Hepatocyte growth factor/scatter factor-induced intracellular signalling

Abstract

Hepatocyte growth factor (HGF) identical to scatter factor (SF) is a glycoprotein involved in the development of a number of cellular phenotypes, including proliferation, mitogenesis, formation of branching tubules and, in the case of tumour cells, invasion and metastasis. This fascinating cytokine transduces its activities via its receptor encoded by the c-met oncogene, coupled to a number of transducers integrating the HGF/SF signal to the cytosol and the nucleus. The downstream transducers coupled to HGF/MET, most of which participate in overlapping pathways, determine the development of the cell's phenotype, which in most cell types is dual.

Hepatocyte regeneration has attracted considerable attention as a possible model for the elucidation of mechanisms governing normal cell proliferation. The remodelling of the liver mass is accompanied by a sequence of events leading to proliferation of the mature hepatocyte compartment, which, during noninduced growth, is composed of quiescent hepatocytes (Fausto 1991; Michalopoulos & Defrances 1997). Exit of primary hepatocytes from quiescence (G₀) (Baserga et al. 1982; Pardee 1989) and their entry into the proliferation cycle (G₁-S) is triggered by many factors, including directly acting major growth modulators such as hepatocyte growth factor (HGF), transforming growth factor alpha (TGF-α) and epidermal growth factor (EGF) (McGowan et al. 1981; Matsumoto & Nakamura 1994).

Induction of hepatocyte proliferation after partial hepatectomy in vivo is believed to require the existence of hepatocytes in G₁ phase and, hence, ‘competent’ to respond to growth factor stimuli. The ‘initiatory’ event in hepatocyte proliferation is unknown but thought to be driven by both ECM and mechanical stressors, in parallel with factors (tumour necrosis factor-α (TNF-α), interleukin-6 (IL-6)) derived from nonparenchymal cells acting in a paracrine or a juxtacrine fashion. In addition, the temporary reduction in liver mass, occurring for several weeks after a regenerative stimulus in vivo (partial hepatectomy), decreases the clearance of many growth-related modulators, including adrenergic agents, insulin, and glucagon (Michalopoulos & DeFrances 1997). It is still uncertain whether signals mediating the hepatocyte's passage from G₀ to G₁ are exclusive of those expressed during early G_1. As the transition from G₀ to G₁ results in a proportion of cells converting their phenotype from a highly differentiated to a proliferating one, some overlap or redundancy of mechanistic pathways would seem advantageous, even if a continuously cycling cell phenotype does not ensue.

The identification and precise functional discrimination of such effectors is complicated. Most factors involved in the initiation and progression of hepatocyte proliferation have been found to activate similar sets of G₀/G₁-related genes, with differences only in the timing and the duration of gene induction. Nonetheless, in vitro studies have shown that short-term culture of hepatocytes under proliferating conditions does not induce significant changes in the cell phenotype towards de-differentiation. Investigators have studied many combinations of growth modulators and extracellular matrix components in an attempt to establish a convincing model of cycling hepatocytes which maintain a differentiated phenotype in long-term culture. To date, this has not been attained. Similarly, the various signals inducing hepatocyte proliferation may trigger intracellular responses which involve overlapping or even contradictory mechanisms. Much in vitro work has been carried out using EGF and TGF-α, factors believed to play important roles in vivo(McGowan 1986; Vesey et al. 1992; Fausto et al. 1995). The recently discovered HGF produced by fibroblasts (Nakamura et al. 1989; Bhargava et al. 1992) and its receptor, the product of the met oncogene, are indisputably significant components of the complex and overlapping pathways operating during rodent and human hepatocyte proliferation following liver injury (Strain et al. 1991; Weidner et al. 1991). The study of HGF has led to an appreciation of the fascinating networks of molecular interactions which may be induced within the cell by a single molecule. Delineation of these intertwining molecular pathways has revealed a finely tuned set of precise interactions between intra- and extra-cellular components that had generally been perceived as distinct entities.

Hepatocyte maintenance in vitro progressively leads to de-differentiation and cell detachment in vitro and to reconstitution of the liver mass in vivo (after partial hepatectomy). Do hepatocytes cease to proliferate as a result of an orchestrated de-activation of the ‘cycling-related genes’ by activated quiescence and/or differentiation-related genes? Microinjection of RNA from quiescent cells inhibits DNA synthesis in cells exposed to mitogenic stimuli. The concept therefore that specific gene transcripts exist in quiescent cells which encode proteins inhibiting cellular proliferation is gaining acceptance (Schneider et al. 1988). Indeed, cell density-dependent growth-inhibition or mitogen deprivation of cells has led to the identification of gas-1, one of a family of growth-arrest-specific (gas) genes (Manfioletti et al. 1990; Del Sal et al. 1992).

HGF: structural and functional characteristics

HGF/SF is derived from a biologically inactive single chain precursor of 728 amino acids (pro-HGF) localized mostly on the cell surface and in the extracellular matrix (Mizuno & Nakamura 1993; Bardelli et al. 1994). The mature form produced following proteolytic cleavage is composed of a 69-kD α-subunit (containing four kringle domains) and the 34 kD β-subunit, similar to the catalytic domain of serine proteases, but with amino acid substitutions in the active site (Nakamura 1991; Mizuno et al. 1992).

The HGF gene is a single copy gene, located on chromosome 7q11.1–21 in humans, and consists of 18 exons and 17 introns which span approximately 70 kb (Seki et al. 1991). Its organization highly resembles that of factors implicated in blood coagulation and fibrinolysis, such as plasminogen and prothrombin (Kuiper et al. 1996). The primary structure of HGF is > 90% homologous in humans and rodents (Zarnegar et al. 1992). Analyses of the intron-exon boundaries and of serine protease domains have revealed that HGF, the macrophage stimulating protein (MSP), plasminogen and apolipoprotein may possess a common genetic precursor, since each encodes an N-terminal domain (corresponding to the plasminogen activation peptide), three copies of the kringle domain and a serine protease domain (Shimamoto et al. 1993; Donate et al. 1994). The N-terminal domains of these similar factors contain four cysteine residues forming a loop within a loop (Donate et al. 1994). The signal peptide (involved with intracellular processing and secretion) and the 5′-untranslated region are contained in the first exon, whereas the α-subunit is encoded by the following 10 exons (Seki et al. 1991; Zarnegar 1995).

The 5′-flanking region of the HGF/SF gene contains a noncanonical TATA box and also a number of putative regulatory elements such as a TGF-β inhibitory element, four IL-6 responsive elements, two potential binding sites for TNF and IL-6, a cyclic adenosine 3′, 5′-monophosphate (cAMP) response element, two oestrogen response elements, a potential vitamin D response element, two liver-specific transcription factor binding sites and a B cell-and macrophage-specific transcriptional factor binding site (Liu et al. 1994). The twelfth exon encodes a short peptide region between the α and the β subunits and the β chain is encoded by the last six exons (Seki et al. 1991). The extracellular urokinase-type plasminogen activator (uPA) complexes with the biologically inactive pro-HGF both in vitro and in vivo, leading to activation of pro-HGF to mature HGF through hydrolysis of the Arg494-Val495 peptide bond (Mars et al. 1993; Naldini et al. 1995). uPA is unable to act on pro-HGF bearing a mutated cleavage site in vitro (Weidner et al. 1993). Macrophages are considered to be the source of uPA, which is produced locally in response to environmental stimuli such as tissue injury (Naldini et al. 1995). Both the N-terminal region and the first two kringle domains are required for biological activity mediated through the MET receptor. Integrity of the HGF/SF-β-chain is also essential for biological activity, contributing to the spatial conformation of HGF/SF (Lee et al. 1995). HGF/SF naturally produced by fibroblasts associates with heparin and heparan sulphate proteoglycan, both on the cell surface and in the extracellular matrix (Matsumoto et al. 1993; Zioncheck et al. 1995) The N-terminal hairpin loop, as well as the second kringle domain are essential for heparin-binding to HGF/SF (Mizuno et al. 1994). In addition, heparin and heparan sulphate stimulate the production of HGF/SF by various cell lines, including MRC-9, IMR-90, WI-38, HL-60 (Matsumoto et al. 1993). The region within the 32–212 N-terminal residues of HGF/SF, termed HGF/NK1, constitutes its binding determinant (Lokker et al. 1994). Within this sequence, F162 is crucial in maintaining the hydrophobic core of the first kringle (Rubin et al. 1991). Recombinant HGF/NK1 competes for binding to the MET receptor on A549 human lung carcinoma cells. However, MET autophosphorylation is inefficient and mitogenesis is not induced, even at very high concentrations of rHGF/NK1 (Lokker et al. 1994).

HGF transcripts and variants

Northern blot analysis of tissue RNA has revealed the existence of a variety of HGF/SF transcripts (6, 3, 2.2 and 1.3 kb) (Miyazawa et al. 1991; Rubin et al. 1991). In leucocyte and fibroblast cDNA libraries, HGF/SF variants encoding 728 and 723-amino-acid forms have been isolated and shown to possess similar mitogenic, cytotoxic and scattering activities as the full length forms of the molecule (Furlong et al. 1991; Weidner et al. 1991).

Clones isolated from a fibroblast cDNA library corresponding to the 1.3 kb transcript encode a variant protein, HGF/NK2, which shares the same receptor as full length HGF/SF (Chan et al. 1991; Miyazawa et al. 1991). This variant includes the signal peptide sequence and the second kringle domain of HGF/SF. HGF/NK2 originating from the SK-LMS-1 human cell line lacks any mitogenic activity in human melanocytes and a human mammary epithelial cell line and, in fact, blocks full length HGF/SF-induced mitogenesis and scattering when used at 10-fold molar excess (Hartmann et al. 1992). Another naturally occurring HGF/SF variant bearing a five amino acid deletion at the N-terminus, dHGF, has also been described (Zarnegar et al. 1990). This deletion alters the solubility and immunological properties of the molecule. Furthermore, dHGF has been shown to possess increased biological activity compared to HGF/SF in inducing DNA synthesis in both cell lines and primary cells (Shima & Higashio 1994; Shima et al. 1994).

HGF/SF is produced by a number of cells (Weidner et al. 1991). While this molecule is associated with modulation of functions related to hepatocyte proliferation, its synthesis within the liver occurs in nonparenchymal cells rather than hepatocytes (Zarnegar et al. 1990; Defrances et al. 1992). Analysis of the liver cell population has revealed that stellate cells are the major source of HGF/SF production (Noji et al. 1990; Ramadori et al. 1992). TGF-β, an inhibitor of hepatocyte proliferation both in vivo and in vitro, inhibits HGF/SF expression in stellate cells, thus regulating in part the balance between liver fibrosis and regeneration (Ramadori et al. 1992).

Among the sites showing HGF/SF immunoreactivity in the developing rat are bronchial and oesophageal epithelia, chondrocytes, somites, haemopoietic cells and the pancreas (Defrances et al. 1992). HGF/SF is also involved in the development of the renal collecting duct via autocrine and/or paracrine mechanisms (Woolf et al. 1995). In mouse embryos homozygous for a mutant HGF/SF gene, placental and liver development is impaired, resulting in death in utero (Schmidt et al. 1995; Uehara et al. 1995). These and other data illustrate the role of HGF/SF as a mesenchymal-epithelial mediator, targeting cellular activities such as mitogenesis, motogenesis and morphogenesis in a variety of normal and transformed cells (Table 1).

Table 1.

Effects of HGF/SF in normal and transformed cells

The HGF/SF receptor

HGF/SF binds with high affinity to a transmembrane receptor encoded by the c-met proto-oncogene (p190^MET) (Bottaro et al. 1991; Weidner et al. 1991). MET belongs to the family of heterodimeric tyrosine kinases, which includes the putative receptors Ron, Ryk and Sea (Huff et al. 1993; Ronsin et al. 1993; Gaudino et al. 1994). Tyrosine phosphorylation of the p185^RON is triggered in epithelial cells by MSP (Tamagnone et al. 1993).The coding sequences of Macrophage stimulatory protein (MSP) and HGF/SF are 45% homologous and MSP is predominantly synthesized in the liver. However, there is no cross-reactivity between HGF/SF and MSP, and treatment of cells with HGF/SF is not inhibitory for MSP-induced activity, or vice versa (Tamagnone et al. 1993). Early erythroid progenitors express a novel tyrosine kinase receptor, a MET-related kinase, whose catalytic domain has amino acid similarities to the MET and V-SEA (Yee et al. 1993). The recently discovered transmembrane protein SEX, also shares significant homology with the extracellular domain of the members of the HGF/SF receptor family (MET, RON, SEA) and is involved in the development of neural and epithelial tissues, although not possessing tyrosine kinase activity (Maestrini et al. 1996).

The mature form of p190^MET is an heterodimer of two disulphide-linked subunits, α and β (Weidner et al. 1991). The α-subunit is extracellular and heavily glycosylated; the β-subunit consists of an extracellular domain involved in ligand binding, a membrane spanning region and a tyrosine kinase catalytic domain (Comoglio 1993). Both subunits are derived by glycosylation and proteolytic cleavage of a common precursor of approximately 170 kD. The MET receptor is exposed in the basolateral plasmalemma and is associated with detergent-insoluble components in polarized cells. Alternative post-translational processing of the mature form may result in the formation of MET isoforms consisting of an intact ligand binding domain but lacking the kinase domain as a result of truncation of the β subunit. One of these truncated forms is soluble and released from the cells (Comoglio 1993). HGF/SF binding triggers tyrosine autophosphorylation of the β-subunit, enhancing its enzymatic activity (Weidner et al. 1991; Comoglio 1993). Upregulation of the kinase activity of the receptor, by concomitant increase of the Vmax of the phosphotransfer reaction, is triggered by autophosphorylation of the major phosphorylation site, Y1235, located in the catalytic domain (Naldini et al. 1991; Ferracini et al. 1991; Longati et al. 1994). This tyrosine residue is part of the three tyrosine motif including Y1230, Y1234 and Y1235. In the absence of Y1235, Y1234 is phosphorylated. Substitution of either Y1235 or Y1234 with phenylalanine significantly reduces the in vitro kinase activity, while replacement of both sites abolishes the ability of the mutated receptor to be autophosphorylated and thereby activated (Longati et al. 1994). Treatment of GTL-16 cells, which over-express the MET receptor, with activators of protein kinase C (PKC) such as phorbol esters, 12–0-tetradecanoylphorbol-13-acetate acetate (TPA) or phorbol myristate acetate (PMA), significantly reduces tyrosine phosphorylation of the β-subunit of the receptor (Gaudino et al. 1990). Depletion of PKC by prolonged treatment with TPA results in increased phosphorylation of the beta-subunit. Ser 985, located in the juxtamembrane domain of the HGF/SF receptor and possessing a PKC consensus phosphorylation amino acid sequence, has been identified as the major phosphorylation site for both PKC and calcium-dependent kinases. These kinases (PKC and Ca²⁺-dependent) are believed to be responsible for inhibition of tyrosine phosphorylation activity of the MET receptor in vitro and in vivo (Comoglio 1993; Bardelli et al. 1994; Gaudino et al. 1994; Lee & Yamada 1994).

HGF/SF stimulation induces a protein tyrosine phosphatase (PTP) activity that coprecipitates with the MET receptor in a time and dose-dependent manner. PKC stimulated phosphorylation of the MET receptor results in decreased PTP activity and effective dephosphorylation of the MET receptor (Villa Moruzzi et al. 1993). Tyrosine phosphorylation of the MET receptor by HGF/SF transduces inhibitory signals regarding mitogenesis in Meth A cells (Komada et al. 1992). The scattering effect of activated MET receptor is transduced through the cytoplasmic domain by activation of its tyrosine kinase activity, as demonstrated in the B16-F1 mouse melanoma cell line transfected with a chimeric receptor (Komada & Kitamura 1993).

Several transcripts of the MET receptor (7.1, 5.9 and 4.6 kb RNA) have been identified. All include in the first exon a 5′ untranslated sequence, suggesting regulation of the various transcripts by a single promoter (Gambarotta et al. 1994). An 8-kb transcript encoding the 190 kD MET protein is expressed predominantly in epithelial tissues and undergoes rapid degradation with a half-life of less than 30 min. IL-1α, IL-6, TNF-α, TGF-β1, EGF, HGF/SF and corticosteroids markedly influence levels of the 8 kb c-MET mRNA in cell lines derived from human cancers of ovary, breast and endometrium (Moghul et al. 1994).

Transfection with the native heterodimeric receptor, the β chain homodimer or a mutant receptor lacking kinase activity does not transform rodent fibroblasts (Zhen et al. 1994). Although the MET transfectants exhibit anchorage-independent growth, they do not form foci in confluent cultures and are not tumorigenic in nude mice (Giordano et al. 1993). However, truncation below the transmembrane region results in the acquisition of constitutive kinase activity by the cytoplasmic domain, leading to the formation of foci and development of tumours (Zhen et al. 1994). Substitution of a lysine residue responsible for ATP binding results in impairment of the catalytic function and transforming capacity (Zhen et al. 1994). In murine studies reported by Vande Wounde and coworkers, the transforming capacity of murine MET was found to be substantially greater than that of the human counterpart and coexpression of MET and HGF/SF led to efficient tumorigenesis (Rong et al. 1992). The first 39 amino acids of the juxtamembrane domain and the regulatory tyrosines in the catalytic domain are essential for maintenance of such transforming capacity (Zhen et al. 1994). It can be concluded that MET-transfected fibroblasts predominantly express a motogenic and invasive, rather than a proliferative, phenotype. Over-expression of the MET receptor has been observed in a variety of human tumours (Table 2).

Table 2.

Human tumours over-expressing c-met

Intracellular responses to hepatocyte growth factor

Activation of cellular adaptors

HGF/SF transduces its proliferating signals via stimulation of the tyrosine kinase activity of the Met receptor, which in turn phosphorylates and activates a number of signal transducers. HGF/SF stimulates formation of inositol-triphosphate (InsP₃), which is inhibited by genistein, a tyrosine phosphorylation inhibitor (Baffy et al. 1992; Okano et al. 1993). Phospholipase C-gamma (PLC-γ) is tyrosine phosphorylated by the tyrosine kinase activity of the MET receptor. Conversely, no formation of InsP₃ or activation of PLC-γ occurs after the addition of HGF/SF to HepG2 cells, the growth of which is suppressed by HGF/SF (Okano et al. 1993). The activation of the InsP₃-PLCγ pathway results in a biphasic increase in the production of 1,2 diacylglycerol, the first corresponding with the peak of InsP₃ and the second, of greater magnitude, related to the formation of phosphatidylcholine through the PLC-γ pathway (Osada et al. 1992). HGF/SF does not affect adenyl cyclase activity or increase intracellular cAMP levels (Marker et al. 1992). Intracellular cAMP levels increase following partial hepatectomy but concentrations of the cAMP regulatory element binding protein (CREB) remain unchanged during a six-day period of liver regeneration (Kwast Welfeld et al. 1991).

The β-subunit of the HGF/SF receptor associates with a number of intracellular transducers which, in turn, may modulate the activity of other adaptors or transcription factors. Sf9, an insect cell derived recombinant Met receptor, has been shown in vitro to associate with and phosphorylate cytoplasmic transducers containing -SH2 domains, such as the 85 kD subunit of phosphoinositide 3-kinase (PI-3 kinase), rasGAP, PLC-γ, p59^FYN, pp60^c–src and Grb-2-SOS (Grb-son of sevenless) (Bardelli et al. 1992). This association is strictly dependent on the phosphorylation state of the receptor, suggesting that it occurs via interaction with the -SH2 domains. NIH3T3 cells transfected with the MET oncogene exhibit constitutive tyrosine phosphorylation of both the MET species and the platelet derived growth factor receptor (PDGF-R). In the absence of PDGF, PDGF-R (phosphorylated on both tyrosine and serine residues) associates with the -SH2 domain of PLC-γ (Kochhar et al. 1994).

The Met receptor also associates with the shc adaptor via its -SH2 domain. Shc is an -SH2 domain protein which may be tyrosine phosphorylated by several kinases (Pelicci et al. 1995). Grb2 is another adaptor protein which contains both -SH2 and -SH3 domains. The -SH3 domains of Grb2 bind to proline-rich motifs within the C-terminal part of SOS, a Ras exchange factor (Cussac et al. 1994). Grb2 binds to Shc via its -SH2 domains, resulting in the recruitment of Grb2/SOS complexes on the cell membrane and Ras activation (Cussac et al. 1994). Shc interacts with tyrosine-phosphorylated proteins by binding to phosphotyrosine (pY) within the NPXpY motif (Batzer et al. 1995; Chardin et al. 1995). Shc also complexes with the tyrosine phosphorylated β4 subunit of the alpha-6 beta-4 integrin (a member of the hemidesmosomes) followed by recruitment of Grb2 (Mainiero et al. 1995). This association provides an additional linkage between cytoskeletal molecules and intracellular signal amplifiers (shc/Grb2) (Mainiero et al. 1995). Over-expression of the shc adaptor results in increased cellular proliferation and migration in response to HGF/SF (Pelicci et al. 1995). The docking sites for shc in the HGF/SF receptor are the phosphotyrosine residues within the motifs Y1349VHV and Y1356VNV. Shc is phosphorylated on Y317VNV, generating a high affinity binding site for Grb2 which duplicates the Grb2 site present on the MET receptor (Pelicci et al. 1995). The importance of the single noncatalytic site Y1356 is demonstrated in experiments using a chimeric receptor consisting of the extracellular domain of the colony stimulating factor-1 (CSF-1) receptor fused to the transmembrane and intracellular domain of the Met receptor and transfected to MDCK cells. Mutation of Y1356 to phenylalanine fails to affect the exogenous kinase activity of the receptor. However, impaired scattering and formation of branching tubules in response to CSF-1 ensues. The resulting chimeric receptor's capacity to associate with Grb2 is also impaired (Zhu et al. 1994).

The association of the autophosphorylated MET receptor with p85/110 PI-3 kinase in vivo or in vitro via the receptor Y-1349 and Y-1356 sites reveals YVXV as a novel recognition motif for PI 3-kinase binding (Ponzetto et al. 1993). In vivo association of PI 3-kinase with the hepatocyte MET receptor has also been demonstrated in cultured hepatocytes where inhibition of PI-3 kinase by wortmannin significantly reduced the proliferating response to HGF/SF (Skouteris & Georgakopoulos 1996). In cells transformed by the BCR/abl oncoprotein (a tyrosine kinase that induces tyrosine phosphorylation of shc), PI 3-kinase associates with shc (Harrison Findik et al. 1995). PI-3 kinase appears to be a transducer of central importance in HGF/SF-induced mito, moto-and morphogenesis, evidenced by the inhibition by wortmannin of branching formation on collagen matrix and chemotaxis in HGF/SF-stimulated inner medullary collecting duct cells. Mitogenesis is also inhibited, although to a lesser extent (Derman et al. 1995).

Observations in MDCK cells suggest that the GTP-binding proteins Ras, Rac and Rho are involved in the regulation of HGF/SF responses (Ridley et al. 1995). HGF/SF treatment activates Ras protein by shifting the equilibrium towards the GTP-bound state and has established the involvement of the Ras pathway in mediating the motility signal of HGF/SF. Microinjection of activated H-Ras protein stimulates spreading and actin reorganization but not scattering, whereas inhibition of endogenous Ras proteins abolishes HGF/SF-induced spreading, scattering and actin reorganization (Ridley et al. 1995). Cellular dissociation due to HGF/SF induction is prevented in MDCK cells that express a dominant-negative N17Ras gene (Hartmann et al. 1994). Treatment of human lung carcinoma A549 cells with HGF/SF increases the level of Ras-bound radiolabelled guanine nucleotides by over 5-fold. After addition of HGF/SF, 50% of Ras is in the GTP-bound, active state (Graziani et al. 1993). Inhibition of tyrosine phosphorylation in GTL-16 cells significantly decreases cytosolic Ras-guanine nucleotide exchange activity (Graziani et al. 1993). Microinjection of a dominant Rac inhibitor abolishes the HGF/SF-and Ras-induced spreading and actin reorganization (Takaishi et al. 1994). Microinjection of Rho inhibits HGF/SF-induced spreading and scattering, while inhibition of Rho function does not diminish motility. The roles played by Rho 21 and Rho GDI in HGF/SF-induced cell motility have also been assessed in mouse keratinocytes (308R cells), in which HGF/SF induced cell motility is mimicked by TPA treatment (Takaishi et al. 1994).

Calcium release and activation of mitogen-activated protein (MAP) kinase(s) by HGF/SF

Hepatocytes stimulated in vitro with HGF/SF respond with a rapid rise in the cytosolic free calcium concentration (${\text{Ca}}_{\text{c}}^{2+}$) (Baffy et al. 1992). HGF/SF elevates intracellular calcium concentrations (${\text{Ca}}_{\text{i}}^{2+}$) in a dose-dependent manner during the early stages of the cell cycle. In calcium-free buffer, the increase in Ca²⁺ is derived from intracellular stores. The effects of EGF and HGF/SF are comparable and partly additive with regard to InsP₃ production and calcium mobilization (Okano et al. 1993). Pre-treatment of cells with genistein abolishes the HGF/SF-induced calcium response. More specifically, HGF/SF induced periodic fluctuations of (Ca²⁺)_c and (Ca²⁺)_i in hepatocytes are suppressed by pretreatment with PMA (Osada et al. 1992a). Administration of PMA blocks the calcium oscillations, an effect reversed by the subsequent addition of H-7, a PKC inhibitor. Pre-treatment with H-7 increases (Ca²⁺)_i peaks elicited by HGF/SF, suggesting negative modulation via PKC of the HGF/SF-induced repetitive (Ca²⁺)_i transients (Osada et al. 1992b). The magnitude of the calcium transients is lower than that induced by EGF and pretreatment of cells with EGF results in a greatly enhanced effect of HGF/SF (Mine et al. 1991). Stimulation of primary hepatocytes with HGF/SF results in a rise in intracellular pH (pHi), sensitive to amiloride and mediated via a tyrosine kinase-calcium and calmodulin-dependent pathway (Kaneko et al. 1995). When hepatocytes, pre-equilibrated in vitro at extracellular calcium concentrations (${\text{Ca}}_{\text{e}}^{2+}$), similar to those corresponding to normal or decreased levels of circulating calcium, are treated with EGF, the (Ca²⁺)_e strongly influences the (Ca²⁺)_i response, with a significantly smaller increase occurring in normocalcaemic than in hypocalcaemic hepatocytes (Bilodeau et al. 1995). By inhibiting receptor autophosphorylation or calcium channels, the EGF-induced increase in (Ca²⁺)_i is also abolished (Zhang & Farrell 1995). In quiescent hepatocytes, the cytosolic calcium is 1.6–2-fold higher than that of nuclear calcium and addition of EGF results in the preservation of the nuclear-to-cytosolic gradient. It has been suggested that the hepatocyte nuclear membrane contains calcium permeability barriers and growth-factor sensitive calcium transport mechanisms (Waybil et al. 1991). An increase in intranuclear calcium may be responsible for activating calcium-dependent endonucleases, nuclear calmodulin, or nuclear PKC, thus regulating cellular proliferation (Waybil et al. 1991).

Treatment of GTL16 gastric carcinoma cells, which express an amplified HGF/SF receptor, with HGF/SF in the presence of A23187 or ionomycin (calcium channel ionophores) results in a rapid and reversible decrease of p145^MET tyrosine phosphorylation, demonstrating that increased serine phosphorylation induced by the rise in intracellular calcium levels negatively modulates the HGF/SF receptor tyrosine kinase activity (Gaudino et al. 1991). This inhibition of tyrosine kinase activity is likely to be effected by a calcium-dependent serine kinase distinct from PKC.

HGF/SF increases Raf and mitogen-activated peptide (MAP) kinase activity in hepatocytes through a protein kinase C and calcium-independent pathway (Gines et al. 1995). PMA is unable to stimulate Raf and MAP kinase activity in hepatocytes despite a marked activation of PKC activity (Gines et al. 1995). The HGF/SF receptor associated with the Grb2-SOS complex is known to interact with and activate the membrane-associated Ras, leading in turn to the activation of Raf (Blenis 1993; Gines et al. 1995). The c-raf gene encodes a serine/threonine kinase regulated by phosphorylation on tyrosine and serine residues. MEK kinase (MAP kinase/Extracellular signal regulated kinase {ERK}, ERK-activating kinase) is activated by Raf phosphorylating activity (Blenis 1993; Hill & Treisman 1995).

The activated MEK in turn phosphorylates MAP kinase, which subsequently activates (by phosphorylation) several nuclear transcription factors, including c-myc and CCAAT/enhancer binding protein (C/EBP) β (Gupta & Davis 1994). MAP kinase also phosphorylates the pp90^rsk and RSK (S6 kinase), which have overlapping roles in phosphorylating a variety of nuclear and cytoplasmic targets and modulating the activity of cellular phosphatases (Blenis 1993; Hunter 1995). The HGF/SF receptor tyrosine kinase phosphorylates p42/p44 MAP kinase in cultured hepatocytes (Stolz & Michalopoulos 1994). MAP (pp54 and pp42/44) serine-kinases, which are regulated by tyrosine as well as by serine/threonine phosphorylation, phosphorylate two serines in positions 62 and 72 of the c-jun protein (Pulverer et al. 1991). Serine phosphorylation of c-jun then results in positive regulation of its transactivating activity. MAP kinase(s) have been reported to phosphorylate (Ser505) and activate cytosolic phopholipase A2 (cPLA₂), a key enzyme in arachidonic acid release (Lin et al. 1993). TGF-α has been reported to induce tyrosine phosphorylation and activation of PLA₂ in rat thymic epithelial cells, leading to subsequent association with the cell membrane (Liu et al. 1993). EGF and TGF-α promote release of prostaglandin E₂ (PGE₂) and PGF_2α in cultured hepatocytes, an event also observed in regenerating liver following partial hepatectomy (Little et al. 1988; Skouteris et al. 1988; Skouteris & Kaser 1991). In the latter circumstance, PGE₂ levels increase within hours, with stimulation of hepatocyte DNA synthesis occurring via a receptor-mediated effect (Little et al. 1988). Primary hepatocytes possess PGE₂ receptors of the EP3-type, a 362 amino acid protein displaying over 95% homology to the EP3 beta receptor from mouse mastocytoma. In isolated hepatocytes, PGF_2α activates glycogen phosphorylase and the release of intracellular InsP₃. Pre-treatment with PMA substantially reduces levels of InsP₃ (Neuschafer Rube et al. 1994). Plasma membranes from hepatocytes possess both high capacity/low affinity and low capacity/high affinity binding sites for PGF_2α. The high capacity/low affinity binding sites are likely to represent the receptor to which binding is unaffected by PMA treatment (Neuschafer Rube et al. 1993).

HGF/SF releases PGE₂ and PGF_2α from hepatocytes (Little et al. 1988; Skouteris & Kaser 1991). Addition of these prostaglandins to hepatocyte cultures induces significant DNA synthesis which is additive to that occurring in response to EGF or HGF/SF alone (Skouteris & Kaser 1991; Neuschafer-Rube et al. 1993). Hepatocyte DNA synthesis induced by either TGF-α or HGF/SF is reduced by inhibitors of cyclo-oxygenase and/or lipo-oxygenase (Skouteris et al. 1988; Adachi et al. 1995; Georgakopoulos et al. 1995; Skouteris & Schröder 1996). It is therefore likely that the release of prostaglandins is sustained by two overlapping mechanisms. In the regenerating liver, DNA synthesis occurs at the same time as increases in lipocortin-1 and 2 levels are observed. In the one week old rat and mouse liver, proliferating cell nuclear antigen-positive hepatocytes contain large amounts of lipocortin-1 and -2, in contrast to livers with no DNA synthetic activity (Masaki et al. 1994). These results further support the concept that lipocortins are involved in the regulatory processes of hepatocyte proliferation. PGI₂ and PGD₂ also increase hepatocyte DNA synthesis, although to a lesser extent (Masaki et al. 1994). All prostaglandins tested as stimulators of hepatocyte DNA synthesis cause a rise in intracellular InsP₃ in proportion to their activity as DNA synthesis stimulants. An involvement of an InsP₃-PLC-specific growth-promoting pathway for prostaglandins in primary hepatocytes is also suggested by the concomitant rise in cytosolic free calcium following eicosanoid treatment (Masaki et al. 1994). Another aspect of prostaglandins as growth modulators is revealed by data showing that PGE₁ and PGD₂ treatment of human skin fibroblasts markedly stimulate production of HGF/SF. Both prostaglandins were also shown to stimulate HGF/SF production in MRC-5 human embryonic lung fibroblasts (Matsumoto et al. 1995b).

Conclusion

Hepatocyte proliferation is triggered by stimuli originating from the liver microenvironment including re-organized ECM and stellate, endothelial and Kupffer cells. The nature of the signals which induce a state of competence in the hepatocyte remain largely unknown. Candidates for this role are ECM-bound growth modulators, such as HGF/SF, TNF-α, interleukins and other presently unknown factors released by neighbouring cell types. Proliferation-arrest signals for the hepatocyte are mostly unknown at present, although it may be envisaged that these two phenomenally distinct categories of signals may be closely related.

Following cell stimulation with HGF/SF and coupling to its receptor, the signal is then coupled to ‘immediate transducers’, namely PI 3-K and Grb2. These two effectors orchestrate the phenotypic response to HGF/SF. From data reported so far, it can be concluded that PI 3-K-and Grb2-driven intracellular pathways may lead to either mitogenesis, scattering or morphogenesis. Most cell types, including both primary cells and cell lines, respond to HGF/SF by developing a mixed phenotype. Therefore, it is anticipated that expression of a particular phenotype requires involvement of specific intracellular effectors unique to, or at least highly abundant in, a particular cell type. In other words, if a cell type responds to HGF/SF mostly by increasing its motility, ¤final acceptors¤ of the HGF/SF signal would be intracellular effectors regulating that specific function. At the same time, effectors implicated in the mitogenesis pathway may be repressed in terms of expression or increase in the synthesis of mitotic inhibitors.

PI 3-K and Grb2 together with Ras and MAPK seem to provide an initiatory signal-platform which alone does not lead to a specific phenotype. At a ‘secondary’ stage, intracellular effectors are recruited and activated, potentially leading to the integration of the initiatory signal into a specific phenotype. An example of such an interlinked network would be that leading to the expression of a motile rather than a proliferating phenotype in response to HGF/SF by NIH3T3 fibroblasts expressing exogenous c-MET. This particular model has not been investigated in depth regarding the effectors responsible for increased motility. Research efforts should be targeted at the identification of cell-specific transducers which integrate the HGF/SF signal to a specific phenotype. By these means, genes encoding intracellular transducers could be exogenously expressed in cells where diversion from a particular phenotype is required. Alternatively, peptides bearing active sites for these effectors might be imported into cells and via competition with the sites in endogenous counterparts could lead to the ‘switching off or enhancement’ of a particular phenotypic response.