation studies or in vitro lineage tracing studies to see what progeny are generated could shed light on unidentified biological characteristics of the vA+ cells.

An interesting question is whether fetal-type HpSTCs are also evident in adult livers. Given the surface markers that we have identified, it is possible to try to immunoselect these cells from adult livers and then further characterize them using the serum-free culture system that we developed and that was found to maintain the unique characteristics of fetal HpSTCs ex vivo. Transplantation of fetal liver-derived HpSTCs or fetal-type HpSTCs is a plausible cell therapy through which to study the alteration of some of the aberrations described for adult fibrogenic livers. In addition, the fetal HpSTCs are likely to express novel factors that may prove regulatory of hepatopoiesis and/or hemopoiesis and therefore offer new directions for novel therapeutic approaches for liver diseases.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.


We thank Dr. M. Cook of the Duke University Cancer Center Flow Cytometry Core for cell sorting, K. McNaughton for frozen section preparations, E. Wauthier for management, L. English for technical support, Dr. Y.W. Rong of the Center for Gastrointestinal and Biliary Disease Biology for hepatocyte preparations, and Dr. T. Tallheden for technical assistance of confocal microscopy. We are especially grateful to Dr. M. Dabeva for the gifts of cryopreserved rat fetuses and Dr. D. Shafritz for a paraffin block of a fetal liver. Funding was provided by Vesta Therapeutics and NIH Grants DK52851, AA014243, and IP30-DK065933. H.K. is currently affiliated with the Department of Animal Science, Kitasato University School of Veterinary Medicine, Aomori, Japan.