Serum components and activated Ha-ras antagonize expression of perivenous marker genes stimulated by β-catenin signaling in mouse hepatocytes

Authors


M. Schwarz, Institute of Pharmacology and Toxicology, Department of Toxicology, University of Tuebingen, Wilhelmstr. 56, 72074 Tuebingen, Germany
Fax: +49 7071 29 2273
Tel: +49 7071 29 77398
E-mail: michael.schwarz@uni-tuebingen.de

Abstract

Hepatocytes of the periportal and perivenous zones of the liver lobule show marked differences in the contents and activities of many enzymes and other proteins. Previous studies from our and other groups have pointed towards an important role of β-catenin-dependent signaling in the regulation of expression of genes encoding proteins with preferential perivenous localization, whereas, in contrast, signaling through Ras-dependent pathway(s) may induce a ‘periportal’ phenotype. We have now conducted a series of experiments to further investigate this hypothesis. In transgenic mice with scattered expression of an activated Ha-ras (Ha-rasG12V) mutant in liver, expression of the perivenous markers glutamine synthetase and two cytochrome P450 isoforms was completely abolished in those hepatocytes demonstrating constitutively activated extracellular signal-regulated kinase activity, even though they were located directly adjacent to central veins. Similarly, incubation of primary hepatocytes or hepatoma cells with increasing amounts of serum caused a concentration-dependent attenuation of expression of perivenous marker mRNAs, whereas the expression of periportal markers was increased. The inhibitory effect of high amounts of serum on the expression of perivenous markers was also observed if their expression was stimulated by activation of β-catenin signaling, and comparable inhibitory effects were seen in cells stably transfected with a T-cell factor/lymphoid-enhancing factor-driven luciferase reporter. Epidermal growth factor could partly mimic serum effects in hepatoma cells, and its effect could be blocked by an inhibitor of extracellular signal-regulated kinase activity. These data suggest that activation of the Ras/mitogen-activated protein kinase (extracellular signal-regulated kinase) pathway favors periportal gene expression while simultaneously antagonizing a perivenous phenotype of hepatocytes.

Abbreviations
Aldh1b1

aldehyde dehydrogenase 1b1

Cyp

cytochrome P450

EGF

epidermal growth factor

ERK

extracellular signal-regulated kinase

Gpr49

G-protein coupled receptor 49

GS

glutamine synthetase

GSK3β

glycogen synthase kinase 3β

MAPK

mitogen-activated protein kinase

TCF/LEF

T-cell factor/lymphoid-enhancing factor.

In mammalian liver, the gene expression profile of an individual hepatocyte is determined by its position relative to the terminal branches of the afferent and efferent blood vessels, the portal and the hepatic (central) veins. Thus, in many processes, hepatocytes from the periportal and perivenous (pericentral) zones of the liver lobule have different or even complementary functions. This phenomenon is called metabolic zonation, and has been described for a variety of metabolic pathways, e.g. the synthesis and/or metabolism of glucose/glycogen, fatty acids, amino acids, and bile acids [1–3]. The metabolism of xenobiotic compounds is also differentially regulated along the portocentral axis, with the main enzymes of xenobiotic detoxification being preferentially located in perivenous hepatocytes [4]. Whereas one group of genes, primarily those encoding enzymes related to the metabolism of carbohydrates, is dynamically regulated in response to changes in nutritional or hormonal status, a second group of genes exhibits a so-called ‘stable zonation’, where the expression of the respective gene is restricted to a certain subpopulation of hepatocytes, either perivenous or periportal, and can hardly be influenced under physiologic conditions [5]. Even though zone-specific expression differences have been reported for a multitude of genes, for a long time very little has been known about the regulation of zonal heterogeneity in gene expression in the liver.

On the basis of striking similarities in the gene expression profiles of Ha-ras and Ctnnb1 mutated mouse liver tumors and those of periportal and perivenous hepatocyte subpopulations, respectively, we have recently put forward a hypothesis to explain differential gene expression within the liver lobule [6]. According to this hypothesis, the zone-specific activity of two antagonistic signaling pathways determines the gene expression profiles of periportal and perivenous hepatocytes: a Ras-activating signal delivered by bloodborne molecules was suggested to prevail in the periportal area, whereas a β-catenin-activating signal, perhaps delivered by endothelial cells of the central veins, was proposed to determine the perivenous gene expression profile by activating β-catenin/T-cell factor (TCF)-dependent gene transcription. Since then, several studies have been published that confirm an essential role for β-catenin signaling in the regulation of perivenous gene expression in mouse liver [7–10], whereas the mechanisms regulating periportal gene expression still remain to be discovered.

In this study, we investigated the effect of Ras-dependent signaling and serum factors on the expression of zonated marker genes and the interaction of serum-activated ‘periportal’ and β-catenin-mediated ‘perivenous’ signaling pathways.

Results

Inhibition of perivenous marker protein expression in hepatocytes and tumor cells carrying an activated form of Ha-ras

We have recently put forward the hypothesis that two opposing signaling pathways, β-catenin-dependent and Ha-ras-dependent, regulate zonal gene expression in mouse liver [6,10]. We have now investigated the effect of the proposed antagonism between the two pathways on hepatocyte gene expression by both in vivo and in vitro approaches. In a first experiment, we studied by immunohistochemistry the zonal distribution of a variety of proteins with known preferential perivenous localization in liver, using transgenic Tglox(pA)H-ras* mice [11]. These mice express a human activated Ha-ras (Ha-rasG12V) protein after Cre-mediated recombination, which is induced by adenoviral transfer of cytomegaly virus (CMV)-cre[11,12]. Upon injection of a construct adenovirus (Ad) CMV-cre into the tail veins of the mutant mice, dysplastic hepatocytes of enlarged size are detectable in the liver tissue as early as 1 week after injection [12]. These dysplastic cells appear mostly as isolated single cells, probably because only a minority of the hepatocytes become infected with the adenovirus during gene transfer. Immunohistochemically, these transgenic cells can be easily detected using an antibody against the phosphorylated form of extracellular signal-regulated kinase (ERK1/2), a downstream target of Ras signaling (Fig. 1A). Control stainings demonstrated the absence of phospho-ERK-positive dysplastic hepatocytes in Cre-infected wild-type mice (Fig. 1B). To study a possible effect of the activated Ha-ras protein in the dysplastic cells on expression of proteins with preferential perivenous localization, we used the markers glutamine synthetase (GS), cytochrome P450(Cyp)1A and Cyp2E1, which were immunohistochemically stained in liver slices from the transgenic mice. Furthermore, GS expression in Ha-ras and B-raf mutated liver tumor cells was examined.

Figure 1.

Figure 1.

 Expression of perivenous marker proteins is abolished by activation of Ha-ras, as demonstrated by immunohistochemical staining of liver sections from transgenic Tglox(pA)H-ras* mice (A, C, D, E, F), wild-type control animals (B, I, J, K), and Ha-ras mutated liver tumors (G, H). (A) Following Cre-mediated activation of the transgene, livers of Tglox(pA)H–ras* mice show large dysplastic hepatocytes that demonstrate activation of ERK1/2, a downstream target of Ras signaling. (B) Cre-infected wild-type mice do not show such abnormal hepatocytes. (C) Liver sections from a transgenic Tglox(pA)H-ras* mouse display GS expression (brown color) in hepatocytes around the branches of the central vein. Some hepatocytes within this area, however, are GS-negative (see arrows). (D) Double-staining of a serial section with antibodies against GS (brown color) and phospho-ERK (red color) reveals that the GS-negative perivenous cells exhibit activation of ERK1/2. (E, F) Hepatocytes with activated Ha-ras also lack expression of the cytochrome P450 isoforms 1A (E) and 2E1 (F) (ERK staining not shown). (G) Ha-ras mutated hepatoma cells do not express GS, even if they are located directly adjacent to a central vein. For more detail, see (H). (I) In normal liver, phosphorylated ERK1/2 is predominantly present in periportal hepatocytes. (J) Perivenous GS expression is shown for comparison. (K) Control biotin/streptavidin staining without primary antibody. Abbreviations: c, central vein; t, tumor; n.l., normal liver tissue.

Figure 1.

Figure 1.

 Expression of perivenous marker proteins is abolished by activation of Ha-ras, as demonstrated by immunohistochemical staining of liver sections from transgenic Tglox(pA)H-ras* mice (A, C, D, E, F), wild-type control animals (B, I, J, K), and Ha-ras mutated liver tumors (G, H). (A) Following Cre-mediated activation of the transgene, livers of Tglox(pA)H–ras* mice show large dysplastic hepatocytes that demonstrate activation of ERK1/2, a downstream target of Ras signaling. (B) Cre-infected wild-type mice do not show such abnormal hepatocytes. (C) Liver sections from a transgenic Tglox(pA)H-ras* mouse display GS expression (brown color) in hepatocytes around the branches of the central vein. Some hepatocytes within this area, however, are GS-negative (see arrows). (D) Double-staining of a serial section with antibodies against GS (brown color) and phospho-ERK (red color) reveals that the GS-negative perivenous cells exhibit activation of ERK1/2. (E, F) Hepatocytes with activated Ha-ras also lack expression of the cytochrome P450 isoforms 1A (E) and 2E1 (F) (ERK staining not shown). (G) Ha-ras mutated hepatoma cells do not express GS, even if they are located directly adjacent to a central vein. For more detail, see (H). (I) In normal liver, phosphorylated ERK1/2 is predominantly present in periportal hepatocytes. (J) Perivenous GS expression is shown for comparison. (K) Control biotin/streptavidin staining without primary antibody. Abbreviations: c, central vein; t, tumor; n.l., normal liver tissue.

In livers of the transgenic Tglox(pA)H–ras* mice, GS was, as expected, expressed exclusively in hepatocytes directly neighboring a branch of the central vein; some of the perivenous cells, however, completely lacked GS expression (Fig. 1C). This is in contrast to wild-type mice, which show homogeneous GS staining in all hepatocytes surrounding a central vein (Fig. 1J). Double staining with phospho-ERK and GS antibodies revealed that the GS-negative cells in transgenic mice strongly express active ERK (Fig. 1D). A similar picture was seen when the expression of the cytochrome P450 isoforms 1A and 2E1 was analyzed. Again, a number of perivenous cells lacked expression of both the perivenous markers Cyp1A and Cyp2E1 (Fig. 1E,F). Just like the GS-negative cells from Fig. 1C,D, the Cyp-negative cells were phospho-ERK-positive (data not shown). Furthermore, GS expression was analyzed in liver slices containing transections of Ha-ras mutated liver tumors. Whereas normal parts of the liver showed the well-known perivenous GS expression, this enzyme was not expressed at all in the Ha-ras mutated tumor cells, even if they were located directly adjacent to a central vein (Fig. 1G,H). Mouse liver tumor cells carrying an activating mutation in the B-raf proto-oncogene also never expressed GS and Cyp1A or Cyp2E1 (data not shown). In summary, both single dysplastic hepatocytes with positive phospho-ERK staining and hepatocytes from Ha-ras or B-raf mutated liver tumors entirely lacked detectable levels of the perivenous marker proteins investigated.

To analyze zonal-specific differences in ERK1/2 activation in normal liver, slices from wild-type mice were analyzed by immunohistochemical staining (Fig. 1I). Even though the detected portocentral gradient of phospho-ERK1/2 was not as prominent as the differences between normal hepatocytes and cells expressing the Ha-ras transgene, the activated kinase was found to be predominantly present in periportal hepatocytes (for comparison, see the perivenous staining of GS in Fig. 1J), indicating physiologic activation of the protein in the areas around the branches of the portal vein. Control slices incubated without primary antibodies did not show any zonal staining differences (Fig. 1K).

Effect of serum concentration on expression of marker genes in primary hepatocytes and cultured hepatoma cells

According to our hypothesis, Ras signaling is activated in hepatocytes by bloodborne molecules that are available at higher concentrations in the periportal area of the liver lobule. Most likely, these hypothetical molecules should be present in the serum fraction. We therefore incubated primary mouse hepatocytes for 96 h with DMEM/F-12 medium containing different concentrations of fetal bovine serum, using 10% of serum to mimic ‘periportal’ conditions and 1% of serum to mimic a more ‘perivenous’ environment. Changes in the gene expression of a subset of periportal marker genes [6,13] were analyzed by real-time RT-PCR. As shown in Fig. 2, the expression of H19, highly expressed in Ha-ras mutated mouse liver tumors [14] and also in periportal hepatocytes [6], was significantly elevated at the higher serum concentration. A significant increase in expression upon incubation with 10% fetal bovine serum was also seen for three other periportal markers investigated, i.e. cathepsin C, reduced in expression 3 (Rex3), and hairy and enhancer of split (Hes1). Serum-dependent upregulation of Cdh1 (encoding E-cadherin) was less pronounced, and aldehyde dehydrogenase 1b1 (Aldh1b1) expression was not affected at all. Activation of β-catenin signaling by addition of the glycogen synthase kinase 3β (GSK3β) inhibitor SB-216763 had no effect on the expression of H19, cathepsin C or Hes1, whereas the expression of Rex3, Cdh1 and Aldh1b1 significantly declined upon treatment with SB-216763.

Figure 2.

 Effect of serum concentration on expression of periportal and perivenous markers in primary mouse hepatocytes. Cells were incubated for 96 h with medium containing 1% or 10% fetal bovine serum in the presence or absence of 5 µm GSK3β inhibitor (SB-216763) (SB). Expression of the periportal markers H19, CathC, Rex3, Hes1, Cdh1, and Aldh1b1, as well as of the perivenous markers Axin2 and Gpr49, was quantified by real-time RT-PCR. Expression is given relative to cells incubated with 1% fetal bovine serum. The top and bottom of the boxes gives the 25th and 75th percentiles, respectively; the centre line within the box gives the median of the distribution. The top and bottom of the whiskers give the minimum and maximum values. The open square denotes the mean. Significant effects (P < 0.05, Student's t-test) are indicated by (*). Each box represents at least eight values.

A completely different picture emerged with regard to perivenous markers. Basal mRNA levels of Axin2, a perivenous marker [6] and β-catenin target [15], were decreased by ∼ 40% at the high serum concentration of 10% as compared to values obtained at 1% serum (Fig. 2). If the GSK3β inhibitor SB-216763 was used to activate β-catenin signaling, the expression of Axin2 was strongly increased at both serum concentrations, but expression of the mRNA was lower at the high fetal bovine serum concentration when compared to 1% fetal bovine serum. Similar behavior was evident with respect to another perivenous marker, the G-protein-coupled receptor 49 (Gpr49), another target of β-catenin signaling [16].

The interaction of high serum concentrations and activation of β-catenin signaling was further analyzed in mouse hepatoma cells from line 70.4, which are wild type with respect to Ha-ras, B-raf and Ctnnb1. To test whether fetal bovine serum would activate the Ras/ERK pathway in this cell system, western blotting for phosphorylated ERK was performed. As depicted in Fig. 3A, strong activation of ERK was achieved at 10% fetal bovine serum as compared to 1% fetal bovine serum. Additionally, protein levels of β-catenin slightly decreased following incubation of cells with higher fetal bovine serum concentrations (Fig. 3A). Furthermore, the ability of SB-216763 to stabilize β-catenin and thereby induce β-catenin/TCF-dependent signaling was estimated by western blot analysis of cell lysates from SB-216763-treated 70.4 cells. In the presence of the GSK3β inhibitor, accumulation of β-catenin protein was clearly visible, as well as induction of Axin2/Conductin protein, which was exclusively detectable in lysates from SB-216763-treated cell cultures (Fig. 3B).

Figure 3.

 Western analysis of ERK1/2 phosphorylation, β-catenin expression and upregulation of the marker protein Axin2/Conductin in 70.4 mouse hepatoma cells in response to treatment with serum, the GSK3β inhibitor SB-216763, or EGF. (A) Elevated concentrations of fetal bovine serum markedly increased the levels of phosphorylated ERK1/2 (p-ERK). At the same time, levels of β-catenin (β-Cat.) were slightly alleviated. (B) Inhibition of GSK3β by 5 µm SB-216763 led to elevated levels of β-catenin. Axin2 expression, which was not discernible in untreated cells, could be clearly detected after incubation with SB-216763. (C) Treatment with 50 ng·mL−1 EGF resulted in the expected strong phosphorylation of ERK1/2. In contrast, β-catenin levels were reduced upon EGF treatment. In all experiments, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control.

At the mRNA expression level, the stimulatory effect of SB-216763 on Axin2 and Gpr49 expression was very pronounced at 1% fetal bovine serum but considerably alleviated at 10% fetal bovine serum (Fig. 4A), which is similar to our observations with primary hepatocyte cultures. The inhibitory effect of fetal bovine serum on both basal and GSK3β inhibitor-induced expression of β-catenin target genes was clearly concentration-dependent, as demonstrated by the data in Fig. 4B. Comparable effects were observed using the mouse hepatoma cell lines 53.2b, 55.1c, PW53T, and Hepa1c1c7, irrespective of the mutational status of the Ha-ras and B-raf proto-oncogenes in these cell lines (data not shown). Similar inhibition of Axin2 expression was seen when horse or adult bovine serum was used instead of fetal bovine serum (data not shown). Charcoal treatment abolished the ability of serum to reduce basal Axin2 levels by approximately 60%, whereas serum-mediated reduction of SB-216763-induced Axin2 mRNA was diminished by 15% as compared to untreated fetal bovine serum (data not shown).

Figure 4.

 Serum attenuates expression of β-catenin target genes in mouse hepatoma cells in culture. (A) Hepatoma cells of line 70.4 were incubated for 48 h with medium containing 1% or 10% fetal bovine serum in the presence or absence of 5 µm GSK3β inhibitor (SB-216763) (SB). The mRNAs for the β-catenin targets Axin2 and Gpr49 were quantified by real-time RT-PCR. Means ± SD of four independent experiments are given relative to cells incubated with 1% fetal bovine serum. (B) Expression of Axin2 in 70.4 mouse hepatoma cells incubated for 48 h with medium containing the indicated serum concentrations (± SB-216763). Means of four independent experiments are given relative to cells incubated with 1% fetal bovine serum. Significant effects (P < 0.05, Student's t-test) are indicated by (*).

Serum contains various thrombocyte-derived components that are absent or present only to a much lesser extent in the blood plasma that cells are physiologically in contact with. To exclude influences of these thrombocyte factors on our model, 70.4 mouse hepatoma cells were incubated in the presence of 1% or 5% human plasma. Comparable to what was seen using fetal bovine serum, plasma inhibited both basal and GSK3β inhibitor-induced levels of Axin2 and Gpr49 mRNAs (data not shown), demonstrating that the observed effect was not due to platelet-derived factors.

Expression of mutant Ha-ras in 70.4 cells led to increased apoptosis and decreased growth of transfected cells (data not shown). Thus, the effects of Ras activation in 70.4 cells could not be studied properly. However, a tendency for a reduced response to SB-216763 was seen in some of the transfected cell clones (data not shown).

Inhibition of β-catenin/TCF-dependent transcription by serum

To analyze whether the downregulation of β-catenin target gene mRNAs by serum was due to serum-mediated inhibition of β-catenin/TCF-dependent transcription, mouse hepatoma cells from cell line 70.4 were stably transfected with the 8xTCF/LEF-driven SuperTopflash luciferase reporter. Seven stably transfected luciferase-expressing sublines were obtained and incubated for 24 h with medium containing either 1% or 20% fetal bovine serum, respectively. Luciferase activity was measured and normalized to the protein contents. Similar to the findings for the mRNA level, basal β-catenin-dependent luciferase activity was lowered in response to the higher fetal bovine serum concentration in all cell lines analyzed, with an average reduction of luciferase activity in the 20% fetal bovine serum-treated cells of ∼ 45%(Fig. 5A). The inducibility of luciferase activity by inhibition of GSK3β was also inhibited by high serum concentrations (Fig. 5B). Fetal bovine serum alleviated β-catenin/TCF-driven luciferase activity in a concentration-dependent manner, comparable to what was seen in the mRNA experiments (Fig. 5C).

Figure 5.

 Serum inhibits basal and inducible β-catenin/TCF-dependent transcription. (A) Mouse hepatoma cells from cell line 70.4 were stably transfected with the 8xTCF/LEF-driven luciferase reporter SuperTopflash and incubated with medium containing either 1% or 20% fetal bovine serum for 24 h. Luciferase activity (normalized to the protein contents) of seven stably transfected sublines at 20% fetal bovine serum is given relative to values obtained with 1% fetal bovine serum. All experiments were performed in quadruplicate. (B) Two selected sublines from (A) were incubated in the presence of 5 µm GSK3β inhibitor (SB-216763) at different serum concentrations for 24 h. Mean luciferase activity of four experiments is given relative to values obtained with 1% fetal bovine serum. Significant effects (P < 0.05, Student's t-test) are indicated by (*). (C) Fetal bovine serum inhibits SuperTopflash-driven luciferase activity in a concentration-dependent manner. Cells stably transfected with SuperTopflash were incubated with the indicated serum concentrations for 24 h. Mean luciferase activity of two experiments is given relative to values obtained with 1% fetal bovine serum. Additionally, for better comparability with the data from Fig. 4B, relative levels of Axin2 mRNA at the respective serum concentrations are also shown in the diagram (dotted line).

Modulation of the Ras/ERK signaling pathway

To obtain information on the nature of the serum-derived factor responsible for the inhibition of β-catenin signaling, 70.4 mouse hepatoma cells were grown in DMEM/F-12 medium containing 1% fetal bovine serum and different compounds that are known to activate signaling pathways that may be potentially relevant for the inhibition of β-catenin signaling and should therefore mimic high fetal bovine serum concentrations. Axin2 mRNA levels were not reduced upon treatment with either 20 nm triiodothyronine, 10 nm dexamethasone, 1 µm retinoic acid, 10 ng·µL−1 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene, 100 nm insulin or 20 ng·mL−1 hepatocyte growth factor (data not shown). In contrast, both basal and GSK3β inhibitor-induced expression of Axin2 was significantly reduced upon incubation of 70.4 cells with 50 ng·mL−1 epidermal growth factor (EGF) (Fig. 6A). Because (a) both hepatocyte growth factor and EGF led to a comparable increase in proliferation of 70.4 cells, but only EGF treatment was able to downregulate β-catenin signaling/Axin2 expression in a serum-like manner, and (b) Axin2 expression in 70.4 cells was not dependent on confluency, the observed EGF effect could not be based on cell growth. Comparable to what was seen with high serum concentrations (Fig. 3A), western blotting revealed a strong induction of ERK phosphorylation upon EGF treatment, whereas β-catenin levels were reduced in response to the growth factor (Fig. 3C). In two other experiments, 70.4 hepatoma cells were incubated (a) in the absence or presence of 50 ng·mL−1 EGF ± 10 µm ERK inhibitor (U0126), and (b) in medium containing different concentrations of fetal bovine serum ± 10 µm ERK inhibitor (Fig. 6B). In EGF-treated cells, U0126 was able to restore Axin2 mRNA levels; however, the ERK inhibitor could only partially antagonize the serum-mediated repression of Axin2.

Figure 6.

 Effects of ERK1/2 activation and inhibition on Axin2 expression in 70.4 mouse hepatoma cells. (A) Hepatoma cells of line 70.4 were incubated for 48 h with medium containing 50 ng·mL−1 EGF and 1% fetal bovine serum in the presence or absence of 5 µm GSK3β inhibitor (SB-216763) (SB). The mRNA for the β-catenin target Axin2, a perivenous marker, was quantified by real-time RT-PCR. Means ± SD of four independent experiments are given relative to cells incubated with 1% fetal bovine serum. (B) Cells were incubated for 48 h with medium containing 1% or 10% fetal bovine serum in the presence or absence of 10 µm ERK inhibitor (U0126) or with 1% fetal bovine serum ± 50 ng·mL−1 EGF in the presence or absence of 10 µm U0126. Axin2 mRNA was quantified by real-time RT-PCR. Means ± SD of six (fetal bovine serum) and four (EGF) experiments are given relative to the values obtained without U0126 treatment. Significant effects (P < 0.05, Student's t-test) are indicated by (*).

Discussion

According to our previous hypothesis, two opposing signaling pathways regulate the zonation of gene expression in mouse liver. A signal presumably delivered by the endothelial cells of the central veins was proposed to activate β-catenin-dependent transcription in perivenous hepatocytes, whereas a second signal, delivered by bloodborne molecules activating Ras-dependent downstream cascades, was postulated to induce the expression of ‘periportal’ genes [6].

Our present data obtained with cultured hepatocytes and mouse hepatoma cell lines demonstrate that high serum concentrations are able to induce the expression of at least a subset of periportal markers, and, interestingly, are also able to diminish the stimulatory effect of β-catenin on expression of perivenous marker genes and to inhibit β-catenin/TCF-driven luciferase reporter activity. Therefore, one or more factor(s) appear(s) to exist in serum that may silence β-catenin/TCF-mediated transcription in liver. On the other hand, β-catenin signaling seems to be able to repress the transcription of certain ‘periportal’ genes, as demonstrated by the decrease of periportal marker mRNAs, e.g. Cdh1 (encoding E-cadherin) and Rex3, in response to treatment with an inhibitor of GSK3β. In summary, these findings support our hypothesis that two signaling pathways − a β-catenin-dependent pathway and a signaling cascade activated by serum factors − regulate gene expression along the portocentral axis in mouse liver by antagonistic interaction. Our present findings, demonstrating (a) inhibition of perivenous marker protein expression by activation of the Ras/ERK cascade in transgenic mice and liver tumors, and (b) physiologic activation of ERK in periportal hepatocytes, are in support of our hypothesis and suggest a significant contribution of Ras/ERK signaling in the regulation of zonal gene expression in mouse liver. The observed preferential activation of Ras/ERK signaling in the periportal areas of the liver lobule is in accordance with previous reports from other groups describing a higher concentration and enhanced ligand affinity of EGF receptor in hepatocytes surrounding a branch of the portal vein, thus leading to predominantly periportal EGF binding to the cells [21–25].

Indirect evidence suggests that serum-dependent attenuation of β-catenin signaling may be caused through activation of Ras. Immunohistochemical stainings showed that (a) hepatocytes of transgenic mice carrying an activated form of Ha-ras and (b) Ha-ras or B-raf mutated mouse liver tumor cells never express GS, even if they are located directly adjacent to central veins. This is also true for certain cytochromes P450, such as Cyp2E1 and Cyp1A isoforms. The inhibitory effect of an activated version of Ha-ras on aryl hydrocarbon receptor function and Cyp1A1 expression in mammary carcinoma cells and keratinocytes [17] further argues for this idea. Additional evidence for an attenuating effect of Ras on β-catenin-dependent signaling comes from the comparison of gene expression profiles of Ctnnb1 and Ha-ras mutated liver tumors. Often, mRNAs that are abundantly expressed in Ctnnb1 mutated hepatomas appear to be downregulated in Ha-ras mutated liver tumors [10,14] and, in a very similar manner, in B-raf mutated tumors [18], indicating repression of β-catenin target genes in tumors with constitutively active Ras/Raf/ERK signaling. However, in contrast to studies suggesting antagonism of signaling by Ha-ras and β-catenin in mouse hepatocytes (this work) and liver tumors [14], synergism of both pathways in liver tumorigenesis has been reported in double-transgenic mice expressing both mutated Ha-ras and activated β-catenin [11]. This apparent discrepancy might well be explained by differences in the experimental settings employed in the different studies. In the case of the chemically induced hepatocellular adenomas, the endogenous Ctnnb1 gene carries point mutations resulting in expression of an activated version of β-catenin. The intracellular concentration of the oncoprotein is not significantly increased in the mutated tumor cells, and neither cytoplasmic nor nuclear staining is detectable by immunohistochemistry [19,20]. By contrast, the liver tumors from the double-transgenic mice express a mutated version of β-catenin with an N-terminal deletion of its regulatory domain. In the tumor cells, comparatively strong expression of β-catenin is detectable by immunohistochemistry, with cytoplasmic and particularly nuclear localization [11]. The same tumors are Ras-positive and demonstrate scattered phospho-ERK staining, indicating activation of Ras signaling in the hepatoma cells (not shown). We speculate that the very high levels of nuclear β-catenin seen under the latter conditions cannot be counterbalanced by the activated Ras/ERK pathway, in contrast to what happens in hepatoma cells harboring only comparatively low levels of the endogenous form of the activated oncoprotein.

According to our data, EGF is capable of mediating the attenuation of the perivenous differentiation program in 70.4 mouse hepatoma cells by activating the Ras/ERK signaling pathway. This is in agreement with observations by others showing that EGF is able to suppress the expression of various ‘perivenous’ Cyp isoforms such as Cyp1a1 and Cyp2b1 in rat liver [26–28]. However, as the inhibition of ERK failed to completely block the inhibitory effect of serum on β-catenin signaling, other serum factors, different from EGF and acting through signaling pathways other than the EGF/Ras/ERK cascade, also seem to play a role in mediating the ‘serum effect’. The fact that serum attenuates Axin2 expression in the mouse hepatoma cell lines 55.1c, PW53T, and Hepa1c1c7, whose ERK signaling pathway is constitutively activated by mutations in either the Ha-ras or B-raf proto-oncogene, strengthens this hypothesis. Further evidence for this idea comes from studies relating the expression of certain zonated markers to other serum factors. For example, the periportal repression of glutathione S-transferase-α was linked to triiodothyronine [29], whereas zonation of Cyp2b1/2b2 and Cyp2c7 seems to be linked to pituitary gland-derived growth hormone [30–32]. These pathways may converge with the Ras pathway further downstream, an idea that has already been discussed in the case of Cyp2b1, where the negative regulation of gene expression by both EGF and growth hormone is mediated via a common distal enhancer region [33]. Thus, although activation of Ras is involved in attenuating perivenous gene expression, the nature of the ultimate effector molecule(s) that potentially interfere(s) with β-catenin signaling remains unknown. A candidate could be Icat/Catnbip, which was found to be preferentially expressed in periportal hepatocytes [13] and is overexpressed in Ha-ras and B-raf mutated liver tumors (own unpublished observation). Icat/Catnbip inhibits the formation of a β-catenin–TCF-4 complex, and thereby negatively regulates β-catenin-mediated gene transcription [34].

Our present data are in favor of the idea that two opposing signals regulate zonal gene expression in mouse liver [6]. Several lines of evidence clearly demonstrate that signal transduction via β-catenin is necessary for induction of the ‘perivenous’ hepatocyte phenotype [6–9]. The second signal, attenuating β-catenin-dependent signal transduction and inducing the ‘periportal’ phenotype, seems to be − at least in part − transduced via the Ras/Raf/ERK pathway to an unknown downstream target that is also affected by other serum-stimulated signaling pathways independent from the Ras cascade. Thus, additional research is required to further elucidate the downstream signal transduction mechanisms that dictate the ‘periportal’ phenotype of hepatocytes and account for the inhibition of β-catenin signaling in the periportal zone of the liver lobule.

Experimental procedures

Tissue samples

Transgenic Tglox(pA)H–ras* mice expressing a mutated Ha-ras gene were generated as previously described [11]. Livers were isolated at 11 weeks of age, 4 weeks after adenoviral cre infection (108 plaque-forming units per mouse). Mouse liver tumors were available to us from an earlier experiment conducted in our laboratory [35]. Tumors had been previously characterized in terms of mutations in the proto-oncogenes Ha-ras, B-raf and Ctnnb1[19,35]. All tissues were fixed with paraformaldehyde. Animals received humane care, and protocols complied with institutional guidelines.

Immunohistochemical staining

Immunostaining was performed using standard protocols as recently described [6]. In brief, 5-µm-thick sections of paraffin-embedded tissue samples were stained using antibodies against GS (1 : 1000 dilution; Sigma, Taufkirchen, Germany), Cyp1A (1 : 1000; gift of R. Wolf, Biochemical Research Centre, University of Dundee, Dundee, UK), Cyp2E1 (1 : 500; Stressgene, Victoria, Canada) and a horseradish peroxidase-conjugated secondary antibody (1 : 20; Dako, Glostrup, Denmark) with 3-amino-9-ethylcarbazole and H2O2 as substrates. For staining of active ERK, a phospho-specific ERK1/ERK2 antibody (1 : 1000; Cell Signaling, Danvers, MA, USA) was used in combination with a biotin-conjugated secondary antibody (1 : 200; Spa, Milan, Italy) and alkaline phosphatase-conjugated streptavidin (1 : 200; Spa) with Fastred (Kementec, Copenhagen, Denmark) as substrate, alone or in combination with GS staining. Nuclei were counterstained by hematoxylin.

Isolation and cultivation of hepatocytes

Hepatocytes were isolated by standard collagenase perfusion and cultivated in DMEM/F-12 medium containing 100 U·mL−1 penicillin and 100 µg·mL−1 streptomycin and different concentrations of fetal bovine serum. In some experiments, hepatocytes were incubated with 5 µm GSK3β inhibitor (SB-216763) (Sigma). Culture medium was changed daily. All experiments were performed in quadruplicate.

Cultivation of mouse hepatoma cell lines

Mouse hepatoma cell lines 70.4, 53.2b, 55.1c [36], PW53T and Hepa1c1c7 were characterized in terms of mutations in the Ha-ras, B-raf and Ctnnb1 proto-oncogenes as previously described [19,35]. Cells were incubated in DMEM/F-12 medium containing the antibiotics mentioned above and different fetal bovine serum concentrations. In one experiment, fresh frozen human plasma (gift of I. Mueller, University Children's Hospital, Tuebingen) was used instead of fetal bovine serum. In this experimental setup, medium was supplemented with 1 U·mL−1 heparin. Culture medium was changed daily. In some experiments, cells of line 70.4 were treated with 5 µm SB-216763, 20 nm tri-iodothyronine, 10 nm dexamethasone, 1 µm retinoic acid, 10 ng·µL−1 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene, 100 nm insulin, 50 ng·mL−1 EGF (all purchased from Sigma), 20 ng·mL−1 hepatocyte growth factor (R&D Systems, Wiesbaden, Germany), and 10 µm ERK inhibitor (U0126) (Promega, Mannheim, Germany).

Western blotting

Whole cell lysates were prepared as previously described [13]. Protein concentrations were estimated using the Bradford assay. Western analysis was carried out as previously described [13] using antibodies against phospho-ERK1/2 (1 : 1000; Cell Signaling), Axin2/Conductin (1 : 1000; Santa Cruz Biotechnology, Santa Cruz, CA, USA), β-catenin (1 : 500; BD, Heidelberg, Germany) and glyceraldehyde-3-phosphate dehydrogenase (1 : 1000; Chemicon, Chandler's Ford, Hampshire, UK), in combination with appropriate alkaline phosphatase-conjugated secondary antibodies (1 : 10000; Tropix, Darmstadt, Germany) with CDP-star as a substrate. Chemoluminescence signals were monitored by use of a CCD camera system.

Quantitative determination of mRNAs by RT-PCR

Total RNA from cultured hepatocytes or hepatoma cell lines was isolated using Trizol (Invitrogen, Karlsruhe, Germany) and reverse transcribed by avian myeloblastosis virus reverse transcriptase (Peqlab, Erlangen, Germany) as previously described [6]. Expression analysis was performed using the LightCycler real-time PCR system (Roche, Mannheim, Germany). Expression of 18S rRNA was used for normalization. The primer pairs used for PCR amplification are given in Table 1.

Table 1.   Sequences of PCR primers used in this study.
TranscriptForward (5′- to 3′)Reverse (5′- to 3′)
Aldh1b1GACCGGAGAACGCTGATACTAGAGGGATTGGGTTCGGGAGA
Axin2CGACGCACTGACCGACGATTTCCAGACTATGGCGGCTTTCC
CathCGAAGTTCCCGAAGCGACATTACACCTTCTTGCCAACAAAGC
Cdh1TCTACCAAAGTGACGCTGAAGCTGATGGGAGGGATGAC
Gpr49AATCGCGGTAGTGGACATTCGATTCGGAAGCAAAAATGGA
H19CTGCTGCTCTCTGGATCCTCGTGGGTGGGTGCTATGAGTC
Hes1GACTGTGAAGCACCTCCGGGGTAGGTCATGGCGTTGATC
Rex3AGGAGGAAGAGCGGAGCAAAGCTGGTAACAGGGAGAGATC
18S rRNACGGCTACCACATCCAAGGAAGCTGGAATTACCGCGGCT

Stable transfection and luciferase assay

Mouse hepatoma cells from cell line 70.4 were transfected with the 8xTCF/LEF-driven luciferase reporter SuperTopflash in combination with pSV2neo (BD) using Lipofectamine 2000 (Invitrogen). After 2–3 weeks of selection with 400 µg·mL−1 G418 (Biochrom, Berlin, Germany), stable transfectants were obtained and screened for basal and GSK3β inhibitor-inducible luciferase activity. Firefly luciferase activity was determined using the Dual Luciferase Kit (Promega, Madison, WI, USA) according to the manufacturer's instructions. Luciferase activity values were normalized to protein contents of the lysates using the Bradford assay.

Statistical analysis

For statistical analysis, Student's t-test was used. Differences were considered significant when P < 0.05.

Acknowledgements

We acknowledge the excellent technical assistance given by Elke Zabinsky and Silvia Vetter. We also thank Dr R. Wolf (Dundee, UK) for gift of Cyp1A antibody and Dr I. Mueller (Tuebingen, Germany) for gift of human plasma. This study was supported by the Deutsche Krebshilfe (grant 106356).

Ancillary