Human livers contain two pluripotent hepatic progenitors, hepatic stem cells and hepatoblasts, with size, morphology, and gene expression profiles distinct from that of mature hepatocytes. Hepatic stem cells, the precursors to hepatoblasts, persist in stable numbers throughout life, and those isolated from the livers of all age donors from fetal to adult are essentially identical in their gene and protein expression profiles. The gene expression profile of hepatic stem cells throughout life consists of high levels of expression of cytokeratin 19 (CK19), neuronal cell adhesion molecule (NCAM), epithelial cell adhesion molecule (EpCAM), and claudin-3 (CLDN-3); low levels of albumin; and a complete absence of expression of α-fetoprotein (AFP) and adult liver-specific proteins. By contrast, hepatoblasts, the dominant cell population in fetal and neonatal livers, decline in numbers with age and are found as <0.1% of normal adult livers. They express high levels of AFP, elevated levels of albumin, low levels of expression of adult liver-specific proteins, low levels of CK19, and a loss of NCAM and CLDN-3. Mature hepatocytes lack expression altogether of EpCAM, NCAM, AFP, CLDN-3, cytokeratin 19, and have acquired the well-known adult-specific profile that includes expression of high levels of albumin, cytochrome P4503A4, connexins, phosphoenolpyruvate carboxykinase, and transferrin. Thus, hepatic stem cells have a unique stem cell phenotype, whereas hepatoblasts have low levels of expression of both stem cell genes and genes expressed in high levels in mature hepatocytes.
During development, the hepatic bud arises from the foregut endoderm [1, –3]. Liver organogenesis takes place by invasion of the endoderm into the mesenchyme of the septum transversum, resulting in the formation of the ductal plates (also called limiting plates), a band of cells around the portal triads. Hepatic stem cells in fetal and neonatal livers have been found recently to be located in the ductal plates  (E. Schmelzer et al., manuscript submitted for publication). These give rise to hepatoblasts, the dominant cell type of fetal and neonatal livers. Hepatoblasts, in turn, give rise to the hepatocytic and biliary lineages, the hepatocytes and cholangiocytes [4, 5] (E. Schmelzer, L. Zhang, A. Melhem, N.G. Moss, E. Wauthier, R.F. McClelland, A. Bruce, H. Yao, W.S. Turner, N. Cheng, M.E. Furth, and L.M. Reid, manuscript submitted for publication). In the adult liver, stem cells have been found associated with the Canals of Hering .
The ratio of hepatic stem cells to hepatoblasts changes depending on developmental stage, with the hepatoblasts being dominant in fetal and neonatal livers and with few if any mature parenchymal cells . In addition, the fetal liver is the primary site of hematopoiesis, consisting of up to 60% erythrocytes at certain developmental stages . By contrast, in pediatric and adult human livers, the hepatic stem cells are the dominant pluripotent progenitor (0.3%–0.7%); the hepatoblasts are few (<0.1%), and the majority (>98%) of the parenchymal cells are diploid and polyploid hepatocytes and biliary epithelia (Schmelzer et al., manuscript submitted for publication). These findings are summarized in a recent review .
Although the identity of hepatic stem cells and hepatoblasts in human livers has been described recently, our knowledge is incomplete on the gene and protein expression profiles of these pluripotent progenitors and whether the profiles change during development. We have shown that epithelial cell adhesion molecule (EpCAM), a 34–40-kDa transmembrane glycoprotein  first described by Spurr et al.  and cloned by Strnad et al.  and Szala et al.  is expressed by hepatic stem cells, hepatoblasts, and committed progenitors but not by mature hepatocytes  (Schmelzer et al., manuscript submitted for publication; Bruce et al., manuscript submitted for publication). Thus, sorting for EpCAM results in only progenitor cells but in distinct ratios of hepatic stem cells to hepatoblasts, depending upon whether the tissue is fetal, neonatal, or adult. In these studies, we analyzed the gene expression profiles in hepatic stem cells and hepatoblasts isolated from fetal, neonatal, pediatric, and adult livers and compared these profiles to those of mature hepatocytes.
Neonatal, Pediatric, and Adult Liver Cell Isolation
Donated livers not suitable for orthotopic liver transplantation were obtained from federally designated organ procurement organizations. Informed consent was obtained from next of kin for use of the livers for research purposes. To isolate cells, adult and pediatric livers were perfused through the portal vein and hepatic artery with EGTA-containing buffer for 15 minutes and 125 mg/l Liberase (Roche Applied Science, Indianapolis, http://www.roche-applied-science.com) for 30 minutes at 34°C. Cells were passed sequentially through filters of pore size 1,000, 500, 250, and 150 μm and centrifuged in OptiPrep density gradients at 500g to select for viable cells.
Fetal Liver Cell Isolation
Fetal livers were obtained from Advanced Biological Resources (Alameda, CA) and were of developmental stages between 16 and 20 weeks of gestation. Nonhepatic tissue was removed, and livers were minced by scalpel and digested at 37°C by 0.06% collagenase (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) in RPMI 1640 (Gibco, Grand Island, NY, http://www.invitrogen.com) supplemented with 0.03% deoxyribonuclease (w/v), 0.1% (w/v) bovine serum albumin fatty acid-free, 30 nM selenium (Sigma-Aldrich), and antibiotic-antimycotic mixture (Gibco). Hematopoietic cells and nonparenchymal cells were separated from the parenchymal cell fraction by slow-speed centrifugations (30g for 5 minutes in 40 ml) until the pellet showed minimal contamination with red blood cells. Cells in suspension were filtered through a 70-μm nylon mesh, and flow through was applied on a Ficoll-Paque (Amersham Biosciences-GE Healthcare, Piscataway, NJ, http://www.amersham.biosciences.com) gradient (cells in a 5-ml suspension applied on 5 ml of 84% Ficoll, placed on 5 ml of 100% Ficoll) and centrifuged for 30 minutes at 4°C at 980g. Cells forming an interphase were collected and washed once in RPMI 1640 supplemented with 0.1% bovine serum albumin fatty acid-free (w/v), 30 nM selenium, 0.054% niacinamide (w/v), insulin (0.135 USP units/ml), bovine apo-transferrin (10 μg/ml), free fatty acid mixture (2.36 μM palmitic acid, 0.21 μM palmitoleic acid, 0.88 μM stearic acid, 1.02 μM oleic acid, 2.71 μM linoleic acid, 0.43 μM linolenic acid), 1 μM M hydrocortizone, 50 μM β-mercaptoethanol (Sigma-Aldrich), 2 mM l-glutamine, and antibiotic-antimycotic mixture (Gibco) and kept on ice for further handling.
Cell sorting for EpCAM+ cells was done using magnetic-activated cell sorting (MACS) according to manufacturer's instructions (Miltenyi Biotec, Auburn, CA, http://www.miltenyibiotec.com). Cells were incubated in the dark at 4°C with fluorescein isothiocyanate (FITC)-conjugated antibody (Miltenyi Biotec) for 20 minutes at a concentration of 50 μl for 50 × 106 total cells in 500 μl of buffer (phosphate-buffered saline [PBS]) containing 0.5% bovine serum albumin and 2 mM EDTA (Sigma-Aldrich). EpCAM+ cells were labeled with magnetic beads using an anti-FITC Multisort Kit (Miltenyi Biotec) and selected by Midi- or MiniMACS columns and separation units (Miltenyi Biotec). All incubation and selection steps were performed on ice or at 4°C with addition of 10% accutase (Innovative Cell Technologies, San Diego, CA) to prevent aggregation of cells.
Cells obtained by Ficoll gradient selection were placed in culture on plastic or collagen III (6.25 μg/cm2) (Sigma-Aldrich)-coated plastic vessels. Cultivation was done in supplemented RPMI 1640 medium (see cell isolation) with an initial 24-hour phase of 10% fetal bovine serum addition. After 24 hours, cells were cultured serum-free, and medium was changed every 4 days. Colonies that formed were picked at 10 and 20 days and collected on ice for further analyses.
Following enzymatic digestion of fetal livers, freshly isolated cells were centrifuged at 600g for 5 minutes at 4°C. Cells were then resuspended in PBS and 4% paraformaldehyde (1:1). Following fixation for 10 minutes at room temperature, cells were centrifuged as previously described. The supernatant was aspirated, and the cells were then blocked in PBS containing 2% Triton X-100 (Sigma-Aldrich), 10% goat serum (Gibco), and 2% cold water fish gelatin (Sigma-Aldrich). Cells were incubated for 1 hour at room temperature and then centrifuged as described above. Primary and secondary antibodies were incubated for 60 minutes at room temperature. Cells were washed twice in PBS plus 2% Triton, centrifuged, resuspended in blocking buffer, and cytospun at 800 rpm for 3 minutes using precoated cytoslides (Shandon Anatomical Pathology, Pittsburgh, PA). Cells were mounted in aqueous glycerol gelatin mounting medium (Sigma-Aldrich) and coverslipped. Images were visualized using a Leica SP2 AOBS laser scanning confocal microscope that was controlled by Leica SP2 TCS software (Leica, Bannockburn, IL, http://www.leica.com). Negative controls were performed by omitting primary antibody from the protocol and revealed no staining (data not shown).
SDS-Polyacrylamide Gel Electrophoresis and Western Blot Analyses
Total proteins were extracted from sorted cells or from cultured cells by washing cells once with ice-cold PBS and resuspending cell pellets in ice-cold protein lysis buffer (1% Triton X-100, 10 mM TrisHCl, 150 mM NaCl, 1% protease inhibitor cocktail [Sigma-Aldrich]). Cell suspensions were spun down at ∼20,000g for 5 minutes at 4°C, and supernatants were frozen in aliquots for further analysis at −20°C.
Denaturing SDS-polyacrylamide gel electrophoresis and Western blots were done as described previously [12, 13] using a Mini-Protean-3-Cell and Mini-Trans-Blot Transfer Cell (Bio-Rad, Hercules, CA, http://www.bio-rad.com). Nitrocellulose membranes were blocked in Odyssey (LI-COR, Lincoln, NE, http://www.licor.com) blocking buffer/PBS (1:1) for 1 hour, and incubated with primary and appropriate secondary antibody in blocking buffer/PBS + 0.1% Tween-20 (Sigma-Aldrich). Membranes were washed four times for 5 minutes with PBS + 0.1% Tween-20 after each incubation with antibody. Dried membranes were scanned using an Odyssey infrared scanner (LI-COR). Quantitative protein expression was done by loading protein standards of known concentration simultaneously on the gel, creating a standard curve that was applied to calculate unknown protein concentrations in samples using the Odyssey version 1.2 software (LI-COR).
Standards for quantitative protein analyzes were as follows: albumin (Research Diagnostics, Flanders, NJ, http://www.researchd.com), α-fetoprotein (AFP) (Fitzgerald Industries International, Concord, MA, http://www.fitzgerald-fii.com), cytokeratin 19 (CK19) (Research Diagnostics), and transferrin (Fitzgerald Industries International).
Primary antibodies were albumin, AFP (Sigma-Aldrich), CK19 (Cymbus Biotechnology, Hampshire, U.K.), transferrin (Rockland, Gilbertsville, PA), and CLDN-3 (Abcam, Cambridge, MA, http://www.abcam.com). Secondary antibodies were Alexa Fluor 680 (Western blots) and AF568 (cytospins) polyclonal goat anti-rabbit IgG (H+L) and Alexa Fluor 680 (Western blots) and AF488 (Cytospins) polyclonal goat anti-mouse IgG (H+L) (Molecular Probes Inc., Eugene, OR, http://probes.invitrogen.com).
Gene expression was analyzed by quantitative real-time reverse transcription-polymerase chain reaction (RT-PCR). Gene-specific mRNAs were created as followed: total RNA from livers was extracted using the RNeasy kit (Qiagen, Valencia, CA, http://www1.qiagen.com) and reverse-transcribed by Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) and oligo-(dT)(12–18) primer (Gene Link, Hawthorne, NY). cDNA was used as template in conventional PCR with gene-specific primers (for sequences, see Table 1) from which the forward primer possessed a 5′ overhang for T7-promotor sequence (5′-gac tcg taa tac gac tca cta tag gg). This amplified gene-specific DNA was used for in vitro transcription with T7-RNA polymerase (Promega, Madison, WI, http://www.promega.com), generating gene-specific RNA(with an additional 5′-ggg included by T7-RNA polymerase) used as standards in quantitative real-time RT-PCR using gene-specific primers without the 5′ overhang; standard ranges were linear from 1 to 108 templates. Primers were selected on different exons wherever possible. Quantitative real-time RT-PCR was done in the LightCycler instrument using the LightCycler RNA Master SYBR Green I kit (Roche). RNA from samples was extracted using RNeasy mini kit (Qiagen).
Table Table 1.. Primer sequences
Developmental Changes in the Percentages of EpCAM+ Cells and of the Ratios of Hepatic Stem Cells Versus Hepatoblasts Within EpCAM+ Cells
Data obtained in this study (Table 2) confirmed our prior finding (Bruce et al., manuscript submitted for publication) that cell suspensions prepared from neonatal and adult livers consisted of low percentages of EpCAM+ cells (1.7%–2.5%), of which less than 0.1% are hepatoblasts. By contrast, EpCAM+-sorted cells from fetal livers accounted for 12% of liver cell suspensions, of which more than 80% are hepatoblasts and 0.1%–0.7% are hepatic stem cells. Thus, the percentage of hepatic stem cells remains at relatively constant ratio of the total liver cells, whereas hepatoblasts exist as the dominant pluripotent cell in fetal and neonatal livers, but their numbers decline with age, occurring in high numbers in adult livers only in disease states, such as cirrhosis. In fetal livers, EpCAM-negative cells comprise nonparenchymal cells, including mainly hematopoietic and mesenchymal cells. By contrast, in postnatal livers, EpCAM-negative cells proved to be more than 95% diploid and polyploid hepatocytes (A. Bruce, L. Zhang, J.W. Ludlow, E. Schmelzer, A. Melhem, M. Kulik, L.M. Reid, and M.E. Furth, manuscript submitted for publication).
Table Table 2.. Relative proportions of hepatic EpCAM+ cells in donors of different ages
Expression Profiles of Hepatic Stem Cells from Livers of Donors of Different Ages
Culture-selected hepatic stem cells from fetal livers were analyzed for quantitative gene expression and used as controls to that of freshly isolated and sorted EpCAM+ liver cells from livers of varying age donors (Tables 3 and 4). In general, culture-selected stem cells, whether isolated on type III collagen or culture plastic and whether from short-term (10 days) or long-term (20 days) cultures, gave similar results. Hepatic stem cells isolated from the livers of all developmental stages assayed expressed very similar gene expression profiles, with high expression of CK19, EpCAM, neuronal cell adhesion molecule (NCAM), CLDN-3, c-kit, and aquaporin 4 but low levels of expression of albumin and very low or no expression of AFP or of any liver-specific genes (e.g., connexins, PEPCK, DPP4, cytochrome P450 3A4, and transferrin). Protein profiles of hepatic stem cells from all developmental stages (Table 5) matched the mRNA expression, showing protein expression for CK19, NCAM, albumin, and CLDN-3 but not AFP. Confirmation of the unique expression of CLDN-3 in hepatic stem cells but not in hepatoblasts or mature hepatocytes could be shown also by Western blot analyses (Fig. 1) and immunocytochemistry (Fig. 2).
Table Table 3.. Comparison of quantitative gene expression from human hepatic stem cells and hepatoblasts
Table Table 4.. Comparison of quantitative gene expression from human hepatic stem cells vs. hepatocytes
Table Table 5.. Comparison of quantitative protein expression from human hepatic stem cells, hepatoblasts, and hepatocytes
Gene Expression Profiles of Hepatoblasts
Gene expression patterns of hepatoblasts (Table 3) are different from those of hepatic stem cells and can be characterized generally as loss of expression (CLDN-3 and NCAM) and low levels of expression (CK19 and c-kit) of the genes characteristic of the stem cell phenotype, activation of expression, albeit at low levels, of adult liver-specific genes (e.g., connexins, PEPCK, and P450s) and intense expression of AFP, a defining feature of hepatoblasts. Albumin gene expression in hepatoblasts is about 1,000-fold higher than in hepatic stem cells and still 50-fold higher in its protein expression (Table 4), which correspondents to that of hepatocytes. Some genes defining mature liver functions (e.g., cytochrome P450 3A4 and α-1-antitrypsin) were not expressed by either hepatic stem cells or hepatoblasts. Although variations between different donors in gene expression profiles could be observed (variations caused not only by individual patient history but influenced by shipping and handling of the donor organ) we were able to manifest distinct trends in gene expression patterns, clearly showing different profiles for hepatic stem cells and hepatoblasts.
Expression Profiles of Hepatocytes from Livers of Donors of Different Ages
Gene expression in hepatocytes was compared with that in hepatic stem cells and is summarized in Table 4 by calculating ratios of the data from EpCAM+ versus EpCAM−-sorted cells. Gene expression data from hepatocytes from postnatal livers proved quite variable due to variation caused by organ handling/shipping (necrosis caused by time delay until processing) and donor history (drug administration, nicotine use, obesity, etc.). Nevertheless, a pattern that proved consistent was the absence of expression of the genes defining the hepatic stem cells (e.g., NCAM, CK19, c-kit, and CLDN-3) or hepatoblasts (e.g., AFP) and high levels of expression of classic liver-specific genes (cytochrome P4503A4 and transferrin).
The gene expression profiles in terms of both RNAs and proteins of two pluripotent human hepatic progenitors, hepatic stem cells and hepatoblasts, have been compared with those of mature human hepatocytes and shown to reveal three distinct patterns: a phenotype for stem cells, a phenotype for mature hepatocytes, and one for hepatoblasts that is a mix of the other two.
The ability to isolate and purify specific subpopulations of human liver cells such as the human hepatic stem cells has been made possible only recently by using immunoselection for a defined antigen such as EpCAM, a transmembrane glycoprotein described by de Boer  and found now to be expressed by hepatic progenitors but not hepatocytes  (E. Schmelzer, L. Zhang, A. Melhem, N.G. Moss, E. Wauthier, R.E. McClelland, A. Bruce, H. Yao, W.S. Turner, N. Cheng, M.E. Furth, and L.M. Reid, manuscript submitted for publication; A. Bruce, L. Zhang, J.W. Ludlow, E. Schmelzer, A. Melhem, M. Kulik, L.M. Reid, and M.E. Furth, manuscript submitted for publication). Subdivision of the hepatic progenitor subpopulations into hepatic stem cells versus hepatoblasts can also be done by culture selection conditions that give rise to clonigenically expanding stem cell colonies (E. Schmelzer, L. Zhang, A. Melhem, N.G. Moss, E. Wauthier, R.E. McClelland, A. Bruce, H. Yao, W.S. Turner, N. Cheng, M.E. Furth, and L.M. Reid, manuscript submitted for publication). The only comparable findings by others are those of Tanimizu et al. [15, 16], who have shown isolation of hepatoblasts from murine and rodent livers by Dlk/Pref-1, a membrane protein with six extracellular epidermal growth factor-like repeats [17, 18].
By immunohistochemical analyses, we have been able to assign EpCAM expression in nonpathologic human livers to hepatic stem cells and hepatoblasts in all developmental stages (i.e., fetal, neonatal, pediatric, and adult livers). Transplantation into the livers of mice of EpCAM+-sorted cells gave rise to human liver tissue expressing human-specific liver proteins (E. Schmelzer, L. Zhang, A. Melhem, N.G. Moss, E. Wauthier, R.E. McClelland, A. Bruce, H. Yao, W.S. Turner, N. Cheng, M.E. Furth, and L.M. Reid, manuscript submitted for publication; A. Bruce, L. Zhang, J.W. Ludlow, E. Schmelzer, A. Melhem, M. Kulik, L.M. Reid, and M.E. Furth, manuscript submitted for publication). Sorting for EpCAM+ cells from pediatric and adult livers resulted in a population mainly comprised of hepatic stem cells, whereas those from fetal and neonatal livers resulted in predominantly hepatoblasts.
We show here that hepatic stem cells from livers at all developmental stages show highly similar expression profiles, defined below as the stem cell phenotype. Phenotypic characterization of gene expression at mRNA and protein levels in pluripotent human hepatic progenitors versus mature hepatocytes indicated two patterns, a stem cell phenotype and a liver-specific phenotype; a gradient from minimal levels was observed in hepatoblasts to maximal levels found in hepatocytes. The stem cell phenotype comprised expression of EpCAM, NCAM, CK19, c-kit, CLDN-3, and weak levels of albumin but no expression of AFP or adult liver-specific proteins such as transferrin, connexins, PEPCK, DPP4, or P450s. The expression of both hepatic (albumin) and biliary (CK19) markers reflects the bipotential nature of the hepatic stem cell toward both lineages of the liver, the hepatic and biliary, which are characterized by albumin and CK19 expression, respectively. In studies published elsewhere, the hepatic stem cells also express components of the Hedgehog signaling pathway, including Indian and Sonic Hedgehog and its receptor, Patched .
The hepatocyte phenotype comprised expression of classic liver-specific genes such as albumin, connexins, transferrin, PEPCK, DPP4, and P450s and loss of expression of the genes typifying stem cells (including EpCAM, NCAM, CK19, c-kit, and CLDN-3). Hepatoblasts proved to be cells in an intermediate position with lowered levels of the stem cell genes and with activation but low levels of adult liver-specific genes and with the unique, defining feature of AFP.
A unique and novel hepatic stem cell surface marker is CLDN-3. Together with occludins, claudins form tight junction proteins  that act as a barrier for the passage of ions and molecules through the paracellular pathway and to the movement of proteins and lipids between the apical and the basolateral domains of the plasma membrane. Claudins comprise a family of more than 20 members (reviewed in ref. ). Morita et al.  demonstrated that CLDN-3 is expressed in mouse lung, liver, kidney, and testis and that the mouse gut shows lineage dependent expression of CLDN-3; liver sections show staining for CLDN-3 in bile canaliculi regions . After partial hepatectomy in rats, CLDN-3 is evident periportally and in sites where regenerative responses are initiated . Combining our results with these previous findings, CLDN-3 is a candidate surface molecule expressed by hepatic stem cells and providing a marker in addition to NCAM, permitting selection uniquely for stem cells.
In summary, this study suggests that the pluripotent hepatic progenitors are evident in both fetal and postnatal livers and remain stable in their phenotypes throughout life, particularly the hepatic stem cells, which persist in relatively constant numbers in livers at all donor ages. The primary changes with age are in the numbers of hepatoblasts that are the dominant cell type in fetal and neonatal livers and then decline in numbers to comprise <0.1% of the parenchyma in normal adult livers. Hepatoblast numbers appear to wax and wane with need for regenerative responses, and the only conditions under which they are observed in high numbers other than in fetal development are in diseased states such as cirrhosis . Their responses suggest parallels with transit-amplifying cells in other tissues.
The authors indicate no potential conflicts of interest.
Funding for this work was derived primarily from a sponsored research grant from Vesta Therapeutics (formerly Incara Pharmaceuticals) (Research Triangle Park, NC); funding to L.M.R. was derived from NIH grants (DK52851, AA014243, IP30-DK065933) and a Department of Energy grant (DE-FG02-02ER-63477). We thank Lucendia English and University of North Carolina Michael Hooker Microscopy Core Facility under the direction of Michael Chua for technical support. We thank National Development and Research Institutes, Inc. for providing available liver tissue. This research was made possible by the willingness of organ donor families to donate livers unsuitable for transplantation to research.