Mature hepatocytes exhibit unexpected plasticity by direct dedifferentiation into liver progenitor cells in culture


  • Yixin Chen,

    1. Departments of Medicine, Cell Biology, and Development, University of Minnesota Medical School, Minneapolis, MN
    Search for more papers by this author
  • Philip P. Wong,

    1. Departments of Medicine, Cell Biology, and Development, University of Minnesota Medical School, Minneapolis, MN
    Search for more papers by this author
  • Lucas Sjeklocha,

    1. Departments of Medicine, Cell Biology, and Development, University of Minnesota Medical School, Minneapolis, MN
    Search for more papers by this author
  • Clifford J. Steer,

    1. Departments of Medicine, Cell Biology, and Development, University of Minnesota Medical School, Minneapolis, MN
    2. Departments of Genetics, Cell Biology, and Development, University of Minnesota Medical School, Minneapolis, MN
    Search for more papers by this author
  • M. Behnan Sahin

    Corresponding author
    1. Departments of Medicine, Cell Biology, and Development, University of Minnesota Medical School, Minneapolis, MN
    • Department of Medicine, University of Minnesota Medical School, 420 Delaware Street Southeast, MMC 480, Minneapolis, MN 55455
    Search for more papers by this author
    • fax: 612 625 6919

  • Potential conflict of interest: Nothing to report.

  • Y.C. and M.B.S. were supported by intramural research funds from the Department of Medicine, University of Minnesota. P.P.W., L.S., and C.J.S. were supported, in part, by National Institutes of Health ARRA Grant R01 DK081865-01.


Although there have been numerous reports describing the isolation of liver progenitor cells from the adult liver, their exact origin has not been clearly defined; and the role played by mature hepatocytes as direct contributors to the hepatic progenitor cell pool has remained largely unknown. Here, we report strong evidence that mature hepatocytes in culture have the capacity to dedifferentiate into a population of adult liver progenitors without genetic or epigenetic manipulations. By using highly purified mature hepatocytes, which were obtained from untreated, healthy rat liver and labeled with fluorescent dye PKH2, we found that hepatocytes in culture gave rise to a population of PKH2-positive liver progenitor cells. These cells, liver-derived progenitor cells, which share phenotypic similarities with oval cells, were previously reported to be capable of forming mature hepatocytes, both in culture and in animals. Studies done at various time points during the course of dedifferentiation cultures revealed that hepatocytes rapidly transformed into liver progenitors within 1 week through a transient oval cell-like stage. This finding was supported by lineage-tracing studies involving double-transgenic AlbuminCreXRosa26 mice expressing β-galactosidase exclusively in hepatocytes. Cultures set up with hepatocytes obtained from these mice resulted in the generation of β-galactosidase-positive liver progenitor cells, demonstrating that they were a direct dedifferentiation product of mature hepatocytes. Additionally, these progenitors differentiated into hepatocytes in vivo when transplanted into rats that had undergone retrorsine pretreatment and partial hepatectomy. Conclusion: Our studies provide strong evidence for the unexpected plasticity of mature hepatocytes to dedifferentiate into progenitor cells in culture, and this may potentially have a significant effect on the treatment of liver diseases requiring liver or hepatocyte transplantation. (HEPATOLOGY 2012;)

The liver is a unique organ with an enormous capacity to regenerate after injury. Up to 75% of liver mass can be regenerated in the rat within 1 week after partial hepatectomy.1 Under normal conditions and in the absence of toxic injury, hepatocytes are predominantly responsible for this process.2 When circumstances do not allow the proliferation of resident hepatocytes, as in previous exposure of the liver to certain toxins or chemicals, a population of stem/progenitor cells is activated. These cells, typically referred to as oval cells, are thought to exist in small numbers in terminal biliary ductules.3, 4 Although a bone marrow origin for oval cells has been suggested,5 more recent studies demonstrated that they were of hepatic origin.6, 7 Oval cells have the capacity to repair the injured liver, giving rise to both functional hepatocytes and cholangiocytes in vivo.8 However, their clinical use is prohibited, in part, because of the methods used to retrieve them in adequate numbers and the high frequency of malignant transformation observed in transplant studies.9, 10

Multipotent stem cell populations, such as embryonic stem cells and the more recently induced pluripotent stem cells, have also been shown to possess hepatic differentiation ability, even though their clinical applicability remains controversial.11-13 Bone-marrow–derived (hematopoietic or mesenchymal) stem cells also have the capacity to produce hepatocytes in vitro and in vivo, as evidenced by human bone marrow transplant studies as well as animal studies.14, 15 However, the frequency of this event appears to be low in the in vivo setting.16, 17 Therefore, there continues to be a major need for a stem/progenitor cell population that is safe and practical for clinical applications and treatment of a variety of human liver diseases. This has resulted in an intensification of efforts to identify and study liver-specific stem/progenitor cells in parallel to those involving embryonic stem cells (ES) and induced pluripotent stem cells (iPS) cells. We, too, have recently reported the isolation and characterization of a unique population of adult liver progenitors, called liver-derived progenitor cells (LDPCs).18

LDPCs are a novel population of bipotential liver progenitor cells whose isolation does not require any chemicals or toxins. In addition to a mixture of hematopoietic and hepatic markers, LDPCs express a number of stem cell markers, including cluster of differentiation (CD)34, stem cell receptor CD117 (c-kit), and cell-surface antigen CD90 (Thy-1), and they are phenotypically very similar to oval cells. More recently, we have shown that LDPCs are capable of differentiating into functional hepatocytes, both in vitro and in vivo.19 However, the origin of these cells, which has both significant scientific and clinical implications, has remained largely unanswered. Our results demonstrated that LDPCs were a direct dedifferentiation product of isolated mature hepatocytes, thus radically altering the concept of lineage relationship between liver progenitors and hepatocytes. The results have significant implications for a variety of stem cell applications.


CD, cluster of differentiation; CK7, cytokeratin 7; c-kit, stem cell receptor CD117; DAPI, 4′6-diamidino-2-phenylindole; EDTA, ethylene diamine tetraacetic acid; EMT, epithelial mesenchymal transition; ES, embryonic stem cells; FISH, fluorescence in situ hybridization; FITC, fluorescein isothiocyanate; GGT, gamma-glutamine transaminase; HepPar1, hepatocyte paraffin-1 antigen; IgG, immunoglobulin G; iPS, induced pluripotent stem cells; HNF, hepatocyte nuclear factor; IF, immunofluorescence; LDPCS, liver-derived progenitor cells; LMO2, Lim domain only-2 transcription factor; MGV, mean gray value; OV-6, oval cell-specific surface antigen; PBS, phosphate-buffered saline; PE, phycoerythrin; RT-PCR, reverse-transcription polymerase chain reaction; Sall4, Sal-like zinc-finger transcription factor 4; SSC, saline sodium citrate buffer; Thy-1, cell-surface antigen CD90; vWF, von Willebrand factor.

Materials and Methods


An Albumin CreXRosa26 double-transgenic mouse strain was generated by cross-breeding Albumin-Cre and Rosa26 mouse strains, both of which were purchased from Jackson Laboratories (Bar Harbor, ME). Sprague-Dawley and Fischer344 rats used in the experiments were 6-8 weeks old and were purchased from Harlan Laboratories (Madison, WI). All animal experiments were performed within the context of institutionally approved animal protocols, and animals were treated humanely in compliance with University of Minnesota regulations.

LDPC Cultures.

Rat LDPCs were generated from Sprague-Dawley and Fischer344 rat livers, as previously described by Sahin et al., 2008,18 with the exception that the liver cell preps used in LDPC cultures were highly enriched for hepatocytes by additional 3 low-gravity-centrifugation steps (50g × 15 seconds). The protocol for the isolation of mouse LDPCs was a modified form for rat LDPCs. Briefly, mouse hepatocytes were cultured in a medium consisting of Dulbecco's modified Eagles medium (Sigma-Aldrich, St Louis, MO) and F12 (Sigma) medium at a ratio of 3:1, supplemented with 15% mouse serum (Equitech, Kerville, TX), 1 mg/mL of bovine serum albumin (Sigma-Aldrich), 100 uM of β-mercaptoethanol (Sigma-Aldrich), 5 mM of nicotinamide (Sigma-Aldrich), and penicillin (100 μg/mL)/streptomycin (100 μg/mL) (Gibco, Carlsbad, CA). Cells were plated at a density of 2 x 103 cells/cm2 on 24-well plates coated with type I collagen (BD Biosciences, Franklin Lakes, NJ) at a concentration of 10 μg/cm2. Unlike rat LDPCs, mouse LDPCs were generated without further media change.

Immunofluorescence Staining.

The expression of each antigen was examined on cells from two independent experiments. Cells were fixed in 2% paraformaldehyde for 10 minutes, permeabilized with 0.1% Triton X-100, washed in phosphate-buffered saline (PBS), preblocked with donkey serum (Sigma-Aldrich), and subsequently stained with sheep antihuman serum albumin (1:1000; Abcam, Inc., Cambridge, MA), mouse anticytokeratin 7 (CK7) (1:100; Abcam), rabbit anti-LMO2 (Lim domain only 2 transcription factor) (1:100; Abcam), mouse anti-HNF-1α (hepatocyte nuclear factor 1 alpha) (1:500; BD Biosciences), phycoerythrin (PE)-rat antimouse cluster of differentiation (CD)45 (1:500; BD Biosciences), PE-mouse anti-rat CD45 (1:20; BioLegend, San Diego, CA), mouse anti-GGT (gamma glutamine transaminase) (1:100; Sigma-Aldrich), mouse anti-β-galactosidase (1:500; Roche, Basel, Switzerland), mouse anti-human/rat oval cell-specific surface antigen (OV-6) (1:100; R&D Systems, Minneapolis, MN), rabbit anti-Sall4 (Sal-like zinc-finger transcription factor 4) (1:100; GeneTex, Irvine, CA), chicken anti-vimentin (1:10,000; Abcam), mouse anti-Liv-2 (1:100; MBL, Nagoya, Japan), rabbit anti-CK19 (1:100; GeneTex), mouse anti-HNF4a (1:100; Abcam), mouse anti-CD44 (1:100; Cell Signaling Technology, Inc., Danvers, MA), and mouse anti-HepPar1 (hepatocyte paraffin-1 antigen) (1:100; Abcam). Appropriate secondary antibodies, DyLight 488 donkey anti-sheep immunoglobulin G (IgG), DyLight 549 donkey antirabbit IgG, and DyLight 594 donkey anti-mouse IgG (1:200; Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) were used. In the control experiments, cells were stained with secondary antibodies only. Nuclei were labeled with 4′,6-diamidino-2-phenylindole (DAPI; Roche). Quantification of immunofluorescence (IF) staining was also performed for a few select markers. For this purpose, National Institutes of Health Image J software was used to measure the mean gray value (MGV) for staining of cells through the culture period. For each cell, the MGV of the area of interest (Sall4 and HNF4a: nuclei; HepPar1: whole cell) was calculated and background was subtracted. An average of the MGV was then calculated for each group. Values were then extracted from that of negative control groups, which were only stained with secondary antibodies.

Histopathology and β-Galactosidase Staining.

Mouse liver tissues were fixed in 2% paraformaldehyde for 30 minutes, then embedded in Tissue-Tek OCT (Sakura Finetek, Torrance, CA) compound. Sections were sliced at a thickness of six microns. After air-drying, sections were incubated in β-galactosidase reporter gene staining solution (β-Galactosidase Reporter Gene Staining Kit; Sigma-Aldrich) at 37°C overnight, then counterstained with nuclear fast-red staining (Sigma-Aldrich) for 5 minutes at room temperature. For IF staining, sections were fixed in cold acetone. Rabbit anti-β-galactosidase (1:20; CEDARLANE Labs, Burlington, NC) was mixed with either sheep anti-albumin antibody (1:500) or mouse anti-CK7 (1:100). After washing, DyLight 549 anti-mouse, DyLight 549 anti-sheep, or DyLight 488 anti-rabbit antibodies, at a dilution of 1:200, were used as secondary antibodies.

RNA Isolation and Reverse-Transcription Polymerase Chain Reaction.

Total RNA was isolated using a miniRNA kit (QIAGEN, Valencia, CA), according to instructions of the supplier, and was subsequently subjected to reverse-transcription polymerase chain reaction (RT-PCR). First, 1 μg of total RNA was reverse-transcribed in a 20-μL volume with the Superscript III Reverse Transcription Kit (Invitrogen, Carlsbad, CA), as per the manufacturer's recommendations. After the reverse transcription, the volume of the reaction was increased to 100 μL, and 2 μL was used in each of the PCR assays with the Red kit (Sigma-Aldrich) on a Bio-Rad thermocycler (Bio-Rad, Hercules, CA), according to the manufacturer's instructions. Reactions were repeated a minimum of three times.

Western Blot Analysis.

Mouse tissues were lysed in 2× total protein buffer containing 10 mM of Tris (pH 7.6), 1% NP40, protease inhibitors (Roche complete mini ethylene diamine tetraacetic acid [EDTA]-Free), and 2 mM of dithiothreitol. Then, cell lysates were sonicated and quantified using the Bio-Rad Protein Assay. Then, 40 μg of protein were loaded into each well and separated by 4%-15% sodium dodecyl sulfate Mini-Protean TGX gel (Bio-Rad). After transfer, the polyvinylidene fluoride membrane was blocked with 5% nonfat dried milk, then cut into two pieces. The upper panel was incubated with rabbit anti-β-galactosidase (CEDARLANE); the lower panel was incubated with mouse monoclonal anti-β-actin (Sigma-Aldrich) at 4°C overnight. After washing, the peroxidase-conjugated secondary antibodies were added for 1 hour at room temperature. Detection was achieved using SuperSignal West Dura Extended Duration Substrate (Pierce, Rockford, IL).

PKH2 Staining and Flow Cytometry.

Hepatocytes obtained from Sprague-Dawley rats were stained with PKH2 for cytoplasmic labeling, according to the manufacturer's instructions (Sigma-Aldrich). After staining, hepatocytes were plated on 24-well plates or T-150 culture flasks at a density of 1 x 104 cells/cm2. Fluorescence images of cells were obtained at predetermined time points on a Leica DMIL fluorescence microscope, using Leica application suite V3.1 software (Leica Microsystems, Buffalo Grove, IL). On days 1 and 14, the cells on flasks were collected and subjected to flow cytometry for quantitative analysis of their fluorescence. For flow cytometric analysis, unlabeled cells were used as negative control. Mean fluorescent intensity was determined for each sample, and total fluorescence was calculated by multiplying mean fluorescent intensity and the total number of cells. The value of total fluorescence on day 1 was given an arbitrary unit of 1. Total fluorescence on day 14 was calculated in an identical manner, then compared with that of day 1. Triplicates were used for statistical calculations. For flow cytometric analysis of hepatocyte purity, cells were fixed and permeabilized by 0.2% Triton X-100. After blocking with donkey IgG, cells were incubated with an anti-albumin/FITC (fluorescein isothiocyanate) antibody (1:20; CEDARLANE) for 30 minutes at room temperature. The negative control was FITC/rabbbit IgG. Fluorescence-activated cell-sorting acquisition was performed, using a FACSCalibur flow cytometer (BD Biosciences), and data were analyzed using FlowJo 6.4. A statistical analysis was done by the Student's t test to identify significant differences. A P value less than 0.05 was considered significant. Flow cytometric analysis of LDPC purity was performed in an identical fashion, except for the antibody, which was PE-conjugated rabbit anti-CD45 antibody (BD Biosciences).

Transplantation of LDPCs.

Five female Fischer344 rats were treated with retrorsine (Sigma-Aldrich) before transplantation to impair the ability of recipient hepatocytes to proliferate. Two intraperitoneal injections were administered 2 weeks apart at a dose of 30 mg/kg. One month after the second dose of retrorsine, recipient animals were anesthetized using isoflurane and subjected to laparatomy and 66% partial hepatectomy under sterile conditions. This was followed by injection into the spleen of 1 x 107 PHK26-labeled male LDPCs obtained from male Fischer344 rats. (PKH26 labeling was performed following the manufacturer's [Sigma-Aldrich] instructions, resulting in the labeling of >90% of the cells.) Two of the rats died from surgical complications on the day after transplantation. The livers of the remaining 3 rats were examined 2 months later for evidence of engraftment. Animal experiments were done within the framework of institutionally approved protocols, and animals were treated and euthanized humanely.

Fluorescence In Situ Hybridization for Y-Chromosome.

Liver sections of 6 μm were prestained with albumin antibody and fixed in 4% formaldehyde at 37°C for 10 minutes. Then, following the manufacturer's protocol, sections were washed with 2x saline sodium citrate (SSC) buffer for 2 minutes at 73°C and treated with 0.005% pepsin for 10 minutes at 37°C. After rinsing in 1x PBS with glycine, slides were dehydrated in ethanol, and rat IDetect Chr-Y Probe (ID Labs, London, Ontario, Canada) was applied. After 2 minutes of denaturation at 69°C, slides were incubated for hybridization at 37°C overnight. After hybridization, slides were washed with 0.4x SSC with 0.3% lgepal (Sigma-Aldrich) for 2 minutes at 73°C, and 2x SSC with 0.1% lgepal for 1 minute at room temperature. After staining with DAPI, samples were examined under a fluorescence microscope and images were obtained.


Rat Liver Preparations Subjected to Low-G Centrifugations Are Highly Enriched for Hepatocytes.

To identify the origin of LDPCs, isolated hepatocytes were highly purified by low-G centifugations. We performed RT-PCR for markers that were specific for various cell types found in the liver, including desmin for stellate cells,20, 21 von Willebrand factor (vWF) for endothelial cells, fucose receptor for Kupffer cells,22 CK7 for biliary epithelial cells,23 and albumin for hepatocytes. Cell prep before low-G spin showed clear signals for all of the markers, except for CK7, indicating that the initial cell population contained hepatocytes, endothelial, Kupffer, and stellate cells. It appears that our standard centrifugation steps before low-G spins eliminated CK7-positive ductal cells, which were present in the whole liver preparation before any manipulation. After low-G spins, we were able to detect only albumin, and the signals for CK7, desmin, vWF, or fucose receptor messenger RNAs were undetectable (Fig. 1A), indicating a virtual absence of other major cell types found in the liver. The purity of the hepatocyte prep was also confirmed morphologically by albumin and HNF-1α staining (Fig. 1B). Additionally, flow cytometric analysis showed that hepatocytes used in LDPC cultures were over 99% pure (Fig. 1C).

Figure 1.

Purity of rat hepatocytes isolated by low-G centrifugation. (A) RT-PCR analysis of the cell purity. Fresh liver cells were subjected to RT-PCR before (lane 1) and after (lane 2) low-G spin for hepatocyte marker albumin, stellate cell marker desmin, endothelial cell marker vWF, Kupffer cell marker fucose receptor, and biliary ductal cell marker CK7. After low-G spin, only hepatocyte marker albumin was detectable. (B) Flow cytometric analysis of hepatocyte purity using an anti-albumin antibody. Over 99% of cells were albumin positive. (C) IF staining of the of the same cell population. Albumin is stained with green, HNF-1α with red, and cell nuclei with blue (DAPI) fluorescence. In the merged image, virtually all cells were triple positive, consistent with a highly pure hepatocyte preparation (original magnification: 40x).

PKH2-Labeled Rat Hepatocytes Give Rise to PKH2-Labeled Liver Progenitor Cells.

To track the fate of cultured hepatocytes, we initially labeled their cytoplasm with a fluorescent marker, PKH2 (Fig. 2A). Hepatocytes underwent drastic morphological changes, including significant cell death, in the first few days of culture. The remaining live cells either became flattened, forming cell clusters with many nuclei (e.g., polykaryons via possible endomitosis), or smaller as if they were undergoing apoptosis (i.e., cell shrinkage or condensation). Between days 5 and 7 of culture, LDPCs began to appear by either shrinkage of hepatocytes or by budding off from multinucleated cell clusters, a mechanism reminiscent of budding yeast (Fig. 2B). By day 14, LDPCs were the only cells left in culture, with the exception of few scattered fibroblast-like cells. Fluorescence images showed that virtually all LDPCs exhibited green fluorescence (i.e., PHK2 positive), which decreased over time. Results were consistent with the hypothesis that they were derived directly from PKH2-labeled hepatocytes and then underwent further cell divisions.

Figure 2.

Dedifferentiation of rat hepatocytes into LDPCs. (A) Morphology of LDPCs cultures initiated with PKH2-stained (cytoplasmic) hepatocytes at various time points during the culture period. The panels on the left are the bright-field images of the cultures on the indicated days, and the panels on the right are the corresponding fluorescence images. PKH2-positive LDPCs began to emerge from hepatocytes starting on day 5. By day 14, virtually all the cells in the culture were LDPCs (original magnification: 100x). (B) Potential mechanisms by which hepatocytes dedifferentiate into LDPCs. The panel on the left shows hepatocytes shrinking or undergoing condensation (arrows) to become LDPCs. The panel on the right demonstrates a single multinucleated dedifferentiating hepatocyte giving rise to LDPCs (cell membranes are forming around the cell nuclei) by what appears to be fragmentation of the cytoplasm and budding off (arrows; original magnification: 200x). (C) Expression of mesenchymal markers CD44 and vimentin during the culture period. These markers, which were not present in hepatocytes on day 0, became detectable around day 4, and on day 12, they were strongly positive in LDPCs, suggesting that hepatocytes underwent an EMT during their transformation to LDPCs (original magnification: 100x).

Morphological changes in LDPC cultures suggested the transformation of hepatocytes (i.e., epithelial) into fibroblast-like cells (i.e., mesenchymal) before the appearance of LDPCs. Thus, we examined the expression of the mesenchymal markers, CD44 and vimentin, in a time-dependent manner by the cells in culture. IF studies revealed that, whereas hepatocytes were negative for these mesenchymal markers on day 0, the cells in the culture began to express both CD44 and vimentin around day 4 and LDPCs were strongly positive for these markers on day 12. This finding suggested that hepatocytes may be undergoing an epithelial mesenchymal transition (EMT) before giving rise to LDPCs, which appear to have a nonepithelial, mesenchymal phenotype.

To confirm our morphological findings and provide quantitative data, we examined the kinetics of LDPC cultures by performing a cell count at certain time points during the culture period. This confirmed our previous observations showing that more than 80% of the plated hepatocytes died by day 6, followed by rapid repopulation of the culture by LDPCs by day 14 nearly restoring the original cell number (Supporting Fig. 1A). Additionally, we performed a quantitative assessment of the total fluorescence of cultured cells by flow cytometry as further evidence for the origin of LDPCs. On days 1 and 14 of LDPC cultures, we collected all the cells cultured within indentical flasks and measured their total fluorescence (Supporting Fig. 1B). We found that nonhepatocyte cells constituted <1% of all cells with a fluorescence intensity of 0.01 units (arbitrary units; total intensity of all cells on day 1 was assigned a value of 1.0). Total fluorescence of LDPCs on day 14 averaged approximately 0.5 (average of three separate experiments), which was at least 50 times greater than the total fluorescence of nonhepatocyte cells on day 1. Considering that PKH staining of particular cells gets distributed among daughter cells and becomes undetectable after 6-8 cell divisions, it was highly improbable that a very small population of progenitors present in the initial cell prep could have expanded many times to give rise to LDPCs with significant fluorescence intensity. This result effectively ruled out the possibility that LDPCs could have originated from the initial nonhepatocyte cell population in culture.

RT-PCR Analysis Confirms Rapid Transition of the Cells from Hepatocyte Phenotype to LDPC Phenotype During the Culture Period.

Next, we wanted to substantiate our PKH staining results by documenting the phenotypic changes taking place during the transformation of hepatocytes into LDPCs. To that end, we performed RT-PCR and IF analyses of hepatocyte- and LDPC-specific markers at predetermined time points during the culture period. On days 0, 4, 8, and 12, cultures were examined for expression of albumin, HNF-1α (hepatocyte specific), CD45, and LMO2 (LDPC specific). RT-PCR studies showed that in the beginning, cells expressed albumin and HNF-1α and no identifiable CD45 and LMO2. By day 4, there was a rapid decline in hepatocyte-specific markers, and LDPC-specific markers became detectable at low levels. Subsequently, on days 8 and 12, hepatocyte markers became undetectable, and LDPC markers were expressed at increasingly higher levels (Fig. 3A). IF studies revealed a similar pattern of marker expression, further confirming our RT-PCR data (Fig. 3B). In addition to these four markers, we examined the expression pattern of several other highly relevant hepatic genes during the culture period to better characterize the transformation process. We looked at the expression of mature hepatocyte markers HepPar1 and HNF-4α, immature hepatocyte marker Liv2,24 biliary ductal/oval cell marker CK19, and liver progenitor/embryonic liver marker Sall425 in a time-dependent manner. IF staining and quantitative analysis of the images revealed a pattern (Supporting Fig. 2A,B), which was consistent with rapid transformation of mature hepatocytes into cells with liver progenitor phenotype, thus supporting our findings shown in Fig. 3. Both the RT-PCR and IF studies correlated well with the morphological changes that took place in the cultures, including temporal appearance of LDPCs. Taken together, the rat studies strongly suggested that LDPCs originated from mature hepatocytes by direct dedifferentiation.

Figure 3.

Analysis of hepatocyte- and LDPC-specific markers at various time points during the LDPC culture period. (A) RT-PCR for hepatocyte markers albumin and HNF-1α, as well as LDPC markers LMO2 and CD45. On day 0, only hepatocyte markers were expressed and no signals for LDPC markers were detectable. Beginning around day 4, hepatocyte markers became weaker and were virtually gone by day 8, whereas LDPC markers showed an opposite trend and became progressively stronger. Day 12 cultures and “pure” LDPCs obtained on day 14 by gentle EDTA treatment of the cultures showed no difference, indicating that the transformation into LDPCs was completed (only the cell number increased after day 12). The lane marked Neg. shows the negative control reaction. (B) IF studies of the LDPC cultures for the same markers confirmed the RT-PCR results, again showing rapid disappearance of hepatocyte markers by day 4 and progressively stronger expression of LDPC markers after day 8 (cell nuclei stained by DAPI in blue; original magnification: 100x). The observed patterns were consistent with the rapid transformation of hepatocytes into LDPCs observed morphologically in Fig. 2.

Hepatocytes Go Through an Oval Cell-Like Stage En Route to Becoming LDPCs.

To gain further insight into the process of dedifferentiation of hepatocytes to LDPCs and to establish a stem/progenitor cell hierarchy, we examined the expression of several oval cell markers during the culture period. We considered the possibility that hepatocytes could be transitioning through an oval cell-like stage en route to becoming LDPCs. This was based on the phenotypic similarities between oval cells and LDPCs, suggesting a potential lineage relationship. Therefore, we studied the expression of OV-6, CK7, and GGT during the dedifferentiation of hepatocytes into LDPCs. For this aim, LDPC cultures initiated with pure hepatocytes were subjected to RT-PCR and IF staining for the aforementioned markers at predetermined time points during the culture period. RT-PCR analysis showed that CK7 expression, which was absent in the beginning, first appeared around day 4, peaked on day 6, and then gradually declined and was undetectable in LDPCs by day 14. GGT first became detectable around day 6 and progressively increased in intensity, only to become undetectable in LDPCs on day 14 (Fig. 4A). IF staining for these markers showed a very similar pattern to that seen with RT-PCR data, with the exception that some GGT protein expression was detectable in LDPCs on day 14. Oval-cell–specific protein OV-6, on the other hand, was first detected by IF staining on day 6 and reached a peak on day 8, after which it rapidly decreased, becoming virtually undetectable in LDPCs (Fig. 4B). The expression pattern of these markers correlated well with the morphological changes we observed in culture. Oval cell markers were up-regulated as hepatocytes were in the process of transforming into progressively smaller cells and down-regulated as the LDPCs became the dominant cell type. To demonstrate that these changes took place in the same cell population, we performed costaining for oval cell marker OV-6 and LDPC markers CD45 and LMO2, and found that on day 8, most of the cells coexpressed oval cell and LDPC markers (Fig. 4C). Taken together, these data strongly suggested that hepatocytes passed through an oval cell-like stage en route to becoming LDPCs.

Figure 4.

Transition of dedifferentiating hepatocytes through an oval cell-like stage en route to LDPCs. (A) RT-PCR analysis of two oval cell markers (also expressed by biliary ductal cells) CK7 and GGT at various time points during the LDPC culture period. Both markers, which were not expressed initially, became detectable between days 4 and 6 at approximately the same time as the loss of hepatocyte-specific markers shown in Fig. 3. They then peaked and subsequently became undetectable in “pure” LDPCs collected on day 14. The lane marked Neg. shows the negative control reaction. (B) IF analysis of the LDPC cultures during the same period for CK7, GGT, and oval cell-specific protein OV-6. The expression pattern for CK7 and GGT confirms the pattern observed with RT-PCR analysis, with the exception that GGT expression persisted in LDPCs at low level (most likely previously transcribed protein). OV-6 expression, on the other hand, became detectable on day 6, reached its peak on day 8, and then gradually declined to undetectable levels in LDPCs on day 12 (original magnification: 100x). (C) Costaining of the cells for LDPC markers CD45 and LMO2 and oval cell marker OV-6 on day 8 of culture. Virtually all cells coexpressed LDPC markers and oval cell marker, indicating a transient overlap between oval cell-like stage and LDPC stage. The combined data from RT-PCR and IF staining strongly suggest that hepatocytes transition through an oval cell-like stage before they turn into LDPCs (original magnification: 100x).

AlbCreXRosa26 Transgenic Mice Express β-Galactosidase Exclusively in Hepatocytes.

To provide additional evidence for the origin of LDPCs from hepatocytes in culture, we generated a double-transgenic mouse strain by crossing AlbCre and Rosa26 mouse strains. As predicted, the resulting AlbCreXRosa26 mice expressed the enzyme, β-galactosidase, only in the liver by western blot analysis (Fig. 5A). The hepatocyte-specific expression of this marker, which labeled albumin-expressing cells permanently, was confirmed by X-gal staining and IF staining for β-galactosidase. Results showed that expression of the reporter construct was restricted to hepatocytes (Fig. 5B).

Figure 5.

β-galactosidase expression in double-transgenic AlbCreXRosa26 mice. (A) Western blot analysis of β-galactosidase expression. The blotting showed that double-transgenic kidney (lane 1), double-transgenic spleen (lane 4), and wild-type liver (lane 3) did not express β-galactosidase, whereas double-transgenic liver (lane 2) was positive. (B) X-gal and β-galactosidase IF staining of the double-transgenic liver. X-gal staining of the wild-type liver did not show any reaction (left upper panel; original magnification: 100x), whereas double-transgenic liver strongly reacted. X-gal positivity appeared to be restricted to hepatocytes of the double-transgenic mice, as some cells, including vascular endothelium (arrow), were X-gal negative (right upper panel; original magnification: 100x). IF costaining for β-galactosidase and albumin (left lower panel; original magnification: 100x) showed complete overlap of the markers. On the other hand, CK7-positive ductal cells (arrow) did not express β-galactosidase (right lower panel; original magnification: 200x). Overall data were consistent with tissue-specific (liver) and cell-specific (hepatocyte) expression of β-galactosidase in AlbCreXRosa26 mice.

Hepatocytes From AlbCreXRosa26 Mice Dedifferentiate Into β-Galactosidase-Positive LDPCs.

The next step was to examine LDPCs generated from AlbCreXRosa26 mice for β-galactosidase expression. LDPC cultures of hepatocytes from double-transgenic mice were subjected to X-gal staining at various time points, which strongly suggested hepatocytes as the source of LDPCs (Fig. 6A). To ensure that the small, round cells that appeared in the cultures were LDPCs, we performed costaining for β-galactosidase and LDPC markers CD45 and LMO2. Virtually all cells coexpressed β-galactosidase and LDPC markers, thus confirming the identity of the mouse hepatocyte-derived LDPCs (Fig. 6B).

Figure 6.

Analysis of β-galactosidase expression in LDPCs derived from double-transgenic AlbCreXRosa26 mice. (A) X-gal staining of LDPC cultures on day 0 showed that nearly all hepatocytes were X-gal positive. On day 4, dedifferentiating hepatocytes began to give rise to small X-gal positive cells, which become more prominent by day 8. On day 15, virtually all cells in culture were X-gal-positive LDPCs (original magnification in all panels: 100x). (B) Confirmation of β-galactosidase-positive cells as LDPCs. The round, small cells that emerged in the cultures set up with hepatocytes from AlbCreXRosa26 mice were subjected to costaining for β-galactosidase and LDPC markers CD45 and LMO2 (original magnification in all panels: 100x). Results showed complete overlap of the β-galactosidase and LDPC markers, confirming that the cells were, indeed, LDPCs expressing β-galactosidase. These findings strongly support the hypothesis that LDPCs directly originate from hepatocytes.

LDPCs Engraft and Differentiate Into Hepatocytes In Vivo.

To underscrore the biological relevance of LDPCs, we performed a transplantation experiment using rat LDPCs generated from male Fischer344 rats. We did a flow cytometric analysis of the harvested LDPCs using CD45 as a marker of LDPC purity, which was >97% (Supporting Fig. 4A). Therefore, we proceeded with the transplantation of these cells into female Fischer344 rats after LDPCs had been labeled with fluorescent dye PHK26 for tracking purposes. After 2 months, we examined the liver samples of the surviving 3 rats under a fluorescence microscope before and after performing IF staining for albumin and fluorescence in situ hybridization (FISH) for Y-chromosome. Unstained sections revealed the presence of PKH+ cell clusters (approximately 1% of all cells) morphologically consistent with biliary ductal cells and hepatocytes (Fig. 7A-C). To confirm this finding, we then proceeded with IF staining for albumin and FISH for Y-chromosome, which showed the presence of male hepatocytes (Fig. 7D-F; Supporting Fig. 4C) in approximately 0.2% of the cells examined. This is not an insignificant number, in view of the fact that even in our positive control, male rat liver, (Supporting Fig. 4B) only 2% of cells were positive for Y-chromosome. Taken together, our findings strongly suggested that LDPCs had engrafted and differentiated into hepatocytes in the recipient animals.

Figure 7.

Hepatic differentiation of LDPCs in vivo. Fluorescent-labeled (PKH26) LDPCs from male Fischer rats were transplanted into female Fischer rats that had undergone retrorsine pretreatment and partial hepatectomy. Frozen liver section obtained from the recipient rats 2 months after transplantation were examined under a fluorescence microscope. (A) Unstained section showing PKH+cells, (B) DAPI nuclear staining, and (C) merged images of (A) and (B) (original magnification: 100x). They showed foci of PKH+cells morphologically consistent with biliary ductal cells and hepatocytes. To confirm that male LDPCs had engrafted and differentiated into hepatocytes in the recipients, we performed IF staining for albumin and FISH for Y-chromosome. (D) Merged images of DAPI nuclear staining and FISH (green dot) for Y-chromosome, (E) merged images of albumin staining (red) and FISH for Y-chromosome, and (F) merged images of (D) and (E) (original magnification: 400x). These findings showed that transplanted LDPCs had engrafted in the liver and differentiated into hepatocytes.


The main aim of this study was to identify the origin of LDPC, which are unique bipotential adult hepatic progenitors that were first isolated and characterized by us.18 LDPCs, which are capable of forming mature hepatocytes both in vitro and in vivo, are generated in culture from normal liver tissues that have not been exposed to chemicals or any type of injury. This is in contrast to the many published protocols used to generate the quintessential hepatic progenitor oval cells. Therefore, LDPCs have a unique clinical application potential in humans. However, the source or the origin of these cells, and, therefore, their lineage relationship to other cells in the liver, is essentially unknown.

It is now well established that many adult tissues harbor stem cells or progenitors, which are capable of generating some or all of the cell types found in that particular tissue. Commonly referred to as “tissue-specific stem cells,” these cells have been identified in tissues including, but not limited to, heart, skin, brain, small intestine, mammary gland, and teeth.26-31 In the adult liver, however, the situation is a bit more complex. This results from an extensive proliferative capacity of mature hepatocytes, which can regenerate the original liver mass even after 90% hepatectomy. However, when the degree of liver injury is very severe or when the liver has been exposed to certain toxins or chemicals, hepatocytes are unable to proliferate. It is under these conditions that the hepatic stem/progenitor compartment is activated. The most widely known and characterized liver progenitors are oval cells. They are believed to have primary hepatic origin and are thought to reside in small numbers in the terminal bile ducts. It is widely accepted, and perhaps even assumed, that hepatocytes do not contribute directly to this progenitor or stem cell compartment.

Our earlier observations on the production of LDPCs, and the rapid replacement of primary hepatocytes by hepatic progenitors in cultures, suggested that hepatocytes might, in fact, be the original source of the hepatic progenitor cells. To test this hypothesis, we designed a series of experiments generating LDPCs from two different species. First, we obtained an extremely pure population of hepatocytes from rat liver. Then, by fluorescently labeling and documenting the evolution of LDPC cultures set up with these hepatocytes, we were able to show that the LDPCs were also fluorescently labeled, strongly suggesting that they were directly derived from hepatocytes. The quantitative flow cytometric analysis of the initial and resultant cell populations indicated that hepatocytes were the only logical source for LDPCs.

Next, we wanted to show the origin of LDPCs in a transgenic mouse model, which would allow us to track the fate of hepatocytes. To accomplish this, we generated a double-transgenic AlbCreXRosa26 mouse strain, which was previously generated and published by others.32 Again, using several different techniques, we showed that these animals activated and expressed β-galactosidase only in hepatocytes. Using purified hepatocytes from these animals, we found that the LDPCs were also β-galactosidase positive, supporting the conclusion that they were derived directly from hepatocytes. These studies also showed two potentially unique mechanisms by which hepatocytes dedifferentiate into hepatic progenitors. One is by cell condensation or shrinkage, which is morphologically similar to cells undergoing apoptosis; the other is budding off from multinucleated cell clusters reminiscent of cell division in yeast.33, 34 It will be important to confirm these observations by future studies. Another interesting finding was the up-regulation of mesenchymal markers CD44 and vimentin during this dedifferentiation process, suggesting an epithelial-mesenchymal transition. This is particularly intriguing, given the known role of EMT in liver development and regeneration.35-37 However, the confirmation of EMT in the generation of LDPCs requires further studies. The other somewhat unexpected finding was the transient expression of some oval cell markers as hepatocytes dedifferentiated into LDPC. Based on the coexpression of OV-6 and LDPC markers by the majority of cells in culture, we believe that there is a single population of hepatic progenitors that are generated through a transient oval cell-like stage from hepatocytes.

Having shown the ability of mature hepatocytes to dedifferentiate into LDPCs in vitro, we tested the potential biological relevance by demonstrating their ability to regenerate hepatocytes in an animal model. Therefore, we proceeded with a well-established transplant protocol consisting of retrorsine pretreatment and partial hepatectomy in rats. The presence of both PKH-labeled and Y-chromosome-positive hepatocytes in the recipient animals convincingly showed that transplanted LDPCs were able to engraft and redifferentiate into hepatocytes in vivo. This finding raises the question of whether, under certain circumstances where the microenvironment is not conducive to the proliferation of hepatocytes, the process of dedifferentiation of hepatocytes into progenitors, followed by proliferation and redifferentiation, is a potential mechanism for organ regeneration in the liver. It is also conceivable that the process of dedifferentiation of hepatocytes via EMT may also be playing a role in tumor formation under pathological conditions.38 Even though we do not have any direct evidence, some published studies provide indirect clues for the presence of LDPC-like cells in vivo. Sell et al. reported on the proliferation of small, OV-6-negative intraportal progenitors, termed “null cells,” as a restitutive response to allyl alcohol-induced periportal necrosis in the rat liver.39, 40 In these studies, null cells later began to express OV-6 and eventually differentiated into mature hepatocytes. Based on morphological similarities between null cells and LDPCs, and the sequence of phenotypic changes null cells undergo to become hepatocytes, it is tempting to speculate that null cells may be the in vivo counterpart of LDPCs.

In summary, the combination of studies using primary rat and transgenic mouse hepatocytes has allowed us to trace the putative origin of LDPCs, a population of liver progenitors. Our results strongly suggest that they arise from direct dedifferention of mature hepatocytes in culture. This finding has a number of major implications. First, it shows that hepatocytes are far more plastic than previously thought and are potentially capable of contributing directly to the stem/progenitor cell pool of the liver. Second, it confirms that a fully mature, terminally differentiated somatic cell can acquire a stem/progenitor cell phenotype without genetic or epigenetic manipulation. Though our studies have demonstrated this in the liver, similar differentiation and dedifferentiation properties may exist in other organs and tissues. Finally, it has a significant potential impact on the treatment of liver diseases requiring liver or hepatocyte transplantation.