2-acetylaminofluorene dose-dependent differentiation of rat oval cells into hepatocytes: Confocal and electron microscopic studies



The 2-acetylaminofluorene (AAF)/partial hepatectomy (PH) model is one of the most extensively studied experimental systems for oval cell proliferation and differentiation. We have previously described the oval cells as forming ductular structures surrounded by basement membrane, representing extensions of the canals of Hering. Herein we analyze the differentiation of oval cells into hepatocytes after varying degrees of liver damage induced by AAF. At a low dose of AAF, most oval cells synchronously differentiate into small hepatocytes by 6 days after the PH, resulting in complete restoration of the liver structure in 10 days. Higher doses of AAF delay the differentiation process and the new hepatocytes form foci, in contrast to what is observed at the low dose. Qualitatively, the differentiation process seems to be identical at the cellular level under both conditions. The transition from the expanding oval cell population into hepatocytes was correlated with the upregulation of hepatocyte nuclear factor 4 and the disappearance of the basement membrane. Also, the differentiation of oval cells into hepatocytes coincided with the loss of alpha-fetoprotein and OV-6 staining, and the replacement of the biliary cell-specific α6 integrin and connexin 43 with the hepatocyte-specific α1 integrin and connexin 32. In addition, bile canaliculi form between the new hepatocytes. In conclusion, these results indicate the rate of oval cell differentiation into hepatocytes is context dependent and suggest that, under favorable conditions, oval cells can complete this process much faster than previously appreciated. (HEPATOLOGY 2004;39:1353–1361.)

The liver has an enormous regenerative capacity best illustrated by the fact that in rodents a two-thirds loss of liver mass can be replaced in a few days by the compensatory hyperplasia of the surviving hepatocytes.1, 2 In addition, two stem cell–fed back-up regenerative systems also exist in the liver.3–6 The activation of these dormant stem cell systems for liver regeneration takes place when residual hepatocytes are functionally compromised, are unable to divide, or both. We and other investigators have provided evidence that epithelial cells of the canals of Hering are the most probable candidates for the resident adult liver stem cells.1, 7 In case of stem cell–fed liver regeneration in rat liver, progeny of the stem cells multiply in an amplification compartment composed of the so-called oval cells.8 Oval cells form ductular structures surrounded by a continuous basement membrane,1 forming elongations of the canals of Hering, and attached their distal end to a hepatocyte of the liver plate.

Recent studies suggest bone marrow cells may be able to transdifferentiate into hepatocytes.5, 6 It is not yet clear if the bone marrow stem cells form hepatocytes via the oval cells. Petersen et al.9 described bone marrow–derived oval cells, but other investigators have detected only the end product, the hepatocyte.10, 11 Regardless of their origin, oval cells express a phenotype that is transitional between the biliary cells and hepatocytes. Although oval cells display several phenotypic characteristics of the mature hepatocytes (e.g., liver-enriched transcriptional factors,12 albumin production13), they differ structurally and functionally from mature hepatocytes.

The final step of the stem cell–fed regenerative process, that is, the differentiation of the oval cells into hepatocytes, is the focus of the present study, in which we use the well-characterized acetylaminofluorene (AAF)/partial hepatectomy (PH) experimental model to induce oval cells in the rat liver. Earlier, we observed that differentiation of oval cells into hepatocytes depends on the dose of AAF.14 For example, a high dose of AAF caused a delay in the differentiation of oval cells into hepatocytes.15 Similar results also were obtained by Alison et al.16, 17 We revisited the issue of AAF dose dependency on oval cell differentiation. In particular, we asked how the dose of AAF modifies oval cell differentiation. Herein, we demonstrate two patterns of oval cell differentiation. At a low dose of AAF (i.e., when the hepatocyte damage is mild), the oval cells rapidly and synchronously differentiate into small hepatocytes. In contrast, at a higher dose of AAF (i.e., more extensive damage to hepatocytes), differentiation of oval cells into hepatocytes is delayed and proceeds via an intermediate stage in which small basophilic hepatocytes accumulate in focal nodules. However, hepatocyte differentiation is eventually identical at the cellular level at both low and high AAF doses, that is, it is correlated with sudden upregulation of hepatocyte nuclear factor 4 (HNF-4) and the disappearance of laminin (basement-membrane) staining. Subsequently, oval cells lost their phenotypic characteristics (alpha-fetoprotein (AFP), OV-6 staining, α6 integrin, connexin 43) and gained hepatocytic features (α1 integrin, connexin 32, bile canalicular formation) as the differentiation process advanced.


AAF, 2-Acetylaminofluorene; PH, partial hepatectomy; HNF, hepatocyte nuclear factor; AFP, α-fetoprotein.

Materials and Methods

Animal Experiments.

Male F-344 rats (160–180 g) were used for all experiments and were kept under standard conditions. The animal study protocols were conducted according to National Institutes of Health guidelines for animal care.

AAF/PH Experiment.

AAF 2 mg/mL suspended in 1% dimethylcellulose (low dose, 2.5 mg/kg daily; high dose, 5 mg/kg daily) was given to the rats on 6 consecutive days by gavage. Traditional two-thirds PH was performed18 on the seventh day, which was followed by six additional AAF treatments. Animals were killed at several time points in pilot experiments to determine the times at which hepatocytes differentiated in the two AAF doses. All of the histological analyses described in the present paper were carried out on livers 6 days after PH in the low-dose model and 12 days after the PH in the high-dose model.

Morphological Analysis.

Cryostat sections (15–20 μm) were fixed in methanol and were incubated overnight with a mixture of the primary antibodies (Table 1); appropriate secondary antibodies were used (Jackson Immunoresearch, West Grove, PA). All samples were analyzed by confocal laser-scanning microscopy using Bio-Rad MRC-1024 system (Bio-Rad, Richmond, CA). Alpha-fetoprotein reaction was visualized by ABC peroxidase method using DAB as chromogen (Elite kit; Vector Laboratories, Burlingame CA). Electron microscopy was performed as described previously.1

Table 1. Antibodies Used for the Immunohistochemical Studies
AntibodySpeciesManufacturerCatalog NumberDilution
  1. Abbreviations: AFP, alpha-fetoprotein; FITC, fluorescein isothicyanate; HNF-4, hepatocyte nuclear factor-4.

AFP Nordic Immunological LabsShARa/AFP1:500
DesminMouse monoclonalDakoM07241:100
Pan Cytokeratin FITC labeled DakoF08591:20
OV6Mouse monoclonalGift from Dr. Hixson 1:100
LamininRabbit polyclonalDakoZ00971:300
Connexin 43Rabbit polyclonalZymed71-07001:100
Connexin 32Mouse monoclonalSanta CruzSC-72581:100
HNF-4Goat polyclonalSanta CruzSc-65561:100
Integrin α1Mouse monoclonalSerotecMCA 17911:100
Integrin α6Mouse monoclonalSerotecMCA 20341:100
CD26Mouse monoclonalSerotecMCA 9241:20

Retrograde Cholangiography and Analysis of the Bile Duct Structure.

The rats were anesthetized with Nembutal. A 30-gauge needle was inserted into the common bile duct, and 1 mL of the 1:10 dilution of fluorescein isothiocyanate-labeled lycopersicon esculentum lectin (Vector Laboratories) was injected slowly. After 15 minutes, the liver was removed and frozen. One hundred-micrometer frozen sections were cut and fixed in 4% paraformaldehyde. Stacks of optical sections (up to 40) were taken at 0.5 to 1-μm intervals. Horizontal views of the images were made and analyzed using the Bio-Rad Laser Sharp software.


Differentiation of Oval Cells After a High Dose of AAF.

Small foci composed of small hepatocytelike cells (Fig. 1A) appear in the liver 10 to 12 days after the partial hepatectomy when 5 mg/kg AAF was administered daily to the rats. The cells have all the ultrastructural characteristics of hepatocytes: abundant round nuclei, cytoplasm rich in rough endoplasmic reticulum, mitochondria, occasionally peroxisomes, and glycogen particles are observed. These features are not present in the oval cells. The foci are not composed of randomly arranged individual hepatocytes, but tightly packed cells that form tubular structures (Figs. 1E and 2A). The foci are relatively well circumscribed and the small basophilic hepatocytelike cells can be distinguished easily from the larger surrounding “old” hepatocytes (Fig. 1A). Oval cells are still numerous in the liver at this stage and are found commonly in contact with the randomly distributed foci.

Figure 1.

(A) High-dose model. Basophilic focus composed of small hepatocytes 12 days after partial hepatectomy (PH). Numerous oval cell cords emanate radially form the portal area. (B) Low-dose model. Small basophilic hepatocytes can be observed around the portal tract 6 days after PH. No typical oval cells can be seen. Arrowheads point at small hepatocytes arranged in ductular fashion. (C) High-dose model. HNF-4–positive (green) small hepatocytes within a focus (arrowheads). Note the high density of the nuclei and the almost complete absence of laminin staining within the focus. Oval cell ductules near the focus are laminin positive. (D) Low-dose model. The differentiating small hepatocytes are HNF-4 positive (small green nuclei). OV6-positive (red) and laminin-positive (blue) oval cell ductules close to the portal area are not stained for HNF-4. The nuclei of interlobular bile ducts (lower left corner) are also negative for HNF-4. (E) High-dose model. There is no OV6 (green) staining in the focus except several entrapped oval cell ductules (arrowheads). The numerous oval cells surrounding the focus are strongly stained for OV6 and laminin (red). Scraps of laminin are present in the focus outlining the tubular architecture of the small hepatocytes (arrows). (F) Low-dose model. Differentiating oval cell ductule close to a portal tract. The proximal part of the ductule has continuous laminin (red) staining, whereas distally the basement membrane is fragmented or absent (small arrows). The old hepatocytes show fine reticular cytoplasmic staining for pan cytokeratin (green; arrowheads). Although a strong membranouslike reaction is present in the proximal oval cells, the cytoplasm of the differentiating cells over the fragmented basement membrane already has reticular cytokeratin staining (large arrowheads). Scale bars: (A, C, E), 100 μm; (B, D), 50 μm; (F), 10 μm.

Figure 2.

Ultrastructure of the focus in the high-dose model. (A) Electron micrograph of the periphery of a focus. Small hepatocytes containing numerous low-density mitochondria and rough endoplasmic reticulum form a tubular structure. From the central lumen (L) bile canaliculi originate (arrows). The basal side of the cells face sinusoids (S). An old hepatocyte with dense mitochondria is visible on the bottom. A large arrow points at the area visible at higher magnification on Fig. 2D. (B) A fragmented basement membrane (arrowheads) is visible at the basal surface of a small hepatocyte lacking microvilli. S, sinusoid. (C) The basement membrane is completely absent at the basal surface of the small hepatocytes (E, endothelial cell). (D) Detail of Fig. 2A. Bile canaliculi sealed by desmosomes can be observed between the small hepatocytes. Numerous microvilli project into the lumen of the canaliculi. Scale bars: (A), 10 μm; (B–D), 2 μm.

Differentiation of Oval Cells After a Low Dose of AAF.

When 2.5 mg/kg AAF was administered daily, fewer oval cells developed, and the oval cells did not infiltrate the hepatic lobule to the same extent as seen after the high dose was administered. The basic difference between the two doses becomes obvious 5 to 6 days after the PH when oval cells rapidly differentiate into small hepatocytes. The structural configuration of the new hepatocytes was completely different from the foci seen after the high dose of AAF: they were arranged in more-or-less straight ducts (Fig. 1B). However on the cellular level, these differentiated ductular cells were similar to the cells of the foci: small polygonal cells with basophilic cytoplasm and ultrastructural features of hepatocytes (Fig. 3A). Hepatocytic differentiation seemed to occur synchronously in the overwhelming majority of the oval cells. The few unchanged oval cells were located at the proximal end of the ducts, whereas the distal part of the same duct was composed of differentiated small hepatocytes (Fig. 1F). The liver structure was almost normal 10 to 12 days after PH in the low-dose model, whereas it took 23 to 25 days after the high-dose treatment.

Figure 3.

Ultrastructure of a differentiating ductule in the low-dose model. (A) A short, differentiating ductule near a portal tract in the low-dose model. Two of the ductular cells (D1, D2) are attached to an old hepatocyte, which contains numerous dense mitochondria. One ductular cell (D3) closely resembles to hepatocytes. It has a round nucleus with prominent nucleolus, numerous mitochondria, and organizing rough endoplasmic reticulum. (B) Detail of Fig. 3A. The basement membrane cannot be observed at the basal side of the small hepatocyte (D3), while it is still present (arrowheads) around the nondifferentiated ductular cell (D4). Scale bars: (A), 10 μm; (B), 2 μm.

Differentiation of Oval Cells Into Hepatocytes HNF-4 is Upregulated in the Differentiating Cells.

HNF-4 is a liver-enriched transcriptional factor that is expressed in the hepatocytes but not in the biliary cells of the normal liver.12 Oval cells were not decorated by the HNF-4 antibody. However, nuclear staining with the HNF-4 antibody was seen in hepatocytes that formed foci in the high-dose model and in hepatocytes in ductlike structures in the low-dose model (Fig. 1C, D). Newly differentiated hepatocytes were recognized after both low and high doses of AAF by the smaller nuclei and higher cell densities compared with the old hepatocytes. Cells faintly HNF-4 positive also were observed occasionally at the distal ends of the oval cell ductules in the low-dose model (Fig. 4C).

Figure 4.

(A) High-dose model. Focus (arrowheads), 12 days after partial hepatectomy. Sinusoids of the liver parenchyma and the focus staining positively for α1 integrin (green). Note the lower density of sinusoids within the focus. The oval cell ductules bordering the focus are stained red (laminin), demonstrating the lack of colocalization of the two signals. (B) High-dose model. The same focus (arrowheads) as on Fig. 4A in a different sectioning plane stained for α6 integrin. The focus and the liver parenchyma are not reacting with this antibody. Yellowish color of the oval cell ductules bordering the focus shows that α6 integrin (green) overlaps with the positive (red) basement membrane. (C) Low-dose model. HNF-4 positive (small red nuclei) small hepatocytes express α1 integrin (green) mainly at their basal side (arrowheads). Weak α1 integrin positivity also can be observed at the lateral side of these cells in some locations (small arrows). Laminin positivity (blue) cannot be observed around the small hepatocytes. Laminin-positive oval cell ductules are negative for this integrin and HNF-4. However, note a weak HNF-4 nuclear staining in a differentiating cell (large arrowhead). (D) Low-dose model. The periportal area, stained for HNF-4 (red), α6 integrin (green), and laminin (blue). Bile and oval cell ductules are positive for this type of integrin and laminin, whereas the small hepatocytes (small red nuclei) are not stained. (E) High-dose model. There are numerous laminin-positive (red) oval cell ductules, also stained positively for desmin (green). The focus (lower left corner) is negative for desmin except some oval cell ductules enclosed into the focus (arrows). (F) Low-dose model. Only a few desmin-positive (red) stellate cells are visible (arrows) around the HNF-4–positive (green) and laminin-negative (blue) small hepatocytes still forming a ductule (arrowheads). Scale bars: (A, C, E), 100 μm; (C, D, F), 50 μm.

The Basement Membrane Disappears During the Process of Differentiation.

Laminin-containing basement membrane surrounds the undifferentiated oval cells, which also have relatively strong cytoplasmic and membranous cytokeratin staining, unlike the weak reticular staining of the hepatocytes (Fig. 1F).

Differentiating oval cells enlarge and acquire a hepatocytelike reticular pattern of cytoplasmic cytokeratin staining. Simultaneously, the basement membrane disintegrates, starting at the distal end of the oval cell ducts in the low-dose model (Fig. 1F). This observation was confirmed by electron microscopy (Figs. 2B, C and 3B). The paucity or complete lack of laminin staining was characteristic of foci (Fig. 1C,E), but occasional entrapped oval cell ducts were surrounded by basement membrane (Fig. 1E). Also, numerous undifferentiated oval cells located outside foci were still outlined brightly by laminin staining (Fig. 1C, E) in the high-dose model.

Changes in Integrin Expression Correlate With Loss of Basement Membrane.

We analyzed the expression of the biliary integrin α6 as well as of integrin α1, which is present on hepatocytes and sinusoidal endothelial cells in the normal liver.19 The α1 integrin was absent from the oval cell ducts, but it is clearly present on the small hepatocytes located in the foci (Fig. 4A). The density of α1 integrin on small hepatocytes is lower than in the surrounding parenchyma, but the sinusoidal-type arrangement already could be recognized. α1 integrin also was expressed on the small hepatocytes (Fig. 4C) in the low-dose model.

Opposite results were obtained with the α6 integrin. Surfaces of the oval cells facing the basement membrane stained positively, whereas the small hepatocytes in the foci (Fig. 4B) and in the ducts of the low-dose model (Fig. 4D) were negative.

Characteristic Changes in Differentiation Markers OV-6, AFP Staining, and Desmin-Positive Stellate Cells Disappear During Differentiation.

OV-620 and AFP13 are among the most frequently used antibodies to identify oval cell rat liver. OV-6, which is a monoclonal antibody recognizing cytokeratin (CK) 14 and 19,21 did not react with the differentiated hepatocytes either in the high-dose (Fig. 1E) or low-dose AAF models (Fig. 1D). AFP staining was lost (data not shown) from the differentiated hepatocytes located in the foci or in differentiated ducts, and desmin-positive stellate cells disappeared simultaneously (Fig. 4E, F).

Changes in Connexin Expression.

Hepatocytes express connexin 32 in normal liver, whereas cells of the biliary system express connexin 43.22 Oval cells were positive for connexin 43, but this protein was not present in the newly differentiated small hepatocytes (Fig. 5B). Connexin 32 was entirely lacking in oval cells (Fig. 5A), but it highlighted very clearly the new hepatocytes in both models (Fig. 5A). (The low-dose data are not shown.)

Figure 5.

(A) High-dose model. Connexin 32 (green) is expressed by the small hepatocytes of the focus (arrowheads). The surrounding liver parenchyma express connexin 32 at a very low level, whereas the laminin-positive (red) oval cell ductules are negative. (B) Low-dose model. Oval cell ductules are stained positively for connexin 43 (red) and OV6 (blue). The small hepatocytes marked by HNF-4 (small green nuclei) do not express this connexin and OV-6. (C) High-dose model. Numerous CD26-positive (green) starlike structures are discernible within the focus. CD26 staining reveals the polygonal organization of the bile canaliculi in the normal liver. This antigen shows a linear arrangement (arrowheads) in the oval cell ductules surrounded by basement membrane (laminin, red). (D) High-dose model. Higher magnification of a focus. Development of bile canaliculi (CD26 staining, green) between the small hepatocytes takes place only where regular basement membrane (laminin, red) is not present. (E) Low-dose model. A differentiating oval cell tube near a portal tract. Basement membrane (laminin, red) cannot be observed around the starlike structure (CD26, green) representing bile canaliculi development. (F) High-dose model. Advanced stage of the bile canaliculi development within a focus. Beside the starlike arrangement of the CD26 staining (green), two polygonal structures (arrows) can be seen, showing close similarity to the bile canaliculi network of the normal liver. Some remnants of the basement membrane can also be observed (laminin, red). (G) High-dose model. Peripheral zone of a focus filled partially with fluorescein isothiocyanate labeled lycopersicon esculentum lectin. The border of the focus is marked by arrowheads. The wide lumen of an oval cell ductule is continuous with the narrow bile canaliculi of the focus. Starlike (large arrowhead) and polygonal (small arrows) structures can be recognized within the focus. Scale bars: (A, C), 100 μm; (B), 50 μm; (D–G), 10 μm.

Bile Canaliculus Formation by the Differentiating Cells.

Oval cells, which extend from the canals of Hering, form ducts that are continuous with bile canaliculi in hepatic plates. We studied the structure of the bile drainage system by CD26 staining and by retrograde infusion of a fluorescein isothiocyanate–labeled lectin into the biliary system via the common bile duct.

CD26 or dipeptidylpeptidase IV23 was expressed at the apical surface or biliary pole on the hepatocytes and on the apical domain of the biliary and oval cells (Fig. 5C). In the low-dose model, the pattern of CD26 staining changed sharply where laminin staining disappeared and the oval cells differentiated. Luminal branching to form starlike structures was observed in this area, representing the formation of bile canaliculi and the polarization of the new hepatocytes (Fig. 5E). The situation was very similar in the high-dose model, where numerous starlike lumens were detected among differentiating hepatocytes within foci (Fig. 5C, D). Electron microscopy of the foci clearly showed the formation of bile canaliculi from a central lumen between the differentiated small hepatocytes (Fig. 2A, D). Occasionally, polygonal canalicular structures were outlined inside the foci by CD26 staining (Fig. 5F), similar to the pattern of bile canaliculi in the normal liver.

Cholangiography supported the results of CD26 staining. Broad lectin-filled ducts led to foci, where the ducts divided into smaller branches that formed polygonal canalicular structures (Fig. 5G). The newly formed bile canaliculi also could be filled between the small hepatocytes in the low-dose model (data not shown).


We report two distinct patterns for the differentiation of oval cells into hepatocytes in the AAF/PH model (Fig. 6). When a low dose of AAF was used, most of the oval cells differentiated synchronously and rapidly into small hepatocytes, even while AAF was being administered. Newly formed hepatocytes maintained a ductular arrangement during the early phase of the differentiation process. In contrast, when a high dose of AAF was used, hepatocyte differentiation was delayed by 7 to 10 days, and even at this later time point, most of the oval cells did not differentiate into hepatocytes. Newly formed hepatocytes formed foci scattered throughout the liver with the high dose of AAF. However, the differentiation process seemed to be identical in the two models at the cellular level. The small basophilic cells were very similar to the adult hepatocytes by light microscopic and electron microscopic examination. New hepatocytes lost several phenotypic characteristics of oval cells, for example, AFP and OV-6 expression. Upregulation of albumin13 and CYP3A116 in the newly formed cells indicated their functional maturation.

Figure 6.

Schematic representation of the two differentiation models. Low-dose model: oval cell ductules penetrate into the liver lobule and differentiate simultaneously into small hepatocytes. High-dose model: oval cells penetrate deep into the liver lobule and most of them do not change during the observation period. The differentiating cells arranged in tortuous and branching ductules, which form the foci. The lumen of these ductules is continuous with the larger oval cell ductules. The developing bile canaliculi between the small hepatocytes invaginate from the central lumen of these ductules in both models. (The days mean the time after the partial hepatectomy).

The disappearance of the basement membrane that surrounds the oval cell ductules is closely associated with initiation of the differentiation process. Oval cells always sit on a well-structured basement membrane1 that is almost completely missing from the foci. Gradual disintegration of the basement membrane was observed even better in the low-dose model. It is well known that the connection with the basement membrane influences the differentiation state of the cell. The close contact with the basement membrane is absolutely required for the maintenance of differentiated tubules in the kidney.24 In the skin, however, the loss of contact is a stimulus for differentiation for the keratinocytes.25 The hepatocytes have no contact with laminin containing structured basement membrane. Yin et al.26 reported that isolated hepatic stem cells expressed biliary or hepatocytic phenotypes in culture, depending on the presence or absence of basement membrane matrix (Matrigel). Matrigel also was found to play an important role in maintaining the biliary phenotype in tissue culture in other experimental systems.27, 28 These observations are in good agreement with our in vivo results.

Integrins are heterodimeric glycoproteins, consisting of α and β subunits, that enable cells to recognize adhesive substrates in the extracellular matrix. The α6 integrin subunit binds laminins exclusively. The expression of the α6 integrin and laminin seem to influence each other. The induction of α6-containing integrins at the surface of developing epithelial cells is strongly correlated with the deposition of laminin.29, 30 The downregulation of the laminin receptor was described in differentiating HBC-3 hepatic stem cells in culture.31 The disappearance of the α6 integrin from differentiating oval cells is in good agreement with the disintegration of the basement membrane. In the normal liver, α6 and α1 integrins are expressed on biliary cells and hepatocytes, respectively.19 Therefore, the α6-α1 switch is another indicator of the hepatocytic differentiation. The opposite change was described when the hepatoblasts differentiated into biliary cells.32 The α1 integrin is present on the basal surface of endothelial cells and on the basolateral surface of small hepatocytes. This polarized expression may promote the establishment of connections with collagenous components of the perisinusoidal space, an important step in the reconstruction of the normal trabecular liver structure.

The upregulation of HNF-4 expression may play a central role in the induction of hepatocytic differentiation of the oval cells. HNF-4 is not present in oval cells, but it clearly decorates the nuclei of small hepatocytes in both models, and a faint staining was sometimes observed in the transitional cells in the low-dose model. In an earlier work, we failed to detect HNF-4 mRNA in oval cells, although other liver enriched transcriptional factors were present.12 However, HNF-4 mRNA was upregulated in foci of small basophilic hepatocytes, suggesting a critical role for this transcription factor in the oval cell differentiation.12 The lack of expression of a functional HNF-4 gene results in embryonic lethality in mice before development of the liver as a result of defects in visceral endoderm function.33 However, HNF-4 null mouse embryos can be rescued until E12.0 by extraembryonic endoderm complementation.34, 35 Livers from these endoderm-complimented embryos were morphologically and histologically indistinguishable from wild-type embryos. When the mRNA level of hepatocyte-specific genes was compared between wild-type and HNF-4 null livers, it was found that the expression of these genes were either downregulated or undetectable in the genetically manipulated mice. Similar results were found in adult hepatocytes that lack HNF-4.36 These data suggest that HNF-4 is not required during ontogenesis for the competency of hepatic precursor cells or for their specification, but it must be present during the final step of hepatocyte differentiation to establish the hepatocytic gene expression pattern. This observation is consistent with our current results that indicate the critical role of HNF-4 in the transition of oval cells into hepatocytes.

HNF-4 also is expressed in the intestinal glands,37 and intestinal metaplasia is another option for the oval cells.38 One of the major differences between the intestinal glands and hepatocytes is that the enterocytes reside on a laminin-containing basement membrane. It is possible that the upregulation of HNF-4 without the disintegration of the basement membrane may be responsible for the “abnormal” differentiation of oval cells into intestinal epithelium.

Connexins form special membrane structures, the gap junctions (connexons), which play crucial role in the intercellular communication. Connexin 32 is expressed on hepatocytes, and connexin 43 is expressed on the biliary epithelial cells in the normal rat liver.39 Zhang and Thorgeirsson40 reported that the mRNA for connexin 43 is expressed in the oval cells and connexin 32 in the small hepatocytes of the foci. Here we confirm this observation at the protein level. The switch from connexin 43 to connexin 32 can be observed in both the low-dose and high-dose models of oval cell differentiation. Because connexin 32 and connexin 43 hemichannels do not form heterotopic patent channels,41 the described switch is required for the newly formed hepatocytes to be integrated into the preexisting liver plates and to communicate with the preexisting hepatocytes. It also may be of importance that the stellate cells also express connexin 43.42 The expression of identical connexins on the oval cells and stellate cells may be important in the establishment of the close communication between these two cell types. The loss of connexin 43 from the oval cells coincides with the disappearance of stellate cells. It is well documented that the stellate cells provide a battery of growth factors that support the proliferation of oval cells.43 The inductive role of mesenchymal cells during the specification of hepatic lineage also is well known.44 Stellate cells may provide support for the growth of oval cells, as does the portal mesenchyme during embryogenesis. This notion is supported by the suggestion that the stellate cells may be derived from the embryonic septum transversum.44

We observed the development of biliary canaliculi among the newly formed hepatocytes in both models by infusing the biliary system with fluorescein isothiocyanate–labeled lectin via the common bile duct. The pentagonal and hexagonal bile canalicular patterns in the foci are similar to the arrangement present in the normal liver. This structural similarity indicates the architectural remodeling in the foci toward the normal hepatic structure. The CD26 epitope is present on the canalicular surface of the normal hepatocytes.23 The similar CD26 staining pattern on the newly formed small hepatocytes marks the functional polarization of these cells.

The morphological and immunophenotypical characteristics clearly show that oval cells differentiate into hepatocytes at both low and high doses of AAF. Although both the rate of differentiation and timing of the process differ with AAF dose, the differentiation process at the cellular level seems to be identical. Furthermore in both models, the differentiating small hepatocytes can regenerate the liver after two-thirds PH, because the old hepatocytes do not show mitotic activity. At this point, we can only speculate about the cause of focus formation, when a high dose of AAF was applied. A plausible explanation may be that the high dose of AAF inhibits the low dose–type differentiation of the oval cells, as suggested by Alison et al.16 However, the inhibitory effect of the high dose of AAF primarily may affect the later stages of oval cell differentiation into hepatocytes, and this in turn can attenuate the necessary remodeling needed to rebuild the liver structure. It is possible that this or a similar scenario can contribute to the focus formation seen after administration of the high dose of AAF. Whether this process involves genetic or adaptive epigenetic mechanisms, or both, is not clear. However, we have not observed liver tumors in the rats subjected to the high AAF doses used here, suggesting that if a mutation(s) is involved in the process, it may not be carcinogenic. Nevertheless, the focus formation pattern of oval cell differentiation after high doses of AAF identifies, independent of the precise mechanism(s), an efficient differentiation process that is functional under adverse conditions. Therefore, further characterization of the high-dose model may be very useful to delineate the factors that can be used to enhance the differentiation efficiency of the hepatic stem cells. This issue may have important implications for the clinical application of adult liver stem cells. Also, the fact that oval cells can, under favorable conditions (e.g., in the low-dose AAF model), rapidly differentiate in new hepatocytes that effectively integrate into liver plates raises the intriguing possibility that adult liver stem cells may contribute to liver regeneration and repair more often than previously anticipated. Finally, this study provides no evidence that hematopoietic bone marrow stem cells are involved in the generation of new hepatocytes through oval cells, which seem to be totally derived from epithelial cells of the canals of Hering.