There are several alternative mechanisms of liver regeneration.1 Hepatocytes can enter the cell cycle or enlarge, and the consequent compensatory hyperplasia or hypertrophy replaces the lost liver mass. We have demonstrated that new hepatic lobules are not formed during this type of regeneration, but the remaining lobes grow exclusively by the enlargement of preexistent hepatic lobules.2 If the hepatocytes are compromised, the progenitor cell compartment is activated and the liver regenerates by means of the so-called oval cells in the rat.3 Oval cells invade the liver parenchyma from the periportal region. They form ductules, which are the extensions of the canals of Hering.4 The distribution of the differentiating cells can be different, depending on the species and etiology. In human liver, singular intermediate hepatobiliary cells can often be observed in proliferating ductules after extensive necrosis,5 but small groups of differentiating cells are also described in chronic hepatitis and cirrhosis.6, 7 The 2-acetaminofluorene/partial hepatectomy (AAF-PH) model8 is one of the most widely applied experiments to study oval cell proliferation and differentiation in rats. We characterized the differentiating progenitor cells in this experimental model. Two alternative patterns of differentiation were observed previously depending on the applied dose of AAF.9 In the case of a low dose (2.5 mg/kg/day of AAF), differentiation occurred earlier and practically all the oval cells transformed into small, newly formed hepatocytes. If the dose of AAF was higher (5 mg/kg/day of AAF), differentiation occurred later and involved only a small portion of oval cells. Sharply circumscribed foci constructed by small hepatocytes appeared in the liver parenchyma. Their distribution seemed to be random on traditional sections and were surrounded by the rest of the oval cells. This latter histological reaction is almost identical with early events of the Solt-Farber hepatocarcinogenesis model,10 where the foci are supposed to be putative, clonal tumor precursor lesions, which derive from the initiated hepatocytes.11 The AAF-PH model does not lead to tumor formation; apparently, the normal architecture is reestablished in a few weeks. In our present work, we investigated the structural aspects of oval-cell–mediated liver regeneration: (1) how the “foci” evolve into lobular hepatic parenchyma; (2) whether there is any difference between the foci of the carcinogenic and noncarcinogenic experiments, and (3) if there are any architectural changes in the regenerated liver.
We have analyzed the architectural aspects of progenitor-cell–driven regenerative growth in rat liver by applying the 2-acetaminofluorene/partial hepatectomy experimental model. The regeneration is initiated by the proliferation of so-called oval cells. The oval cells at the proximal tips of the ductules have a more differentiated phenotype and higher proliferative rate. This preferential growth results in the formation of a seemingly random collection of small hepatocytes, called foci. These foci have no clonal origin, but possess a highly organized structure, which shows similarities to normal hepatic parenchyma. Therefore, they can easily remodel into the lobular structure. Eventually, the regenerated liver is constructed by enlarged hepatic lobules; no new lobules are formed during this process. The foci of the Solt-Farber experimental hepatocarcinogenesis model have identical morphological features; accordingly, they also represent only regenerative, not neoplastic, growth. Conclusion: Progenitor-cell–driven liver regeneration is a well-designed, highly organized tissue reaction, and better comprehension of the architectural events may help us to recognize this process and understand its role in physiological and pathological reactions. (HEPATOLOGY 2012)
Materials and Methods
Male F-344 rats (160-180 g) were used for all experiments and were kept under standard conditions. At least 4 animals were used for each experimental time point (unless otherwise marked). Animal study protocols were conducted according to the National Institutes of Health (NIH) guidelines for animal care.
AAF (2 mg/mL suspended in 1% methylcellulose) in a dose of 5 mg/kg was administered to rats daily for 4 consecutive days by gavage. A traditional two thirds PH was performed12 on day 5, which was followed by four additional AAF treatments. Animals were sacrificed at time points indicated below.
Diethyl nitrosamine (DEN) (200 mg/kg) was intraperitoneally injected into rats.10 Two weeks later, the AAF-PH protocol was performed as described above.
Determination of the Size of the Lobules.
The procedure was described in detail previously.2 In brief, central veins and sinusoids were filled up retrograde through a cannula inserted into the vena cava inferior by a polystyrol resin containing fluorescent dye. The liver surface was monitored by a stereomicroscope, and the filling was stopped when the resin filled the hepatic sinusoids partially. At this stage, the “negatives” of the interlobular borders were outlined by the areas not filled by the resin. The surface image of the lobe was captured by a Bio-Rad (MRC1024; Bio-Rad, Richmond, CA) confocal system (excitation, 488 nm; emission, 520 ± 16 nm). Circumference and surface area of the lobules (Table 1) were always determined on the right lateral liver lobe by the ImageJ program (NIH, Bethesda, MD).
|Parameter||Control||28 Days After PH||3 Months After AAF-PH|
|Weight (g)||1.48 ± 0.004||5.11 ± 0.24*||6.92 ± 0.59†|
|Area of surface lobules (mm2)||0.369 ± 0.0367||0.851 ± 0.026*||0.965 ± 0.039†|
|Circumference of surface lobules (mm)||2.42 ± 0.122||3.67 ± 0.052*||3.87 ± 0.044†|
|Number of surface lobules||794.5 ± 37.1||719.33 ± 31.8||736 ± 39|
|Number of portal vein branches around central veins||6.04 ± 0.2||8.13 ± 0.38*||8.4 ± 0.2†|
Counting of the Surface Liver Lobules.
In another set of rats, the portal venous system was filled up by blue-stained resin, in addition to the red-resin–outlined hepatic veins. These animals were used to count the absolute number of liver lobules on the convex surface of the right lateral lobe. The number of portal vein branches around a central vein was also counted on these specimens2 (Table 1; Fig. 1A,B).
Zonality of Liver Lobules.
The right lateral liver lobe was frozen under slight pressure to produce a flat surface for cutting. The section was stained for cytochrome p450 (CYP450) IIE1 (Supporting Table 2) for 1 hour at room temperature, and the reaction was visualized by a fluorescein-isothiocyanate–labeled secondary antibody (30 minutes at room temperature) (Fig. 1 C,D).
Quantitative Real-Time Polymerase Chain Reaction.
Frozen sections (10 μm) were fixed in acetone, dried at room temperature, and stained with RNase-free hematoxylin. Laser microdissection of oval cells, foci, and hepatocytes (pericentral and periportal hepatocytes from healthy liver and pericentral hepatocytes from AAF-PH liver) was performed by using the PALM MicroBeam system (Carl Zeiss Microimaging GmbH, Göttingen, Germany). The area of oval cell proliferation was divided into three concentric regions. Oval cells dissected from the pericentral regions were called proximal and those dissected from periportal regions were called distal oval cells (Supporting Fig. 3). Total RNA was isolated by the RNA Aqueous Micro Kit (catalog no.: AM 1931; Ambion, Austin, TX). A high-capacity complementary DNA (cDNA) reverse-transcription kit (catalog no.: 4368814; Applied Biosystems, Weiterstadt, Germany) was used for cDNA synthesis, as recommended by the supplier. Polymerase chain reaction (PCR) was performed by the ABI Prism 7300 Sequence Detection System (Applied Biosystems), using ABI TaqMan gene expression assays (Supporting Table 1) according to the manufacturer's instructions. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH; part no.: 4352338E) was used as an endogenous control, because GAPDH expression was quite uniform in all examined samples. All samples were run in triplicate in a 20-μL reaction volume. Results were obtained as threshold cycle (CT) values. Expression levels were calculated using the ΔCTmethod.13 Values were calculated as the mean value of three independent measurements, and expression levels of messenger RNA (mRNA) in all samples were defined as the ratio to GAPDH expression (Fig. 2 and Supporting Figs. 1 and 2).
Immunohistochemistry and Three-Dimensional Reconstruction.
Frozen sections (10-20 μm) were fixed in methanol and were incubated at room temperature (1 hour) with a mixture of the primary antibodies (Supporting Table 2) and then with appropriate secondary antibodies (Jackson ImmunoResearch, West Grove, PA). All samples were analyzed by confocal laser scanning microscopy using the Bio-Rad MRC-1024 system (Bio-Rad) (Figs. 3 and 4). Three-dimensional (3D) reconstruction was made from serial frozen sections stained for laminin and OV-6 (Supporting Fig. 4). The former was used to recognize the outline of the focus (ductularly arranged small hepatocytes) and vessels, and the latter was used to detect the oval cells. As a result of the extremely large number of individual oval-cell–formed ductules, only the outline of the area occupied by oval cells was reconstructed. Large-caliber structures with strong basement membrane and OV-6 staining was recognized as interlobular bile ducts. The 3D reconstruction was performed by using the BioVis3D program (Biovis 3D, Montevideo, Uruguay) (Fig. 5).
We used the amphotropic retroviral vector (kindly provided by Dr. Nicholas Ferry, Nantes, France), containing the Escherichia coli beta-galactosidase (β-gal) gene coupled to the nuclear localization signal from a SV40 large T antigen, produced by the TELCeB6-producing cell line.14 Retroviral titer determination and beta-galactosidase (LacZ) staining was performed as described before.15 The viruses were injected into the liver of the AAF-PH-treated animals 2 days after the PH through the common bile duct. The first set of rats was sacrificed 4 days (pulse), the second and third sets 12 and 23 days after PH (chase) (Figs. 6 and 7). Glucose-6 phosphatase (G6P) background staining16 of old hepatocytes outlined the foci.
To confirm the presence of the virus in foci, “blue” and “white” components (LacZ-positive and -negative areas) of the same focus was microdissected (8 foci were analyzed from 3 different livers), RNA was isolated and transcribed in cDNA as described before. Traditional PCR was performed (Verity 96-well thermal cycler; Applied Biosystems) using Red Taq Ready Mix (catalog no.: R2648; Sigma-Aldrich, St. Louis, MO). On the basis of the virus sequences, one pair of primers (Supporting Table 1) was designed with Primer Express V3.0 (ABI). GAPDH was used as the internal control and the PCR products were analyzed by 2% agarose gel electrophoresis (Bio-Rad) (Fig. 6).
Corrosion Casting of the Foci.
The vascularization of the foci was analyzed in AAF-PH (Fig. 3) and the Solt-Farber hepatocarcinogenesis (Fig. 7) model. The portal venous system was filled up by blue-stained resin, in addition to the red resin, which outlined hepatic veins. Livers were frozen, and to detect the foci deep in the parenchyma, cryostat sections were cut and stained with toluidine blue. Frozen sections were photographed (Figs. 3C and 7C), and the remaining part of the liver lobes were placed overnight in 35% KOH at 60°C. The casts (Fig. 3D) were washed in running tap water. The positions of the foci were indentified by comparing the surface of the cast with the toluidine-blue–stained section. Other livers were removed and cut in lobes, and then superficially situated foci were photographed (Olympus SZ61 dissecting microscope; Olympus Optical Co., Tokyo, Japan) (Fig. 7A,B).
Data are represented as the mean ± standard error (SE) of at least three independent experiments. Statistical significance of difference between groups was analyzed by the Student's t test. Values of P < 0.05 were considered statistically significant.
Morphometric Analysis of the Surface Liver Lobules in the AAF-PH Regeneration Model.
We have developed a new technique enabling us to perform accurate morphometric studies on surface hepatic lobules. This technique was applied to measure different parameters of lobules on the right lateral lobes of rat livers 3 months after the AAF-PH experiment. By this time point, the lobular structure was reestablished. Results were compared to the dimensions of the healthy control rat livers and to the regenerated livers 28 days after two thirds PH. This latter data derive from our previous published study.2 The weight of the investigated liver lobe grew 3.29-fold in the PH model and 4.68-fold in the 3 months after the AAF-PH experiment. The area and circumference of the individual surface lobules enlarged 2.3- and 2.6-fold or 1.5- and 1.6-fold, respectively, in the two alternative regenerative models. The number of the surface lobules was practically constant in all investigated samples (Table 1). The preexistent lobules enlarged after PH, whereas in the AAF-PH model, the small hepatocytes of the foci remodeled into lobules. Further similarity between the outcome of the two regenerative reactions was that the number of the portal vein branches increased around each individual lobule (Fig. 1A,B), and the expression of CYP450 IIE1 (Fig. 1C,D) showed a kind of arborescent pattern, compared with the regular concentric pericentral arrangement, in healthy liver tissue.
Zonal Heterogeneity of Oval Cells.
The heterogeneity of oval cells is well known. There is no generally accepted terminology of oval cells located in different regions of hepatic lobules; oval cells closer to the central vein will be called “proximal,” whereas the ones closer to the portal tract will be called “distal” further on. Higher alpha-fetoprotein (AFP) and delta-like protein expression in the proximal oval cells has already been described.17, 18 Similar zonal distribution could be demonstrated for the expression of other differentiation-related genes. Distal and proximal oval cell groups have been microdissected separately from frozen sections, and the relative gene expression has been analyzed by real-time quantitative RT-PCR. Higher AFP expression in the proximal oval cells supported the previous immunohistochemical (IHC) results and appropriate microdissection. The relative gene expression level of other differentiation-related genes, such as transthyretin (TT) and tyrosin aminotransferase (TAT), was also significantly higher in the proximal samples (Fig. 2B,C). Tryptophan 2,3-dioxygenase (TDO2) (Fig. 2D) and alpha-1 antitrypsine (AAT) (Supporting Fig. 1A) expression was lower in the distal oval cells, but was not significantly different from the expression in the proximal oval cells. AAT expression of the periportal hepatocytes (healthy liver) and pericentral hepatocytes (healthy liver and AAF-PH-treated liver) proved to be quite uniform (Supporting Fig. 1B). TT, TAT, and TDO2 expression was significantly higher in the periportal versus pericentral hepatocytes and decreased after AAF treatment (Supporting Fig. 2), as is known from the literature.19 Consequently, the polarized expression pattern of these enzymes cannot be explained by hepatocytic contamination during microdissection, because the gradient of expression level is opposite in the hepatocyte plates and oval cells. The expression of these “differentiation markers” further increased in the foci (Fig. 2). No such expression gradient could be demonstrated in the oval cells for cytokeratin (CK)19; in fact, the steady-state mRNA level of this biliary marker was slightly higher in the distal oval cells (Fig. 2E).
The proliferation activity between proximal and distal oval cells has also been compared. Both bromodeoxyuridine (BrdU) incorporation (Fig. 3A) and cyclin A (Fig. 2F) expression indicated higher cell proliferation in proximal oval cells.
The Architecture of Regenerating Foci in the AAF-PH Experiment.
The oval cells differentiate into hepatocytes in the presently investigated model by forming well-defined groups of small hepatocytes; these structures are called small hepatocyte foci8, 20 (Fig. 3B). These foci grow and gradually merge into the structure of the liver. However, it is not clear how the lobular structure is reconstructed from the foci, which look like a randomly arranged collection of hepatocytes on traditional histological sections. To study the relationship of the foci with the liver vasculature, the portal and hepatic venous system was filled by blue- and red-colored resin 12 days after PH, at the peak of focus formation. Surprisingly, the blue resin injected into the portal system always appeared inside the foci, whereas the red resin injected into the central venular system became visible in vessels outside of the foci (Fig. 3C). Central veins were never observed inside the foci. The blue portal resin sometimes filled dilated sinuses inside the foci, but in larger ones, the incorporation of portal veins could occasionally be observed. After the corrosion of the liver tissue, stereo microscopic examination of casts revealed that each focus is supplied by one terminal portal venule and the resin is distributed by this vessel into the sinusoidal system (Fig. 3D).
Careful examination of the foci confirmed that they were not simply a collection of randomly arranged cells. Remnants of basement membrane outline ductularly arranged structures with apical cluster of differentiation 26 (CD26) staining (Fig. 3E) and there were well-formed Lyve-1-positive (Fig. 3F) sinusoids among them, which were continuous with sinuses of the adjacent liver parenchyma. Furthermore, occasionally entrapped OV-6-positive oval-cell–formed ductules could be observed inside the foci, which ended up with an open “U”-shaped basement membrane on the small hepatocytes of the foci, because the canals of Hering are connected to the hepatic plate in the healthy liver (Fig. 4A,B and Supporting Fig. 5).
To reveal the relationship of the foci with their neighboring structures, serial immunostained frozen sections were reconstructed to create the 3D image of a focus. The focus replaces a part of a classical lobule and is surrounded by neighboring fields of oval cells situated around portal venules. The focus is supplied by a single branch of a portal vein, whereas biliary ducts, which also enter the focus, are responsible for drainage of the bile that is produced (Fig. 5). The blood is collected by a central vein outside the focus. This arrangement supports that the formation of foci is not a random event. They fit perfectly into the preexistent hepatic structure, and therefore the lobular structure can be easily reestablished, because the convoluted ductular tissue structure of small hepatocytes “unfolds.”
Clonality Examination of the Foci in the AAF-PH Experiment.
A pulse-chase experiment was performed after the retroviral marking of the proliferating oval cells. Two days after PH, retroviruses were injected into the liver of AAF-PH-treated rats through the common bile duct. The pulse animals were killed 2 days later. β-gal staining demonstrated the preferential transduction of oval cells (usually, no more than 5% of the oval cells were labeled), no hepatocytes were marked, and no foci were present at this time point. The chase animals were sacrificed 12 and 23 days after PH. The majority of the foci remained white, because no retrovirally transduced oval cells contributed to their formation. However, approximately 25% of the foci (21 of 89) contained β-gal-stained cells, which mostly formed differently sized clusters inside the foci, but not a single completely blue focus could be observed in any of the rat livers that were studied (Fig. 6A). When the blue and white parts of the same focus were microdissected separately, the viral RNA could be demonstrated in the blue part, but not in the white one by traditional PCR (Fig. 6B). Consequently, we could not demonstrate any focus of monoclonal origin. Twenty-three days after the hepatectomy, when the nearly normal liver histology was reestablished, small groups of β-gal- and G6P-positive hepatocytes were observed as incorporated in the hepatic plates (Fig. 6C,D).
Structural and Clonality Examination of the Foci in the Solt-Farber Hepatocarcinogenesis Experiment.
The early histological events are almost identical if the AAF-PH treatment is preceded by a single treatment of 200-mg/kg DEN administration. This protocol is one of the most widely analyzed chemical hepatocarcinogenesis models, called the Solt-Farber or resistant hepatocyte model.10, 11 Using this carcinogenic protocol, the analysis of the foci could not reveal any difference to the foci developed without DEN treatment. The foci of the Solt-Farber model were also arranged around the terminal branches of the portal veins, whereas the central veins were pushed aside (Fig. 7A-C). The same retroviral pulse-chase experiment used in the case of the AAF-PH model revealed no completely blue-stained foci, but blue cells in small or larger clusters were present in 62 of the investigated 345 foci, suggesting their nonclonal origin (Fig. 7D,E).
In our present experiment, we have demonstrated that the liver regenerated after the AAF-PH protocol is constructed by the same number of lobules as the healthy liver, but these lobules are enlarged and have a convoluted lateral surface (Supporting Fig. 6). The outcome of this complex process is the same as that of the regeneration induced by simple PH.2 It is easy to explain the preservation of lobular structure in the latter case. The elongation of the hepatic plates might be responsible for the enlargement of the lobules. However, the original liver structure is seemingly disintegrated during the first few weeks of the AAF-PH experiment. The parenchyma is infiltrated by oval cell ductules; later, oval-cell–derived small hepatocytic foci are scattered in the liver and they have to reestablish the original structure.
Oval cells are not a homogenous population. The higher proliferative rate and the more differentiated (i.e., higher TDO2, TAT, AAT, and TT expression) phenotype of the proximal oval cells suggest that the differentiation process starts in these segments, probably involving a small subset of oval cells. The rapid proliferation of the small differentiating hepatocytes, as indicated by the high cyclin A expression (Fig. 2F), results in a quickly increasing cellular cluster (i.e., focus) pushing the neighboring parenchyma and mimicking clonal growth. However, the result of the retroviral pulse-chase experiment proves that this growth reaction is not clonal, because no completely blue, one-cell–derived foci can be demonstrated. There seems to be a close relationship between the proliferation and differentiation of hepatocyte precursors. The reverse-transcription (RT)-PCR experiments revealed the gradually increasing expression of hepatocytic differentiation markers in the distal-proximal oval cells and small hepatocytes of foci. A similar correlation between proliferation and hepatocytic differentiation has been described in other experimental models.15, 21
We have observed that the foci are neither just a hazard pile of newly formed hepatocytes nor randomly distributed in the highly organized liver parenchyma. The “organoid” structure of the foci is supported by (1) clues of ductular arrangement of the small hepatocytes inside the foci, (2) well-formed sinusoids, communicating with neighboring parenchyma, and (3) a permanently maintained connection with the biliary system through intrafocal ductules (canal of Hering–like). This organoid focus eventually can easily evolve into healthy lobular liver parenchyma by “unfolding” of the convoluted ductules and by reorganization into trabecules (Supporting Figs. 7 and 8). This process is also indicated by the transition of retrovirally marked small hepatocytes of foci into plate-forming G6P-positive hepatocytes.
At present, we do not know exactly how large a portion of the enlarged lobules is derived from the foci. There is an important difference between foci and lobules. Though the classical lobule is organized around a central vein and the portal branches are at the periphery, a fully developed focus is more similar to the so-called portal lobule.22 It should be kept in mind that the healthy liver lobule is demarcated by approximately six portal tracts. Thereby, an individual focus does not turn into a hepatic lobule, but more foci can contribute to the build-up of a regenerated liver lobule. Conversely, depending on its location and size, one focus can participate in the construction of more neighboring lobules.
Though the liver parenchyma is permanently reconstructed in our experimental model, the “backbone” of the liver, the vasculature remains stable. For that reason, we have investigated the relationship between the foci and blood vessels. The terminal branch(es) of portal veins can be demonstrated inside larger foci, and the central veins are pushed aside. This is not surprising, considering that others22, 23 and we demonstrated that the foci are the descendant of canals of Hering–derived oval cells, which are always arranged in a radial fashion around the portal tracts. It seems that oval cells preserving their original periportal position and connections transform into small hepatocytes and constitute the foci. Endothelial cells24 and branches of the portal vein25 play an essential role in liver organogenesis; they seem to be equally important for regeneration, mediating the so-called “hepatocyte-sinusoid alignment,”26 and the growth of oval cells formed ductules.4
The vascular structure of these differentiating foci is sharply different from the one we observed during the development of primary (unpublished) or metastatic experimental hepatic tumors.27 These “tumorous foci” are nourished exclusively by hepatic arteries, and the compressed portal and central vein branches are located around them. We are hypothesizing that this disagreement can be explained by particular pressure ratios. Regardless of the exact explanation, this basic difference in the vasculature also underlines the “friendly” growth pattern of the regenerative foci, contrary to aggressive neoplastic processes.
The oval cell origin of the “foci” and their participation in the regeneration in the AAF-PH experiment is widely accepted. However, the genesis and function of morphologically similar structures in the Solt-Farber hepatocarcinogenesis model are still debated.28-30 According to the original explanation, foci are clonal progenies of the initiated “resistant” hepatocytes.11 Some of them are thought to remodel into healthy parenchyma, whereas others progress into liver carcinomas. We have demonstrated substantial similarity in the distribution and structure of the foci in the AAF-PH and carcinogenesis models, suggesting that the foci of the carcinogenesis model also depict organoid, regenerative, oval-cell–derived organization. The tracing of retrovirally transduced oval cells also supports that the overwhelming majority of the foci have polyclonal, oval cell origin. Because oval cells were preferentially labeled by viruses, hepatocyte-derived foci might be potentially present among the β-gal negative ones. However, these hepatocyte-derived foci cannot be distinguished morphologically and structurally from the oval-cell–derived ones, or their number is too low to detect them. The precursor-product relationship between the foci and later tumors is not clear. There are histochemical markers, such as gamma-glutamyl transferase (GGT), a placental form of glutathione S-transferase (GST-P), which might label the putatively preneoplastic foci.31 The ratio of these marker-positive foci is much higher (data not shown) than the number of potentially hepatocyte-derived foci. Therefore, a large portion of GGT-GST-P-positive foci must derive from oval cells; they might only represent a regenerative growth or some of them may be preneoplastic foci, but in that case, tumors derive from oval cells.
The close correlation between structure and function is one of the basic facts of biology. Therefore, it is not surprising that the progenitor-cell–driven liver regeneration is a well “designed” tissue reaction resulting in almost completely healthy liver structure. It seems that the portal tree provides the “backbone” of efficient liver regeneration, but the maintenance of the direct cellular contact between the old and regenerating parenchyma may also be a requirement.