Hepatocarcinogenesis in mice with a conditional knockout of the hepatocyte growth factor receptor c-Met



The receptor for the hepatocyte growth factor/scatter factor (HGF/SF), c-Met, plays a role in tumour promotion, progression and metastasis. In this study, we analysed chemically induced hepatocarcinogenesis in mice lacking a functional HGF receptor in their liver. Control and c-Met deficient mice were injected with a single dose of N-nitrosodiethylamine (DEN, 90 μg/g b.wt.) at 6 weeks of age and mice were subsequently kept on a phenobarbital (PB) containing diet (0.05%) for 35 weeks or on control diet. At the end of the experiment, the carcinogenic response in liver of the animals was monitored. Conditional c-met knockout (KO) mice showed a higher prevalence of macroscopically visible liver tumours and of glutamine synthetase positive and glucose-6-phosphatase deficient lesions in liver. Tumour promotion by PB led to significant increases in the number of preneoplastic and neoplastic lesions in liver of both wild-type and c-met knockout mice, with only minor differences in response. Our results indicate that a defect in c-Met-mediated signaling increases chemically induced tumour initiation in liver but does not significantly affect PB-mediated tumour promotion. © 2008 Wiley-Liss, Inc.

c-Met, the receptor for the hepatocyte growth factor/scatter factor (HGF/SF), plays an important role during early development, but also in tumour promotion, progression and metastasis [reviewed in Ref.1–3]. Although c-Met was originally reported to be an oncoprotein,4 it turned out that it could have both, oncogenic and tumour suppressive potential. On one hand, it can transform NIH3T3 cells5 and knockdown of c-met by siRNA inhibits the proliferation of hepatocellular carcinoma cells in vitro and in vivo.6 On the other hand c-Met can induce apoptosis and inhibit cell proliferation in several cancer cell lines including those derived from primary liver cancers.7

c-Met is a tyrosine kinase receptor consisting of a 50 kDa extracellular α-chain and a 145 kDa β-chain. The β-chain consists of an extracellular part, a juxtamembrane domain and an intracellular part. After stimulation of the c-Met receptor by HGF/SF autophosphorylation occurs at tyrosine residues Y1234 and Y1235 of the cytosolic part of the β-chain. By recruiting proteins like Grb2 or Gab1 to its c-terminal docking site, c-Met can activate a variety of downstream signaling cascades among which the Ras/mitogen activated protein kinase (MAPK) signaling pathway, the phosphatidylinositol-3 kinase (PI3K/Akt), the signal transducer and activator of transcription 3 (STAT3) and the Rac/p21 activated kinase (Rac/Pak) signaling pathways are most prominent.3, 8

Transgenic mice which overexpress HGF and therefore have a constitutively activated c-Met receptor are resistant to tumour promotion by phenobarbital (PB).9 Since promotion of mouse hepatocarcinogenesis by PB consists in a selection for hepatocytes carrying activating mutations in Ctnnb1, which encodes β-catenin,10 some kind of cross-talk may exist between the HGF/Met and the Wnt/β-catenin/TCF signal transduction pathways. Moreover, β-catenin has been shown to physically interact with c-Met at the inner side of the outer cell membrane.11, 12 To further examine the role of c-Met in DEN-induced tumour initiation and PB dependent tumour promotion we used a c-met conditional knockout model.2 In these mice the floxed exon 15 of the c-met gene, which codes for the ATP binding site, was deleted by activating the Mx-promoter controlled Cre recombinase. Livers from mice generated in this way show no autophosphorylation and therefore no activation of the c-Met receptor; delayed hepatocyte proliferation after partial hepatectomy and increased lipid-vesicle formation in liver.2

In agreement with data from a very recently published study13 our present results suggest that a c-Met defect in hepatocytes significantly enhances DEN dependent tumour initiation in liver. The number of GS-positive lesions was significantly increased in liver of c-met knockout (KO) mice in the absence of PB, which is remarkable, since considerable numbers of GS positive liver lesions are normally only seen when using a regimen including PB or a PB-like agent as tumour promoter. PB was effective as a tumour promoter in both c-met wildtype (Wt) and KO mice, with only minor differences in response, suggesting that a deficiency in functional c-Met has only a minor—if at all—effect during the promotional phase of hepatocarcinogenesis.

Material and methods

Initiation–promotion experiment

Male Metfl/ko/MxCre mice (phenotype: Met KO) and Metfl/Wt/MxCre mice (phenotype: Met Wt) received a single intraperitoneal injection of DEN (90 μg/g body weight dissolved in water) at 6–7 weeks of age. Two weeks later, all mice were given 3 injections at 2 day intervals of 300 ng poly inosylic/poly citidylic acid (poly I/C, Sigma) to induce Mx-promoter dependent Cre recombinase as previously described.2 One week thereafter, 1 group of mice of each genotype was kept on a 0.05% PB containing diet, whereas the respective control groups were kept on a PB-free diet for 35 to 36 weeks (see Fig. 1 for a treatment scheme). The numbers of mice in the 4 experimental groups were: group 1 (Met Wt minus PB), 12; group 2 (Met Wt plus PB), 13; group 3 (Met KO minus PB), 14; group 4 (Met KO plus PB), 14. Animals were singly housed in macrolon cages and kept on a 12 hr dark/light cycle. Animals received humane care and protocols complied with institutional guidelines.

Figure 1.

Treatment scheme of mice in the tumour initiation–promotion experiment. Mice were injected with a single dose of the tumour initiator DEN (90 μg/g body weight) at 6 weeks of age. Induction of the Mx-promoter controlled cre-recombinase and subsequent induction of the c-Met-KO-genotype was achieved by injection of 3 times 300 ng polyinosylic/polycitidylic acid (poly I/C) in a 2 day interval starting at 8 weeks of age. Mice were kept on a PB-containing diet or a control-diet from week 9 to week 35.

Tissue preparation

Mice were killed and livers were carefully removed and weighed. Tumours visible on the surface of the liver upon macroscopic examination were counted. Aliquots of the liver lobes were then frozen on dry ice for immunohistochemical analyses or fixed in Carnoy solution for subsequent paraffin embedding. Paraffin embedding was performed with Histowax (Reichert-Jung, Nußloch, Germany) according to standard procedures.

Enzyme and immunohistochemistry

At least 2 independent frozen liver tissue sections per animal were prepared and used for immunohistochemical detection of GS expression as recently described.14 The primary rabbit anti-GS antibody (1:1,000) was obtained from Sigma, Saint Louis and the HRP-conjugated anti-rabbit secondary antibody (1:100) from Dako, Hamburg. When GS was double stained with Ki67 (see below), a β-Gal conjugated anti-rabbit secondary antibody (1:50; American Qualex, San Clemente, USA) was used. G6Pase-staining was performed according to the method of Wachstein and Meisel.15 Stained sections were analyzed by use of a Zeiss Axio Imager microscope equipped with an Axiocam MRl (Zeiss, Oberkochen). For quantitative analysis of enzyme-altered lesions, numbers and areas of transections from G6Pase negative (G6Pase-) and GS positive (GS+) focal lesions visible in the stained sections were determined using Axiovision Rel. 4.5 software (Zeiss). Lesion numbers per cm3 liver tissue, volume fractions (%) and three-dimensional size class distributions were calculated by standard stereological procedures.16

Ki-67 labeling and TUNEL staining

Paraffin embedded liver sections were stained for Ki-67, a marker for cell proliferation. After deparaffination and rehydratization, cells were disrupted by microwave pulping for 25 min at 720 Watt. Sections were blocked with pig normal serum in PBS (1:30) for 1hr before incubation with a rat anti-mouse Ki67 primary antibody (1:20, Dianova, Hamburg) over night at 4°C in a humid chamber. After washing with PBS and subsequent blocking with pig normal serum in PBS (1:30) for 5 min, sections were incubated with HRP conjugated rabbit anti-rat secondary antibody (1:20, Dako, Hamburg) for 60 min at room temperature. Sections were washed and equilibrated with acetate buffer before staining of sections with AEC staining solution. Sections were counterstained with Haematoxilin Eosin as previously described.14 For quantitative analysis, numbers of Ki67 positive nuclei and total number of nuclei were counted per section area on a Zeiss Axio Imager microscope equipped with an Axiocam MRl (Zeiss) and Ki67 positive nuclei were than related as ‰ of total nuclei. An average of 155 × 103 nuclei was counted in total per group. The numbers of GS+ tumours counted in the respective groups were Wt: 3, KO: 10, WtPB: 27, KOPB: 37. TUNEL staining was performed with the In-Situ-Cell-Death-Detection-Kit (Roche, Mannheim). TUNEL positive nuclei per area were counted in normal tissue and GS+ tumours of both genotypes by use of a Leitz Labolux fluorescence microscope.

Statistical analysis

Statistical analysis was performed by Students t test for comparison of 2 groups and ANOVA analysis for comparing 4 groups of mice. Results were analyzed for outliers by Grubbs test (p < 0.01). Data were collected in MS-Excel, calculations were performed with MS-Excel (t test) or the software-package R version 2.5.1 (ANOVA analysis).17 Differences were considered as significant if p < 0.05 and a trend was assumed if p < 0.1. Three different types of effects were analyzed by ANOVA: 1. effect of genotype (met Wt versusmet KO), 2. effect of treatment (plus PB versus minus PB) and 3. interaction.


Phenotype of mice

At the end of the initiation promotion experiment, liver and body weights of the animals were determined. In animals without PB treatment, there were no significant differences in liver/body weight ratios between c-met Wt and c-met KO mice. This indicates that inactivation of c-Met function does not affect liver size in adult mice where the liver is fully grown. PB treatment induced significant increases in relative liver weights (p < 0.001), mostly because of a higher tumour burden both in c-met Wt and KO mice, without difference between genotypes (Fig. 2a). To determine cell proliferation we assayed the Ki67 labeling index in liver. The labeling indices of hepatocytes from normal liver were slightly higher in the PB treatment groups and in c-met KO mice when compared with their c-met Wt littermates, but these effects were not significant (Fig. 2b). Interestingly, GS+ lesions from c-met KO mice treated solely with DEN showed an increased Ki67 labeling index when compared with their counterparts from the respective wildtype mice (Fig. 2c). TUNEL stainings for apoptotic cells showed a slightly increased rate of TUNEL positive nuclei in GS+ tumours from c-met KO mice as compared with their Wt counterparts (4.01 ± 0.15 TUNEL positive nuclei/area [mm2] in GS+ tumours from c-met KO mice as compared with a value of 3.46 ± 0.04 in GS+ tumours from c-met Wt mice). Additionally, normal liver tissues from PB-treated mice showed slightly lowered rates of apoptosis when compared with nontreated tissues (0.25 ± 0.02 and 0.28 ± 0.01) TUNEL positive nuclei/area [mm2] in livers from c-met Wt and KO mice, respectively, as compared to 0.23 ± 0.01 and 0.18 ± 0.03 positive nuclei/area [mm2] in livers from c-met Wt and KO mice treated with PB. However, none of these rather small differences was statistically significant at a significance level of p < 0.05.

Figure 2.

Relative liver weight of mice and Ki67 labelling indices of hepatocytes. (a) Relative liver weights of c-met Wt and KO mice ± PB treatment. PB-treated mice showed a significantly (p < 0.001) increased relative liver weight, whereas there was no significant effect of genotype. (b) Ki67 labelling index in normal liver and liver tumours. In normal liver, there were no significant differences between groups whereas the Ki67 labelling index was significantly increased in GS positive (GS+) lesions from c-met KO mice (n = 10) when compared with GS+ lesions from Wt (n = 5) mice (p = 0.025). For normal tissues, an average of 155 × 103 nuclei was counted per group.

Tumour response

After removal of the livers, all macroscopically visible liver tumours > approximately 1 mm in diameter were counted. In animals treated with DEN only, we observed significant differences between c-met KO and Wt mice: whereas 50% of the Wt mice had liver tumours, the tumour prevalence was increased to approximately 80% in KO mice (Fig. 3b) and the average number of tumours per animal was approximately 4-fold higher (p = 0.035) in KO than in Wt mice (Fig. 3a). PB treatment strongly increased the number of liver tumours per animal (Fig. 3a), irrespective of genotype of mice (c-met KO and Wt p < 0.001) and 100% of animals in these 2 groups showed liver tumours (Fig. 3b).

Figure 3.

Tumour response. (a) Average number of tumours macroscopically detectable on the surface of the livers of mice of the different groups. For the sake of clarity, data from c-Met Wt and KO mice without PB treatment are shown in the insert at a different scaling of y-axis. (b) Percentage of tumour-bearing mice in the respective genotype and treatment groups.

Enzyme-altered lesions

Early liver lesions were analyzed by use of the 2 marker enzymes glucose-6-phosphatase (G6Pase) and glutamine synthetase (GS). G6Pase is a commonly used marker which is absent or lowered in activity in neoplastically transformed hepatocytes.15 GS, on the other hand, is a transcriptional target of β-catenin and is therefore specifically overexpressed in liver tumours with mutations in the β-catenin gene.18–20 The number, volume fraction, and size distribution of G6Pase- and GS+ lesions in liver was quantified on stained liver sections. The numbers of G6Pase and GS+ lesions per cm3 liver tissue were increased 3.3-fold (p = 0.030) and 3.1-fold (p = 0.038), respectively, in c-met KO as compared to Wt mice (Figs. 4a and 4b). The mean volume fraction of G6Pase- and GS+ lesions were also increased by a factor of about 2.5, but the effects were not statistically significant (Figs. 4c and 4d). Interestingly, the effect on lesion number produced by the gene knockout was most pronounced in the smallest diameter classes (up to class 10), both with regard to G6Pase and GS+ lesions (Fig. 5).

Figure 4.

Number and volume fraction in liver of enzyme-altered lesions. (a) Number per volume unit [cm3] and (c) volume fraction in liver [%] of G6Pase deficient lesions, (b, d) respective parameters for GS positive lesions. For the sake of clarity, data from c-Met Wt and KO mice without PB treatment are shown in the inserts at a different scaling of y-axis.

Figure 5.

Effect of genotype (c-met Wt or KO) and PB treatment on size class distribution of enzyme-altered lesions. The numbers of G6Pase negative and GS positive lesions were calculated by stereological procedures from 2-dimensional observations. The upper limits of the diameter classes increase logarithmically from class 1, antilog 1.8 (63 μm), class 2, antilog 1.9 (79 μm)… to antilog 3.9 (7.9 mm) in size class 22. Because of technical limitations of our evaluation, there were no lesions detectable smaller than diameter class 7.

PB treatment strongly enhanced both the number (Figs. 4a and 4b) of G6Pase and GS+ lesions and their volume fraction in liver (Figs. 4c and 4d). This promoting effect on occurrence and growth of enzyme-altered liver lesions was highly significant for both c-met Wt and KO mice (p < 0.001) but did not statistically significantly differ between genotypes. The analysis of the size class distribution of G6Pase and GS+ lesions demonstrated a PB-mediated increase in lesion numbers in all size classes, with the largest diameter classes (19-22) being occupied only by lesions from PB-treated animals (Fig. 5).


In our study we observed a 3 to 4-fold increase in the number of tumours and G6Pase and GS+ lesions in DEN-treated c-met KO mice when compared with their wildtype counterparts. This effect is in agreement with results from a very recent study which was conducted with a comparable but different strain of c-met KO mice and was published during the final course of our experimental study.13 The analysis of the size distribution of the enzyme-altered lesions indicated that the increase in lesion number produced by the gene knockout was primarily due to larger numbers of comparatively small lesions present in the livers of the c-met KO mice. This evidence suggests that silencing of c-Met dependent signaling affects primarily the process of tumour initiation and to a lesser extent tumour promotion. In fact, we did not observe any significant differences between c-met Wt and KO mice in their neoplastic response in the PB treatment groups. The observed increase in lesion number in the c-met KO mice could have different causes: The metabolism of the carcinogen DEN, which was used for initiation, could be altered in mice with liver-specific c-Met deficiency, e.g., as a consequence of (hypothetical) differences between c-met Wt and KO mice in the activity of enzymes required for metabolic activation of DEN. Interestingly Takami et al. found an increased expression of several cytochrome P 450 enzymes in c-Met deficient mice. We can exclude this possibility, however, since the inactivation of the functional “floxed” c-met gene in the Metfl/ko/MxCre mice was performed 2–3 weeks after administration of DEN and therefore at a time point when the metabolism of the carcinogen is history. We have some evidence, however, from other studies that the process of initiation, which follows the primary DNA alkylation reaction, may take several weeks,21 possibly because a critical mutation has to be fixed during 2 subsequent rounds of cell division. Therefore a temporary increase in hepatocyte division rate, if it were present, mediated by the deficiency in functional c-Met, could explain an increased initiation rate in the gene knockout mice. A second critical parameter is the rate of apoptosis. At the very early stages (1, 2, 4… cell stage) during formation of a (pre)neoplastic clone, clones are at high risk to become extinct—and therefore uncountable—because of the comparatively high apoptotic rates of the neoplastically transformed cells (e.g., see Ref.22). Since c-Met has been shown to have proapoptotic properties in other systems,7 elimination of c-Met receptor function may be beneficial for the survival of initiated cells and consequently lead to increased numbers of enzyme-altered lesions in the KO mice. The results of the TUNEL stainings of this study contradict this hypothesis but the measurements were performed on manifest tumours from c-Met Wt and KO mice and may not reflect the situation at the very early stage of the carcinogenic process.

Treatment of animals with the tumour promoter PB strongly increased the formation of liver tumours and enzyme-altered liver lesions. There were, however, no significant differences in tumour response between c-met Wt and c-met KO mice, suggesting that c-Met dependent signaling is not required for the promoting activity of PB. The enzyme-altered lesions observed in livers of PB-treated animals were in their vast majority GS positive (and negative for G6Pase). The GS+ phenotype of lesions is in agreement with results from earlier studies from our group which showed that tumour promotion by PB in liver consists of a positive selection for hepatocytes that express an activated version of β-catenin, because of Ctnnb1 mutations10 and therefore overexpress GS, which is controlled by β-catenin signaling.18 Since such GS+ lesions are only relatively rarely found in the absence of PB-mediated tumour promotion,10 we were surprised to see a comparatively large number of GS+ lesions in the livers of c-met KO mice treated with DEN but not with PB. Therefore, elimination of signaling through the c-Met receptor seems to create a cellular environment favoring the outgrowth of Ctnnb1 mutated, GS+ lesions and thus mimics to some extent the promotional effect of PB.

In the absence of PB, mouse liver lesions are frequently ras mutated (e.g., see Ref.10) or carry mutations in the gene encoding the ras downstream target B-raf.10, 23, 24 These tumours with constitutively activated Ras/Raf/Erk signaling pathway cannot be promoted by PB, in contrast to their Ctnnb1 mutated cousins.10 The observation that the promotional effect of PB is markedly attenuated in mice where c-Met-dependent downstream signaling, including signaling through the MAP kinase cascade, is constitutively activated by transgenic overexpression of the c-Met ligand HGF,9 fits well into this picture.

The promotional activity of PB in mouse liver requires the functional constitutive androstane receptor (CAR) as recently shown by studies with CAR knockout mice.25 PB is known to activate the protein phosphatase PP2A26 which is required for nuclear translocation and activation of CAR (for review see Ref.27). PP2A, on the other hand, is an inhibitor of Erk-dependent signaling, since it dephosphorylates the activated form of Erk (p-Erk).28 Interestingly, p-Erk levels are significantly lowered in livers from PB-treated mice (T. Knorpp, unpublished results). Inversely, c-Met is one of the important stimulators of the Ras/Erk signaling pathway in liver.2 Activation of Erk signaling, e.g., by serum components, inhibits the expression of CAR target genes, including many drug metabolizing enzymes of both phase I and II.29 Furthermore, activation of c-Met by HGF represses PB induced translocation of CAR to the nucleus and expression of CAR target genes in a MEK/Erk dependent manner.30 In synopsis, presence of PB and elimination of functional c-Met might both attenuate the negative effect of activated Erk on the expression of CAR target genes amongst which we would expect the ones, which directly or indirectly mediate the positive selection for Ctnnb1 mutated, GS+ liver lesions.


We thank Mrs. Johanna Mahr and Mrs. Elke Zabinsky for excellent technical support.