Oval cell proliferation in p16INK4a expressing mouse liver is triggered by chronic growth stimuli.

Terminal differentiation requires molecules also involved in aging such as the cell cycle inhibitor p16(INK4a). Like other organs, the adult liver represents a quiescent organ with terminal differentiated cells, hepatocytes and cholangiocytes. These cells retain the ability to proliferate in response to liver injury or reduction of liver mass. However, under conditions which prevent mitotic activation of hepatocytes, regeneration can occur instead from facultative hepatic stem cells.For therapeutic application a non-toxic activation of this stem cell compartment is required. We have established transgenic mice with conditional overexpression of the cell cycle inhibitor p16(INK4a) in hepatocytes and have provoked and examined oval cell activation in adult liver in response to a range of proliferative stimuli. We could show that the liver specific expression of p16(INK4a) leads to a faster differentiation of hepatocytes and an activation of oval cells already in postnatal mice without negative consequences on liver function.

Introduction

P16INK4a is an inhibitor of cell cycle progression that binds specifically to cyclin-dependent kinases (CDKs) 4 and 6, preventing the formation of the CDK/cyclin D1 complexes which normally phospho-rylate retinoblastoma protein (RB) to allow progression through the G1/S checkpoint [1]. There is no evidence of an essential role for p16INK4a in development, with expression being undetectable in mouse embryos [2] and development apparently normal in knock out mice [3, 4]. However, expression increases with age in both mice and humans [5], consistent with a role in terminal differentiation and senescence. Indeed, p16INK4a contributes to the impaired cellular regeneration of an aging organism and acts as a marker in this process [6].

Consistent with a decisive role in tissue homeosta-sis, alterations in expression of p16INK4a have been implicated in a variety of disease processes including cardiomyopathy [7], Morbus Alzheimer [8, 9] and a wide range of cancers [10, 11].Here, classic tumour suppressor function has been confirmed by the increased susceptibility of the knock-out mice to a range of cancers [3, 4]. Furthermore, the specific involvement of p16INK4a in liver cancer is indicated both by frequent deletion of the locus in hepatocarci-noma [12, 13], and by an association between p16INK4a expression level and hepatocarcinoma susceptibility in rats [14, 15].

Adult liver is a quiescent organ incorporating highly differentiated epithelial cells (hepatocytes and cholangiocytes) and mesenchymal cells (Kupffer cells, endothelial cells and hepatic stellate cells). The hepatocytes account for approximately 70% of liver cells and are arrested in the G0 phase of cell cycle [16], but retain the unusual ability to re-enter cycle in order to regenerate large amounts of tissue in response to liver injury [17]. Under conditions which prevent the mitotic activation of hepatocytes, regeneration can occur instead from a distinct population of hepatic stem cells (oval cells) which reside in the canal of hering, a well-defined zone between hepatocytes and cholangiocytes of the biliary epithelium [18, 19]. Oval cells are considered to be bi-potential epithelial stem cells capable of regenerating both hepatocytes and cholangiocytes (for review see [20]).

While the potential use of somatic stem cells for therapeutic applications has raised interest in mechanisms of liver regeneration in recent years, models of oval cell recruitment to date have all involved chemically induced injury which compromises therapeutic potential [21]. In order to generate a non-cytotoxic model of oval cell recruitment, we have established transgenic mice with conditional overexpression of the cell cycle inhibitor p16INK4a in hepatocytes, thereby permitting the investigation of p16INK4a overexpression in liver in all phases of development. Using these mice, we have examined oval cell activation in adult liver in response to a range of proliferative stimuli.

Materials and methods

Cloning of human p16INK4a cDNA and generation of transgenic mice

For cloning p16INK4a cDNA human fibroblast RNA was isolated, reverse transcribed using random pdN6-primers and Superscript II RT (Gibco, Eggenstein, Germany), and the obtained cDNA was amplified using the following specific primer pairs containing MluI and HindIII sites allowing sub-cloning into pBI vector (p16-Mlu-F: ctcacgcgtagcgggagcagcat ggagccggcg; p16-Hind-R: atcaagcttgctctggttctttcaatcggggat). Transgenic mice with inducible hepatocyte-specific expression of p16INK4a were generated using the heterologous tTA system [22] as illustrated in Figure 1. Briefly, individuals of transgenic line ptetp16INK4a (C57Bl/6-DBA background) carrying the bi-directional transcription unit for luciferase and p16INK4a were interbred with individuals of the transactivator line TALAP-2 (C57Bl/6-NMRI background) [23]. Double transgenic mice ptetp16INK4a/TALAP-2 were interbred with C57Bl/ 6-Tg(ACT-EGFP)1Osb/J mice expressing the green fluorescence protein [24] to obtain triple transgenic mice ptetp16INK4a/TALAP-2/EGFP for isolating labelled cells.

1

Conditional liver-specific expression of p16INK4a. (A) The transactivator (tTA), a fusion protein of an Escherichia coli derived tet repressor (tetR) DNA binding domain and the transacti-vation domain of VP16 protein derived from herpes simplex virus [62] is placed under the control of a C/EBPβ (LAP) promoter, which allows a hepatocyte-specific expression of the tTA protein. The tTA protein can specifically bind to the tet operator (tetO) sequence and subsequently induces the transcription from the adjacent cytomegalovirus (CMV) minimal promoter which is combined with a transgene (p16INK4a). Tetracycline or its derivative doxycycline (Dox) can prevent binding of tTA to tetO and the transactivation of any transgene cloned behind the CMV promoter is stopped. In contrast, removal of Dox allows the induction of transgene expression. Modified from Gossen et al.[63] (B) Western blot analysis of human p16INK4a in liver extracts of ptetp16INK4aTALAP-2 mice.Liver extracts of induced male double transgenic ptetp16INK4aTALAP-2 mice (minus Dox) were compared to control liver extracts (plus Dox). Three different p16INK4aantibodies were utilized as indicated and blots were normalized by scanning β-actin as loading control.(C) Immunohistochemical staining of mouse liver sections with anti-human p16INK4a antibody. In induced ptetp16INK4aTALAP-2 mice (left photomicrograph) hepatocytes express p16INK4a, detected by the brown colour (DAB staining), which is localized in both cytoplasm and nucleus. Non-parenchymal liver cells and endothelial cells do not express any transgenic protein. No p16INK4a protein was detected in repressed ptetp16INK4aTALAP-2 mice (right photomicrograph). Bar represents 20 μm.

Animal treatment

Animals were housed under a constant day-night cycle of 12:12 hours and fed a standard chow diet (Altromin 1324, Altromin Gesellschaft für Tierernährung, Lage, Germany), with access to water/doxycycline hydrochloride solution ad libitum under all conditions. Animal experiments were carried out in accordance with the European Council Directive of November 24, 1986 (86/609/EEC) and were approved by the local authorities.

Doxycycline hydrochlorid (Dox; Sigma, Deisenhofen, Germany) was dissolved to 50 μg/ml in water and given in brown bottles, which were exchanged twice a week, to prevent transcription. Expression of transgenic proteins was induced by substituting plain water for Dox.P16INK4a expressing mice and controls were treated with nodularin (25 μg/kg body weight, MP Biomedicals, Aurora, OH, USA) thrice weekly by i.p. injections over 4 to 16 weeks and killed 1 day (NDLN) or 4 weeks (NDLN+) after this treatment.

Starvation and re-feeding (interval feeding) was performed as previously described [25]. Briefly, animals were given no food over a period of 48 hrs, but water/Dox ad libitum and then re-feeded 24 hrs. This procedure was repeated ten times.

Partial hepatectomy (PH) was performed according to the method of Greene and Puder (2003) [26] with slight modifications under ketamine/xylazine anaesthesia. Briefly, food was withdrawn at 6 a.m. and surgery was done between 1 to 3 p.m. Total body weight and weight of removed liver lobes was recorded and the removed portion was calculated using the formula (removed liver pieces in g/0.045*body weight)*100%, in which 0.045 is the mean index of liver weight to body weight of 12-week-old male double transgenic mice of the ptetp16INK4a/TALAP-2 line.

Fifty to 60% of liver was resected and after closing the abdomen 5 ml sterile saline was given subcutaneously on the back. Animals were monitored post-operatively and methamizol-sodium (5 mg/mouse/day) was given in drinking water for analgesia.

Isolation of liver cells, flow cytometric analysis and cell culture

Hepatocytes and oval cells were isolated using an in vitro perfusion technique [27].

For preparation of hepatocytes liver was perfused with calcium-free buffered saline and subsequently with colla-genase (1 mg/ml, 240 U/mg, Biochrom AG, Berlin, Germany). Cell suspension was centrifuged thrice at 70 x g, 5 min. Hepatocytes were counted and 200,000 cells were re-suspended in 100 μl wash buffer (PBS, 5% FCS) and incubated with an rat anti-mouse CD45 antibody (FITC) (1:40, Serotec, Oxford, UK) 30 min at room temperature to identify contaminating cells. After washing 500 μl ice-cold ethanol (–20°C) were added slowly with continuous mixing avoiding agglutination. Hepatocytes were kept 10 min on ice, washed twice with 1 ml PBS/5% FCS and incubated 1 hr with RNAse A solution (1mg/ml in PBS) at 37°C.Suspension was filled up to 200 μl with PBS/5% FCS and 20 μl propidium iodide (1 mg/ml) were added. This hepatocyte suspension was measured on a flow cytometer (FACScan, Becton Dickinson, Mountain View, CA, USA) for determining the ploidy grade.

Oval cells were isolated by perfusing liver consecutively with calcium-free buffered saline, pronase (1 mg/ml) and collagenase (1 mg/ml) for 10 min each. Cell suspension was centrifuged twice at 70 x g disposing the hepatocytes and twice at 250 x g for washing and collecting non-parenchymal cells. Cell suspension was than loaded on a discontinuous Percoll gradient to separate oval cells from both other non-parenchymal cells and cellular debris.After density gradient centrifugation (1hr at 500 x g) cells banded at a density of 1.07 were harvested and washed three times in William's E medium. Cells were seeded in either 90-mm culture dishes or flaks and were cultured in William's E medium supplemented with 10% FCS, 2 mM glutamine as clones after trypsination in clonal rings or as whole cell preparation after differential trypsination.

Microarray

Murine Genome Arrays U74Av2 (Affymetrix®) were applied. Following Affymetrix recommendation first strand synthesis was performed using Superscript II RT (Gibco) and second strand synthesis using T4 polymerase. Phenol/chloroform extracted cDNA was used as template for synthesis of biotinylated anti-sense cRNA with the Enzo Bioarray High Yield RNA Transcript Labeling kit. Hybridization and washing were performed according Affymetrix recommendation with fluidics station (Affymetrix®) and scanned with a laser scanner (Agilent Tech®) at 570 nm.

RNA isolation and real time RT-PCR

Total RNA was isolated using the PeqGOLD RNA Pure isolation system (Peqlab, Erlangen, Germany). A cleanup of total RNA was performed using the Qiagen RNeasy Total RNA isolation kit (Qiagen, Hilden, Germany). Quality of RNA was assessed by electrophoresis in denaturing formaldehyde agarose gels and purity was estimated by ratio A260/A280 nm spectrophotometrically. Concentration was adjusted to 1 mg/ml and RNA was stored at –80°C until use. RT-PCR for real time quantification was performed using cyclin D1 specific primers as previously described [28].

Luciferase assay

Liver tissue (100 mg) was homogenized and incubated 5min in luciferase lysis buffer (500 μl, 77 mM K2HPO4, 23 μM KH2PO4, 0.2% Triton, 1 mM DTT). The homogenate was centrifuged 5 min in a microcentrifuge.Ten microlitres of the supernatant was applied for measurement of luciferase activity using Luciferase-Assay Reagenz (Promega, Heidelberg, Germany).

Western analysis

Liver tissue was homogenized in cold extraction buffer (10 mM Tris, 150 mM NaCl, 1% NP-40, 0.1% SDS, 1 mM EDTA, 1 x protease inhibitor cocktail (Roche, Mannheim, Germany)) at 4°C followed by sonication. Lysates were cleared by centrifugation and protein was quantified using standard Bradford assay. Samples were separated by SDS-PAGE and transferred to PVDF membranes (GE Healthcare, Freiburg, Germany).Membranes were washed in Tris-buffered saline (TBS)/Tween and blocked in 5% nonfat dry milk and subsequently incubated with the primary antibody listed in Table 1 together with an anti-actin antibody as loading control following the multiplex detection protocol of ECL Plex Western blotting system (GE Healthcare, Freiburg, Germany). Secondary CyDye labelled antibodies were diluted 1:2500 and detected after wash steps by scanning the membrane using a fluorescent laser scanner (Typhoon 9410 Imager, GE Healthcare). Western blots detecting the mouse p16 expression levels were performed using the Chemiluminescence detection system (Pierce, Bonn, Germany).

1

Antibodies used in western analysis (WB) and immunohistochemistry (IHC)

Antibody	Dilution	Supplier
Human-p16INK4a (C20)	1:2000 (WB), 1:300- to 1:1000 (IHC)	Santa Cruz Biotechnology (Heidelberg, Germany)
Human-p16INK4a (N-20)	1:2000 (WB), 1:300- to 1:1000 (IHC)	Santa Cruz Biotechnology
Mouse-p16INK4a (M-156)	1:2000 (WB), 1:300 (IHC)	Santa Cruz Biotechnology
Human- p16INK4a (clone G175-405)	1:5000 (WB)	Pharmingen (BD Biosciences Pharmingen, Erembodegem, Belgium)
p27Kip1	1:1000 (WB)	Neo Markers (Lab Vision Corp., Fremont, CA, USA)
CDK4	1:1000 (WB)	Santa Cruz Biotechnology
CDK6	1:1000 (WB)	Santa Cruz Biotechnology
Cyclin D1	1:1000 (WB)	Stressgen (Ann Arbor, MI, USA)
Cyclin E	1:1000 (WB)	Santa Cruz Biotechnology
PCNA	1:1000 (WB), 1:50 (IHC)	Santa Cruz Biotechnology
glutamine synthetase	1:1000 (IHC)	BD Transduction Lab
carbamoyl phosphatase	1:2000 (IHC)	W.Lamers (Amsterdam, Netherlands)
Cytochrome-P450 IIE1	1:1000 (IHC)	Stressgene
A6	1:100 (IHC)	N.Engelhardt (Moscow, Russia)
E-Cadherin	1:1000 (IHC)	BD Transduction Lab (Heidelberg, Germany)
β-actin	1:2500 (WB)	Abcam (Cambridge, UK)
β-actin	1:10,000 (WB)	Sigma (Schnelldorf, Germany)

Histology and immunohistochemistry

Liver samples were either quick-frozen in liquid nitrogen and stored at –80°C or fixed in 4% paraformaldehyde and routinely embedded in paraffin.Frozen liver samples were used for A6-immunohistochemistry. Liver tissue was cut in 5-μm -thick sections and mounted on Superfrost plus slides (Menzel, Braunschweig, Germany). After drying sections were fixed in ice-cold aceton and dried again. Sections were rehydrated in distilled water, rinsed with PBS and either quenched with H2O2 or directly blocked with 10% FCS/Avidin (Vector Laboratories, CA, USA) and than incubated with anti-A6 antibody produced by immunization with an oval-cell enriched liver fraction [29]. A6 antibody was detected with a biotinylated rabbit anti-rat antibody (DAKO, Hamburg, Germany).Sections were further processed with Extravidin-peroxidase conjugate (Sigma).For all other antibodies (Table 1) and for periodic-acid-Schiff (PAS) staining and haematoxylin–eosin (HE) staining 2-μm paraffin sections were dewaxed in Xylol and descending ethanol sequence and rehydrated in distilled water. Antigens were unmasked by microwave treatment and antigen-antibody complexes were detected by peroxidase conjugated secondary antibodies as previously described [25, 30].

Sections used for morphometric analysis of size distribution of nuclei were dewaxed and embedded with Prolong Gold anti-fade reagent containing DAPI (Invitrogen, Karlsruhe, Germany).For PAS staining, used for detection of glycogen in liver tissue, slides were oxidized in 0.5% periodic acid 5 min, rinsed in distilled water, placed in Schiff reagent for 15 min and counterstained after washing in lukewarm tap water in haematoxylin.

Morphometrical analysis

Liver slides were dewaxed, washed in distilled water and equilibrated in TBS. Nuclei of cells were stained with 4′-6-Diamidino-2-phenylindole (DAPI) using Prolong anti-fade reagent containing DAPI (Invitrogen GmbH, Karlsruhe, Germany). Twenty randomly selected visual fields per slide were captured automatically with sensitive greyscale camera (Hamamatsu, Orca 285) using an inverted microscope (Nikon, TE2000) with 20X objective (CFI Planfluor ELWD 20X/0.45). Focus and brightness parameter were calculated by software reducing human influence on image acquisition [31].

The number of cell nuclei was identified using a histogram based mixture model threshold algorithm. Hepatocytes were described on their size between 8 μm x 8 μm and 25 μm x 25 μm and circular shape with index of convexity >0.9 (index of convexity: IC(a) = Area(a)/ConvexArea(a)) and roundness <1.5 (roundness: R(a) = long axis length(a)/short axis length(a)) and quantified in 20 visual fields of 293,260 μm2 (width: 620 μm, height:473 μm).

For determining the mitotic index mitotic figures were counted in the right middle lobe and omental lobe of each section of four mice per time point. Area of the lobes was quantified using Axiovision Release 4.5 software after scanning the sections with an Axiovert 200M microscope (Zeiss, Oberkochen, Germany).

Senescence-associated β-galactosidase (SA-β-Gal) staining

Frozen liver samples were cut and mounted on Superfrost plus slides (Menzel). Slides were dried 1 hr at room temperature and fixed with 0.5% glutaraldehyde in PBS 5 min at room temperature. After washing in PBS slides were incubated in citrate buffer (40 mM Na2HPO4, 40 mM citrate pH 6,0) stained overnight at 37°C with SA-β-GAL staining solution (1 mg/ml 5-bromo-4-chloro-3-indolyl-β-D-galactoside (X-Gal) in 40 mM citric acid, sodium phosphate, pH 6,0, 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 150 mM NaCl and 2 mM MgCl2).Slides were washed in PBS and counterstained with Mayer's haematoxylin.

Results

Inducible expression of transgenic p16INK4a in livers of adult mice

Human p16INK4a cDNA was cloned into the tetracycline-inducible, bi-directional vector pBI-L (Clontech) and used to generate transgenic mice (service of transgenic mouse facility, F. Zimmermann, ZMBH Heidelberg), leading to the production of three lines which stably transmitted the transgene ptetp16INK4a. P16INK4a expression was subsequently targeted to the liver by interbreeding each ptetp16INK4a line with the liver-specific TALAP-2 mouse [23] (Fig. 1A). Transgenic p16INK4a expression was prevented during early development by the administration of Dox during mating and throughout pregnancy. Since the pBI-L vector links the expression of p16INK4a to that of luciferase, luciferase activity was used to monitor the level and inducibility of transgene expression in the three lines, revealing that only one of them (U59) was stably inducible on Dox withdrawal (10,000 r.l.u./μg protein compared to 0.1 to 1.0 r.l.u. per μg protein in Dox-treated controls).

On western blots of liver protein, an antibody directed against the C-terminus of human p16INK4a (clone C-20) detected the expected species of 16kDa in livers of induced transgenic mice but not of Dox-treated mice (Fig. 1B). Two further antibodies directed against the amino terminus (clone N-20) or full length protein (clone G175-405) confirmed these observations but also revealed an additional 14 kD species, which therefore appears to be a proteolytic product of p16INK4a lacking the C-terminus.Interestingly, levels of endogenous p16INK4a slightly decrease upon induction of the transgene, indicative of compensatory regulation.

Immunocytochemical studies revealed transgenic human p16INK4a protein to be localized to both the cytoplasm and the nuclei of expressing cells (Fig.1C) similar to results described for other cell cycle proteins (for review see [32]). The expression of trans-genic p16INK4a did not result in any gross phenotypical changes either in liver or other organs. However, distinct changes were found at the histological level. The hepatocyte nuclei of p16INK4a expressing male mice were both fewer in number and larger in size than those of the controls (Fig. 2), while p16INK4a expressing female mice accumulated fat in hepato-cytes and rarely possessed hypertrophic hepato-cytes. All subsequent experiments were performed using male mice only. Flow cytometric analysis of propidium iodide stained hepatocytes [33] subsequently confirmed that p16INK4a expressing male mice had an increased proportion of hepatocytes with ploidy > 4N (Fig. 3A and B). Interestingly, the hypertrophic hepatocytes did not themselves express detectable levels of transgenic p16INK4a (Fig. 4A1). However, they clearly did demonstrate metabolic activity appropriate to their environment, producing enzymes of nitrogen metabolism (including glutamine synthetase, Fig. 4A2) if localized in the pericentral region of liver lobulus, and carbamoyl phosphate synthetase (Fig. 4A3) if located periportally or midzonally. The hypertrophic hepatocytes also accumulated glycogen in a manner comparable to normal hepatocytes (Fig. 4A4) andexpressed both the pericentral-specific detoxification enzyme P450 IIE1 (Fig. 4A5) and E-cadherin (Fig.4A6) which is an additional marker of periportal hepatocytes [34].

2

Size distribution of hepatocyte nuclei in liver slides of ptetp16INK4aTALAP-2 mice and control mice with and without nodularin treatment. Number of hepatocyte nuclei (DAPI stained) identified on their circular shape were quantified. The absolute number of nuclei in this area was 121 ± 30 for controls (with Dox, untreated) and 105±40 for p16INK4a expressing mice (without Dox, untreated) and 74±39 for controls and 62±40 for p16INK4a expressing mice after nodularin treatment (NDLN), respectively. All differences were statistically significant (P < 0.05, Student's t-test). For graphical illustration the total number of cells counted in each area was set 100% and the proportion of cells with nuclei of sizes indicated is shown. Liver slides from six control mice and six p16INK4a-expressing mice sacrificed at the age of 8 weeks from different litters were quantified. Livers from four control and six p16INK4a expressing mice injected 12 weeks i.p. with nodularin and sacrificed 4 weeks after treatment were used for the quantification.

3

Flow cytometric analysis of propidium iodide stained hepatocytes. (A) The proportions of cells with diploid (2N), tetraploid (4N), octaploid (8N) and hexade-caploid and higher (16N and >) DNA content were analysed as percentage of cells from 100% gated. Significant differences P < 0.02 between p16INK4a expressing livers and controls were labelled with an asterisk.Representative histograms of FACS analysis of a unique p16INK4a expressing liver (B) and control (C) are depicted.

4

Expression of cell cycle proteins in liver extracts of ptetp16INK4aTALAP-2 mice and histological features of livers of p16INK4a expressing mice. (A) Photomicrographs of liver slides. P16INK4a protein was detected in hepatocytes of induced ptetp16INK4aTALAP-2 mice with an anti-human p16INK4a antibody (C-20) (A1, brown colour). In livers of male p16INK4a expressing mice many of enlarged hepatocytes were observed (A1-A6, black arrows). These hepatocytes do not express p16INK4a (A1) but express liver region-specific proteins like glutamine synthetase (A2) and cytochrome-P450 IIE (A5) representing pericentrally specific enzymes and carbamoyl phosphate synthase (A3), and E-cadherin (A6) being marker proteins for periportal and midzonal hepatocytes and also store glycogen (A4, pink colour).Proteins were detected with primary antibodies and immunocomplexes were visualized by DAB staining (brown). Bars represent 50 μm; CV, central vein; PV, portal vein (B, C) Western blots with antibodies against cell cycle proteins in liver extracts of induced (minus Dox) respectively switched off (+Dox) ptetp16INK4aTALAP-2 mice.

The induction of p16NK4a expression resulted in associated changes in the expression of a number of other genes involved in cell cycle control. Specifically, the levels of p27Kip1, cyclin E and CDK6 protein were decreased, while those of cyclin D1 and CDK4 proteins were increased as determined by western blot analysis (Fig. 4B and C). Differences between p16INK4a expressing mice and controls on transcriptional level, measured by microarray analysis were only marginal (data available at http://www.ebi.ac.uk/arrayexpress/experiments/E-MEXP-1264). Despite the only slight increase of cyclin D1 expression in microarray analysis it was also determined by quantitative RT-PCR (Fig. 7A) to be significantly increased at the level of mRNA, whereas CDK4 mRNA levels quantified also with RT-PCR were equal to controls (data not shown).

7

Liver-specific stimulation of proliferation by nodularin.Mice were treated with nodularin thrice weekly and killed next day after the last injection (NDLN) or 4 weeks after the last injection (NDLN+). (A) Cyclin D1 mRNA levels were determined by real time RT-PCR. Differences between untreated p16INK4a expressing and control mice, between controls and NDLN-treated controls (NDLN), and between nodularin treated controls that were killed 4 weeks after NDLN treatment were statistically significant (P < 0.05, Student's t-test). The number of different animals tested was: p16(untreated) = 7, control(untreated) = 5, p16(NDLN) = 8, control(NDLN) = 12, p16(NDLN+) = 21, control (NDLN+) = 12.(B) in situ immunodetection of PCNA (B1, B2), A6 antigen (B3, B4) and E-Cadherin (B5, B6) in nodularin treated mice.80-90% of hepatocytes nuclei of control mice were immunostained with the PCNA antibody (brown colour, B2). In p16INK4a expressing mice predominantly small cells with oval nuclei were positively stained with the PCNA antibody (white arrow) and some of the hypertrophic hepatocytes (B1, black arrows), but less intensive than hepatocytes in control mice (B2, black arrows). The anti-A6 antibody identifies these cells as facultative liver stem cells, oval cells, forming ductular structures (B3, white polygon). These bipotent facultative stem cells of liver express the A6 antigen like cells of the biliary system (cholangiocytes, B3, B4, black polygon). A common feature of epithelial cells, hepatocytes, cholangiocytes and oval cells, in liver is the expression of E-cadherin.On the histological level E-cadherin is detected in periportal hepatocytes (B6, black arrowhead) biliary ductuli (B5, B6, black polygon) and oval cells (B5, white polygons), whereas the latter form the same ductular structures detected with the A6 antibody (B3).(C) Nodularin treated p16INK4a expressing mice (C1, C2) develop p16INK4a negative loci (C2) consisting of accumulating hypertrophic hepa-tocytes. Some of the hypertrophic hepatocytes were PCNA positive (C1, brown). The black lines mark the two foci in the PCNA stained liver slide (C1). (D) Oval cells were isolated from nodularin treated p16INK4a expressing mice by two-step collagenase perfusion and subsequently density gradient centrifugation. A6-positive cells (red colour) of passage 0 (D1) growing clonally were trypsinated and cultured to passage 10 (D2) without loss of A6 antigen. In higher passages (D3, passage 16) a loss or down-regulation of A6 expression was observed.Representative photomicrographs of OVUE265 are displayed. Bar represents 50 μm.

P16INK4a expression during early mouse development

The adult mice described above were bred under Dox to avoid the potentially disruptive influence of p16INK4a expression on the rapid developmental expansion and differentiation of liver which starts around embryonic day 10 with expression of C/EBPβ (LAP) [35]. To test whether this is in fact an issue, ptetp16INK4a and TALAP-2 mice were interbred in the absence of Dox. Luciferase activity was detectable by embryonic day 12, being the earliest time point tested.Pups were born apparently normally, and the expression levels of both luciferase and p16INK4a in neonates (Fig. 5A, B, and C) were comparable to those found in adult mice following discontinuation of Dox. The livers of adult mice bred and maintained in the absence of Dox displayed the same histological features as those of mice withdrawn from Dox either at birth or after weaning, demonstrating that the embryonic overexpression of p16INK4a has no marked consequences for liver development. However, the livers of neonatal and weaning mice (14 days) expressing p16INK4a were characterised by the overexpression around portal vessels of E-cadherin (Fig. 5D1, D3), which might correspond to progenitor cells [36].

5

Induction of gene expression in early liver development. Inter-breedings of ptetp16INK4a mice with the liver specific TALAP-2 mice were performed without Dox or with Dox as indicated.(A) Luciferase activity was measured in liver extracts of foetal mice (e-12d, 12 days postinsemination), neonatal mice (neo) and postnatal mice (post14d, post13d, post17d and post18d). The solid line represents the mean values of luciferase measured in adult mice. (B) P16INK4a expression was checked by western blotting in neonatal and postnatal mice with the anti-human p16 antibody (C-20). Anti- -actin antibody was used as loading control. (C) Paraffin sections of livers of neonatal (C1, C2) and 14-day- postnatal mice (C3, C4) were incubated with anti-p16 antibody (C-20) and immunocomplexes were visualized with an peroxidase conjugated secondary antibody and DAB staining (brown). Expression levels of the four representatively depicted p16INK4a expressing mice vary individually, despite comparatively high luciferase activities. (D) Anti-E-cad-herin immunoreactivity on paraffin sections of livers of neonatal (D1, D2) and postnatal (14d) mice. P16INK4a expressing mice (D1, D3) were compared to controls (D2, D4). Pronounced expression of E-cadherin around portal vessels were detected in p16INK4a expressing mice (D1, D3, white polygon), but not observed in liver slides of control mice (D2, D4). E-cadherin was expressed in periportal hepatocytes (arrow heads) and cholangiocytes (black polygon) like in adult mice. Bar represents 50 μm.

Actions of proliferative stimuli on p16INK4a expressing adult livers

Partial hepatectomy

(PH) firstly described by Higgins and Anderson [37] is the strongest known stimulus for liver regeneration, with a two-third PH recruiting almost all hepatocytes into cycle within 48 hrs [17], so that an average of less than two doublings is sufficient to restore hepatocyte numbers. As expected, a 2/3 hepatectomy of control mice resulted in rapid activation of hepatocytes and the regeneration of 100% over 14 days. Interestingly, the mitotic index (MI) of hepatocytes in mice expressing p16INK4a was slightly higher than that in the controls from 48 hrs after PH, equal in both groups 72 hrs after PH, but significantly increased in livers of induced mice 14 days after PH (p16INK4a expressing mice display a MI of 9.9 per 50 mm2 and controls show 1 per 50 mm2). The mitotic figures, few of them visible as triphasic spindles, were restricted to cells which showed no detectable p16INK4a staining (Fig. 6). Bearing in mind the reduced regenerative capacity in these animals, it seems likely that these cells undergo either a retarded mitosis or most probably a prolonged proliferative response compensating for the low frequency of recruitable (non-expressing) hepatocytes similar as shown in the telomerase knock-out model [38]. This proliferation of p16INK4a non-expressing hepatocytes was obviously sufficient to restore liver mass after the unique proliferative stimulus by PH. Livers of induced mice werechecked up on oval cell proliferation at several time points (48 hrs, 3, 4, 6, 7, 12 and 14 days after PH) but never a recruitment of facultative stem cell compartment was observed (data not shown).

6

Mitotic figures inmouse liver sections after partial hepatectomy. (A) Sections of mouse livers of induced ptetp16INK4aTALAP-2 mice (A, C) and control mice (B, D) were stained with the anti-p16INK4a antibody (C-20, yellowish-brown colour) 72 hrs (A, B) and 14 days (C, D) after PH. Mitotic figures were identified by chromosomes visible as tangled, dark-staining threads (arrows).Number of mitotic figures in p16INK4a expressing and non-expressing livers was similar 72 hrs after PH, but differed significantly 14 days after PH (with control mice having no but p16INK4a expressing mice displaying some mitotic figures 14 days after PH (C)). Only in p16INK4a expressing mice abnormal mitotic figures like triphasic spindles were observed. Bar represents 50 μm.

Nodularin treatment

Nodularin, a natural heptapeptide produced by Cyanobacteria, is an inhibitor of protein phosphatases 1/2A (PP1/2A) and has strong liver specificity resulting from the route of uptake via the multi-specific bile acid transport system which occurs only in hepatocytes [39-41]. Since PP1/2A activity is required for the de-phosphorylation of RB at the end of mitosis [42], inhibition of PP1/2A should prolong the proliferative response in all cells capable of phospho-rylating RB. Indeed, the promoting effect of nodularin on hepatocyte proliferation has previously been shown in rat liver by proliferating cell nuclear antigen (PCNA) staining [43]. In our inducible transgenic model, those hepatocytes which express p16INK4a should not phosphorylate RB, and should therefore be indifferent to PP1/2A inhibition. Here, the proliferation of non-p16INK4a expressing cells, including the liver-specific stem cells would be expected to be stimulated preferentially by nodularin treatment. Indeed after treatment livers of induced mice show large loci of p16INK4a negative hypertrophic hepato-cytes (Fig.7C2) never seen in untreated livers and a lot of small oval cells which were also p16INK4a negative in immunohistochemical stainings of transgenic p16INKa.Nodularin increases the proportion of larger nuclei in control mice, but slightly shifts the ratios in those expressing p16INK4a (Fig.2) resulting in a total number of hepatocytes per area of 74±39 for controls and 62±40 in p16INK4a expressing mice. The differences between all for groups, control-untreated, p16INK4a-untreated and the two nodularin treated groups were statistical significant (P < 0.05).

The ability of nodularin to promote proliferation in the livers of control and induced transgenic mice was assessed by quantitative analysis of cyclin D1 mRNA levels and by PCNA staining of liver sections. As expected, nodularin increased hepatocyte proliferation in control mice, leading to increases in the levels of both cyclin D1 mRNA (Fig. 7A) and PCNA staining of hepatocytes (Fig. 7B2). In the induced mice, the comparatively high baseline level of cyclin D1 mRNA was not increased further by nodularin treatment. The PCNA staining pattern in induced mice was clearly different to that of the controls, with weak staining of occasional hypertrophic hepatocytes (Fig. 7B1) and loci of hypertroph hepatocytes (Fig.7C1) and predominant staining of small, oval shaped cells which we also found to express the oval cell marker A6 (Fig. 7B1 and B3), suggesting that nodularin treatment of thetransgenic mice does indeed recruit a population of hepatic stem cells to proliferation. Further immunocytochemical analysis revealed the population of activated oval cells to over-express the epithelial cell marker E-cadherin (Fig.7B5) that represents another surface marker of oval cells [36]. Oval cells were isolated from p16INK4a expressing mice by collagenase perfusion followed by Percoll gradient centrifugation. The developing cell clones, which were A6 positive to 90% (Fig. 7D1), immediately following isolation, were removed from the culture flask by trypsin- treatment and expanded. These cells remained the high level expression of A6-antigen over approximately 10 passages (Fig. 7D2). Further culture resulted in a progressive reduction in reactivity (Fig. 7D3). Four lines (OVUE265, OVUE867, OVUE869 and OVUE871) individually isolated from independent preparations were expanded and cryopreserved. The same experimental procedure was used in parallel for control mice but did not result in expandable cells.

Three of the four lines (OVUE867, 869 and 871) were derived from triple transgenic mice carrying an EGFP marker appropriate for subsequent transplantation experiments. None of the lines express either p16INK4a protein or luciferase activity, consistent with the complete silencing of C/EBPβ (LAP) in oval cells. These lines should therefore provide an excellent tool for monitoring activation of the C/EBPβ (LAP) promoter via reporter enzyme luciferase induction during differentiation of oval cells into mature hepatocytes under in vitro and in vivo conditions.

Interval feeding

The proliferation of hepatocytes is strongly responsive to feeding rhythms [44] and a repeated interval-feeding (2d-starvation, 1d-refeeding, 10-cycles) leads to synchronized cell cycling and increased hepatocyte renewal to replace cells lost through frequent apoptotic events during the starvation phases [25, 45]. It was therefore of interest to determine whether interval feeding resulted in the recruitment of hepatic stem cells to proliferation. No activation of oval cells was detected either in control (uninduced) mice or in young (less than 8 weeks) induced mice subject to interval feeding. However, oval cell activation was detected by both A6 and E-cadherin immunohistochemistry (data not shown) in 100% of interval-fed induced mice over 8 months of age. This frequency was reduced to 66% when Dox was re-administered during the period of interval feeding. The recruitment of oval cells to proliferation is therefore a function of both p16INK4a expression and age, associated most probably with cumulative proliferative demand.

Expression of senescence-associated β-galactosidase

Physiological p16INK4a overexpression is associated with senescence [46] and is frequently accompanied by expression of senescence-associated β-galactosidase (SA-β-Gal) both in cultured cells [47, 48] and in liver in situ[49]. The age-related effects of p16INK4a expression revealed by interval feeding raise the question of whether p16INK4a is merely a marker of senescence or may actively promote senescence processes. To test this, we determined the expression level of SA-β-Gal in situ. We found equal levels of SA-β-Gal in the livers of p16INK4a expressing and control mice, either in the quiescent state (without any treatment) or following PH. In contrast, strong up-regulation of SA-β-Gal was detected in p16INK4a expressing livers both in nodularin-treated and interval-starved mice, both treatments being associated with recruitment of oval cells. Indeed, the most intensive staining for SA-β-Gal was detected in hepatocytes of livers with the highest numbers of oval cells (Fig.8A).

8

Enzyme histochemical detection of senescence-associated β-galactosidase. Eight weeks old p16INK4a expressing mice (A) and control mice (B) were treated with nodularin for a period of 12 weeks and killed 5 days after treatment. SA-β-Gal (light blue colour) was expressed mainly in hepatocytes. Bar represents 50μm.

Discussion

We report the derivation and characterisation of transgenic mice in which the physiological cell cycle inhibitor p16INK4a is expressed in hepatocytes in an inducible fashion. The aim of this exercise was to increase the proliferative demand on hepatic stem cells by blocking the proliferation of hepatocytes which occurs as a first-line response to liver damage.

The induction of p16INK4a expression in adult liver hepatocytes had no obvious pathological conse-quences.However, at the histological level there was an increase in polyploidy and the appearance of hypertrophic hepatocytes, being indicative of increased maturation and compromised cell division [50] respectively. Interestingly, the hypertrophic hepa-tocytes characteristic of p16INK4a expressing livers did not themselves express detectable levels of p16INK4a protein. This suggests that hypertrophy occurs through a pathway dependent on p16INK4a, but that expression is inactivated prior to or during hypertrophic growth, possibly reflecting a selection for non-expressing cells. The hyperploidy in these hepatocytes is a sign of premature senescence and points to the former expression of p16INK4a. A second rather unlikely hypothesis might suppose a generation of p16-negative hepatocytes from oval cells in a direct way avoiding any expression of transgenic p16INK4a during their lifetime. This strategy would exclude the C/EBPβ (LAP) promoter activation, which (i) is a prerequisite for expression of transgenic p16INK4a, but (ii) is also an essential event during normal hepatocytes differentiation process. But in our study the hypertroph/hyperploid p16-negative hepa-tocytes display normal hepatocytic markers, indicating a normal hepatocytic development. Against the background of compromised hepatocyte proliferation in adult, we found that hepatic stem (oval) cells could indeed be recruited by two independent stimuli: nodularin treatment (which should extend the proliferative response of replication competent cells that do not express p16INK4a);and interval feeding (which both increases and synchronizes proliferative demand due to hepatocyte apoptosis during the starvation phases). We found the recruitment of oval cells by an interval-feeding regime to be dependent not only on the expression of p16INK4a in hepatocytes but also on advancing age of the mouse. This suggests that the mechanisms of regeneration may differ markedly between young and old mice and/or that oval cells tend to respond to cumulative proliferative demand over extended periods. We favour the latter interpretation firstly since the response to nodularin confirms the presence and responsivity of oval cells in young mice, and secondly because of our surprising finding that neither the acute proliferative response of adult hepatocytes to PH, nor the rapid growth of liver during foetal development were strongly compromised by p16INK4a expression. There may therefore be fundamental differences in the mechanisms of regeneration in response to chronic/extensive and acute/intensive damage, perhaps reflecting the different requirements for the replacement of isolated cells on the one hand and of entire functional units on the other.

We show herein that accumulation of transgenic p16INK4a restricts the regenerative capacity of hepa-tocytes and enforces the activation of facultative stem cell compartment of liver, being p16INK4a negative, when repeated proliferation is required like in the nodularin- and interval-feeding model.This verifies the limited role of p16INK4a in normal development as shown in p16INK4a null mice [3, 4] and its pivotal role in critical situations influencing the tissue homeostasis. Otherwise it was shown that growth-constrained environment can select for outgrowth and malignant transformation of hepatocytes [51, 52] and p16INK4a overexpression was described in invasion front of carcinomas [53]. This could result in selection of cells with malignant potential. However, our study show that neither the p16INK4a negative hepatocytes nor oval cells that developed in liver after proliferative stimuli possess an increased proliferative potential after removing the stimulus. Both livers of nodularin treated and untreated transgenic mice weighed slightly lesser than control livers and tumours were never observed actually after expression of p16INK4a over several month up to one year.

It is remarkable, that during liver development overexpression of p16INK4a did not cause dramatic effects. This would hardly be explainable on the basis of the classical understanding of cell cycle where such increased p16INK4a expression should completely arrest hepatocytic proliferation. However, recently, results obtained from CDK4 and CDK6 null mouse embryos [54] show normal organogenesis and cell proliferation levels, indicating, that the function of CDK4 and CDK6 can be accomplished at least temporarily (or locally in definite compartments) by other CDKs, i.e. CDK2, supporting a hypothesis that lacking of specific CDKs, cyclins or inhibitors can be compensated for by alternative mechanisms during various phases of development [55-58]. This also could explain the rather normal development of livers in our transgenic mice, with only limited involvement of alternative strategies.

Nevertheless, the fact that the overexpression of p16INK4a during development results in increased polyploidy and E-cadherin overexpression even in neonates suggests that oval cells are recruited in these mice from an early stage. The lack of pathology in these mice even at advanced age therefore supports the robust, normal function of oval cell derived hepatocytes in the long term and their potential relevance for regenerative therapy.

With a view to defining strategies for the cell therapy of chronic liver damage in humans it will therefore be important to define the mechanisms underlying the recruitment, proliferation and differentiation of oval cells. In this respect, our results provide an important proof of principle that oval cells can indeed be recruited by a combination of CDK inhibition (in our case via the expression of p16INK4a) and a non-cytotoxic interval-feeding regime. The silencing of C/EBP promoter in oval cells in our mouse model confirms the strong hepatocyte specificity of the used promoter [23, 30] and data obtained during differentiation of a transformed liver progenitor cell line of foetal mouse liver [59], but are in contrast to in situ data from rat experiments that show the expression of C/EBPβ in a subpopulation of oval cells [60]. The reason of the inconsistent data might be differences in expression pattern of oval cells in rat and mouse, recently strikingly shown by Jelnes and coworkers [61].

The oval cell lines isolated in this study, all of which carry both luciferase and p16INK4a under the control of the hepatocyte-specific C/EBP promoter (which is obviously inactive in the progenitor oval cell stage) and three of which are tagged additionally with GFP, will provide a valuable tool to analyse the regenerative potential of the recruited population both in vitro and in vivo.