Potential conflict of interest: Nothing to report.
We examined the response of murine oval cells, that is, the putative liver progenitor cells, to acetaminophen. Female C57BL/6J mice were injected intraperitoneally with varying doses of N-acetyl-paraaminophen (APAP) (250, 500, 750, and 1,000 mg/kg of weight) and sacrificed at 3, 6, 9, 24, and 48 hours. In preliminary studies, we showed that anticytokeratin antibodies detected A6-positive cells with a sensitivity and specificity of greater than 99%. The oval cell reaction was quantified, on immunostaining for biliary-type cytokeratins, as both number and density of oval cells per portal tract, analyzed by size of portal tract. Acetaminophen injury was followed by periportal oval cell accumulation displaying a moderate degree of morphological homogeneity. Oval cell response was biphasic, not temporally correlating with the single wave of injury seen histologically. Increases in oval cells were largely confined to the smallest portal tracts, in keeping with their primary derivation from the canals of Hering, and increased in a dose-dependent fashion. The timing of the two peaks of the oval cell reaction also changed with increasing dose, the first becoming earlier and the second later. In conclusion, our studies indicate a marked oval cell activation during the height of hepatic injury. Oval cells appear to be resistant to acetaminophen injury. The close fidelity of mechanism and histology of acetaminophen injury between mouse and human livers makes it a useful model for investigating liver regeneration and the participation of stem/progenitor cells in that process. (HEPATOLOGY 2005.)
Experimental models currently available to study hepatic progenitor cells (HPCs) in animals include liver injury caused by partial hepatectomy, various chemical hepatotoxins, irradiation, toxic diets, and generation of transgenic (urokinase-type plasminogen activator, hepatitis B surface antigen, suicide genes) and knockout (fumaryl acetoacetate hydroxylase–deficient) mice.1–6 Nonetheless, except for partial hepatectomy, none of the above-mentioned models shows close fidelity to the pathological alterations found in human livers.
However, acetaminophen (N-acetyl-paraaminophen, tylenol, paracetamol, APAP) is a predictable hepatotoxin that reproducibly demonstrates close fidelity of histological injury between animals and humans.7–10 Because APAP is the most frequent cause of fulminant liver failure in both the United States and the United Kingdom, with a mortality rate of approximately 90%, exploration of HPC functioning in this model may have direct clinical applicability.11–13
The main loss of liver mass after high-dose APAP ingestion occurs because of hepatocyte necrosis in the centrilobular areas. Appearance of necrosis is preceded by hepatic microvascular injury and congestion.14, 15 In the massive injury characteristic of high-dose APAP intoxication, hepatocytes may be unable to accomplish full parenchymal reconstitution, and, thus, regeneration must include activation and hepatocyte differentiation of HPCs. In animal models of liver regeneration and hepatocarcinogenesis, these HPCs are referred to as oval cells (OVc), which constitute a heterogeneous cell population, characterized by an ovoid nucleus, small size, and scant basophilic cytoplasm.16 They express phenotypical markers of both the biliary epithelium (cytokeratins CK7, CK19), and hepatocyte lineages (α-fetoprotein, albumin).17–20 OVc cytoplasm is also strongly immunoreactive for rat oval cell marker OV-6, targeting an epitope shared by CK14 and 19.21 Monoclonal antibody A6, developed by Valentina Faktor and colleagues, targets an uncharacterized protein that is also widely considered a reliable marker of OVc.22, 23
In mice and in humans, HPC proliferations derive largely from the canals of Hering.23, 24 We first demonstrated the derivation of HPCs from the canals of Hering in humans in the setting of APAP-induced massive human liver necrosis.25 However, the time- and dose-dependent response of local HPCs to the massive APAP-induced liver injury has not yet been carefully documented. Moreover, methods of detection and reliable quantification of OVc in murine livers have not been well established.
Here we used the murine model of APAP-induced injury to study time course, dose–response, and distribution of the HPC reaction secondary to APAP-induced liver injury. The results show that in the earliest period after APAP challenge the HPC reaction is time- and dose-dependent, biphasic in sublethal doses, and prevails in the smallest, most proximal portions of the biliary tree. These data support the concept of the contribution of HPCs to hepatocellular regeneration.
All animal experiments were carried out after obtaining permission from the Division of Laboratory Animal Resources, New York University School of Medicine. Female C57BL/6J mice (Taconic, Germantown, NJ) each weighing approximately 20 g were provided free access to chow and water for at least 1 week before study. Before APAP administration, animals were fasted for 8 hours with free access to water. Fresh suspensions of APAP (Sigma Chemical Co., St. Louis, MO) were prepared in warm (40°C) phosphate-buffered saline (PBS), pH = 7.2, and given in a volume of 1 mL intraperitoneally at doses of 250, 500, 750, and 1,000 mg/kg body weight. Control littermate mice received vehicle (PBS at the same volume) only. Mice from each group were killed for organ retrieval at 3, 6, 9, 24, and 48 hours after acetaminophen administration.
In an additional experiment, mice received injections of APAP at doses of 750 and 1,000 mg/kg body weight, as well as injections of PBS and sham injections (without administration of any fluid), and then were sacrificed after 1 and 2 hours.
Tissue Preparation and Histological Grading of Hepatic Injury.
Liver tissues for permanent sections were fixed in 10% phosphate-buffered formalin for 4 hours, embedded in paraffin, sectioned at 4-μ–thick sections, and stained with hematoxylin-eosin. Hepatic injury was evaluated as a percentage of necrotic tissue in the 5 randomly selected areas at a magnification of ×40. Separate samples of liver were snap frozen in liquid nitrogen and stored at −20°C. Five-micrometer cryostat sections of these were prepared on charged slides and air dried for immunofluorescent staining.
Detection of OVc by Immunofluorescence and Immunoperoxidase Staining.
Evaluation of murine OVc reactions is hampered by the lack of commercially available antibodies against OVc differentiation markers. OV-6, a widely accepted human and rat OVc marker, is known to target a shared epitope of cytokeratins 14 and 19; yet anti–OV-6 antibodies (R&D Systems, Minneapolis, MN; catalogue no. MAB2020) failed to detect OVc in mouse tissues (data not shown). Another widely accepted marker of OVc found predominantly in the epithelia is recognized by rat monoclonal antibody A6.22, 23 The A6 epitope has been shown to be less detectable in formalin-fixed tissues26; therefore, we selected commercially available antibodies that target wide-spectrum cytokeratins, including murine biliary-type cytokeratins: rabbit anti-cow cytokeratin polyclonal antibodies (catalogue no. Z0622, DAKO, Carpinteria, CA). These polyclonal antibodies were raised against bovine epidermal keratin units of 58, 56, and 52 kd, among others, and cross-reacted with a wide range of cytokeratins (CKs), including the high–molecular weight CKs present in cholangiocytes (DakoCytomation, Carpinteria, CA; data sheet).
In the preliminary studies, to validate the use of these anti-CK antibodies, we immunofluorescently double-stained frozen liver tissue from normal and APAP-treated mice at various times after APAP injection. The frozen sections were air dried and incubated at 37°C for 2 hours with the rabbit polyclonal anti-CK antibodies and with rat monoclonal A6 antibodies. After a PBS wash, fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit antibodies (Molecular Probes, Eugene, OR; catalogue no. F2775), and Cy5-conjugated goat antirat antibodies (Abcam, Cambridge, MA; catalogue no. ab6565) were then layered on the tissue and incubated for another 2 hours at room temperature. The slides were then counterstained with 4′,6-diamidine-2-phenylindole and coverslipped with Vectashield mounting medium (Vector Laboratories, Burlingame, CA; catalogue no. H1200). Immunofluorescence was detected using the 4′,6-diamidine-2-phenylindole, FITC, and Cy5 filters of an Olympus (Melville, NY) fluorescent microscope and captured, pseudocolored, and combined using imaging software IPLab 3.9 (Scanalytics Inc., Fairfax, VA) as previously described.27 Immunoperoxidase with the same anti-CK antibodies for OVc light-microscopic detection on all experimental tissues was then performed. Formalin-fixed liver sections were deparaffinized, and endogenous peroxidase was quenched with 3% H2O2 for 15 minutes. Slides were then treated with Proteinase K 100 μL (DAKO) for 10 minutes at room temperature, thoroughly rinsed, blocked with serum-free protein in PBS (DAKO Catalized Signal Amplification System) for 20 minutes, and incubated with rabbit anti-CK antibodies previously described for 2 hours at 37°C. After washing with PBS, the secondary antibody (anti-rabbit biotinylated immunoglobulins) (DAKO) was applied. Antigen localization occurred after repeat washing with avidin-biotin horseradish peroxidase macromoleclar complex (ABC reagent, VECTASTAIN, Vector Laboratories) according to the manufacturer's instructions and counterstained with Mayer's hematoxylin (0.1%; Sigma Chemical Co.).
Evaluation of OVc Reaction.
In each fixed liver tissue sample, we examined up to 25 portal tracts (PTs) of the appropriate size (see below) with visible bile duct(s) (BD) and portal vein (PV) and counted OVc, defined as densely cytokeratin-positive cells with oval/cuboidal morphology and high nuclear to cytoplasmic ratio. Care was taken to identify those cells that did not border on lumina and were therefore not part of the preexisting normal BD. Intermediate hepatocyte-like cells, that is, cells with hepatocyte-like morphology but cytokeratin-positive staining, were rarely present in the regenerating mouse livers and were not included in quantifications.
In normal mouse liver, the PV occupies nearly the entire PT area, with scant stroma. Thus, to evaluate distribution of OVc according to PT size, we used the easily definable wall of the PV as a surrogate marker for PT size. Because the largest portal tracts in any section (one or two) exhibited extensive branching, cross-sectional measurements could not be assessed; thus, these largest units were also excluded from analysis.
The longest and shortest diameters of the sectioned PV were measured using an eyepiece with measuring grid at the magnification 40×. At this magnification, one scale unit of the measuring grid corresponds to 2.5 μm. The PV area was calculated according to the formula S (μm2) = 2.52πab/4, where a and b were the major and minor PV diameters expressed in the eyepiece scale units. Thus, we calculated not only the number of OVc per portal tract, but also the density (D) of OVc per portal tract, using the formula D = N/ab, where N was the number of OVc per portal tract, and a and b were the major and minor PV diameters expressed in the eyepiece scale units.
Results were expressed as means for each group of animals. Parametric (Pearson) correlation and the two-tailed P value were calculated. Results were considered statistically significant at P ≤ .05.
Histology of APAP Injury.
Progressive centrilobular vacuolization, congestion, necrosis, and inflammation could be observed in all APAP-treated animals. Administration of APAP at the highest dose (1,000 mg/kg) resulted in massive destruction of liver tissue (up to 90%-100%) and infiltration of blood cells into the space of Disse with the highest level of destruction at 9 and 24 hours. At 48 hours, despite the presence of significant residual hemorrhagic areas, well-defined areas of restoration could be identified in surviving animals. The lower doses of APAP (750, 500, and 250 mg/kg body weight) also caused significant damage (30%-70%), with the peak at the 9-hour time point. At every time and at every APAP dose administrated, viable hepatocytes were concentrated in the periportal areas around the CK cells, which appeared to form foci of regeneration of the liver tissue (Fig. 1). The 1,000-mg/kg APAP dose represented the median lethal dose in these experiments.
Validation of Anti-CK Antibodies for Detection of OVc.
A6 antibody stained bile ducts, ductules, and morphologically recognizable OVc, as previously described,22, 28 although authors also reported some A6-negative epithelial cells in terminal bile ductules. In the preliminary studies, portal tracts and surrounding tissue were examined by fluorescence microscopy. Our analysis included 22 portal tracts and periportal areas, identifying 746 associated OVc. The anti-CK antibody was positive in all but one A6-positive cell (which appeared morphologically intermediate between OVc and hepatocytes). Four CK-positive, A6-negative morphological OVc were also identified. Hepatocytes (1,022) in the captured images were negative with both antibodies. Using A6 as the standard, anti-CK antibodies detected OVc with a sensitivity and specificity of greater than 99%, validating their use in murine OVc detection (Fig. 2). Interestingly, in A6/CK-positive cells, the antigens detected had only partially overlapping subcellular localization.
Expansion of CK-Positive OVc After APAP-Induced Injury.
Activation of HPC compartment under the various pathological conditions is reflected by the expansion of OVc. Systemic administration of APAP at different doses led to accumulation of OVc that appeared singly or in irregular strings without lumens (Fig. 3). OVc sometimes appeared to emerge from the BD. OVc displayed a moderate degree of morphological homogeneity: all of the cells were strongly positive for cytokeratin, though the intensity of staining varied from very strong (more characteristic to cholangiocyte-like, “true”oval cells) to rather weak (in cells with slightly more hepatocyte-like morphology).
Quantification of OVc around the portal tracts with portal veins of cross-sectional areas less than 5,000 μm2 showed a statistically significant increase (P < .005) in OVc after APAP treatment compared with PBS- and sham-treated animals (Fig. 4). For the sham group (n = 85; the number of counted portal tracts) the average number of OVc per portal tract was 8.07 ± 5.02; for PBS-injected animals (n = 160), 7.60 ± 5.53; and for the APAP group (n = 940), 9.66 ± 6.1. Standard errors of the sample means were 0.54, 0.44, and 0.2 for the sham, PBS, and APAP groups, respectively. These data for APAP-injured livers, as well as those from sham and PBS-injected animals, are presented cumulatively, including all time points and, for APAP-injected mice, all doses.
OVc Reaction Is APAP Dose- and Time-Dependent.
Figure 5 summarizes changes in OVc numbers at different times depending on the APAP dose. OVc response did not correlate well with the development of the morphological signs of liver injury (i.e., necrosis, hemorrhagic infiltration). Whereas the intensity of liver damage was proportional to APAP dosage and appeared in a single wave extending outward from the central vein (Fig. 1), OVc dynamics were biphasic in response to the intermediate APAP (250-, 500-, 750-mg/kg) doses. Injection of APAP at the lowest concentration, 250 mg/kg resulted in a two-wave dynamic, with the highest points at 9 and 48 hours after APAP injection. APAP at both 500 mg/kg and 750 mg/kg resulted in OVc reaction characterized by the same detectable two-wave time course, but with peaks at 6 and 24 hours. APAP 750 mg/kg caused the highest increase in OVc, at 6 hours.
In a separate experiment, at 1 and 2 hours after intraperitoneal injection of APAP, we also detected an increase in the amount of OVc compared with sham- and PBS-injected animals. However, the difference between the groups was statistically insignificant (data not shown). Within the first 3 hours after PBS injection, the levels of OVc increased somewhat compared with sham-injected mice (data not shown).
In summary, in the first 2 days after APAP-induced liver injury, 2 peaks of OVc appearance characterized the local progenitor response at doses of 250, 500, and 750 mg/kg. The biphasic nature of these responses is confirmed by statistically significant differences between values of each peak and the trough value between them. The 1,000-mg/kg response did not share this statistically significant biphasic distribution. APAP concentration at 750 mg/kg elicited the highest OVc reaction. No statistically significant changes were found for the cells constituting bile ducts (data not shown).
Association of OVc With the Smallest Portal Tracts.
To address the question of whether OVc reaction was uniform around all PTs over their complete size range, we first calculated the correlation between the average OVc numbers around the PT and PT sizes in normal and injured livers (Fig. 6). The Pearson correlation coefficient (r) for these 2 parameters was statistically significant for all groups (r = 0.41 for sham group, r = 0.34 for PBS-injected, and r = 0.46 for APAP-treated animals). The correlation levels were approximately the same within all the ranges of PT sizes divided into 5 size ranges for ease of calculation: less than 1,000 μm2, between 1,000 and 2,000 μm2, between 2,000 and 3,000 μm2, between 3,000 and 4,000 μm2, more than 4,000 μm2 (Fig. 6A). To characterize more precisely the relationships between OVc generated around the PT and PT size, we next calculated the density of OVc around each PT (see Materials and Methods). In the experimental group (APAP 750 mg/kg), OVc density significantly (P < .05) exceeded that in the PBS group (Fig. 7A). Whereas in PBS-injected animals density remained basically the same over time, 6 hours after 750 mg/kg APAP injection, density rose almost 3 times (P < .02), yet quickly returned to the previous levels later (Fig. 7B). The scatter analysis of the density and portal vein size in APAP group showed a statistically significant negative correlation between these 2 parameters in all of the groups (r = −0.60 for the normal group, r = −0.37 for PBS-injected, and r = −0.31 for APAP-treated animals, see Fig. 8A). Although OVc were also present in significant amounts within the larger PTs, their density in large PTs was much lower, and it was much higher in small PTs. In other words, APAP treatment caused a greater increase in density in the smallest PTs (Fig. 8B). Interestingly, intraperitoneal injection of PBS itself was associated with the higher density compared with normal animals (P < .05).
OVc Reaction and Liver Regeneration.
APAP is metabolized by cytochrome P-450 into the toxic metabolite, N-acetyl-p-benzoquinone imine (NAPQI). NAPQI is normally conjugated in the liver with glutathione to yield a harmless mercaportal tracturic acid; however, overdose of APAP depletes hepatic glutathione, and NAPQI covalently binds to DNA and cysteine residues on numerous hepatic proteins,7 resulting in the formation of 3-(cysteine-S-yl) APAP adducts. The first pathological changes in the liver may be detected within minutes after APAP injection.29–31
Our findings confirm that APAP-induced liver toxicity in the mouse elicits an HPC response as it does in humans. The appearance of OVc could first be recognized within hours; this is different from the several days reported in murine cocaine injury.32 The difference may arise in part because of the nature of the injury: cocaine yields periportal necrosis. The disruption in stroma, nonparenchymal cells, and perhaps the canals of Hering and ductules themselves, is likely to require recovery before the organized OVc reaction can take place. In APAP injury, the lesion is central and, thus, the zone wherein OVc reactions take place is anatomically preserved. Our findings are in accordance with the literature's data pointing to the rapid biochemical and morphological changes in the liver in response to APAP; we detected statistically significant OVc reaction as early as 3 hours after injection of APAP at the highest dose. It remains unclear whether the APAP-induced OVc are entirely derived from intrabiliary HPCs or else some engraft from the circulation.33
Biphasic OVc Responses to APAP Injury.
The time course of the OVc response did not directly correlate with the histologically defined liver damage. This may be explained by the fact that excess NAPQI is probably cyclically reduced back to APAP and again to NAPQI, thus resulting in the progressively growing liver damage. Also, whereas OVc are cytochrome P-450 negative and would have a survival advantage, OVc-derived hepatocytes, which express P-450, would then become sensitive to residual APAP. This may, in part, explain the second wave of OVc proliferation in response to APAP.
The biphasic time course of OVc activation after APAP at 250 mg/kg precisely matches the previously reported biphasic, temporal fluctuations of serum stem cell factor (SCF) in mice after APAP liver injury.34 SCF is a transmembrane protein in all epithelia of the liver, enzymatically cleaved from the cell surface during injury.35 The receptor for SCF, c-kit, plays a fundamental role in stem cell functioning and, in the liver, is expressed by OVc, the canals of Hering, and the intralobular bile ducts.25, 33 SCF cleavage and release from damaged hepatocytes may lead to the first wave of OVc activation. The second wave may come with de novo production and secretion of new SCF. Importantly, exogenously administered SCF rescues mice from lethal APAP administration,34 suggesting the importance of this signaling pathway in HPC-mediated hepatic regeneration. The loss of this biphasic response in the median lethal dose (1,000 mg/kg) may relate to the significant lethality at this dose level.
Quantification of OVc: Absolute Number Versus Density per Portal Tract.
The methods used to quantify OVc, once detected, also have been historically problematic. It has previously been approached in 3 ways: (1) semiquantitative grading, (2) quantifying numbers of cells in randomly selected medium- or high-power fields, (3) quantification of OVc per portal tract. The first approach is imprecise and interferes with statistical analysis. Quantification by randomly selected microscopic fields is inappropriate for phenomena that are not themselves randomly distributed, such as OVc. The third approach rests on the assumption that OVc proliferation in response to injury is uniform around all portal tracts. Here we show that the overwhelming mass of increasing oval cells is found around the smallest portal tracts (Fig. 8). Following on this, it then makes sense to limit assessment of OVc expansion to analysis of those smallest portal tracts, excluding the larger ones.
The effects of such an analysis are clear in comparing the data presented in Figs. 4 and 6 with those presented in the leftmost portion of Fig. 7B. Trends with indistinct or low-level significance when assessed by number of OVc per portal tract, including portal tracts of every size, become much more statistically meaningful when confined to the smallest portal tracts, approximately 1,000 μm2. Thus, our method represents a rational and reproducible system for quantitative evaluation of OVc reactions in acute liver injury. The data also support our prior contention that the proximal biliary tree is a primary source of HPCs.24, 25
A6 is considered a differentiation marker of both epithelial and erythroid cell lineages in the developing mouse and mouse liver.22 The pattern of A6 expression in the epithelial cells of the mouse fetal liver precisely represents that of cytokeratin 8 in the developing bile duct of the rat liver.36 In human tissues, the most widely used markers of intermediate hepatobiliary cells, the human OVc equivalent, has been anti-CK antibodies targeting biliary-type cytokeratins (CK7 and CK19).37 This corresponds to the use of OV-6, which recognizes a shared epitope of CK14 and CK19. Thus, we chose to validate and use anticytokeratin antibodies for OVc detection, but focusing on antibodies that were both commercially available and easily used to decorate formalin-fixed, paraffin-embedded tissues for the extensive light microscopic evaluation and quantification required. The extensive colocalization of the employed anti-CK antibodies with A6 confirms that it is a valid tool for detection of murine OVc.
OVc Activation in Response to Intraperitoneal Fluid Injection.
Intraperitoneal PBS injection also caused a small, but statistically significant, increase in OVc density compared with sham-treated animals. The reasons for this observation remain unclear. A direct irritation of the intra-abdominal branches of the vagus nerve by intraperitoneal fluid or reciprocal activation of the vagal nerve system after the stress caused by the injection could lead to autonomic-based activation of the hepatic progenitor cell compartment via muscarinic acetylcholine receptor.38 Another possibility is that increased intraabdominal fluid pressure after injection affects lymphatic or blood flow into the liver that, in turn, may be recognized as a damage signal by some other mechanism and induce cell proliferation as a part of regenerative response. Because the anatomical relationships of the canals of Hering to lymphatic drainage or to the finest branches of the vessels supplying the liver are, as yet, unknown, these and other possibilities cannot be excluded. The absence of the biphasic OVc activation suggests that it is distinct and separate from the toxic impetus to additional proliferation.
The use of anti-CK antibodies for OVc detection in formalin-fixed mouse livers, validated here by its extremely high sensitivity and specificity compared with A6 antibody, provides a methodological link between the approaches customarily used by investigators relying on animal experimentation and those exploring HPC activation in human tissues. Using a mouse model of APAP hepatic injury, we show that in the first 2 days after APAP administration the OVc response is both time- and dose-dependent. OVc proliferate most around the smallest PTs, in keeping with their primary origin in the most proximal biliary tree, in particular the canals of Hering. Our analysis of OVc distribution suggests a rational, anatomy-based approach for precise OVc quantification in injured mouse livers, which can allow for statistically meaningful comparison of OVc response within and between models of acute injury.
The marked OVc activation during the height of hepatocyte injury suggests that they are resistant to the toxic effects of APAP metabolites, perhaps through minimal presence of cytochrome P450 metabolic pathway in their resting and early proliferative stages. The OVc activation is biphasic, mirroring previously reported biphasic SCF release, suggesting the importance of this chemokine for OVc activation. Also, perhaps, the second wave might reflect a gain of susceptibility to APAP-toxicity when they begin to differentiate down an hepatocyte lineage while the drug and its metabolites are still present, leading to a second (though lesser) wave of hepatocyte injury triggering a further OVc proliferation.
In summary, in view of its close fidelity with human APAP-related injury, we suggest that the murine model of APAP toxicity provides a most useful tool for investigation of hepatocellular regeneration, in particular from intra-organ and extra-organ stem cell compartments.
A6 antibody was supplied to us by Dr. V. Faktor (Laboratory of Experimental Carcinogenesis, NCI, NIH).