We have used single- and double-label immunocytochemistry to examine the distribution of AGpllO, integrin α5β1 and fibronectin in adult rat liver during carcinogenesis induced by aflatoxin B2 or diethyl-nitrosamine. In normal liver fibronectin and the fibronectin integrin receptor α5β1 are localized on all three domains of the parenchymal cell surface: sinusoidal, lateral and canalicular. In contrast, AGpllO, a non-integrin monomeric glycoprotein with fibronectin receptor properties, is confined to the bile canalicular (apical) plasma membrane of hepatocytes. Hepatocarcinogenesis induced by aflatoxin B2 causes altered cell foci to form in the parenchyma, followed by enlargement of these foci to form pre-neoplastic nodules and finally hepatocellular carcinomas of either poorly differentiated, trabecular or adenocarcinoma morphology. Expression of AGpllO decreased to a minimal level, at first selectively in altered cell foci, from the 9th week of treatment, and then indiscriminately in poorly differentiated carcinomas. The same lesions that were deficient in AGpllO also displayed a reduced level of fibronectin and 25βi, although the observed change in AGpllO demarcated altered foci and poorly differentiated tumour lesions more sharply, since expression of α5β1 and fibronectin, though substantially reduced, was still faintly apparent on the cell surface. Small acinar structures, observed in late hyperplastic nodules and in trabecular carcino mas, exhibited even, pericellular staining of fibronectin and α5β1 including prominent staining of the lumen area, whereas staining of AGpllO appeared to be confined to the lumen. In larger ducts of overt adenocarcinomas, fibronectin and α5β1 were distributed along the basal surface of the epithelium and AGpllO on the apical domain. Tumours induced by diethylnitrosamine and promoted with ethinyl estradiol displayed similar histology and staining patterns for all three proteins as that described for aflatoxin B2. Finally, comparisons between AGpllO and cytokeratin 19, a selective tumour marker, indicated that whereas loss of AGpllO occurs in poorly differentiated lesions and tumours, expression of cytokeratin 19 is associated with acinar and glandular structures found in late hyperplasia and with trabecular and pseudoglandular tumours. The results indicate that loss of differentiation in either hyperplastic or neoplastic lesions correlates with reduced expression of fibronectin and of its receptors α5β1 and AGpllO. On the basis of morphological similarities and staining patterns in the pre-neoplastic and neoplastic state, we deduce that hepatocellular carcinomas derive from differentiated hepatocytes.

Experimental carcinogenesis in the liver is a multistage process involving a complex structural rearrangement of the parenchyma that results in either morphologically poorly differentiated tumours or in clearly distinct, pseudo-differentiated structures referred to as adenomas and trabecular carcinomas (Stewart and Williams, 1980; Williams, 1980). This transition in the cellular architecture of the liver can be triggered experimentally by chemical carcinogens such as aflatoxin B1 and diethylnitrosamine and is most probably mediated by alterations in cell-matrix and cell-cell adhesion mechanisms. Changes in the expression of extracellular matrix proteins such as fibronectin and laminin (Sell and Ruoslahti, 1982; Jagir-dar et al. 1985; Szendröi and Lapis, 1985) and of intracellular adhesion molecules such as cell CAM105 (Hixson et al. 1985) during hepatic carcinogenesis strongly implicate adhesive interactions in phenotypic neoplastic alterations. Modifications in the composition of the extracellular matrix are particularly significant in this respect, since they directly affect the metabolic activity of parenchymal cells: in primary cultures of rat hepatocytes synthesis of liver-specific proteins can be manipulated by inoculation on different extracellular matrix substrata (Sudhakaran et al. 1986; Reid et al. 1988; Ben-Ze’ev et al. 1988) . These cellular interactions with matrix components are mediated by specific receptors, most of which belong to the integrin family of macromolecules (Hynes, 1987), and transformation-associated changes in integrins have already been described in various cells (Buck et al. 1990; Dedhar, 1990; Plantefaber and Hynes, 1989; Virtanen et al. 1990).

In this study our aim has been to investigate changes in fibronectin and fibronectin receptors during hepatocarci-nogenesis. So far, two cell surface glycoproteins have been described that mediate adhesion of hepatocytes on fibronectin: integrin α5β1 (Johansson etal. 1987) and AGpllO, a non-integrin glycoprotein (Stamatoglou et al. 1990a). Under certain experimental conditions these two receptors may act in synergy (Stamatoglou et al. 19900). In this study we demonstrate that poorly differentiated hepatic tumours are deficient in both receptors and fibronectin whereas adenomas maintain a polarized expression of these proteins. Absence of AGpllO was found to be the most distinct marker for poorly differentiated hepatomas and presence of cytokeratin 19 in parenchyma cells demarcated pseudoglandular carcinomas (adenocarcinomas).

Chemicals

Aflatoxin B, was obtained from Makor Chemicals Inc. (Jerusalem, Israel). Diethylnytrosamyne was from Merck (Dagenham, Essex, UK). Ethinyl estradiol, Naphthol AS-BI phosphate, Fast Red TR, ethyl carbazole, BCIP, NBT and Accustain (Gills’s haematoxylin No. 2) were purchased from Sigma Chemical Co. (Poole, Dorset, UK). Apathy’s mounting medium was from BDH (Poole, Dorset, UK).

Antibodies

Antisera against rat AGpllO and fibronectin were raised in rabbits as described (Stamatoglou et al. 1990a). Mouse monoclonal against the heparin-binding domain of rat fibronectin was a generous gift from Dr R. O. Hynes (MIT, MA, USA); this was provided as culture medium and IgG was isolated on a protein A column as described (Stamatoglou et al. l990b). Affinity-purified hen IgG against α5β1 was kindly donated by Dr Staffan Johansson (Biomedical Centre, Uppsala, Sweden). This antiserum recognizes both the α5and β1 subunits of fibronectin receptor but, since α5 is far more abundant than other alpha subunits in rat liver (Stamatoglou et al. 1990a), we assumed that α5β1 was the predominant β1 integrin immunolocalized with the antiserum. Mouse monoclonal against cytokeratin 19 was a kind gift from Dr E. B. Lane (University of Dundee, UK). Goat anti-rabbit IgG/alkaline phosphatase (2.4mgml-1), rabbit anti-chicken IgG/alkaline phosphatase and extravidin/peroxidase (2mgml-1) were purchased from Sigma Chemical Co. (Poole, Dorset, UK). ABC kits (‘standard’ grade for rabbit antibodies for use with either alkaline phosphatase or peroxidase and ‘elite’ grade for mouse monoclonals with peroxidase detection) were purchased from Vector (Peterborough, UK).

Carcinogenesis protocols

Male Fischer 344 rats, 6-8 weeks of age, were fed a diet of powdered MRC 41B, contaminated artificially with 2 p.p.m. (parts per million) aflatoxin B2, for times varying from 2 weeks to 10 months.

Diethylntrosamne was administered to young adult male Sprague-Dawley rats as a single i.p. injection (200 mg kg-1 body weight). After 2 weeks the animals were treated with ethinyl estradiol incorporated into the diet at 10 p.p.m. until death. After 1 month from the first diethylnitrosamine injection animals received either a weekly injection of diethylnitrosamine (80 mg kg-1 body weight) for 4 weeks or a weekly injection (40 mg kg-1 body injection) for 8 weeks. Animals were killed 9-12 months after the beginning of treatment.

Immunocytochemistry

Specimens were fixed in ice-cold acetone and embedded in paraffin wax. Sections were dewaxed in xylene, equilibrated in an ethanol series of decreasing concentrations and hydrated in distilled water. Quenching of endogenous peroxidase or alkaline phosphatase was achieved by incubating for 30 min in 0.3 % H2O2or by 15 min in 15% acetic acid, respectively. Washes and antibody dilutions were in phosphate-buffered saline, pH 7.5, containing 0.05% Tween 80. After endogenous enzyme quenching, sections were incubated in this buffer for 30 min and then in 1 % non-immune goat (or rabbit) serum for 30 min.

AGp11O staining

This was performed by incubating sections for lh at room temperature with anti-AGp11O (1:200), washing and then incubating with goat anti-rabbit IgG/alkaline phosphatase. Sections were developed with Fast Red (O.5mgml-1) and Napthol AS-BI phosphate (O.5mgml-1) in 50mM veronal acetate buffer, pH9.2. Sections were counterstained by immersion in haematoxylin for 20 s and mounted in Apathy’s mounting medium.

Integrin α5β1 staining

This was accomplished by incubation with anti-α5β1 IgG (l5µgml-1) followed by rabbit anti-chicken/ alkaline phosphatase. Developing and counterstaining were as described above.

Fibronectin single label

Polyclonal anti-fibronectin (1:200) was used, followed by goat anti-rabbit/phosphatase. Identical results were obtained with monoclonal anti-fibronectin (100 µg IgG ml-1) using the ‘elite’ ABC reagents from Vector.

AGp110/fibronectin double label

Vector ABC reagents were used according to the manufacturer’s specifications. AGpllO was detected using our anti-AGpllO serum and the ABC/alkaline phosphatase kit. For fibronectin we used the mouse monoclonal antiserum and the Vector ‘elite’ peroxidase kit. In outline the sections were first processed for AGpllO (1:200), then with anti-rabbit/biotin and finally with ABC reagent (freshly prepared complex of biotin and streptavidin/phosphatase). Development was in a substrate solution prepared by adding 33 µl of BCIP (50 mg ml-1 in dimethyl formamide) and 66 µl of NBT (75 mg ml-1 in 70% dimethyl formamide) in 10 ml of 100 mM Tris-HCI, pH 9.5, 100 mM NaCl, 5 mM MgCl2. Sections were then processed for fibronectin staining, using mouse monoclonal IgG (100 µgml-1), following the same order as in the procedure above. The ABC reagent contained avidin linked to peroxidase and the developing solution was 10 ml of 20mM acetate buffer, pH 4.0, to which 50 µl of carbazole (20 mg ml-1 stock in DMSO) and 30 µl of 30 % H2O2 were added. No counterstain was used.

AGp110/α5β1 double label

AGpl10 was localized using the Vector ABC alkaline phosphatase kit as described in the previous section. Integrin α5β1 immunostaining was achieved by incubating first with normal rabbit serum, then with anti-α5β1 (l5µgml-1), followed by rabbit anti-chicken IgG/biotin (1:500) and, finally, with extravidin/peroxidase (1:500). Development was as described for AGp11O/fibronectin labelling.

AGpl 10/cytokeratin 19 double label

This is described in detail elsewhere (Green and Manson, 1991). Briefly, sections were incubated with anti-AGp11O and anti-ckl9 and then with anti-rabbit/peroxidase and anti-mouse/alkaline phosphatase. Development was as in AGp110/Fn double label.

Animals fed aflatoxin B3 were killed at different times after initiation of dietary administration (2 weeks to 10 months). Animals subjected to the diethylnitrosamine regimen were killed 9-12 months after the first i.p. injection.

AGpllO

In adult rat liver AGpllO was found in canalicular plasma membranes (Fig. la) as previously described (Stamato-glou et al. 1990a,b). The expression and distribution of this glycoprotein was drastically altered during carcinogenesis (Fig. lb–f). Approximately 8–9 weeks after initiation of the aflatoxin B3 regimen, altered cell foci with markedly diminished expression of AGpllO were noted (Fig. lb). Such foci were characterized by parenchymal disorganization, closer packing of cells and augmentation of cell size (Fig. lb and c). Continuation of aflatoxin treatment over longer periods of time resulted in an increase in the number and size of these hyperplastic lesions that lacked significant amounts of AGpllO (Fig. lc: 24 weeks of aflatoxin in the diet). At the same time, however, from the 5th week onwards, we occasionally observed increased expression in periportal areas, often manifested as pericellular staining (Fig. Id). The cells that exhibited this apparently non-polarized staining were closely packed together with no discernible plate structure but appeared indistinguishable from hepatocytes in the surrounding parenchyma and did not have the appearance of oval cells. In poorly differentiated hepatic tumours AGpllO was virtually absent (Fig. le). The boundaries of such tumours were distinctly demarcated by the lack of AGpllO stain from adjacent morphologically normal parenchyma, as was evident in areas of normal tissue being invaded by carcinoma (Fig. le, arrows). Pseudoglan-dular, adenoma-like tumours exhibited apical membrane staining only (Fig. If, arrows). The intensity of that apical stain varied, but mostly appeared weaker than canalicular stain in normal liver. In tumours induced by diethylnitro-samine AGpllO was similarly lacking in poorly differentiated carcinomas but persisted on the apical domain of pseudoglandular hepatocarcinomas, as described for aflatoxin-induced tumours (Fig. le and f). Ductular cholangio-mas in either aflatoxin B, or diethylnitrosamine-induced tumours showed variable expression: most were positive on the apical domain, but occasionally minimal staining was observed. Proliferating bile duct cells appeared negative (results not shown).

Fibronectin

Normal liver was intensely stained for fibronectin (Fig. 2al), this matrix protein being particularly prominent on sinusoidal cell surfaces. As previously documented, fibronectin can also be detected on canalicular and lateral surfaces of hepatocytes (Hughes and Stama-toglou, 1987; Enrich et al. 1988), but this is more evident using immunofluorescence on frozen sections (Hughes and Stamatoglou, 1987) rather than enzyme immunocytochemistry on paraffin wax-embedded tissue sections (Fig. 2al). Strong staining was also observed in the cytoplasm of control parenchymal cells (Fig. 2al; Hughes and Stamatoglou, 1987) in accordance with results indicating that hepatocytes are the main source of plasma fibronectin (Tamkun and Hynes, 1983). During aflatoxin administration, an overall decline in cytoplasmic staining was observed 3–4 weeks after the initiation of the carcinogenesis regimen and, additionally, marked reductions in both cell surface and cytoplasmic fibronectin occurred selectively in altered foci after the 8th-9th week of treatment (Fig. 2a2: 22 weeks of aflatoxin in the diet). In poorly differentiated hepatocellular carcinomas (Fig. 2a) the expression of fibronectin was similarly reduced but cell surface staining was frequently conspicuous, particularly in areas that maintained the normal liver cord appearance (Fig. 2a, arrows). Pseudoglandular hepatocellular carci nomas were invariably weakly positive for fibronectin along the basal surface of the cells lining the lumen (Fig. 2a4, arrowheads) but pericellular staining could, occasionally, be detected (Fig. 2a4, arrows).

Fig. 1.

Distribution of AGpllO in rat liver during AFB3-induced hepatocarcinogenesis. Paraffin-embedded acetone-fixed liver sections from rats fed AFB3 for various times were stained with anti-AGpllO serum followed by secondary antiserum conjugated to alkaline phosphatase (see Materials and methods). Red deposit of alkaline phosphatase substrate shows the localization of AGpllO. (a) Control male Fischer rat. AGpllO is confined to bile canaliculi. (b) 9 weeks of AFB3 at 2 p.p.m. in the diet. Decreased expression of AGpllO in some altered cell foci (central, delineated area) but normal distribution in most areas, (c) 24 weeks of treatment. Pre-neoplastic lesion with markedly reduced expression of AGpllO. (d) Treatment as in c. Periportal region showing increased expression of AGpllO and apparent pericellular, non-polarized localization of the protein, (e) 10 months of AFB1 at 1 p.p.m. in the diet, followed by 2 months on control diet. Non-differentiated hepatocellular carcinoma negative for AGpllO. Carcinoma invades AGp 110-positive tissue with normal morphology; the invading cellular front is marked by arrows, (0 Treatment as in e. AGpllO is localized on the lumen of pseudoglandular tumour (arrows). Bar, 100 µm.

Fig. 1.

Distribution of AGpllO in rat liver during AFB3-induced hepatocarcinogenesis. Paraffin-embedded acetone-fixed liver sections from rats fed AFB3 for various times were stained with anti-AGpllO serum followed by secondary antiserum conjugated to alkaline phosphatase (see Materials and methods). Red deposit of alkaline phosphatase substrate shows the localization of AGpllO. (a) Control male Fischer rat. AGpllO is confined to bile canaliculi. (b) 9 weeks of AFB3 at 2 p.p.m. in the diet. Decreased expression of AGpllO in some altered cell foci (central, delineated area) but normal distribution in most areas, (c) 24 weeks of treatment. Pre-neoplastic lesion with markedly reduced expression of AGpllO. (d) Treatment as in c. Periportal region showing increased expression of AGpllO and apparent pericellular, non-polarized localization of the protein, (e) 10 months of AFB1 at 1 p.p.m. in the diet, followed by 2 months on control diet. Non-differentiated hepatocellular carcinoma negative for AGpllO. Carcinoma invades AGp 110-positive tissue with normal morphology; the invading cellular front is marked by arrows, (0 Treatment as in e. AGpllO is localized on the lumen of pseudoglandular tumour (arrows). Bar, 100 µm.

Fig. 2.

Distribution of fibronectin (a) and fibronectin receptor, integrin α5β1 (b) during AFB3-induced hepatocarcinogenesis. (al) Fibronectin (Fn) in control male Fischer rat. Strong intracellular and cell surface staining, (a2) 22 weeks of AFB3 in the diet. Overall reduction in fibronectin synthesis (area marked 2) with further decrease shown in altered cell focus (area marked 1). Expression on cell surface may persist. (a3) 9 months of treatment. Hepatocellular carcinomas exhibit reduced expression of fibronectin but cell surface staining still present, especially in regions that display normal morphology (in between arrows), (a4) 9 months of treatment. Pseudoglandular hepatic carcinoma. Variable expression of fibronectin on cells lining the lumen: mostly negative (open arrow), but occasionally pericellular staining, apparent on basolateral surfaces can be distinguished (arrows). Sub-epithelial fibrillar-like staining is also shown (arrowheads), (bl) integrin α5β1 distribution in control male Fischer rat. Protein is expressed on all surface domains of hepatocytes and intracellularly, as fibronectin (al). (B2) 5 weeks of AFB3 in the diet. Overall reduction in α5β1 intracellular expression in liver but surface staining appears equally strong as in control (bl). Pericellular nonpolarized staining is clearly shown: α5β1 is found on sinusoidal (s), lateral (1) and canalicular (c) surface domains. (B3) 8 months of AFB3 in the diet. Hepatocellular carcinoma with reduced expression of α5β1. Arrow points to acinar formation where α5β1 persists in the lumen. (a4) Pseudoglandular hepatic carcinoma. Note absence of α5β1 in lumen (arrow). Bar, 100 µm.

Fig. 2.

Distribution of fibronectin (a) and fibronectin receptor, integrin α5β1 (b) during AFB3-induced hepatocarcinogenesis. (al) Fibronectin (Fn) in control male Fischer rat. Strong intracellular and cell surface staining, (a2) 22 weeks of AFB3 in the diet. Overall reduction in fibronectin synthesis (area marked 2) with further decrease shown in altered cell focus (area marked 1). Expression on cell surface may persist. (a3) 9 months of treatment. Hepatocellular carcinomas exhibit reduced expression of fibronectin but cell surface staining still present, especially in regions that display normal morphology (in between arrows), (a4) 9 months of treatment. Pseudoglandular hepatic carcinoma. Variable expression of fibronectin on cells lining the lumen: mostly negative (open arrow), but occasionally pericellular staining, apparent on basolateral surfaces can be distinguished (arrows). Sub-epithelial fibrillar-like staining is also shown (arrowheads), (bl) integrin α5β1 distribution in control male Fischer rat. Protein is expressed on all surface domains of hepatocytes and intracellularly, as fibronectin (al). (B2) 5 weeks of AFB3 in the diet. Overall reduction in α5β1 intracellular expression in liver but surface staining appears equally strong as in control (bl). Pericellular nonpolarized staining is clearly shown: α5β1 is found on sinusoidal (s), lateral (1) and canalicular (c) surface domains. (B3) 8 months of AFB3 in the diet. Hepatocellular carcinoma with reduced expression of α5β1. Arrow points to acinar formation where α5β1 persists in the lumen. (a4) Pseudoglandular hepatic carcinoma. Note absence of α5β1 in lumen (arrow). Bar, 100 µm.

Integrin α5β1

This fibronectin receptor was distributed in a manner similar to fibronectin, in this case the pericellular, nonpolarized localization of the protein being more distinct (Fig. 2; bl, control liver; B2, liver from animal fed aflatoxin for 5 weeks, with marked membrane domains, sinusoidal (s), canalicular (c) and lateral (1)). As with fibronectin, overall expression of α5β1 noticeably declined approximately 4 weeks after initiation of aflatoxin dietary administration (Fig. 2b2). Further marked reductions in ßi expression were detected after 9 weeks of treatment in altered foci and then in poorly differentiated tumours (not shown: see section on double staining) although cell surface staining was usually retained, albeit with reduced intensity.

Small acinar, duct-like structures, frequently seen within trabecular carcinomas (Fig. 2b), were positive for integrin, the protein being quite prominent in lumina (Fig. 2b3, arrow) that resembled enlarged canaliculi, as well as along basolateral cell surfaces. Adenomas with larger pseudoglandular structures (Fig. 2b4), perhaps emanating from the small acinar formations in trabecular carcinomas (Fig. 2b3), were negative for α5β1 on apical (Fig. 2b4, arrow) and lateral epithelial surfaces; expression on the basal surface was also barely discernible.

Comparative distribution of AGpllO and fibronectin

The spacial and temporal expression of AGpllO during aflatoxin-induced hepatocarcinogenesis was compared with that of fibronectin in double-label experiments. As described in control liver AGpllO was detected in bile canaliculi whereas fibronectin was most conspicuous in sinusoids (Fig. 1a). AGpllO was also visible on the apical surface of bile duct epithelium (Fig. 1a, arrow). By the 6th–8th week after initiation of the aflatoxin treatment, expression of both proteins became slightly reduced in small foci that initially did not appear overtly hyperplastic, apart from a slight increase in individual cell size and concurrently increased expression of both proteins was noticed in periportal areas (Fig. 1b). On longer exposure to aflatoxin (from the 9th week onwards), distinct altered or hyperplastic foci appeared (Fig. 1c; 14-week treatment) that, presumably, gave rise to larger pre-neoplastic lesions observed later on (Fig. 3d; 32-week treatment). Such hyperplastic and pre-neoplastic lesions were clearly deficient in both AGpllO and fibronectin (Fig. ?c and d). In altered foci, which were beginning to take on the appearance of adenomas, AGpllO was localized on the apical surface of the luminal epithelium whereas fibronectin was sparsely distributed along the basal cell surface (Fig. ?e). In poorly differentiated hepatocellular carcinomas little staining for either protein was observed (Fig. ?f), although fibronectin persisted, albeit weakly, on the cell surface. Furthermore, fibrillar accumulations of fibronectin were sparsely dispersed in both adenomas and poorly differentiated carcinomas (Fig. ?e and f). Some of these areas were encapsulated by extracellular matrix intensely positive for fibronectin, or by rows of hepatocytes over-producing fibronectin (Fig. ?e and g). It is worth noting that groups of cells in tumours that appear to preserve a differentiated morphology, maintain a normal pattern of staining for both AGpllO and fibronectin (Fig. ?h).

Fig. 3.

Double-label staining of AGpllO and fibronectin in livers of AFB?-fed rats. AGpllO localization is shown by blue colour and fibronectin by red colour (see Materials and methods), (a) Control liver. AGpllO staining along the biliary tree, whereas fibronectin is mostly obvious on sinusoidal surface. Note AGpllO stain on apical domain of bile duct epithelium (arrow), (b) 8 weeks of AFB3 in the diet. Normal cell surface distribution for AGpllO and fibronectin is maintained in most areas but focal reductions in expression can occasionally be observed as well as enhanced expression mainly in periportal areas (see text), (c) 14 weeks of treatment. Concurrent reduction of AGpllO and fibronectin in altered foci, (d) 32 weeks of AFB3 in the diet. Pre-neoplastic nodule with markedly decreased staining for both AGpllO and fibronectin, (e) 10 months of AFB3 in the diet. Large altered focus with adenoma-like (pseudoglandular) lesions. Overall reduction of AGpllO and fibronectin in this area. Fibronectin staining encapsulating th3/4area and occasionally scattered within it. AGpllO present only in lumen of ducts, (f) 10 months of AFB3 in the diet. Non-differentiated hepatocellular carcinoma with minimal expression of AGpllO and fibronectin, (g) 10 months of AFB3 in the diet. Stratum of cells at border of tumour (right) overproducing fibronectin and expressing normal levels of polarized AGpllO. (h) 10 months of AFB3 in the diet. Cord of hepatocytes, traversing morphologically poorly differentiated tumour, that expresses normal levels of AGpllO and fibronectin. Bar, 100 fan.

Fig. 3.

Double-label staining of AGpllO and fibronectin in livers of AFB?-fed rats. AGpllO localization is shown by blue colour and fibronectin by red colour (see Materials and methods), (a) Control liver. AGpllO staining along the biliary tree, whereas fibronectin is mostly obvious on sinusoidal surface. Note AGpllO stain on apical domain of bile duct epithelium (arrow), (b) 8 weeks of AFB3 in the diet. Normal cell surface distribution for AGpllO and fibronectin is maintained in most areas but focal reductions in expression can occasionally be observed as well as enhanced expression mainly in periportal areas (see text), (c) 14 weeks of treatment. Concurrent reduction of AGpllO and fibronectin in altered foci, (d) 32 weeks of AFB3 in the diet. Pre-neoplastic nodule with markedly decreased staining for both AGpllO and fibronectin, (e) 10 months of AFB3 in the diet. Large altered focus with adenoma-like (pseudoglandular) lesions. Overall reduction of AGpllO and fibronectin in this area. Fibronectin staining encapsulating th3/4area and occasionally scattered within it. AGpllO present only in lumen of ducts, (f) 10 months of AFB3 in the diet. Non-differentiated hepatocellular carcinoma with minimal expression of AGpllO and fibronectin, (g) 10 months of AFB3 in the diet. Stratum of cells at border of tumour (right) overproducing fibronectin and expressing normal levels of polarized AGpllO. (h) 10 months of AFB3 in the diet. Cord of hepatocytes, traversing morphologically poorly differentiated tumour, that expresses normal levels of AGpllO and fibronectin. Bar, 100 fan.

Comparative distribution of AGpllO and integrin α5β1

In double-label experiments of AGpllO and integrin α5β1 we were able to confirm the results we obtained in singlelabel experiments and to establish that the alterations in the expression of AGpllO during hepatocarcinogenesis coincide with those observed for α5β1 (Fig. 4). Altered cell foci (Fig. 4c and d) and poorly differentiated carcinomas (Fig. 4e, area marked 1) were deficient in both proteins, although α5β1, could still be detected on the cell surface. The normal distribution pattern and expression of AGpllO (Fig. 4a and b) was lost in such lesions. In adenomas (Fig. 4e, area marked 3, and f) the pseudoglandular lumen (Fig. 4e and f, arrows) was AGpllO-positive whereas could only be detected on the basal surface of the cells (Fig. 4e and f, arrowheads). Areas in tumours sustaining a morphology reminiscent of normal parenchyma were positive for both antigens (Fig. 4e, area marked 2). Our overall impression from these double-label experiments was that the difference in expression between normal and neoplastic liver was significantly more pronounced for AGpllO than for α5β1 . Furthermore, the observed reduction of α5β1 on tumour cell surfaces was less than that seen for fibronectin (Figs 3 and 4).

Fig. 4.

Double-label staining of AGpllO and integrin α5β1 in livers of AFBl-fed rats, α5β1 is marked by red colour and AGpllO by blue colour (see Materials and methods). Dark mauve colour of canaliculi (as in b) results from co-localization of the antigens, (a) Liver from rat fed AFB3 for 8 weeks. Area showing normal distribution of both proteins as in control livers but a reduction of α5β1 intracellular stain has already occurred (cf. Fig. 2bl and B2). (b) Detail from a. α5β1 is evenly distributed on all hepatocyte membrane domains but AGpllO is found only on canalicular surfaces, (c) 22 weeks of AFB1 in the diet. Altered cell foci. Decrease in the stain intensity for both α5β1 and AGpllO is observed, (d) Detail from c. In comparison to morphologically normal parenchyma (top left), hepatocytes of the altered focus (bottom right) exhibit reduced α5β1 and minimal AGpllO. Note that α5β1 though reduced, persists on cell surfaces in the focus, (e) 10 months of treatment. Hepatocellular carcinoma. Adenoma-like region (3) shows apical AGpllO stain (arrows) and basal α5β1 stain (arrowheads). In areas where some plate structure is preserved both proteins appear to be present (2) but poorly differentiated carcinomas lack AGpllO and show deficiency in α5β1 (1). (f) 9 months of treatment. Pseudoglandular carcinoma where cells are virtually negative for α5β1 (arrowheads) but remain weakly positive for AGpllO on their apical domain (arrow). Bar, 100 µm.

Fig. 4.

Double-label staining of AGpllO and integrin α5β1 in livers of AFBl-fed rats, α5β1 is marked by red colour and AGpllO by blue colour (see Materials and methods). Dark mauve colour of canaliculi (as in b) results from co-localization of the antigens, (a) Liver from rat fed AFB3 for 8 weeks. Area showing normal distribution of both proteins as in control livers but a reduction of α5β1 intracellular stain has already occurred (cf. Fig. 2bl and B2). (b) Detail from a. α5β1 is evenly distributed on all hepatocyte membrane domains but AGpllO is found only on canalicular surfaces, (c) 22 weeks of AFB1 in the diet. Altered cell foci. Decrease in the stain intensity for both α5β1 and AGpllO is observed, (d) Detail from c. In comparison to morphologically normal parenchyma (top left), hepatocytes of the altered focus (bottom right) exhibit reduced α5β1 and minimal AGpllO. Note that α5β1 though reduced, persists on cell surfaces in the focus, (e) 10 months of treatment. Hepatocellular carcinoma. Adenoma-like region (3) shows apical AGpllO stain (arrows) and basal α5β1 stain (arrowheads). In areas where some plate structure is preserved both proteins appear to be present (2) but poorly differentiated carcinomas lack AGpllO and show deficiency in α5β1 (1). (f) 9 months of treatment. Pseudoglandular carcinoma where cells are virtually negative for α5β1 (arrowheads) but remain weakly positive for AGpllO on their apical domain (arrow). Bar, 100 µm.

Comparative distribution of AGpllO and cytokeratin 19

The expression of AGpllO during chemically induced carcinogenesis was most conspicuously altered in poorly differentiated foci and tumours. To assess the potential usefulness of AGpllO as an histological marker of such lesions in transformation we compared the distribution of the proteins with that of cytokeratin 19 (ckl9), a selective marker for hepatic transformation. In normal liver ckl9 is confined to bile duct cells but in altered foci, groups of hepatocytes synthesize this cytokeratin (Green et al. 1990). Loss of AGpllO in hepatocellular foci occurred earlier (from 9 weeks of treatment) than appearance of ckl9 (from 14 weeks of treatment). This change in ckl9 expression was restricted to a few foci initially but, as the treatment progressed, more and larger lesions were observed. The various patterns of double-staining of AGpllO and ckl9 are shown in Fig. 5: (i) in parenchyma with normal morphology AGpllO was located on canalicular membranes and ckl9 in bile duct cells (Fig. 5a); (ii) some altered cell foci were AGp 110-negative but ckl9 was still restricted to bile duct cells (Fig. 5b); (iii) in other foci with minimal AGpllO expression, however, ckl9 was detected in hepatocytes being organized around lumina (Fig. 5c) that remained AGpllO-positive (see previous sections), although in this case this could not be conclusively confirmed or refuted, since the strong ckl9 reaction in the luminal area masked AGpllO stain; (iv) finally, there were foci positive for both proteins (Fig. 5d).

Fig. 5.

Double-label immunocytochemistry to localize AGpllO and cytokeratin 19. Blue colour corresponds to cytokeratin 19 (ckl9) and red to AGpllO (see Materials and methods). The acetone-fixed sections illustrate staining patterns in liver from animal fed on aflatoxin B1-containing diet for 24 weeks. (a) Area of normal parenchymal morphology comparable to physiologically healthy control liver. AGpllO is on canalicular membranes of hepatocytes and ckl9 only in bile duct epithelial cells, (b) Altered cell focus with loss of AGpllO and normal (bile duct) localization of ckl9. (c) Altered cell focus showing loss of AGpllO and positive staining of part of the focus with ckl9. Note that hepatocytes positive for ckl9 are organizing around lumina in acinar formations. Adjacent focus (to the right) without acini is negative for both AGpllO and ckl9. (d) Altered focus positive for both AGpllO and ckl9. Bar, lOOµm.

Fig. 5.

Double-label immunocytochemistry to localize AGpllO and cytokeratin 19. Blue colour corresponds to cytokeratin 19 (ckl9) and red to AGpllO (see Materials and methods). The acetone-fixed sections illustrate staining patterns in liver from animal fed on aflatoxin B1-containing diet for 24 weeks. (a) Area of normal parenchymal morphology comparable to physiologically healthy control liver. AGpllO is on canalicular membranes of hepatocytes and ckl9 only in bile duct epithelial cells, (b) Altered cell focus with loss of AGpllO and normal (bile duct) localization of ckl9. (c) Altered cell focus showing loss of AGpllO and positive staining of part of the focus with ckl9. Note that hepatocytes positive for ckl9 are organizing around lumina in acinar formations. Adjacent focus (to the right) without acini is negative for both AGpllO and ckl9. (d) Altered focus positive for both AGpllO and ckl9. Bar, lOOµm.

From the 9th to the 14th week of treatment virtually none of the AGpllO-negative altered foci showed any ckl9 in hepatocytes (as in Fig. 5b). From then on, the patterns portrayed in Fig. 5c and d appeared, the pattern in Fig. 5c becoming progressively more frequent although we could still observe lesions being negative for both proteins (Fig. 5b). In tumours, ckl9 was expressed only in pseudoglandular hepatomas (not shown) as described (Green et al. 1990). In poorly differentiated carcinomas ckl9 was occasionally present, although these tumours were usually histologically heterogeneous and ckl9 was found in pseudoacinar areas.

Distinct histological changes were observed during hepatocarcinogenesis, first at the pre-neoplastic stage, early on during aflatoxin B3 dietary administration, and then in tumours induced by either aflatoxin B3 or diethylnitros-amine. The histopathology of induced hepatocyte lesions was as described previously by other investigators (Ogawa et al. 1979; Williams, 1980). In brief, hyperplastic foci that appeared early during AFB3 treatment increased in number and size to form so-called pre-neoplastic nodules with a morphology similar to that of foci. At a later stage, approximately 9 months after initiation of either carcinogenesis regimen, tumours appeared that could be classified as poorly differentiated, trabecular or pseudoglandular (adenoma). In general, the one-cell-thick plate structure of the normal parenchyma would adopt either of the following three morphologies in carcinogenesis: (i) poorly differentiated, apparently non-polarized groups of cells appearing first in altered cell foci and nodules and later in poorly differentiated hepatomas; (ii) small acini of hepatocytes organizing around a dilated, stellate canaliculus were present in late (14th week onwards) altered foci and nodules and in trabecular carcinomas; (iii) large ducts in pseudoglandular carcinomas (adenomas). Furthermore, a certain degree of structural heterogeneity was observed in most tumours and areas in between tumours often maintained the plate structure of normal liver. Adenomas may arise from small acinar configurations, since the presence of cytokeratin 19 in both structures (see Results, and Green et al. 1990) suggests a common origin. Moreover, the number of acini and the cross-sectional area of the lumen increased progressively at the pre-neoplastic stage. Use of the antiserum against the apical antigen AGpllO indicated a frequent occurrence of acini in trabecular carcinomas where these glandular structures cannot usually be easily recognized with conventional histochemical stains. It is worth noting that similar glandular formations also occur during non-pathological liver growth, in late embryogenesis and during regeneration following partial hepatectomy (Ogawa et al. 1979). In this study, cytokeratin 19, normally absent in parenchymal cells, was always expressed in hepatocytes organizing around a lumen and could be the causal factor in initiating and/or maintaining this rearrangement, perhaps mediated by an association of this intermediate filament protein with desmosomal components.

Significant alterations in the pattern of expression of AGpllO, integrin α5β1 and fibronectin were observed during chemical induction of hepatocarcinogenesis. Poorly differentiated hyperplastic, pre-neoplastic and tumour lesions were deficient in all three antigens, particularly AGpllO. In small acinar structures, found in hyperplastic lesions and trabecular carcinomas, AGpllO was maintained on the apical (luminal) surface whereas fibronectin and a’ξßγ were present on all surfaces. In pseudoglandular carcinomas, however, all three proteins appeared polarized: AGpllO persisted on the luminal cell surface but fibronectin and α5β1, (the latter noticeably reduced in expression) were redistributed along the basal surface. Overall therefore, the results indicate that the pattern of distribution and the expression of fibronectin and its receptors relate to, or perhaps induce, distinct structural arrangements of the parenchyma: poor differentiation correlates with deficiency in AGpllO, α5β1 and fibronectin, whereas pseudo-differentiated structures such as acini or adenomas with polarized cells preserved the proteins. The unpolarized state of cells in undifferentiated hepatomas has been documented (Scoazec et al. 1988).

In addition to AGpllO, expression of other canalicular glycoproteins has also been shown to be altered by cell transformation (Becker et al. 1985; Hixson et al. 1985; Scoazec et al. 1988). Furthermore, depolarization of a canalicular antigen may precede its eventual disappearance in undifferentiated tumours (Scoazec et al. 1988) and evidence for such depolarization is also presented here (Fig. Id). The absence of AGpllO in poorly differentiated carcinomas may enhance their tumorigenicity and invasive potential by reducing stabilizing cell-fibronectin interactions.

Our findings concerning the reduction of fibronectin synthesis in hepatocytes, especially in poorly differentiated tumours, corroborate earlier observations on liver (Sell and Ruoslahti, 1982; Jagirdar et al. 1985; Szendröi and Lapis, 1985) and other cells (Hynes and Yamada, 1982). The expression of integrin α5β1 has not, to our knowledge, been studied previously in liver tumours, but investigations on virally transformed rodent cells (Plante-faber and Hynes, 1989) and on highly tumorigenic human cells (Dedhar and Saulnier, 1990) have produced results in line with our own, i.e. down-regulation of α5β1. Transformation may not only affect expression but function too. Transformation of chicken embryo fibroblasts by oncogenes encoding tyrosine kinases results in phosphorylation of the chicken fibronectin receptor and consequent impairment of its binding affinity for fibronectin and talin (Hirst et al. 1986; Topley et al. 1989). Reduced expression of α5β1 has been correlated to high tumorigenicity (Giancotti and Ruoslahti, 1990; Schreiner et al. 1991). Synthetic peptides containing the RGD sequence of fibronectin recognized by some integrins, including α5β1 (Ruoslahti and Pierschbacher, 1987), can inhibit malignant invasion and metastasis (Gehlsen et al. 1988; Humphries et al. 1986; Saiki et al. 1989). Transformation affects other integrins too (Buck et al. 1990; Virtanen et al. 1990; Dedhar, 1990): for instance, chemical transformation of human cells into highly tumorigenic cells results in increased expression of α6β1 α2β1 and α1β1 receptors for laminin, collagen and type IV collagen and laminin, respectively (Dedhar and Saulnier, 1990).

Our study focused on pre-neoplastic and neoplastic lesions that appeared to be of hepatocellular origin. Only a fraction of altered cell foci appear to develop into tumours (Farber, 1980; Williams, 1980; Pitot and Sirica, 1980; Emmelot and Scherer, 1980) and no definitive single marker for these pre-malignant cells has been identified. An alternative possibility is a de novo genesis of carcinomas from hepatocyte stem cells (Sell and Dunsford, 1989). In our study early hyperplastic foci appeared to originate from parenchymal cells and loss of differentiation correlated with deficiency in AGpllO, α5β1 and fibronectin in early and later hyperplasia and, finally, in tumours. Adenomas, positive for ckl9, seemed to emanate from small pseudoacinar structures, also positive for ckl9. These acini were formed in foci from hepatocytes organizing around dilated canaliculi. Furthermore, loss of AGpllO and appearance of ckl9 in hepatocytes was an event of comparatively infrequent incidence among early hyperplastic foci but occurred in all undifferentiated tumours and adenomas, respectively. AGpllO and ckl9 appear therefore to be good cell markers for progenitors of two different types of carcinomas and our study supports the hypothesis that at least some tumours arise from differentiated hepatocytes.

We are grateful to Dr R. O. Hynes (MIT, USA) for the antifibronectin monoclonal antibody, to Dr Staffan Johansson (Biomedical Centre, Uppsala, Sweden) for the anti-a3β antibody and to Dr E. B. Lane (University of Dundee, UK) for the anti-cytokeratin 19 antibody. We also thank Dr P. Carthew for much helpful discussion, Mr Neil Papworth for advice on photography and Mr J. Morgan for printing the photographs. The excellent secretarial assistance of Ms Marilyn Brennan is gratefully acknowledged.

Becker
,
A.
,
Neumeier
,
R.
,
Park
,
C.-S.
,
Gossrau
,
R.
and
Reutter
,
W.
(
1985
) .
Identification of a transformation-sensitive 110-KDa plasma membrane glycoprotein of rat hepatocytes
.
Eur. J. Cell Biol
.
39
,
417
423
.
Ben-Ze’ev
,
A.
,
Robinson
,
G. S.
,
Bucher
,
N. L. R.
and
Farmer
,
S. R.
(
1988
).
Cell-cell and cell-matrix interactions differentially regulate the expression of hepatic and cytoskeletal genes in primary cultures of rat hepatocytes
.
Proc. natn. Acad. Sci. U.S.A
.
85
,
2161
2165
.
Buck
,
C.
,
Albelda
,
S.
,
Damjanovich
,
L.
,
Edelman
,
J.
,
Shih
,
D.-T.
and
Solowska
,
J.
(
1990
).
Immunohistochemical and molecular analysis of β1 and β3 integrins
.
Cell Differ. Dev
.
32
,
189
202
.
Dedhar
,
S.
(
1990
).
Integrins and tumour invasion
.
BioEssays
12
,
583
590
.
Dedhar
,
S.
and
Saulnier
,
R.
(
1990
).
Alterations in integrin receptor expression on chemically transformed human cells: specific enhancement of laminin and collagen receptor complexes
.
J. Cell Biol
.
110
,
481
489
.
Emmelot
,
P.
and
Scherer
,
E.
(
1980
).
The first relevant cell stage in rat liver carcinogenesis: a quantitative approach
.
Biochim. biophys. Acta
605
,
247
304
.
Enrich
,
C.
,
Evans
,
W. H.
and
Gahmberg
,
C. G.
(
1988
).
Fibronectin isoforms in plasma membrane domains of normal and regenerating rat liver
.
FEBS Lett
.
228
,
135
138
.
Farber
,
E.
(
1980
).
The sequential analysis of liver cancer induction
.
Biochim. biophys. Acta
605
,
149
166
.
Gehlsen K”Argraves
,
W. S.
,
Piersbacher
,
M. D.
and
Ruoslahti
,
E.
(
1988
).
Inhibition of in vitro tumour cell invasion by Arg-Gly-Asp containing synthetic peptides
.
J. Cell Biol
.
106
,
925
930
.
Giancotti
,
F.
and
Ruoslahti
,
E.
(
1990
).
Elevated levels of c?/Ø? fibronectin receptor suppress the transformed phenotype in CHO cells
.
Cell
60
,
849
859
.
Green
,
J. A.
,
Carthew
,
P.
,
Heuillet
,
E.
,
Simpson
,
J. L.
and
Manson
,
M. M.
(
1990
).
Cytokeratin expression during AFB?-induced carcinogenesis
.
Carcinogenesis
11
,
1175
1182
.
Green
,
J. A.
and
Manson
,
M. M.
(
1991
).
Double label immunohistochemistry of tissue sections with alkaline phosphatase and peroxidase conjugates
.
In Methods in Molecular Biology - Immunochemical Protocols
, vol.
10
(
M. M.
Manson
, ed.).
Humana Press
,
New Jersey
.
Hirst
,
R.
,
Horwitz
,
R.
,
Buck
,
C.
and
Rohrschneider
,
L.
(
1986
).
Phosphorylation of the fibronectin receptor complex in cells transformed by oncogenes that encode tyrosine kinases
.
Proc. natn. Acad. Sci. U.S.A
.
70
,
3170
3174
.
Hixson
,
D. C.
,
McEntire
,
K. D.
and
Öbrink
,
B.
(
1985
).
Alterations in the expression of a hepatocyte cell adhesion molecule by transplantable rat hepatocellular carcinomas
.
Cancer Res
.
45
,
3742
3749
.
Hughes
,
R. C.
and
Stamatoglou
,
S. C.
(
1987
).
Adhesive interactions and the metabolic activity of hepatocytes
.
J. Cell Sci. Suppl
.
8
,
273
291
.
Humphries
,
M. J.
,
Olden
,
K.
and
Yamada
,
K. M.
(
1986
).
A synthetic peptide from fibronectin inhibits experimental metastasis of murine melanoma cells
.
Science
233
,
467
470
.
Hynes
,
R. O.
(
1987
).
Integrins: a family of cell surface receptors
.
Cell
48
,
549
554
.
Hynes
,
R. O.
and
Yamada
,
K. M.
(
1982
).
Fibronectins: multifunctional modular glycoproteins
.
J. Cell Biol
.
95
,
369
377
.
Jagirdar
,
J.
, ?
SHAK
,
K. G.
,
Columbo
,
M.
,
Brambilla
,
C.
and
Paronetto
,
F.
(
1985
).
Fibronectin patterns in hepatocellular carcinoma and its clinical significance
.
Cancer
56
,
1643
1648
.
Johansson
,
S.
,
Forsberg
,
E.
and
Lundgren
,
B.
(
1987
).
Comparison of fibronectin receptors from rat hepatocytes and fibroblasts
.
J. biol. Chem
.
262
,
7819
7824
.
Ogawa
,
K.
,
Medline
,
A.
and
Farber
,
E.
(
1979
).
Sequential analysis of hepatic carcinogenesis. A comparative study of the ultrastructure of preneoplastic, malignant, prenatal, postnatal and regenerating liver
.
Lab. Invest
.
41
,
22
31
.
Pitot
,
H. C.
and
Sirica
,
A. E.
(
1980
).
The stages of initiation and promotion of hepatocarcinogenesis
.
Biochim. biophys. Acta
605
,
191
215
.
Plantefaber
,
L. C.
and
Hynes
,
R. O.
(
1989
).
Changes in integrin receptors on oncogenically transformed cells
.
Cell
56
,
281
290
.
Reid
,
L. M.
,
Abreu
,
S. L.
and
Montgomery
,
K.
(
1988
).
Extracellular matrix and hormonal regulation of synthesis and abundance of messenger RNAs in cultured liver cells
.
In The Liver: Biology and Pathobiology, second edition
(
I. M.
Arias
,
W. B.
Jakoby
,
H.
Popper
,
D.
Schachter
and
D. A.
Shafritz
, ed
.).
Raven Press Ltd
,
New York
. Pp.
717
737
.
Ruoslahti
,
E.
and
Piersbacher
,
M. D.
(
1987
).
New perspectives in cell adhesion: RGD and integrins
.
Science
238
,
491
497
.
Saiki
,
I.
,
Lida
,
J.
Muranta
,
J.
,
Ogawa
,
R.
,
Nishi
,
N.
,
Sugimura
,
K.
,
Tokura
,
S.
and
Azuma
,
I.
(
1989
).
Inhibition of metastases of murine malignant melanoma by synthetic polymeric peptides containing core sequences of cell adhesive molecules
.
Cancer Res
.
49
,
3815
3822
.
Schreiner
,
C.
,
Fisher
,
M.
,
Hussein
,
S.
and
Juliano
,
R. L.
(
1991
).
Increased tumorigenicity of fibronectin receptor deficient Chinese hamster ovary cell variants
.
Cancer Res
.
51
,
1738
1740
.
Scoazec
,
J.-Y.
,
Maurice
,
M.
,
Moreau
,
A.
and
Feldmann
,
G.
(
1988
).
Analysis of hepatocyte plasma membrane polarity during rat azo dye hepatocarcinogenesis using monoclonal antibodies directed against domain-associated antigens
.
Cancer Res
.
48
,
6882
6890
.
Sell
,
S.
and
Dunsford
,
H. A.
(
1989
).
Evidence for the stem cell origin of hepatocellular carcinoma and cholangiocarcinoma
.
Am. J. Path
.
134
,
1347
1363
.
Sell
,
S.
and
Ruoslahti
,
E.
(
1982
).
Expression of fibronectin and laminin in the rat liver after partial hepatectomy, during carcinogenesis, and in transplantable hepatocellular carcinoma
.
J. natn. Cancer Inst
.
69
,
1105
1114
.
Stamatoglou
,
S. C.
,
Ge
,
R. C.
,
Mills
,
G.
,
Butters
,
T. D.
,
Zaidi
,
F.
and
Hughes
,
R. C.
(
1990a
).
Identification of a novel glycoprotein (AGpllO) involved in interactions of rat liver parenchymal cells with fibronectin
.
J. Cell Biol
.
111
,
2117
2127
.
Stamatoglou
,
S. C.
,
Sullivan
,
K. H.
,
Johansson
,
S.
,
Bayley
,
P. M.
,
Burdett
,
I. D. J.
and
Hughes
,
R. C.
(
1990b
).
Localization of two fibronectin-binding glycoproteins in rat liver and primary hepatocytes
.
J. Cell Sci
.
97
,
595
606
.
Stewart
,
H. L.
and
Williams
,
G.
(
1980
).
Co-chairmen, sub-committee on rat liver tumors, Institute of Laboratory Animal Resources. 1980. Histologic typing of liver tumours of the rat
.
J. natn. Cancer Inst
.
64
,
177
206
.
Sudhakaran
,
P. R.
,
Stamatoglou
,
S. C.
and
Hughes
,
R. C.
(
1986
).
Modulation of protein synthesis and secretion by substratum in primary cultures of rat hepatocytes
.
Expl Cell Res
.
167
,
505
516
.
Szendröi
,
M.
and
Lapis
,
K.
(
1985
).
Distribution of fibronectin and laminin in human liver tumors
.
J. Cancer Res. clin. Oncol
.
109
,
60
64
.
Tamkun
,
J. W.
and
Hynes
,
R. O.
(
1983
).
Plasma fibronectin is synthesized and secreted by hepatocytes
.
J. ?ol. Chem
.
258
,
4641
4647
.
Topley
,
P.
,
Horwitz
,
A.
,
Buck
,
C.
,
Duggan
,
K.
and
Rohrschneider
,
L.
(
1989
).
Integrins isolated from Rous Sarcoma Virus-transformed chicken embryo fibroblasts
.
Oncogene
4
,
325
333
.
Virtanen
,
L
,
Korhonen
,
M.
,
Kariniemi
,
A.-L.
,
Gould
,
V. E.
,
Laitinen
,
L.
and
Ylänne
,
J.
(
1990
).
Integrins in human cells and tumors
.
Cell Differ. Dev
.
32
,
215
228
.
Williams
,
G. M.
(
1980
).
The pathogenesis of rat liver cancer caused by chemical carcinogenesis
.
Biochim. biophys. Acata
605
,
167
189
.