Inherent growth advantage of (pre)malignant hepatocytes associated with nuclear translocation of pro-transforming growth factor α

The pro-peptide of transforming growth factor α (proTGFα) was recently found in hepatocyte nuclei preparing for DNA replication, which suggests a role of nuclear proTGFα for mitogenic signalling. This study investigates whether the nuclear occurrence of the pro-peptide is involved in the altered growth regulation of (pre)malignant hepatocytes. In human hepatocarcinogenesis, the incidence of proTGFα-positive and replicating nuclei gradually increased from normal liver, to dysplastic nodules, to hepatocellular carcinoma. ProTGFα-positive nuclei almost always were in DNA synthesis. Also, in rat hepatocarcinogenesis, proTGFα-positive nuclei occurred in (pre)malignant hepatocytes at significantly higher incidences than in unaltered hepatocytes. For functional studies unaltered (GSTp−) and premalignant (GSTp+) rat hepatocytes were isolated by collagenase perfusion and cultivated. Again, DNA synthesis occurred almost exclusively in proTGFα-positive nuclei. GSTp+ hepatocytes showed an ∼3-fold higher frequency of proTGFα-positive nuclei and DNA replication than GSTp− cells. Treatment of cultures with the mitogen cyproterone acetate (CPA) elevated the incidence of proTGFα-positive nuclei and DNA synthesis in parallel. Conversely, transforming growth factor β1 (TGFβ1) lowered both. These effects of CPA and TGFβ1 were significantly more pronounced in GSTp+ than in GSTp− hepatocytes. In conclusion, nuclear translocation of proTGFα increases in the course of hepatocarcinogenesis and appears to be involved in the inherent growth advantage of (pre)malignant hepatocytes.

Hepatocellular carcinoma (HCC) is one of the most common cancers worldwide, particularly in Asia and Africa, accounting for about 1 million deaths per year (Parkin et al, 2001). Recently, its incidence has substantially increased in Europe and the United States (El-Serag and Mason, 1999;Parkin et al, 2001). Transforming growth factor a (TGFa) is one of the cytokines, causally involved in the pathogenesis of liver cancer (Kiss et al, 1997;Grisham, 2001). It is produced by hepatocytes and nonparenchymal liver cells for paracrine, autocrine and/or juxtacrine stimulation, as shown in developing, regenerating, preneoplastic and neoplastic livers of rodents and humans (Mead and Fausto, 1989;Miller et al, 1995;Grisham, 1997;Kiss et al, 1997;Yarden and Sliwkowski, 2001). Transforming growth factor a appears to be upregulated in all stages of liver cancer development. Single hepatocytes, infected with the hepatitis-B virus, overexpress TGFa due to transactivation of the TGFa gene by the virus (Schirmacher et al, 1996). Hepatocellular adenomas and carcinomas as well as childhood hepatoblastomas have been found to be rich in TGFa, leading to elevated plasma levels of this cytokine (Yamaguchi et al, 1995;Kiss et al, 1997;Grisham, 2001).
According to textbook knowledge, TGFa is produced as a precursor transmembrane molecule (proTGFa). The ectodomain of the pro-peptide may be shed from the cell surface, where it may bind to and activate the erbb-1 receptor, that confers the growth signal via phosphorylation cascades to the nucleus (Massagué, 1990;Yarden and Sliwkowski, 2001). Considering the upregulation of TGFa in human malignancies, including liver cancer, hope focuses on the possible therapeutic benefit of blocking TGFaevoked signal transduction on the cell surface, for example, by blockade of the receptor or of ligand -receptor interactions (Levitzki and Gazit, 1995;Mendelsohn, 1997). In a recent study, however, we have shown that hepatocytes in the intact liver and in primary culture synthesise proTGFa that translocates to the nucleus, where it appears to be involved in the mitogenic response of the cell (Grasl-Kraupp et al, 2002). This proposed novel pathway was induced by various different growth stimuli and is active in three different mammalian species, including humans. In mouse hepatocytes, almost all of the proTGFa-pos nuclei were also positive for erbb-1 (Schausberger et al, 2003). Moreover, several very recent papers suggest that the erbb receptors 1, 3 and 4 may bypass the protein phosphorylation cascades for transducing mitogenic stimuli (Lin et al, 2001;Ni et al, 2001;Offterdinger et al, 2002). Thus, there is considerable evidence of a direct action of growth factors/growth factor receptors from the EGF/erbbreceptor family in the nucleus (Wells and Marti, 2002). The question emerges whether the nuclear occurrence of proTGFa is involved in the altered growth regulation of (pre)malignant cells.
Rodent liver provides excellent tools for functional studies on hepatocarcinogenesis (Pitot, 1990;Grasl-Kraupp et al, 1997;Grisham, 1997). Treatment of rats with genotoxic carcinogens, such as N-nitrosomorpholine (NNM), induces single initiated hepatocytes that are detectable by their selective immunoreactivity for placental glutathione-S-transferase (GSTp þ cells); a considerable fraction of these cells develops to GSTp þ (pre)malignancy (Grasl-Kraupp et al, 2000). Although human HCC often do not express GSTp due to epigenetic silencing of the gene, GSTp þ lesions of rats and (pre)malignant lesions in human liver show significant similarities, such as mutations of the wnt-pathway, overexpression of TGFa, IGF-I and -II and other growth factors (Miller et al, 1995;Grisham, 1997;Yamada et al, 1999;Yang et al, 2003). Rates of replication and death of GSTp þ cells, and thus overall cell turnover, are somewhat reduced in the single-cell stage, but are significantly increased from the two-cell stage onwards (Grasl-Kraupp et al, 1997. Thus, initiation causes a change in the growth-regulatory network that becomes evident after the first replication cylce of GSTp þ cells. Treatment with tumour promoters, such as the progestin cyproterone acetate (CPA), or increased food intake further increases cell replication in preneoplasia, which is analogous to human hepatocarcinogenesis driven by anabolic steroids or overnutrition (Schulte-Hermann et al, 1990;Grasl-Kraupp et al, 1994;Fiel et al, 1996;Nair et al, 2002). This enhanced sensitivity of liver preneoplasia towards the various growth stimuli could result from an altered uptake, production and/or processing of endogeneous growth regulatory factors by the premalignant cell compartment.
In the present study, we asked whether the nuclear occurrence of proTGFá is involved (i) in the altered growth regulation of (pre)malignant hepatocytes and (ii) in the enhanced sensitivity of these cells towards induction of DNA replication by known growth stimulators. Recently, preneoplastic rat liver cells have become available for investigation in an ex vivo culture model (Löw-Baselli et al, 2000b). We applied this model, in combination with studies on human livers, and found that proTGFa-positive nuclei increased in the course of hepatocarcinogenesis. The possible role of nuclear proTGFa for the growth advantage of (pre)malignant cells is discussed with regard to tumour-therapeutic strategies targeted at TGFa/erbb-1 interactions on the cell surface.

Human liver samples
Patients suffering from dysplastic liver nodules (n ¼ 9), hepatocellular adenoma (n ¼ 3), or HCC (n ¼ 10) were resected; chemotherapy had not been applied before surgery. Tissue samples were immediately fixed in 10% buffered formaldehyde. Classification of liver lesions and stage of disease followed published guidelines (Edmondson and Steiner, 1954;Hermank et al, 1993;International Working Party, 1995). For further details, see Table 1. Informed consent was obtained from all patients.

Human hepatocyte and hepatoma cell lines
The human hepatoma cell lines HepG2 (ATCC-No HB-8065), Hep 2B2.1 -7. , and WRL 68  were obtained from the American Type Culture Collection (Rockville, MD, USA). The cells were maintained in vitro at 371C and 5% CO 2 in Dulbecco's minimum essential medium (DMEM) supplemented with 5% foetal calf serum, 100 U ml À1 penicillin and 100 mg ml À1 streptomycin (all obtained from Gibco, Life Technologies Inc., Gaitherburg, MD, USA). Once per week cells were passaged at a seed density of 1 Â 10 6 cells per 25 cm 2 plate.

Animals and treatment
Male SPF Wistar rats, about 3 weeks old, were obtained from the Institut für Versuchstierkunde und Genetik (Himberg, Austria). Animals were kept under standardised conditions and were fed powder diet (Altromin 1321N, Altromin, Lage, FRG). At 3 weeks before treatment, animals were adapted to rhythmic feeding (from 0900 to 1400 h). This procedure synchronises DNA synthesis in the liver to a single peak per day (Grasl-Kraupp et al, 1994. After adaptation, animals were treated with a single dose of NNM (Sigma, St Louis, MO, USA; 250 mg per 10 ml phosphate-buffered saline per kg body weight). Phenobarbital (PB) was admixed to the powder diet and was fed to a subgroup of rats from day 4 to 17 months post-NNM. Concentrations of PB were adjusted every 14 days to provide a daily dose of 50 mg kg À1 body weight (Löw-Baselli et al, 2000b). Animals were killed by decapitation under CO 2 asphyxation. For further details, see Grasl-Kraupp et al (1994 and Löw-Baselli et al (2000a). All experiments were performed according to the 'Austrian Guidelines for Animal Care and Protection', which meet the standards required by the 'UKCCCR Guidelines for the Welfare of Animals in Experimental Neoplasia' (2nd edn.) (Workman et al, 1998).
Löw-Baselli et al, 2000a); two serial sections, 1 mm thick, were cut; one of the sections was stained for GSTp (rat only) or Ki-67 (human only), the second one was stained for TGFa.
Immunostaining for TGFa, GSTp and Ki-67 The primary antibodies used were rabbit polyclonal IgG against rat Yp-subunit of GSTp (Biotrin International, Dublin, Eire); mouse monoclonal IgG against recombinant mature TGFa encompassing amino acids 39 -88 (clone 213-4.4, Oncogene Science, Uniondale, NY, USA); mouse monoclonal IgG against a synthetic peptide encompassing amino acids 144 -160 of the C-terminus of proTGFa (Ab-3; InnoGenex, San Ramon, CA, USA); mouse monoclonal antibody against full-length recombinant Ki-67 protein (Dianova, Hamburg, FRG). For TGFa or Ki-67 staining, formaldehyde-fixed and de-waxed tissue sections were placed in a glass Coplin jar filled with 0.01 M sodium citrate buffer, pH 6.0. Slides were heated for periods of 2 min at a maximal power setting (about 800 W) and for 2 Â 2 min at the submaximal power setting (600 W). The citrate buffer reached boiling point within 2 min and the fluid level in the Coplin jar was topped up with distilled water between heating periods to prevent drying of the sections.

Determination of DNA synthesis
Rat liver 3 H-thymidine (6.7 Ci mmol À1 ; NEN, Frankfurt, FRG) was injected into the peritoneal cavity as a single dose of 0.2 mCi kg À1 body weight at the daily peak of DNA synthesis between 2000 and 2100 h (see above). After 36 h, animals were killed. GSTp-stained sections were coated with a solution of 1% gelatine (BioRad, Richmond, CA, USA) and 0.05% chromalaun (Merck, Darmstadt, FRG) and were air-dried. After autoradiography, the percentage of hepatocyte nuclei in DNA synthesis was determined for at least 1000 nuclei of unaltered cells in each liver and in all nucleated cells within individual GSTp þ foci (labelling index (LI)). Since interindividual variations were small, LIs obtained from different livers or foci of the same experimental group were pooled.
Human liver In order to identify human hepatocyte nuclei in the S-phase of the cell cycle, serial sections were stained for TGFa and Ki-67 (see above). Individual Ki-67-positive nuclei were followed in the consecutive TGFa-stained serial section by overlaying the two images in two microscopes linked by a bridge for overprojection (Zeiss, Germany).

Primary hepatocytes
Male SPF Wistar rats were obtained from the animal facilities of the Medical University of Vienna at the age of 3 -4 weeks and were treated with a single dose of NNM (250 mg kg À1 body weight), as described above. After 21 days, livers were perfused with collagenase as described (Parzefall et al, 1989;Löw-Baselli et al, 2000b).
Treatment of primary hepatocyte cultures Cells were seeded and kept under serum-free conditions, as described (Parzefall et al, 1989;Löw-Baselli et al, 2000b). Treatment commenced 4 h after plating (time point 0). A stock of 10 mg (10 ml) À1 of 10 mM acetic acid of human recombinant mature TGFa (UBI, Lake Placid, NY, USA) was prepared and was added to the medium for a final concentration of 10 ng ml À1 . Tyrphostin A25 (synonym: tyrphostin AG82; Calbiochem, La Jolla, CA, USA) was dissolved in dimethylsulphoxide (DMSO) to obtain a stock of 10 mg ml À1 ; 1 ml of this stock was added per ml medium. Cyproterone acetate, a gift from Schering AG (Berlin, FRG) was dissolved in DMSO. In all experiments, the final concentration of CPA was 10 mmol in 0.2% solvent. Recombinant mature TGFb1 synthesised by CHO transfectants was supplied gratuitously by Bristol-Myers Squibb (Seattle, WA, USA). TGFb1 was dissolved as described (Oberhammer et al, 1992).
Determination of DNA replication in cultured hepatocytes Immunohistochemically stained plates were coated with 1% gelatine/ 0.05% chromalaun, air-dried, dipped into photo-emulsion (Ilford K5, Dreieich, FRG), and exposed for about 14 h. The plates were processed with a photographical developer and fixative, and were finally dried at room temperature and mounted in Kayser's glycerine gelatine (Merck, Darmstadt, FRG). The LI was calculated as percentage of labelled hepatocyte nuclei per total number of hepatocyte nuclei counted.

Two-dimensional gel electrophoresis, immunoblotting and identification of protein spots
Nuclei and cytoplasm (postnuclear supernatant) of human hepatoma cells and of rat liver were separated according to the method of Tata, applying 2.0 M sucrose for purification (Tata, 1974). This was followed by a nuclear matrix preparation, as has been described in detail (Gerner et al, 1998).
For mass spectrometry fingerprinting, Coomassie Blue-stained proteins were directly cut out of preparative gels. Matrix-assisted laser desorption ionisation-time-of-flight (maldi-tof) of tryptic protein hydrolysates and protein identification were carried out essentially as described (Fountoulakis and Langen, 1997;Grasl-Kraupp et al, 2002). Proteins were considered identified by means of mass spectrometry fingerprinting when at least 15% of the whole sequence gave hits and when the molecular mass/pI values were identical to the ones calculated or published in 2D databases.

Statistics
If not stated otherwise, data of at least three animals per time point and treatment group are given. Where indicated, the significance of differences of means was calculated by Kruskal -Wallis test or Wilcoxon test. For incidences, confidence intervals were calculated for Po0.05.

RESULTS
The pro-peptide of TGFa is present in the nuclei of human and rat hepatocytes and human hepatoma cell lines In anti-TGFa-stained liver sections, immunoreaction was found within the cytoplasm, cell membranes and, most prominent, the nuclei of hepatocytes ( Figure 1B, C and E). Likewise, the nuclei of the hepatoma cell lines HepG2 (Figure 1F), WRL68 and Hep2B (not shown) displayed a strong immunoreaction. Nuclear matrices, prepared from human HepG2 cells (Figure 2A) and from rat liver ( Figure 2B), were separated by two-dimensional gel electrophoresis and were subsequently subjected to anti-TGFa immunoblotting; four spots around 17 kDa and an isoelectric point of 7.5 appeared. Spots 1 -4 of HepG2 cells were subjected to maldi-tof analysis: spot 1 was verified to be the wild-type form of proTGFa; spot 2 was recently found to be wild-type proTGFa as well (Grasl-Kraupp et al, 2002). Spots 3 and 4 could not yet be identified. Immunostaining with antisera against the C-terminus of the pro-peptide also confirmed the presence of proTGFa in hepatocyte nuclei (see below). The mature form of TGFa at about 5.6 kDa was not detected.

Occurrence of DNA replication and nuclear proTGFa in the different stages of human hepatocarcinogenesis
The percentage of both, replicating nuclei and proTGFa-pos nuclei, was gradually increased from normal liver, to hepatocellular adenoma, to dysplastic nodules, to HCC (Table 1, Figure 3). In unaltered liver, replicating nuclei and proTGFa-pos nuclei occurred at a similar frequency. Premalignant and malignant liver lesions, however, revealed a much higher percentage of proTGFapos nuclei when compared to nuclei in S-phase ( Figure 1E and 3). Staining for Ki-67 was used to identify hepatocyte nuclei in the Sphase of the cell cycle (for details, see Materials and methods); five HCC were studied: 77.379.2% of 343 Ki-67-positive hepatocyte nuclei evaluated displayed proTGFa. A high co-incidence of DNA synthesis and proTGFa was recently shown also for hepatocyte nuclei in cirrhotic liver (Grasl-Kraupp et al, 2002). Taken together, these findings indicate that a hepatocyte nucleus undergoing DNA replication almost always contains proTGFa. Further expression of proTGFa in nuclei not immediately undergoing DNA synthesis occurs predominantly in premalignant and malignant hepatocytes.

Occurrence of DNA replication and nuclear proTGFa in the different stages of rat hepatocarcinogenesis
Hepatocarcinogenesis was induced in rats by application of NNM. This leads to the formation of single GSTp þ cells and small preneoplastic GSTp þ lesions, followed by expansive growth of some of the lesions. To enhance the formation of liver tumours, rats were treated with the tumour promoter PB (Löw-Baselli et al, 2000b). Analogous to the findings in human liver, proTGFapositive nuclei occurred in (pre)malignant hepatocytes at a significantly higher incidence than in unaltered hepatocytes ( Figures 1A -C and 3). This substantiates that the experimental model fits the human situation.

Inherent growth advantage of cultured preneoplastic GSTp þ rat hepatocytes associated with nuclear proTGFa
To study the functional significance of nuclear proTGFa for hepatocarcinogenesis, unaltered GSTp À and preneoplastic GSTp þ cells were isolated from the livers by collagenase perfusion and cultivated. Cell isolation was performed on day 21 post-NNM treatment (no PB promotion). At this time point, there is the maximal occurrence of GSTp þ cell clones in the liver. Based on stereological calculations of the size distribution in the third dimension, only 23711% of the isolated GSTp þ cells in culture derive from single-cell clones and 77729% derive from small foci (three cells on average), with an inherently elevated cell turnover (DeGunst and Luebeck, 1998;Grasl-Kraupp et al, 2000).
Replicative DNA synthesis was generally high in the isolated hepatocytes ( Figure 4); this may be explained by the fact that the livers still underwent regeneration, as observed in vivo (Grasl-Kraupp et al, 2002). ProTGFa was present in the nuclei of about 10% of the hepatocytes in primary culture ( Figure 4 and Table 2). In any case, DNA replication occurred preferentially in proTGFapos nuclei in both GSTp À and GSTp þ hepatocytes (Table 2). This was also confirmed by immunostaining of parallel culture plates with two different antisera: DNA was synthesised by 76.576.8% of the nuclei being positive for amino acids 39 -88 of proTGFa (encompassing the mature form) and by 65.3719.2% of the nuclei being positive for amino acids 144 -160 of the C-terminus of the pro-peptide. In both stains, DNA synthesis in negative nuclei was a rare event (see also Table 2).
GSTp þ cells showed an about three-fold higher frequency of proTGFa-pos nuclei and DNA replication than GSTp À cells ( Figure 4 and Table 2). Thus, the inherent growth advantage of the cultured preneoplastic cell population appears to be highly associated with the enhanced nuclear translocation of proTGFa.

Different signal transduction pathways of TGFa gene products in GSTp À and GSTp þ hepatocytes
We have confirmed recently that in our system mature TGFa acts via erbb-1 in the cellular membrane of proTGFa-neg cells according to classical concepts of growth signal transduction (Grasl-Kraupp et al, 2002). Accordingly, in the present study, addition of mature TGFa increased DNA synthesis almost exclusively in the GSTp À cells that do not express proTGFa. This increase in DNA synthesis was blocked by the erbb-1-tyrosine kinase inhibitor tyrphostin A25 ( Figure 5). However, mature TGFa exerted no significant effect on any cell with nuclear proTGFa, which was most evident for the proTGFa-rich GSTp þ population ( Figure 5). At present, it is unclear why TGFa exerted no effect on the GSTp þ cells with proTGFa-neg nuclei. The expression of yet unidentified growth factors may confer autonomy from exogeneous growth stimuli to this subpopulation of premalignant cells. Taken together, these findings suggest that two different TGFamediated signal transduction pathways are operative in different cell populations: the 'classical' erbb-1-mediated pathway of mature TGFa becomes active mostly in the proTGFa-poor GSTp À hepatocytes, while the second one, triggered by nuclear proTGFa, is effective in the proTGFa-rich preneoplastic GSTp þ cells.
Transforming growth factor-b1 reduces DNA synthesis and nuclear proTGFa in cultured GSTp À and GSTp þ hepatocytes DNA replication was suppressed by 1 ng of TGFb1 and even more by 3 ng of TGFb1 ml À1 medium in GSTp À and GSTp þ cells ( Figure 6). DNA replication was inhibited at the most effective concentration of the cytokine in about 66% of the GSTp À cells, but in at least 90% of the GSTp þ cells. These data suggest that TGFb1 acts more strongly on GSTp þ than on GSTp À cells. Furthermore, TGFb1 diminished the fraction of hepatocytes expressing nuclear proTGFa and synthesising DNA ( Figure 6 and Table 2). Apparently, the suppression of DNA synthesis by TGFb1 in primary hepatocytes involves downregulation of nuclear proTGFa. This effect was evident for both, GSTp À and GSTp þ cells.   The hepatomitogen CPA induces DNA synthesis and nuclear proTGFa in GSTp À and GSTp þ hepatocytes Treatment with the hepatomitogen CPA doubled the LI in cultured GSTp À and GSTp þ cells (Figure 7). Thus, every fifth GSTp þ cell was stimulated to DNA synthesis by CPA, while only every tenth GSTp À cell was recruited to the pool of replicating hepatocytes. We asked whether nuclear proTGFa might be involved in this enhanced growth stimulation of the preneoplastic cell population. In fact, the CPA-induced increases in DNA synthesis were closely paralleled by an equal elevation of proTGFa-pos nuclei (Figure7 and Table 2). This may indicate that the enhanced occurrence of nuclear proTGFa in liver preneoplasia may commit the preneoplastic cells for preferential response towards growth stimulation.

DISCUSSION
The present paper describes that the pro-form of TGFa occurs in the nucleus of (pre)malignant rat and human hepatocytes, while mature TGFa was not detected. The translocation of proTGFa to the nucleus may contribute to the inherent growth advantage of early and late stages of hepatocarcinogenesis. The present observations extend recent findings on normal hepatocytes from rat, mouse and humans, which revealed that the translocation of proTGFa to the nucleus in G1 of the cell cycle is almost always followed by replicative DNA synthesis (Grasl-Kraupp et al, 2002). Nuclear translocation of proTGFa is induced by various hepatomitogenic stimuli, such as regenerative growth after partial hepatectomy or intoxication with CCl 4 and hyperplastic growth induced by CPA in the intact animal. In culture, the incidence of proTGFa-pos hepatocyte nuclei is elevated by treatment with hepatomitogenic CPA, prostaglandins E2 and F2a, and hepatocyte growth factor (Grasl-Kraupp et al, 2002;Schausberger et al, 2003). Thus, a great variety of growth stimuli all involve nuclear translocation of proTGFa. This peptide may therefore serve as a kind of intracellular shortcut in mediating autocrine growth stimulation of normal, premalignant and malignant liver cells.
The possible interactions of proTGFa with the growth-regulatory machinery in the nucleus of unaltered and (pre)neoplastic hepatocytes are still unclear. Soluble or membrane-bound precursors of TGFa are biologically active, suggesting that the pro-form may attach to the binding site of erbb-1 as known for the mature form (Ignotz et al, 1986). We found that proTGFa and erbb-1 almost always co-localise within the nucleus of mouse hepatocytes, as shown by confocal laser-scanning microscopy (Schausberger et al, 2003). It is currently under investigation by FRET technology whether the large TGFa precursor attaches to erbb-1 and may be co-targeted to the nucleus as a receptor-bound ligand. This may provide the clue for the function of this propeptide for DNA synthesis, considering that erbb-1 may act as a transcription factor for cyclin D1 (Lin et al, 2001). Our data also suggest that the activity of proTGFa in the nucleus does not depend on an erbb-1 receptor tyrosine kinase activity, since DNA replication of proTGFa-pos nuclei was not affected by the tyrosine kinase inhibitor tyrphostin ( Figure 5). The cell culture system used in the present study allows to investigate the functional significance of nuclear proTGFa in premalignant cells; cultured GSTp þ hepatocytes showed significantly higher basal rates of DNA replication than GSTp À hepatocytes. These characteristics closely reflect those described for GSTp þ cells in the intact liver in vivo. Thus, this defect in growth regulation persists under culture conditions and therefore appears to be independent of intercellular contacts within the intact organ, and of cytokines, growth factors or hormones circulating in the whole body. The present study shows that considerably more GSTp þ cells synthesise and transport proTGFa to the nucleus than GSTp À cells. The nuclear import of this propeptide almost always is followed by DNA replication. The enhanced probability of nuclear translocation of proTGFa may therefore be essential for the intrinsic growth advantage of the preneoplastic cell population. Compared to the current concepts of signal transduction, nuclear proTGFa may not depend on intact erbb-1 receptors on the cell surface. It circumvents the secretion and possible loss of TGFa to the outside of the cell and may confer autonomy and an inherent growth advantage, a pathway preferentially used by the preneoplastic cell population.
In an untreated, healthy liver, almost all of the hepatocytes are in the Go-phase of the cell cycle. In the present study, the incidence of nuclei positive for proTGFa increased in the course of hepatocarcinogenesis. Since proTGFa translocates to the nucleus in the G1-phase of the cell cycle (Grasl-Kraupp et al, 2002), the elevated presence of nuclear proTGFa in (pre)malignant hepatocytes may be evidence for a G1-status of these cells. Several groups reported that premalignant liver cells show an increased expression of c-myc and cyclin D1, known inducers of the transition from G0 to G1 of the cell cylce (Deguchi and Pitot, 1995;Ramljak et al, 2000). On the other hand, the expression of the WAF1/CIP1 gene product, p21 and the c-myc antagonist mad were decreased in hepatocarcinogenesis (Martens et al, 1996). Many growth-stimulating factors exert their activity, provided that the target cell is in the G1-phase of the cell cycle. Thus, the enhanced presence of nuclear proTGFa in the (pre)neoplastic cell compartment appears to be involved in the overcoming of critical checkpoints of the cell cycle and in a facilitated response of liver (pre)neoplasia towards various growth stimuli. Transforming growth factor a and erbb-1 are upregulated in malignancies of many different organs, including HCC. Novel therapeutic approaches have been focusing on the possible benefit of blocking TGFa-evoked signal transduction on the cell surface, for example, by erbb-1 blockade (Levitzki and Gazit, 1995;Mendelsohn, 1997). The present study shows that treatment with mature TGFa stimulated DNA synthesis rather in proTGFa-neg than in proTGFa-pos hepatocytes, which was abrogated by an erbb-1-specific tyrosine kinase inhibitor. In GSTp þ hepatocytes, however, mature TGFa and tyrphostin exerted no significant effect. Thus, the 'classical' signal transduction pathway of mature TGFa via erbb-1 seems to be active in the unaltered cell population, while the novel pathway seems to operate preferentially in the (pre)malignant cell compartment. It is therefore tempting to speculate that liver tumours may use alternative pathways for growth stimulation by proTGFa. Then, they may be resistant against therapeutic strategies targeted at TGFa/erbb-1 interactions on the cell surface.
In conclusion, the present work shows that the incidence of nuclei expressing proTGFa is elevated in the course of rat hepatocarcinogenesis, which may reflect and contribute to the inherent growth advantage of (pre)neoplastic hepatocytes. Further research is necessary to elucidate the mechanisms that regulate the different intracellular routes of proTGFa and that may link nuclear proTGFa to DNA replication.