Global Increase of p16^INK4a in APC-Deficient Mouse Liver Drives Clonal Growth of p16^INK4a-Negative Tumors Free

Genesis of hepatocellular carcinoma is not fully understood, though several carcinogenic pathways involved in this process were identified (1) among them the Wnt/β-catenin pathway. Activated β-catenin (CTNNB1) signaling contributes to approximately 30% of hepatocellular carcinomas (2, 3) and is characterized by nuclear and/or cytoplasmic staining of β-catenin (4) which contrasts to cell membranous staining of unaffected liver. Consequently, Wnt/β-catenin target genes, i.e., glutamine synthetase (Glul; ref. 5), are upregulated in β-catenin–positive hepatocellular carcinomas. Induction of β-catenin signaling in hepatocellular carcinomas is caused by gain-of-function mutations in β-catenin (6) or mutations in genes coding for components of the β-catenin destruction complex, i.e., AXIN1 (7) or AXIN2 (8). Though mutations of adenomatous polyposis coli (APC) gene, a further constituent of β-catenin destruction complex, are rarely detected in primary liver cancer (9), methylation of APC promoter seems significant in hepatocellular carcinoma suggesting functional importance of altered APC levels (10). In contrast, other malignancies of gastrointestinal cancer, i.e., colon and rectal cancers, are strongly predisposed to APC germline mutations, as found in familial adenomatous polyposis (FAP; ref. 11). Recently, context-specific responsiveness for Wnt/β-catenin signaling has been suggested necessary to develop cancer following APC reduction (12). Buchert reported development of hepatocellular carcinomas in livers of aged homozygous Apc^580S mice expressing reduced APC levels (13). Unexpectedly, hepatocellular carcinomas of this model display a downregulation of the Wnt target gene Axin2 (14), suggesting its reduction in tumors might have promoted hepatocellular carcinoma formation at the long latencies observed in these mice.

However, the tumors lacked hypermethylation of Axin2 promoter (12). Moreover, additional events causing tumorigenesis in this model were still unclear as mutations in oncogenes, i.e., Hras, were not observed. Therefore, the mechanism of how APC reduction promotes hepatocellular carcinoma development remains ambiguous and important to elucidate.

Here, we investigate potential mechanisms supporting development of hepatocellular carcinomas in livers with reduced APC levels. To this end, we utilize Apc^580S mice (13) exhibiting elevated β-catenin levels and investigate deregulation of Wnt signaling during all stages of hepatocellular carcinoma development. By comparing expression pattern in isolated hepatocytes and liver tissue of transgenic mice with different levels of β-catenin, which were generated with the help of the tet-inducible expression system, we inferred a concerted action of a senescence-inducing program in hepatocytes with increased β-catenin levels and a positive selection for hepatocytes with loss of cell-cycle inhibitor p16^INK4a (CDKN2A; ref. 15) as driver for the development of liver tumors in Apc^580S mice.

An overview of transgenic mice is presented in Supplementary Table S1. Apc^580S mice (13), termed Apc^homo, carry a homozygous floxed exon 14 Apc allele. Ctnnb1^flox/flox (16) mice, termed Ctnnb1^homo, were purchased from Jackson Laboratories. Three knockout (KO) mice, Apc^KO, Ctnnb1^KO, and Apc^KO/Ctnnb1^KO were obtained by interbreeding of floxed mice with liver-specific inducible Cre mice (17) carrying inducible P_tetCre-recombinase (LC-1) and tetracycline controlled transactivator (TA^LAP2) transgenes (18). Homozygous P_tetCre-Apc^homo mice were bred with homozygous TA^LAP2-Apc^homo mice to obtain P_tetCre-TA^LAP2-Apc^homo, which after induction by doxycycline withdrawal result in Apc^KO (for breeding schema see Supplementary Fig. S1). Accordingly, interbreeding was performed with Ctnnb1^homo and combined Apc^homo/Ctnnb1^homo mice leading to Ctnnb1^KO or Apc^KO/Ctnnb1^KO mice (Supplementary Table S1). Conductin^+/lacZ mice in which the reporter enzyme β-galactosidase is controlled by endogenous Conductin (Axin2) promoter were interbred with Apc^homo mice and other lines specified above to obtain Apc^hetero/Conductin^+/lacZ, Apc^homo/Conductin^+/lacZ, and Apc^KO/Conductin^+/lacZ mice, respectively. Mice used in proliferation tests received a unique bromodeoxyuridine (BrdU) injection (i.p.,10 mmol/L, 0.01 mL/g body weight) 2 hours before killing.

Human specimens were provided by the Tissue Bank of the National Center for Tumour Diseases (University of Heidelberg, Heidelberg, Germany, ethics proposal 206/2005). Fourteen formalin-fixed, paraffin-embedded liver specimens mostly taken from partial hepatectomies for liver metastases of colorectal cancer excised far from the metastases were investigated. Seven patients had clinically known FAP.

Immunohistochemistry was performed as described (19, 20). Briefly, 5 μm paraffin sections were dewaxed with xylol and hydrated through downward alcohol series. Antigen retrieval was assessed by microwaving in citrate buffer (pH 6.0). Slides were equilibrated in Tris-buffered saline (pH 7.4), quenched with H₂O₂, blocked with biotin/avidin, and goat serum and blocking reagent (VECTOR M.O.M. Immunodetection Kit) for mouse antibodies, respectively, and incubated with primary antibodies (listed in Supplementary Table S2) overnight at 4°C. Corresponding biotinylated secondary antibodies were coated for 1 hour at room temperature. After washing slides were incubated with extravidin–POD conjugate and washed three times before staining with 3,3′-diaminobenzidine-tetrahydrochloride (DAB). POD oxidizes DAB producing a brown precipitate. For p16^INK4a immunohistochemistry human liver sections were stained automatically by a Ventana Nexes autostainer (Ventana) using CINtec p16 (E6H4) immunohistochemistry.

Frozen liver samples were cut and mounted on Superfrost plus slides (Menzel). Slides were dried 1 hour at room temperature and fixed with 0.5% glutaraldehyde in PBS 5 minutes at room temperature. After washing in PBS, slides were incubated in 40 mmol/L Na₂HPO₄, 40 mmol/L citrate pH 6.0, stained overnight at 37°C with 1 mg/mL 5-bromo-4-chloro-3-indolyl-β-d-galactoside (X-Gal) in 40 mmol/L citric acid, sodium phosphate, pH 6.0, 5 mmol/L potassium ferrocyanide, 5 mmol/L potassium ferricyanide, 150 mmol/L, NaCl and 2 mmol/L MgCl₂ and counterstained with hematoxylin.

Sections were fixed in 2% paraformaldehyde and 0.2% glutaraldehyde in PBS pH 7.4 for 30 minutes at room temperature, washed three times for 15 minutes with 2 mmol/L MgCl₂, 0.01% Nonidet P-40 in PBS and stained with 1 mg/mL X-Gal, 5 mmol/L K₃Fe(CN)₆, 5 mmol/L K₄Fe(CN)₆ for 36 to 40 hours at 37°C in the dark. Slides were counterstained with nuclear fast red.

Primary hepatocytes were isolated and cultivated as described (20) by collagenase perfusion (21). Pericentral and periportal hepatocytes were isolated by a modified digitonin/collagenase perfusion technique (22, 23) and transfected with TOP-Gau reporter plasmids (24) using Effectene (Qiagen). Apc silencing in vitro was achieved by transfection with Apc siRNA (APCMSS202103, APCMSS202104, APCMSS202105, Invitrogen, 4 nmol/L) using Interferin (PeqLab).

Gaussia luciferase activity was measured in 10 μL supernatant of transfected hepatocytes with 50 μL assay reagent on Orion-II microplate luminometer (Titertek-Berthold).

Total RNA was isolated using PeqGOLD RNA Pure isolation system (Peqlab). RNA quality was assessed by gel electrophoresis and purity was estimated using the A260/A280 ratio. Concentration was adjusted to 0.5 mg/mL. qRT-PCR was performed as described (20) using primers listed in Supplementary Table S3. RNA load was normalized with the housekeeping gene cyclophilin (Ppia). Standard curves of serial dilutions from total RNA were used for transforming C_t values to concentration values depicted as arbitrary units.

DNA from tumor and surrounding liver tissue was isolated using NucleoSpin Tissue Kit (Macherey & Nagel). DNA methylation was tested with the bisulfite conversion method using the EZ DNA-Methylation-Gold kit (Zymo Research) and p16^Ink4a promoter-specific primers (ref. 25; Supplementary Table S3).

Alanine aminotransferase (ALT) and glutamate dehydrogenase (GLDH) were measured on an automated clinical chemistry analyzer (Modular PPE, Roche).

Hepatocytes were counted and 200,000 cells were resuspended in 100 μL wash buffer (PBS, 5% FCS). After washing, 500 μL prechilled ethanol (−20°C) was added slowly with continuous mixing avoiding agglutination. Hepatocytes were kept on ice, washed twice with 1 mL PBS/5% FCS, and incubated for 1 hour with RNAse A (1 mg/mL PBS) at 37°C. Suspensions were brought up to 200 μL with PBS/5% FCS and 20 μL propidium iodide (1 mg/mL) was added. This hepatocyte suspension was measured on a flow cytometer (FACScan, BD Biosciences) for determining the ploidy grade.

All data are expressed as mean ± SEM. Statistical analysis was performed by the Student t test or Mann–Whitney test using SigmaPlot 11 (SSP Science). The accepted level of significance was set at P < 0.05.

Using liver sections, β-catenin signaling was investigated by immunohistochemistry in hepatocytes of young (8 weeks) and aged (10 months) Apc^homo and compared with Apc^KO mice representing a positive control for β-catenin activation (Fig. 1A–D). This staining displayed only few hepatocytes with activated β-catenin (nuclear expression) both in liver tissue from young Apc^homo mice and tissue adjacent to tumors in aged mice. Liver tumors and their prestages in aged Apc^homo mice exhibit, unexpectedly, a highly heterogeneous β-catenin staining in the same animal. Both β-catenin–negative (Fig. 1C, C') and β-catenin–positive (Fig. 1D, D') tumors were detected. Moreover, if tumors contained activated β-catenin, they always displayed membrane-bound β-catenin providing a mixed phenotype, while an exclusively nuclear or cytoplasmic staining was never detectable.

Quantification of the universal and direct Wnt/β-catenin target gene, Axin2 mRNA level in liver tissue extracts (Fig. 1E), expectedly revealed a significant elevation in Apc^homo and Apc^KO and a decrease by trend in Ctnnb1^KO mice (only 2 animals were investigated due to sudden early death of Ctnnb1^KO mice) and a significant decrease in Apc^KO/Ctnnb1^KO mice compared with controls (both heterozygous and wild-type mice). In contrast, Axin2 mRNA levels in tumor extract (containing several tumors) were not significantly altered (Fig. 1E). However, re-evaluation of macroscopically visible tumors and subsequent resection of individual tumors of each liver revealed different individual Axin2 mRNA levels (Fig. 1F). Thereby, in the same animal either up- or downregulation of the β-catenin–responsive target Axin2 occur in individual hepatocellular carcinomas. To functionally validate activation of β-catenin signaling in vitro, we transfected a TCF reporter, which allows monitoring activity of β-catenin triggers, in isolated primary hepatocytes (Fig. 1G). Hepatocytes from Apc^homo mice show a 5-fold induction of Gaussia luciferase activity compared with controls, and in turn represent about one—fifth of luciferase activity obtained in Apc^KO mice (Fig. 1G), thus supporting Axin2 mRNA data.

In vivo, we also confirmed the elevated activation of β-catenin signaling in Apc^homo mice utilizing the Wnt reporter mouse Conductin^+/lacZ in which the reporter gene β-galactosidase is expressed in response to the endogenous Conductin (Axin2) promoter. In Apc^hetero/Conductin^+/lacZ mice β-galactosidase was detected only in pericentral hepatocytes as a weak spot-like staining within the nucleus of the first row of cells around central veins and rarely in the second row, while midzonal or periportal hepatocytes were never positively marked (Fig. 1H). In Apc^homo mice (Fig. 1J) the expression zone expanded and staining intensity increased. Here, several individual midzonal hepatocytes show β-galactosidase activity.

We examined expression of common β-catenin target genes, Glul and Cyp2e1 (Cytochrome P450 2E1; Fig. 2A), representing classical pericentrally (pc) expressed proteins within livers of Apc^homo mice at different ages. In addition, we inspected the pattern of typical periportally (pp) expressed proteins, carbamoyl phosphate synthetase I (CPS) and E-cadherin, which are expected to be conversely localized to pc-specific proteins (Fig. 2A). Liver tissue of Apc^homo mice showed age-dependent upregulation of β-catenin target genes, Cyp2e1 and Glul, supporting data from Buchert and colleagues (12), who investigated five-month old mice. The cellular proportion of the pericentral expression type in the zonal expression pattern starts to increase at age 3 weeks and is nearly completed at age 8 weeks. Later on, in mice, over 5 months old (Fig. 2A), first cancerous lesions appear. Although there was a slight increase of hepatocytes exhibiting nuclear/cytoplasmic β-catenin in aged Apc^homo mice, GLUL expression never spread over the whole lobulus.

The shifted zonal expression pattern of pericentral and periportal proteins in liver parenchyma of Apc^homo mice as a result of a continuous moderate activation of β-catenin signaling raised the question whether zonal shifting in liver parenchyma also occurs in patients with APC-repressing mutations. Thus, we examined the zonal expression pattern of GLUL in paraffin sections of livers of patients with FAP (Fig. 2B). In six of seven FAP liver samples, enlargement of the GLUL-positive zone was detected. In two of these six samples, GLUL-positive focal nodular hyperplasia was demonstrated (Fig. 2B, top right), whereas GLUL-positive nodules were not present in controls.

Precancerous lesions and tumors of aged Apc^homo mice, however, feature surprising heterogeneity regarding the expression of the β-catenin target genes, Glul and Cyp2e1 (Fig. 3). As expected from nuclear/cytoplasmic β-catenin staining pattern and Axin2 mRNA quantification tumors with pericentral, periportal, or the mixed expression program were detected within an individual animal (Fig. 3).

The mechanism driving carcinogenesis in Apc^homo mice seems obscure, because β-catenin–dependent hepatocyte proliferation is supposed not to be cell-autonomous (26) and tumors with homogenously nuclear β-catenin staining were not found in our experiments. Moreover, hyperproliferation neither of hepatocytes carrying nuclear or cytoplasmic β-catenin nor of pc- reprogrammed GLUL-positive hepatocytes occurs in Apc^homo mice (Supplementary Fig. S2). A pronounced proliferation in nontumorous liver parenchyma was detected in livers of Apc^KO mice only (Supplementary Fig. S2C) compared with coeval Apc^homo and control mice (Supplementary Fig. S2A and S2B). However, the hyperproliferation in Apc^KO mice is accompanied with ruin of hepatocytes by necrosis (Fig. 7A and B; black arrows). However, preliminary microarray experiments comparing the mRNA expression pattern of pericentral hepatocytes from Apc^homo mice to periportal hepatocytes from Apc^hetero mice indicate alterations of tumor suppressor levels in pericentral hepatocytes. Furthermore, tumor suppressor p16^INK4a was recently identified as a β-catenin target gene (27). Both results suggest p16^INK4a as a possible key regulator for tumorigenesis in Apc^homo mice. Consequently, we analyzed p16^INK4a immunohistochemically and detected a very high cytoplasmic amount of p16^INK4a in nearly all hepatocytes in aged Apc^homo mice independently of their acinar location (Supplementary Fig. S3 and S4). Livers of patients with FAP were also examined concerning expression of p16^INK4a. Hereby, both stainings manually performed with monoclonal antibody D7D7 used for detection of p16^INK4a-negative lesions in mice (not shown) and automatically performed with monoclononal antibody E6H4 (CINtec p16; Supplementary Fig. S3) detected a pronounced incidence of p16^INK4a in all livers samples of patients with FAP compared with normal liver. Thereby, a preference of p16^INK4a expression in pericentral areas was observed in both FAP livers and controls (frames in Supplementary Fig. S3J and S3K and the corresponding magnifications J' and K').

Simultaneously, loss of p16^INK4a protein in all mouse liver tumors and precancerous lesions (Fig. 4A) was detected. Serial sections of liver tissue from mice at age 10 months or older showing macroscopically visible tumors corroborate the p16^INK4a negativity of tumors and prestages independently on whether a periportal or pericentral expression program was followed (Fig. 4A). Cells within the p16^INK4a-negative tumors proliferate which is shown by BrdU incorporation (Supplementary Fig. S2E, E'). In contrast tumor, surrounding hepatocytes which are characterized by high p16^INKa expression do not proliferate (Supplementary Fig. S2E′).

To confirm p16^INK4a protein deficiency as a general hallmark of hepatocellular carcinomas in mice with abnormal β-catenin signaling, we stained liver sections harboring tumors, in which initial causative event had been identified as activating mutations in exon 3 of the CTNNB1 proto-oncogene, leading to constitutively active Wnt/β-catenin signaling, (ref. 28; Fig. 4C) and tumors, which developed after transfection of Apc^flox/flox mice using a virally encoded Cre-recombinase (ref. 29; Fig. 4B). All tumors including those with different genesis were p16^INK4a protein negative. In contrast, liver tumors of aged Apc^homo mice displayed no reduction of other tumor suppressors, i.e., p19^ARF, p15^INK4b, or p21^CIP1 (data not shown). Higher p16^Ink4a mRNA expression level was found in pericentral hepatocytes of Apc^homo mice (moderate activation of β-catenin signaling) and hepatocytes of Apc^KO mice (highest β-catenin signaling) compared with hepatocytes of Apc^hetero mice (low β-catenin signaling; Fig. 5B).

The downregulation of p16^INK4a protein in tumors of Apc^homo mice matches data on kidney (30), showing that only bypassing senescence caused by p21^CIP1 triggers renal tumors in Apc^homo mice. In liver, p16^Ink4a silencing seems to meet this function.

To validate whether p16^INK4a reduction is epigenetically regulated p16^Ink4a promoter methylation was examined. Only 1 of 7 tumors displayed a methylation-specific PCR product (Fig. 5A). Next, we investigated the transcriptional level by quantifying p16^Ink4a mRNA by qRT-PCR. No reduction of p16^Ink4a mRNA was detected in tumors compared with surrounding normal liver tissue of Apc^homo mice (Fig. 5B, right). In contrast, paradoxically, an increase of p16^Ink4a mRNA level was detected in all tumors lacking p16^Ink4a promoter methylation (Fig. 5B and Supplementary Fig. S5). The p16^INK4a reduction seems posttranscriptionally regulated, probably by increased degradation of p16^INK4a protein which would explain increased p16^Ink4a mRNA levels as compensatory mechanism. Recently, p16^INK4a degradation was suggested to occur ubiquitin independently by PSME3 proteasome (31). Hence, we quantified Psme3 in tumor and liver extracts and found an upregulation exclusively in tumor tissue (Fig. 5C). Tumor T2–2 in which the p16^Ink4a promoter is methylated shows no upregulation of Psme3 but demonstrates alternatives for p16^INK4a reduction in certain cases. Accordingly, this tumor was excluded for statistical analysis of Psme3 mRNA.

Overexpression of tumor suppressors, e.g. p16^INK4a, leads to induction of SA-β-Gal (GLB1; refs. 32, 33), activation of the facultative stem cell compartment, oval cells, in phases with proliferation demand (33) and polyploidization of hepatocytes (33). In all stages of life, more SA-β-Gal was detected in cryosections of Apc^homo compared with Apc^hetero mice (Fig. 6B). However, stronger upregulation of SA-β-Gal was found in livers of Apc^KO mice (Fig. 6C) and the strongest reactivity was observed in tissue surrounding tumors and precancerous lesions (Fig. 6D). Isolated primary hepatocytes of aged mice revealed in FACS analysis an increase of >16N ploidy in Apc^homo and Apc^KO mice (Fig. 6E), whereas hepatocytes with 4N ploidy decreased significantly in Apc^KO mice and by trend in Apc^homo mice. Immunohistologic stainings with anti–pan-cytokeratin antibody specific for oval cells (34) show their activation in Apc^homo mice starting at age 5 months. In livers of mice older than 10 months, both around macroscopically identifiable tumors and areas of cancer prestages, a border of oval cells was visible (Fig. 6J and K, arrows). Hence, the question arises what trigger, preferably produced by tumors, could activate oval cell compartment in Apc^homo mice.

Recently, Indian hedgehog (IHH) was confirmed as an activator of hepatic stem cells produced by dying hepatocytes (35). We found Ihh upregulated both in cultured hepatocytes after Apc siRNA treatment (Fig. 6F) and in Apc^KO mice (Fig. 6F) thus confirming Ihh increase in livers of AhCre-Apc^KO mice (36). These data combined with a trend of Ihh reduction in hepatocytes of Ctnnb1^KO mice confirm Ihh as direct target of β-catenin (37). The upregulation of Ihh mRNA level in tumors (Fig. 6F) supports a role as death signal of hepatocytes and cell damage should be expected.

Necrotic cell death occurs in Apc^KO mice of every age but also in old Apc^homo mice with tumors and prestages (Fig. 7A and B). In addition, apoptotic death featured by numerous Councilman bodies (Fig. 7C), seems to play a role in hepatocyte loss of Apc^homo mice, at least in aged mice, even though caspase-3 detection failed in all liver slides (not shown). As elevated liver enzymes in serum clearly indicate necrotic loss of hepatocytes, we measured characteristic liver parameters in serum of Apc^homo, Apc^hetero, and Apc^KO mice. Although a slight, nonsignificant raise of ALT and GLDH activities was detected age dependently in Apc^homo up to the age of 12 months (Fig. 7D and F), a significant difference of these enzyme activities as shown for old mice (Fig. 7B and D) between Apc^homo and Apc^hetero mice (controls) could be measured at all these ages (not shown). In contrast, serum levels of tumor harboring mice and Apc^KO mice were increased up to 10-fold compared with age-related Apc^homo mice (Fig. 7A–D).

In addition, DNA damage was observed by anti–γ-H2AX staining in all tumors and their prestages (Fig. 7K). Moreover, large quantities of γ-H2AX–positive nuclei were detected in Apc^KO mice (Fig. 7J).

Reduced APC expression leads to permanent activation of β-catenin and consequently, to upregulation of Wnt/β-catenin target genes and a pericentral expression program (12, 28, 38). Conversely, Hras mutations cause a periportal expression program (28). Unexpectedly, livers of Apc^homo mice, providing a model of half-maximal β-catenin activation, develop phenotypically different tumors. Some of them are distinguished by the expression of the Wnt target gene Glul and others by detection of E-cadherin, which is actually downregulated by sufficient β-catenin activation and belongs to the periportal expression program. Other Wnt target genes, i.e., Axin2, are reduced or unchanged in such tumors (ref. 12; current study). Likewise, the periportally expressed protein CPS is heterogeneously expressed.

The tumor suppressor p16^INK4a, recently identified as a β-catenin target gene (27), was completely lost at protein levels in all tumors and precancerous lesions investigated here and results in bypassing senescence as recently also found in human hepatocellular carcinoma and hepatoblastoma (39, 40).

Our data suggest the epigenetic silencing of p16^Ink4a in selected liver cells is a common factor for development of liver tumors. First, APC reduction leads to overexpression of p16^INK4a in the unaffected healthy liver of Apc^homo mice and in hepatocytes of Apc^KO mice with highly activated β-catenin signaling. Subsequently, elevated p16^INK4a induces senescence and protects from tumorigenesis as recently shown by others for premalignant hepatocytes in mice (41). Necessarily, increased ALT and GLDH levels in Apc^homo mice indicate loss of pericentrally programmed hepatocytes, which culminates in permanent proliferative stress of residual hepatocytes, being still senescent by p16^INK4a overexpression. Therefore, no fully replicative senescence seems to occur, because growth and physiologic function of liver of Apc^homo mice are not limited up to a critical age of about 10 months when tumors develop. Because the replicative capacity of hepatocytes is impaired by p16^INK4a overexpression, a physiologically required hepatocyte replacement generates a selective pressure, both forcing the initiation of p16^Ink4a silencing in selected populations of hepatocytes and promoting subsequent proliferation of clusters of p16^INK4a-negative hepatocytes. We hypothesize this sequence of events because no single p16^INK4a-negative hepatocyte was detected in younger animals before precancerous lesions occur and hyperproliferation does also not occur by moderate/half-maximal activated hepatocytes.

It seems irrelevant which metabolic program (pericentral/periportal phenotype) is followed by p16^Ink4a-silenced cells. The most relevant process responsible for silencing of p16^Ink4a function is the specific p16^INK4a removal, most likely by PSME3 mediated proteasomal digestion.

The p16^INK4a overexpression induced by diminished APC levels supports faster ageing and death of hepatocytes, which therefore contain a diminished proliferative capacity. A similar explanation can be supposed to pericentral hepatocytes, which have a continuous β-catenin signaling (42), and might also be relevant in liver parenchyma of patients with FAP as shown here by GLUL expression. P16^INK4a upregulation as a consequence of APC-repressing mutations also occurs in hepatocytes of patients with FAP and likely protects from liver cancer in early stages of life in the majority of cases. The higher incidence of hepatoblastoma in children with FAP (43) and the frequent occurrence of p16^INK4a loss in hepatoblastoma (40) underscores the significance of the tumor suppressor p16^INK4a in β-catenin–activated liver parenchyma.

The escalation of ALT and GLDH concentrations in Apc^KO mice indicates a massive decay of hepatocytes probably causing sudden death of Apc^KO mice 10 to 14 days after conditional knockout. The strong γ-H2AX staining in Apc^KO mice verifying the replicative stress in Apc^KO hepatocytes supports this suggestion.

The IHH signal delivered by dying hepatocytes is obviously not sufficient to activate an adequate number of oval cells to replace lost cells. This also applies to tumors whose elevated Ihh levels might follow, additionally to activation by β-catenin, damage-induced signals, similarly as reported on radiated hepatocytes (44). Even if sufficient oval cells were activated, their subsequent differentiation into hepatocytes would ultimately end in their decline.

Summarizing, our data suggest the epigenetic silencing of p16^Ink4a in selected liver cells bypassing senescence is a common principle for the development of tumors in mouse liver with β-catenin involvement independent of the initial stimulus.

No potential conflicts of interest were disclosed.

Conception and design: E. Ueberham, R. Gebhardt, U. Ueberham

Development of methodology: E. Ueberham, P. Glockner, K. Schonig, U. Ueberham

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Ueberham, P. Glockner, C. Gohler, B.K. Straub, D. Teupser, B. Jerchow, F. Gaunitz, U. Ueberham

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Ueberham, C. Gohler, B.K. Straub, R. Gebhardt, U. Ueberham

Writing, review, and/or revision of the manuscript: E. Ueberham, B.K. Straub, D. Teupser, K. Schonig, A. Braeuning, W. Birchmeier, F. Gaunitz, T. Arendt, O. Sansom, R. Gebhardt, U. Ueberham

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Gohler, B.K. Straub, K. Schonig, A. Braeuning, A.K. Hohn, W. Birchmeier, F. Gaunitz, O. Sansom, U. Ueberham

Study supervision: E. Ueberham, R. Gebhardt, U. Ueberham

The authors thank Doris Mahn, Frank Struck, Elisabeth Specht, and Helga Stache for technical assistance, Dr. Sabine Colnot (Departement Genetique Developpement et Pathologie Moleculaire, Institut Cochin-INSERM CNRS, University of Paris, Paris, France) for the gift of paraffin slides. The authors specially thank Dr. Helen Scott (University of Bristol, Bristol, United Kingdom) for critically reviewing the manuscript.

This work was supported in part financially by the European Union (grant CancerSys 223188), by the BMBF (01EW0907) in the frame of ERA-Net Neuron and the AFI-Project 984000–150 as well as the DFG (STR 1160/1–1).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.