Identification of adult hepatic progenitor cells capable of repopulating injured rat liver


  • Potential conflict of interest: Nothing to report.


Oval cells appear and expand in the liver when hepatocyte proliferation is compromised. Many different markers have been attributed to these cells, but their nature still remains obscure. This study is a detailed gene expression analysis aimed at revealing their identity and repopulating in vivo capacity. Oval cells were activated in 2-acetylaminofluorene–treated rats subjected to partial hepatectomy or in D-galactosamine–treated rats. Two surface markers [epithelial cell adhesion molecule (EpCAM) and thymus cell antigen 1 (Thy-1)] were used for purification of freshly isolated cells. Their gene expression analysis was studied with Affymetrix Rat Expression Array 230 2.0, reverse-transcriptase polymerase chain reaction, and immunofluorescent microscopy. We found that EpCAM+ and Thy-1+ cells represent two different populations of cells in the oval cell niche. EpCAM+ cells express the classical oval cell markers (alpha-fetoprotein, cytokeratin-19, OV-1 antigen, a6 integrin, and connexin 43), cell surface markers recently identified by us (CD44, CD24, EpCAM, aquaporin 5, claudin-4, secretin receptor, claudin-7, V-ros sarcoma virus oncogene homolog 1, cadherin 22, mucin-1, and CD133), and liver-enriched transcription factors (forkhead box q, forkhead box a2, onecut 1, and transcription factor 2). Oval cells do not express previously reported hematopoietic stem cell markers Thy-1, c-kit, and CD34 or the neuroepithelial marker neural cell adhesion molecule 1. However, oval cells express a number of mesenchymal markers including vimentin, mesothelin, bone morphogenetic protein 7, and Tweak receptor (tumor necrosis factor receptor superfamily, member 12A). A group of novel differentially expressed oval cell genes is also presented. It is shown that Thy-1+ cells are mesenchymal cells with characteristics of myofibroblasts/activated stellate cells. Transplantation experiments reveal that EpCAM+ cells are true progenitors capable of repopulating injured rat liver. Conclusion: We have shown that EpCAM+ oval cells are bipotential adult hepatic epithelial progenitors. These cells display a mixed epithelial/mesenchymal phenotype that has not been recognized previously. They are valuable candidates for liver cell therapy. (HEPATOLOGY 2007.)

The existence of adult hepatic stem cells still remains elusive. A unique specific marker for adult hepatic stem cells has not yet been identified, and because these cells are not under constant renewal, in contrast to epithelial cells of the intestine or skin, they escape detection. On the other hand, over the years, substantial evidence has accumulated demonstrating the existence of adult hepatic epithelial progenitor cells, which are named oval cells (OC) or transit amplifying cells (for recent reviews, see Santoni-Rugiu et al.1 and Shafritz et al.2). These cells can be activated to proliferate and differentiate into hepatocytes and bile duct epithelial cells when the regenerative capacity of terminally differentiated hepatocytes is blocked. Activation and proliferation of OC were observed after rodents were fed a choline-deficient diet supplemented with ethionine or treated with azo dyes, N-2-fluorenyl-acetamide, N-2-fluorenyl-acetamide in combination with partial hepatectomy (2-AAF/PH), pyrrolizidine alkaloids, D-galactosamine (D-gal), allyl alcohol, or other protocols.3, 4

Morphologically, OC are small, with an oval-shaped nucleus and scant cytoplasm, and for this reason, they have been named oval cells.5 In the quiescent liver, OC reside in the canals of Hering (lined by both hepatocytes and cholangiocytes).6 When activated to proliferate, OC form tortuous, ductlike structures (pseudoducts), emanating from the portal zone and expanding into the parenchyma.7 These pseudoducts proliferate in close proximity to hepatic stellate cells.7, 8

It has been reported that OC share gene expression characteristics with fetal hepatoblasts [alpha-fetoprotein (AFP), gamma glutamyl transpeptidase (GGT), muscle type 2 pyruvate kinase, and delta homolog-like 1 (Dlk1)], biliary epithelial cells [cytokeratin-7 (CK-7), CK-8, CK-18, CK-19, oval cell antigen 6 (OV-6), glutathione S-transferase π, connexin 43, and mouse A6 antigen], hematopoietic stem cells [thymus cell antigen 1 (Thy-1), c-kit, CD34, and stem cell antigen-1 (Sca-1)], and neuroepithelial cells [chromogranin-A, neural cell adhesion molecule (NCAM), and parathyroid hormone-related peptide].1, 3, 9 These data strongly suggest that OC are a heterogeneous population of cells. In recent studies, we identified novel OC surface markers: CD44, CD24, epithelial cell adhesion molecule (EpCAM), claudin-7, V-ros sarcoma virus oncogene homolog 1 (Ros-1), cadherin 22, and CD133.10 However, we were unable to detect expression of some previously identified OC markers. The broad spectrum of reported OC-specific markers and the lack of some of them in purified OC isolates prompted us to study in more detail the gene expression pattern of OC, reinvestigate their heterogeneity, and determine their identity.

The gold standard for defining a cell population as a progenitor population is its ability to multiply many times in the recipient liver, to differentiate into hepatocytes and cholangiocytes, and to repopulate the injured liver. Successful reconstitution of rat liver by transplanted adult hepatic progenitor cells has not yet been achieved. To identify the cells that can repopulate the injured liver, we first transplanted crude OC isolates into retrorsine (Rs)/PH-preconditioned rat liver. Subsequently, we purified the cells from these isolates and determined the identity of OC that could reconstitute the injured liver.


2-AAF, 2-acetylaminofluorene; 2-AAF/PH, 2-acetylaminofluorene/partial hepatectomy; AFP, alpha-fetoprotein; BMP7, bone morphogenetic protein 7; CK, cytokeratin; D-gal, d-galactosamine; Dlk1, delta homolog-like 1; DPP4, dipeptidyl peptidase 4; EpCAM, epithelial cell adhesion molecule; F344, Fisher 344; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; GGT, gamma glutamyl transpeptidase; HRP, horseradish peroxidase; IF, immunofluorescent; IHC, immunohistochemical; mRNA, messenger RNA; OC, oval cell(s); OV, oval cell antigen; NPC, nonparenchymal cell(s); PH, partial hepatectomy; Ros-1, V-ros sarcoma virus oncogene homolog 1; RT-PCR, reverse-transcriptase polymerase chain reaction; Q-RT-PCR, quantitative real-time reverse-transcriptase polymerase chain reaction; Rs, retrorsine; Sca-1, stem cell antigen-1; αSMA, smooth muscle alpha actin; Thy-1, thymus cell antigen 1.

Materials and Methods

Animals and Animal Treatment.

Male, 180-g Fisher 344 (F344) rats were purchased from Taconic Farms (German Town, NY). Dipeptidyl peptidase 4–negative (DPP4) F344 rats were provided by the Special Animal Core of the Liver Research Center, Albert Einstein College of Medicine. All studies with animals were conducted under the protocol approved by the Animal Care Institute of Albert Einstein College of Medicine in accordance with National Institutes of Health guidelines.

Two models for activation and proliferation of OC were used: 2-acetylaminofluorene treatment in conjunction with partial hepatectomy (2-AAF/PH)11 [2-acetylaminofluorene (2-AAF) pellets were purchased form Innovative Research of America, Sarasota, FL] and D-gal (Sigma, St. Louis, MO)–induced liver injury.12 One 2-AAF pellet (35 mg, 14-day release) was implanted subcutaneously, and 7 days later, two-thirds PH was performed. D-gal was injected intraperitoneally at a dose of 130 mg/100 g of body weight. For transplantation, DPP4 recipient animals were preconditioned as described previously.13, 14


Primary antibodies are listed in Supplementary Table 1. All secondary antibodies were purchased from Jackson Immunoresearch Laboratories (West Grove, PA) and used at a dilution of 1:100. Horseradish peroxidase (HRP)–conjugated sheep anti-mouse immunoglobulin G (NA931V) was from Amersham Biosciences, Piscataway, NJ.

Isolation of Nonparenchymal Cells (NPC).

NPC fractions were isolated from D-gal–treated rats on day 5 and from 2-AAF/PH–treated animals on day 10, when the highest number of transient amplifying OC accumulates in these models. Cell fractionation was performed as described previously.10, 15 NPC were equilibrated with OptiPrep (Accurate Chemicals & Scientific, Westbury, NY) with an 11% final concentration (diluted with Hank's balanced salt solution) and overlaid on 15% and 13% OptiPrep cushions. Cells at the 13%-15% interphase were collected and washed. Cell viability was at least 96% as determined by trypan blue exclusion. Isolation of EpCam+ and Thy-1+ cells was performed with magnetic beads and a MidiMACS separation unit (Miltenyi Biotec, Germany), which collected the positive fractions.

RNA Isolation, Affymetrix Chips, and Microarray Data Analysis.

Total RNA from EpCam+ and Thy-1+ cells was isolated with Trizol reagent (Invitrogen, Carlsbad, CA) and additionally purified with the RNeasy mini kit (Qiagen Inc., Valencia, CA). For microarray analyses, normal and 2-AAF/PH livers were used. For each cell fraction, RNA from two animals was used; 5 μg of RNA was labeled according to Affymetrix Expression Analysis Manual (701021 Rev. 5) and hybridized to 2 Rat Genome 230 2.0 GeneChips from Affymetrix (Santa Clara, CA) containing 680,000 oligo probes representing 31,000 genes and transcript variants. Hybridization procedures, scanning, and normalization (global scaling) were performed following the Affymetrix manual at the Genomics Facility (Albert Einstein College of Medicine). Overall (global method) normalized Affymetrix gene chip.cel files were analyzed by ArrayAssist software (Stratagene, La Jolla, CA). Corresponding replicates from the experimental samples were aggregated in hybridization groups. The hybridization group of the virtual chip of untreated livers (samples of normal liver NPC fraction) was defined as the baseline reference. After the conversion of the data into a logarithmic scale (logarithm base 2), genes across the experimental samples were chosen manually. Significance analysis was done by multiple methods (analysis of variance). Significantly changed gene expression was considered with P values equal to or less than 0.05. The fold change in messenger RNA (mRNA) expression of each gene from the studied groups was compared to that of NPC from normal liver. Gene expression changes were verified by reverse-transcriptase polymerase chain reaction (RT-PCR). Genes showing at least a 2-fold increase in expression level in the pure fractions are shown.

RT-PCR and Quantitative Real-Time Reverse-Transcriptase Polymerase Chain Reaction (Q-RT-PCR) Analysis.

For RT-PCR analyses, RNA from normal, D-gal–treated, and 2-AAF/PH–treated animals was used. All reverse-transcriptase reactions were carried out with SuperScript II reverse transcriptase (Invitrogen) according to the manufacturer's protocol. Rat-specific primers for different genes were chosen with the Primer3 program (Supplementary Table 2). Expression of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as an internal control.

Q-RT-PCR was performed in three repeats with normal and 2-AAF/PH animals with the ABI-Prism 7900 sequence detection system (Applied Biosystems, Foster City, CA) in a two-step RT-PCR. Rat-specific sequences for polymerase chain reaction primers (Supplementary Table 2) were designed to generate amplicons of 50 to 150 base pairs for quantitative real-time detection with SYBR Green master mix (Applied Biosystems). mRNA abundance was determined by normalization of data to the expression level of GAPDH mRNA. mRNA expression in normal liver was taken as the baseline and considered equal to 1.

Immunofluorescent (IF) and Immunohistochemical (IHC) Analysis.

Five-micrometer rat liver frozen sections from normal, D-gal, and 2-AAF/PH animals and cytospins were fixed for 10 minutes in ice-cold methanol, acetone, or paraformaldehyde, washed twice with phosphate-buffered saline, and blocked with 5% normal serum/1% bovine serum albumin. Incubation with antibodies and color development were performed as described previously.10, 16

Liver Cell Transplantation and Repopulation.

An OC-enriched NPC fraction (13%-15% OptiPrep interphase) and pure EpCAM+ and Thy-1+ cells from D-gal–treated F344 rats were isolated as described above. Cells were transplanted into the spleen of Rs-treated DPP4 F344 rats as described previously.13 Animals were sacrificed 15 days and 1 month later, and liver repopulation was analyzed. Liver sections from the recipients were stained histochemically for DPP4 and GGT enzyme activity.12, 15 The percentage of liver repopulation was determined through the scanning of liver sections with a Polaroid CS-600 high-resolution scanner (Polaroid Corp., Cambridge, MA) and the measurement of the red-stained areas (DPP4+) versus the total area of the section with Adobe Photoshop, as described.17


Proliferation of OC in the 2-AAF/PH Model.

To activate OC, we used the 2-AAF/PH model first described by Tatematsu et al.18 Maximum accumulation of OC was observed around day 10. As shown in Fig. 1, OC form tortuous pseudoducts that expand from the periportal zone into the liver parenchyma, as demonstrated by the established OC markers, AFP (Fig. 1A) and CK19 (Fig. 1B), and by three new OC markers recently identified in our laboratory, CD44 (Fig. 1C), EpCAM (Fig. 1D), and claudin-7 (Fig. 1E). Accumulation of OC was accompanied by a substantial increase in the number of cells expressing the hematopoietic stem cell marker Thy-119 (Fig. 1F).

Figure 1.

(A-F) Proliferation of OC in the liver of rats treated with 2-AAF/PH. Frozen liver sections were processed for IHC detection of (A) AFP (red color) developed with HRP and 3-amino-9-ethylcarbazole as the substrate and (B) CK19, (C) CD44, (D) EpCAM, (E) claudin-7, and (F) Thy-1 with HRP-conjugated secondary antibody and DAB as the substrate (brown color). Original magnification: ×200. (G-J) Proliferation of OC in the liver after transplantation. Crude OC (2 × 107) were transplanted into the spleen of F344 DPP4 recipients preconditioned with (G-I) Rs/PH or (J) normal regenerating liver. The tissue was taken 1 month later. Transplanted cells were detected by histochemical staining for DPP4. (G) Whole liver section from a DPP4 F344 rat transplanted with the OC-enriched fraction from 2-AAF/PH DPP4+ F344 rats. (H) and (I) Representative area of the liver section in panel G at a higher magnification. Transplanted cells were detected as hepatocytes and bile ducts (arrows). The liver almost restored its normal lobular structure; only a few megalocytes could be detected (arrowheads). (J) Histochemical staining for DPP4 of liver sections from a normal PH rat transplanted with OC-enriched fraction. Only small groups of DPP4+ hepatocytes can be observed. Original magnification: (G) ×4, (H,J) ×200, and (I) ×400.

Proliferation and Differentiation of OC In Vivo After Transplantation into Injured Liver.

To identify the OC-enriched fraction that can proliferate and differentiate in vivo after transplantation, we isolated NPC from D-gal–treated wild-type F344 rats and fractionated them on a discontinuous OptiPrep density gradient. The abundance of CK19+ and GGT+ cells in fractions collected at the interphase of three different OptiPrep densities (11%-13%, 13%-15%, and 15%-17%) was estimated by the labeling of cells with CK19 antibody and histochemical staining for GGT (see Material and Methods). The highest number of GGT+/CK19+ cells (approximately 40%) was found in the 13%-15% fraction (data not shown). Cells from this fraction were collected, and their repopulating capacity in vivo was determined after transplantation into a mutant, DPP4-deficient F344 rat.20 Recipient animals were preconditioned with Rs, a naturally occurring pyrrolizidine alkaloid that causes a persistent block in the hepatocyte cell cycle. PH was performed immediately before transplantation to induce liver regeneration.13 Because endogenous hepatocytes could not divide, liver regeneration was mediated through OC proliferation. Twenty million cells from the 13%-15% fraction were transplanted into Rs/PH animals, and liver reconstitution with DPP4+ donor cells was evaluated after 1 month. As shown in Fig. 1G, 90% of the recipient liver parenchyma was substituted with donor progenitor cells (red color; average substitution was 70%-90%). Most of the transplanted cells differentiated into hepatocytes (Fig. 1H,I), but cholangiocytes were also present (bile ducts in Fig. 1H,I, indicated by small arrows). The number of megalocytes (polyploid hepatocytes, not capable of dividing) that accumulated in the liver of Rs/PH rats was markedly reduced (arrowheads in Fig. 1H), and clusters of small endogenous hepatocytes were rarely observed. These data show that in 1 month OC proliferated in the recipient liver, differentiated into hepatocytes and bile duct cells, and nearly restored the normal lobular structure. However, when the animals were not preconditioned with Rs/PH, only small clusters of transplanted cells were observed (Fig. 1J).

Identification of 2 Distinct Cell Types Activated in the Nonparenchymal Fraction.

To identify the cells in the OptiPrep-enriched fraction that amplify and differentiate in the liver after transplantation, we purified these cells using magnetic microbeads and antibodies against EpCAM and Thy-1. The purity of the EpCAM+ cells in 3 isolates ranged from 82% to 94%, and that of Thy-1+ cells ranged from 70% to 84%, as determined from the percentage of EpCAM-positive and Thy-1–positive cells, respectively, collected on cytospins. The gene expression profile of EpCAM+ and Thy-1+ cells was compared with Affymetrix Rat Genome 230 2.0 microarrays. Even though the two cell populations were <95% pure, the data presented in Fig. 2 clearly show that EpCAM+ and Thy-1+ cells represent two different cell types with distinct gene expression profiles. EpCAM+ cells express mRNAs for known OC markers (Afp, Ck19, Dlk1, and Ggt) and other markers reported by us recently (aquaporin 5, Cd24, Cd44, claudin-4, secretin receptor, Cd133, cadherin 22, claudin-7, and mucin-1). These mRNAs were not expressed in the Thy-1+ cells (Fig. 2A). In contrast, Thy-1+ cells express mRNAs for mesenchymal and activated stellate cells: collagen type 1 (a1 and a2), desmin, matrix metalloproteinase-2, Pdgf-receptor β, and Cd105, among others. These genes were not expressed in EpCAM+ cells (Fig. 2B). Several growth factors, receptors, and liver-enriched transcription factors are expressed only in EpCAM+ cells (Fig.2C).

Figure 2.

Gene expression analysis of purified EpCAM+ and Thy-1+ cells. Comparison of the mRNA expression level of (A) OC and (B) mesenchymal/stellate cell marker genes and (C) growth and transcription factors in purified EpCam+ and Thy-1+ cells. The cells were isolated from 2-AAF/PH–treated animals at day 10 after PH. Gene expression was analyzed with Affymetrix gene chips. The signal intensities of mRNAs for different genes and mRNA for Gapdh were used for calculating the relative mRNA expression. The signal intensity value of Gapdh used as a reference was taken as 1 [the ratios of the signal intensity of mRNAs for Gapdh, beta-glucuronidase, and β-actin are equal (Thy-1+/EpCAM+= 1.2/1)]. The P value is less than 0.05 (n = 2). *OC unique genes.

IF Microscopic Analysis of the Two Distinct Cell Types Activated During OC-Mediated Liver Regeneration.

In further studies, we determined whether the EpCAM surface marker, which we used for isolation of OC,10 is coexpressed with classical OC markers AFP, CK19, and OV-1.21 All three markers were coexpressed with EpCAM in OC (Fig. 3): AFP (Fig. 3A-C), CK19 (Fig. 3D-F), and OV-1 (Fig. 3G-I). Integrin α6, a marker for OC and cholangiocytes,22 was also coexpressed with CD44 (Fig. 3J-L), confirming the validity of these OC markers. Complete overlap of EpCAM and claudin-7 expression was also observed in normal liver, where the two antigens are coexpressed in cholangiocytes (not shown) and in 2-AAF/PH liver (Fig. 3M-O).

Figure 3.

Coexpression of EpCAM with other OC markers. Frozen liver sections from 2-AAF/PH rats were processed for double IF microscopy. Double IF labeling with (A,D,G) anti-EpCAM antibody and antibodies against three established OC markers: (B) AFP, (E) CK19, and (H) OV-1. (C,F,I) Yellow staining on merged pictures indicates coexpression of the two antigens in OC. Coexpression of (J) α6 integrin and (K) CD44 [(L) merged picture] and (M) EpCAM and (N) claudin-7 [(O) merged picture] shows complete overlap of expression. 4′,6-Diamidino-2-phenylindole counterstaining, blue. Original magnification: (A-F,J-L) ×400 and (G-I,M-O) ×600.

Until now, Thy-1 antigen was considered a marker for OC and was used for identification and isolation of these cells. To confirm that Thy-1 is not expressed in OC, double-label IF microscopy was performed with Thy-1 antibody and the OC markers AFP, CK19, and CD44. As shown in Fig. 4, Thy1-expressing cells have a pattern of distribution similar to that of OC, but there is no coexpression of Thy-1 antigen with AFP (Fig. 4A-C), CK19 (Fig. 4D-F), and CD44 (Fig. 4G-I). However, double IF labeling of Thy-1 and two markers expressed in activated stellate cells, desmin and smooth muscle alpha actin (αSMA), clearly showed coexpression of Thy-1 with desmin (Fig. 4J-L) and αSMA (Fig. 4M-O).

Figure 4.

Coexpression of Thy1 with OC and stellate cell markers. IF labeling of liver sections from 2-AAF/PH–treated animals with anti–Thy-1 and antibodies recognizing OC markers. Double labeling with (A,D,G) Thy-1 and (B) AFP and (E) CK19 and (H) CD44 antibodies. Thy-1–positive cells do not express any of the analyzed OC markers in (C,F,I) the merged pictures. Double IF labeling of liver sections from 2-AAF/PH rats for (J) Thy-1 and (K) desmin and (M) Thy-1 and (N) αSMA. Thy-1 cells express both stellate cell markers, desmin and αSMA, as seen in (L,O) the merged pictures. 4′,6-Diamidino-2-phenylindole counterstaining. Original magnification: ×600.

Proliferation and Differentiation of OC After Transplantation into a Recipient Animal.

The above data show that the two major populations of cells in the OC compartment are epithelial cells expressing EpCAM and mesenchymal cells expressing Thy-1. To determine, with certainty, which cells proliferate and differentiate into hepatocytes in the injured rat liver, we isolated EpCAM+ and Thy-1+ cells and transplanted them into DPP4 Rs/PH rats. As a control, we also transplanted the Thy-1–negative fraction. GGT histochemical staining of the three cell fractions used for transplantation is shown in Fig. 5A-C. The EpCAM+ cells were GGT+. In the Thy-1+ fraction, there were very few GGT+ cells, whereas the Thy-1 fraction contained many GGT+ cells. Three to five million cells were transplanted, and the animals were sacrificed 2 weeks later. The histochemical staining for DPP4, presented in Fig. 5D, shows that the EpCAM+ cells formed multiple hepatocytic clusters (red canalicular staining). Liver repopulation in some areas reached 10.7% (average repopulation was 5.4%, as determined on 20 sections from 3 different liver lobes). Thy-1+ cells did not differentiate into hepatocytes and did not form clusters of DPP4+ hepatocytes (Fig. 5E), whereas cells from the Thy-1 fraction differentiated into hepatocytes and repopulated the liver (Fig. 5F).

Figure 5.

Proliferation of DPP4+ OC after transplantation into preconditioned liver. (A-C) GGT staining of EpCAM+, Thy-1+, and Thy-1 cells on cytospins before transplantation. Note that most EpCAM+ and few Thy-1+ cells are GGT+. (D-F) Liver sections from DPP4 recipients 15 days after transplantation with cells shown in panels A-C. Clusters of DPP4+ cells were observed in the parenchyma of the livers transplanted with (D) EpCam+ and (F) Thy-1 but not with (E) Thy-1+ cells. (G-J) Histochemical staining of serial liver sections 15 days after transplantation of EpCAM+ cells. (G) DPP4 and (H) GGT staining of the newly formed hepatocytic clusters, showing that transplanted DPP4+ OC express GGT. Endogenous OC (arrowheads), which are DPP4, also express GGT. (I) DPP4 and (J) Ki-67 staining of serial liver sections show that the two labels are localized in the same area, indicating that the transplanted cells (arrows) are proliferating. (K) Transplanted cell-forming bile ducts (arrows). (L) Another bile duct at a higher magnification. Original magnification: (D-F) ×40, (A-C,G-J) ×200, (K) ×400, and (L) ×600.

To confirm that transplanted progenitors proliferate and differentiate in vivo after transplantation, recipient rats were sacrificed 2 weeks after transplantation, and the DPP4-positive cells were tested for expression of GGT, a marker of immature hepatoblasts that is not expressed in normal adult hepatocytes. As shown on serial sections (Fig. 5G,H), DPP4+ cells are also positive for GGT, confirming that the emerging clusters of transplanted cells originate from OC. The proliferation of OC was documented by labeling with Ki-67 antibody. The labeled cells (brown nuclear staining) are localized in DPP4-positive clusters of transplanted OC (Fig. 5I,J). In addition, some of the transplanted cells differentiated into bile ducts (Fig. 5K,L). These results show that OC proliferate and differentiate in the recipient liver. They also directly prove that OC are bipotential hepatic progenitors.

The Identity of OC.

The identity of rat OC was revealed through microarray analyses performed with isolated EpCAM+ cells from 2-AAF/PH livers on day 10 after PH. OC markers common with bile duct cells and not expressed in the Thy-1+ fraction are presented in Fig. 2A. This figure also presents the unique OC markers cadherin 22, CD133, discoidin domain receptor family member 1, mucin-1, Ros1,10 deleted in malignant brain tumor 1,23 and Dlk1.24, 25 The expression profiles of genes expressed in early hepatoblasts, hepatocytes, and mesenchymal cells were also compared between OC and crude NPC from normal (reference) and 2-AAF/PH–treated animals (Table 1). As expected, several genes present on the chips and expressed in hepatoblasts were up-regulated in OC, such as aldolase A (the fetal form), alpha 1 antitrypsin, hepatocyte growth factor receptor, and E-cadherin. Expression of Afp mRNA in normal hepatocytes could be attributed to the presence of variant forms of Afp in adult hepatocytes.26 Expression of genes induced later in fetal liver development, such as tyrosine aminotransferase and carbamoyl phosphate synthetase 1, was also evident, whereas expression of genes characteristic for adult hepatocytes was low or not detected (asialoglycoprotein receptor 1, tryptophan 2,3-dioxygenase, and others); this was consistent with our previous data on expression of these genes during fetal liver development.27 A noteworthy finding was expression in OC of mesenchymal markers such as vimentin, glypican-1, glypican-3, CD44, and mesothelin (Table 1).

Table 1. Gene Expression of OC
Gene NameNetAffx IDNL/HcNPC FractionEpCam+P Value
NL2-AAF/PH (10 Days)
  • Experimental samples were divided into several groups: normal liver hepatocytes (NL/Hc), nonparenchymal cells (NPC) from normal liver (NL) and 2-AAF/PH liver, and EpCAM+ OC isolated from 2-AAF/PH liver at day 10. Oligonucleotide hybridization, normalization of the chips, and statistical analysis were performed as described in Materials and Methods. Fold changes in mRNA expression are shown in bold. The signal intensity values are presented in parentheses. Genes are listed in alphabetical order.

  • *

    C-kit, CD34, and Ncam1 mRNA expression of these genes was also analyzed by quantitative RT-PCR.

  • NetAffx ID is the gene number in the Affymetrix gene database.

Mesenchymal genes      
 Fos-like antigen 11368489_at(10.07)−0.02 (72.76)0.91 (141.81)2.23 (215.41)6.06E−05
 Glypican 11387039_at(14)−0.01 (16.18)0.95 (32.38)1.15 (39.37)1.05E−05
 Mesothelin1368441_at(29.46)0.00 (73.04)0.88 (145.9)4.09 (1241.86)5.55E−07
 Vimentin1367574_at(19.92)−0.03 (2098.28)0.62 (3262.07)1.04 (4376.4)3.22E−12
Hepatocyte-specific genes      
 Albumin1367556_s_at(8437.38)0.00 (4147.73)0.22 (4860)1.04 (8733.58)4.12E−08
 Asialoglycoprotein receptor 11370149_at(1911.3)−0.42 (69.98)−0.43 (53.17)−0.16 (69.38)3.06E−07
 Glucose-6-phosphatase, catalytic1386944_a_at(5710.15)−0.42 (473.48)−0.91 (286.24)−1.03 (101.85)1.2E−04
 Tryptophan 2,3-dioxygenase1368720_at(6364.88)−0.11 (1083)−1.41 (304)−0.42 (1001.9)6.41E−09
Genes expressed in hepatoblasts      
 Aldolase A1367617_at(169.78)0.00 (1691)−0.1 (1594.38)0.53 (2525.37)3.2E−04
 Alpha fetoprotein1367758_at(341.82)−0.1 (22.6)2.71 (150)8.16 (6731.9)3.03E−10
 Alpha 1 antitrypsin1367647_at(9372.33)−0.1 (2215.84)0.66 (3510)1.37 (6175.06)1.13E−07
 E-cadherin1386947_at(587.77)−0.05 (781.57)0.08 (845.9)1.39 (2087.89)4.85E−09
Genes expressed in nervous tissues      
 Activity regulated cytoskeletal-associated protein1387068_at(27.8)−0.01 (34.22)0.08 (37.28)2.56 (234)5.93E−05
 Brain-enriched SH3-domain protein Besh31387924_at(82.73)−0.05 (15.34)2.59 (104.55)4.39 (328.08)2.06E−06
 Neogenin1392770_at(60.25)0.00 (51.83)0.45 (67.95)1.67 (173.73)5.03−08
 Ng22 protein1376117_at(30.67)−0.01 (21.77)1.82 (83.09)3.92 (328.88)1.46E−06
 Ncam1*1368320_at(1.6)−0.63 (3.26)3.15 (30.82)−0.4 (2.58)1.2E−04
 Protein tyrosine phosphatase, receptor-type, Z polypeptide 11368350_at(6.71)−0.04 (14.09)2.42 (78.79)4.03 (267.23)1.61E−06
 Reticulon 11368097_a_at(13.22)−0.01 (85.27)1.25 (216.02)2.08 (399.76)3.55E−09
Hematopoietic markers      
 C-kit*1369822_at(5.75)−0.23 (24.4)−0.22 (22)−1.9 (11.6)4E−02
 Lymphocyte antigen 6 complex, predicted (Sca1)1378690_at(30.38)−0.22 (279.87)−0.48 (268.48)−1.09 (146.19)4.2E−03
 Similar to Cd34*1372481_at(7.92)−0.02 (65.64)0.23 (77.07)−2.55 (11.5)1.24E−09
Other differentially expressed genes      
 Aldehyde dehydrogenase family 1, subfamily A21368003_at(13.59)−0.07 (48.93)0.71 (90.5)2.95 (392.66)9.31E−06
 ATP-binding cassette, subfamily B, member 1A1370465_at(31.76)−0.18 (8.53)0.84 (15.5)3.7 (119.25)4.81E−05
 Bradykinin receptor b21370650_s_at(9.12)−0.08 (14.19)1.06 (32.53)2.63 (103.35)2.92E−05
 Cytokeratin 211388074_at(13.72)0.00 (49.14)1.54 (157.84)3.02 (462.55)2.41E−05
 Fucosyltransferase 2 (secretor status included)1387097_at(18.18)−0.18 (6.86)2.23 (34.02)4.82 (196.80)8.2E−0.4
 Glutathione peroxidase 21374070_at(6.55)0.00 (51.72)2.54 (348.58)5.16 (1960.48)1.17E−06
 Growth arrest specific 61369735_at(19.98)−0.08 (53.43)1.74 (193.6)3.26 (579.12)1.76E−06
 Membrane-associated protein 171368209_at(2.18)−0.01 (25.64)2.64 (179.18)4.09 (437.9)4.87E−06
 Microtubule-associated protein 21388152_at(2.33)−0.13 (27.55)2.49 (175.6)4.21 (519.87)3.93E−07
 Protease, serine, 8 (prostasin)1370379_at(8.67)−0.56 (17.34)2.32 (94.16)4.11 (314.69)3.7E−04
 Suppression of tumorigenicity 141387195_at(14.16)0.00 (112.2)1.45 (316.04)3.04 (929.05)1.33E−9
 Transmembrane protease, serine 21373329_at(19.96)−0.01 (19.08)2.04 (87)3.83 (276.31)1.23E−06
 Zetin 11379366_a_at(15.76)−0.13 (17.85)1.23 (44.16)4.75 (289.16)4.83E−0.5

Expression of several hematopoietic genes considered common with OC and present on the chips was not detected (c-kit, CD34, CD43, and fms-related tyrosine kinase 1), but CD133, a marker for hematopoietic and mesenchymal progenitors, was expressed (Table 1). In contrast, c-kit mRNA expression was increased in total 2-AAF/PH liver (Fig. 6A). To evaluate the expression of c-kit and CD34 mRNA in the NPC fractions from normal and 2-AAF/PH-treated rats and in the isolated EpCAM+ and Thy-1+ cells by an independent method, we performed quantitative RT-PCR. These data show that c-kit and CD34 mRNA expression in isolated OC and Thy-1+ cells was lower than c-kit expression in NPC from normal and treated livers (Fig. 6B,C), and this reveals that c-kit and CD34 mRNAs are expressed in another cell population and not in OC. As previously reported, expression of Ncam1 mRNA was not detected in EpCAM+ cells.10 However, our data from Q-RT-PCR analysis showed increased Ncam1 mRNA expression in the Thy-1+ population of cells in comparison with the total NPC fraction from normal liver (Fig. 6D). Interestingly, expression of Ncam1 was reported in activated stellate cells after bile duct ligation.28 Table 1 also presents several differentially expressed genes in OC not reported previously.

Figure 6.

mRNA expression levels of c-kit, CD34, and Ncam1. (A) Expression level of c-kit mRNA in total liver homogenate of normal and 2-AAF/PH–treated animals obtained with RT-PCR (38 cycles). (B-D) Comparative mRNA expression of (B) c-kit, (C) CD34, and (D) Ncam1 in NPC fractions from normal and 2-AAF/PH livers and from pure EpCAM+ and Thy-1+ cells determined by quantitative RT-PCR. Values are means ± standard deviation of three repeats and are expressed as fold differences with respect to the NPC from normal liver. (E) OC markers and their coexpression in hepatoblasts, cholangiocytes, hepatocytes, and mesenchymal cells.


Identification of Progenitor Cells Capable of Repopulating Injured Rat Liver.

In these studies, we analyzed the gene expression profile of OC in adult rat liver. To our knowledge, complete gene expression analysis of these cells has not been reported previously, although many different markers attributed to rat liver OC have been described. Recently, we described new OC surface markers, some of which are also expressed in cholangiocytes (CD44, CD24, EpCAM, aquaporin 5, claudin-4, secretin receptor, and claudin-7), and others not detected in other liver cells (Ros1, cadherin 22, mucin-1, and CD133). In the present work, we studied the gene expression of previously reported OC markers in highly enriched OC isolates, using antibody against the surface antigen EpCAM, and proved that these cells are epithelial progenitor cells:

  • 1EpCAM+ cells express all classical liver epithelial OC marker mRNA and mRNA for markers reported recently by us (Fig. 2A and Table 1).
  • 2They express liver-enriched transcription factors (Fig. 2C).
  • 3IF microscopy showed that EpCAM colocalizes with classical OC markers AFP, OV-1, CK19, a6 integrin, and connexin 43.
  • 4We did not detect any epithelial markers (hepatocytic or cholangiocytic) present in the nonparenchymal OC-enriched fraction and absent from the EpCAM purified fraction.

For this reason, we consider the EpCAM+ cells as representative of the OC population, which is composed of epithelial progenitors at different stages of maturation. Some markers appear earlier, others appear later, and some are species-specific,10 which is reflected in the heterogeneity of these cells.29 Unexpectedly, we did not detect expression of mRNA for c-kit receptor or its ligand, stem cell factor, in purified OC or other hematopoietic markers present on the Affymetrix chips. It has been reported that c-kit and its ligand are expressed in OC in the 2-AAF/PH model during the early stages of their proliferation.30, 31 We cannot exclude the possibility that during OC maturation, expression of c-kit and other hematopoietic genes is quickly down-regulated, and because our gene expression analyses were performed on day 10, we did not detect their expression. An alternative explanation is that the hematopoietic genes are expressed in another cell population, which is supported by the data in Fig. 6A.

Our results demonstrate further that rat OC, isolated with antibodies against one of their surface markers, EpCAM, are the cells that proliferate and repopulate the recipient liver. That the clusters of transplanted cells emerging in the recipient liver originate from OC is evident from the expression of GGT, a marker of immature hepatocytes. This result resembles the transplantation data with fetal hepatoblasts, which did not turn off GGT expression in the recipient liver 2 weeks after transplantation.14 We also observed differentiation of transplanted OC into cholangiocytes (Fig. 2B, C and Fig. 5K, L), further supporting the bipotential capabilities of these cells. Our experiments also show that for OC to repopulate the recipient liver; the proliferation of endogenous hepatocytes needs to be blocked. If the liver is not preconditioned, the endogenous hepatocytes have a growth advantage over OC and reconstitute the liver remnant.

Dual Epithelial-Mesenchymal Nature of OC.

Another very interesting characteristic of OC revealed in this work is their mixed epithelial/mesenchymal phenotype, which has not been previously recognized. Because gene expression analysis was carried out with freshly isolated OC and not with cells in culture, their gene expression pattern described in this study represents their real expression pattern in situ and not the changes observed after in vitro culturing. Epithelial/mesenchymal transition of liver cells in culture has been clearly documented.32 We have found that some but not all tested mesenchymal markers were significantly overexpressed in OC (Table 1), and this places them in an intermediate epithelial-mesenchymal lineage. Recent data from nematodes and zebrafish suggest that endoderm and some portion of the mesoderm may derive from a bipotential layer of cells called the mesoendodermal.33 It is very plausible that OC have a mesoendodermal origin. As these cells differentiate, they lose mesenchymal markers and gradually acquire a full epithelial phenotype: hepatocytic or cholangiocytic. In a very recent article, Schmelzer at al.34 discuss a possibility of coisolation of the hepatic progenitor cells with mesenchymal cells. We cannot rule out the possibility of coselection of mesenchymal cells with EpCAM+ cells. However, the specific mesenchymal markers identified in this work are expressed predominantly in the OC and not in companion contaminating cells: (1) EpCAM+ OC coexpress vimentin (see reference10, supplementary fig. 2I-L), as determined by double IF microscopy; (2) EpCAM OC express the typical mesenchymal marker CD44 (see 10, Fig. 5D-L); and (3) only a restricted number of mesenchymal genes are expressed in OC [such as bone morphogenetic protein 7 (BMP7), CD44, and mesothelin], and these markers are not expressed in the Thy1+ mesenchymal cells or stellate cells, which are closely associated with OC in the OC niche or in the rest of the NPC.

The dual nature of OC explains their activation by Tweak (tumor necrosis factor ligand superfamily, member 12A), a novel regulator of mesenchymal progenitor cells,35, 36 and the high expression of its receptor, tumor necrosis factor receptor superfamily member 12A (Fn 14; Fig. 2A). Another very interesting finding demonstrating their dual phenotype is the unique expression of BMP7 and its receptor, activin receptor-like kinase 3, in OC. Very recently, it has been reported that this growth factor facilitates liver regeneration, and it has been suggested that BMP7 serves as a hormone released from the kidney that regulates hepatocyte proliferation.37 BMP7 expression is not detected in other liver cell types in normal or 2-AAF/PH-treated animals, and it seems that it affects OC growth by an autocrine mechanism. It has also been reported that BMP7 induces a mesenchymal to epithelial transition in adult renal fibroblasts and facilitates regeneration of injured kidney.38, 39 It is likely that BMP7 exerts the same effect on OC and that BMP7 is a guardian of the epithelial phenotype.

Identification of Thy-1+ Cells as a Distinct Mesenchymal/Stellate Cell Subpopulation.

Most noteworthy are the lack of expression of Thy-1 in OC and the evidence that Thy-1–expressing cells are myofibroblasts or a subpopulation of activated stellate cells: (1) gene expression analysis showed that Thy-1+ cells have not the OC phenotype but the phenotype of mesenchymal cells; (2) double IF microscopy showed coexpression of Thy-1 with stellate cell markers and a lack of coexpression with OC markers; and (3) Thy-1+ cells do not repopulate the liver after transplantation. Upon activation, Thy-1+ cells accumulate in the liver in parallel with OC. That Thy-1+ cells are not OC but liver myofibroblasts and that Thy-1+ cells coexpress desmin and a-SMA were also reported by another group.40 A summary of our gene expression data (coexpression of some marker genes in OC and hepatoblasts, hepatocytes, cholangiocytes, and mesenchymal cells) is presented in Fig. 6E.


The authors thank Dr. S. Sell for OV1 antibody, Dr. D. Reynolds for help with microarrays, Dr. D. A. Shafritz for critical comments on the article, Dr. D. Neufeld for liver perfusions, and Ethel Hurston for technical assistance.