HGF was identified in 1984 as a long-sought hepato-trophic factor. HGF was purified from serum of rats after 70%-partial hepatectomy [2], and this factor potently stimulated DNA synthesis and proliferation of primary culture of adult rat hepatocytes. In the mid-1980s, we purified HGF to homogeneity from blood platelets of 3000 rats and demonstrated that it was a new growth factor [18]. Based on complete purification of native HGF, cloning of HGF cDNAs was successfully performed, followed by the complete amino acid sequence of human HGF determined from the nucleotide sequences of cDNAs [3,4] (Figure 1A).

Scatter factor (SF) was identified as an embryonic lung fibroblast (MRC5)-derived factor that enhances motility of renal cells (MDCK) [19]. Such SF-induced motogenic activities were abolished by anti-HGF antibody, while recombinant HGF mimicked the scatter activity in the culture of MDCK [20]. Indeed, MRC5-fibroblasts expressed 6 kb mRNA that was hybridized with a HGF cDNA probe, and SF cDNA cloned from the MRC5 cDNA library had the same sequence as HGF of cDNA from human leukocytes [20]. Overall, it was shown that SF is identical to HGF [20,21], suggesting roles of HGF/SF during cancer metastasis.

In the mid-1980s, MET was identified as a “mutant” oncogenic gene. Dean et al. isolated MET from carcinogen-induced osteosarcoma cells (MNNG-HOS), which induced NIH-3T3 transformation in soft agar [17]. The MET proto-oncogene is localized to the seventh chromosome (7q21–q31) in humans. The MET-encoding protein has a tyrosine kinase activity, suggesting that MET is an orphan receptor of growth factors. On the other hand, HGF was identified as a mitogen of hepatocytes [2–4], but in the late 1980s, its receptor was still unknown.

Bottaro et al. reported in 1991 that MET may be a signaling receptor for HGF, along with the tyrosine phosphorylations [9]. The protein band was identified as an antigen of anti-MET antibody in the cross-linking of ¹²⁵I-labeled NK1 (an HGF-variant) to cellular protein extraction. High-affinity binding cellular sites specific to HGF were detected in MET cDNA-transfected COS7 cells, with a Kd value of 30 pM. Only MET-transfected COS7 showed the mitogenic response to HGF, thus indicating that “wild-type” MET is an HGF-receptor [10].

In the extra-cellular domain of MET, the first 519 amino aids form a 7-bladed β-propeller domain, called Sema domain, which is necessary for binding to the HGF β-chain and activation of the receptor by ligation. The MET immunoglobulin-like domain is required for binding to the first kringle domain (NK1) of HGF α-chain. HGF induces MET activation via the formation of 2:2 complex where MET–MET dimerization is mediated via dimmer-formed HGF [7,8] (Figure 1B).

The accurate binding of HGF to MET triggers signaling transduction. The ATP-dependent phosphorylation at three residues in the MET active loop kinase domain, Tyr-1230/34/35, is an initial step for activating MET. Phosphorylation at Tyr-1349/56 in the C-terminal docking site is required for various bio-functions via recruiting downstream adopters. For example, phospho-Tyr1349/56-dependent recruitment of Grb2-SOS activates Ras-ERK cascades, leading to cell proliferation. Association and tyrosine phosphorylation of Gab-1, a docking protein that couples MET with multiple signaling proteins such as PI-3kinase, PLC-γ, Shp-2, and Crk-2, plays definite roles in HGF-induced morphogenesis and motility. Various adaptor molecules, recruited at MET docking-site, explain multi-faceted functions of HGF [7,8] (Figure 1B).

In embryonic organs, HGF is mainly produced in stroma cells such as fibroblasts, while MET is expressed by parenchymal cells [4–8], suggesting that the paracrine loop of HGF–MET is critical for organ development. Indeed, HGF acts as a mesenchyme-derived morphogenic factor during fetal lung development. When HGF-antisense oligo-DNA was added to embryonic lung cultures, stromal HGF production became negligible, followed by impairment in lung branching morphogenesis [22]. In vivo, lung alveoli-specific MET gene destruction led to a decrease in alveologenesis in mice. Organ-specific MET deletion techniques revealed pivotal roles for HGF in development of various organs, such as liver, kidney muscle, etc.[6,8].

Endogenous HGF is also important for tissue repair and protection in vivo. We provided the first evidence that recombinant HGF enhanced hepatic regeneration in mice with hepatitis [23]. Therapeutic effects of recombinant HGF were also seen in injured organs, such as kidneys, lung, stomach–intestine, skin, heart, brain etc.[8]. Blood HGF levels markedly increase in patients and rodents during tissue injuries. When anti-HGF antibody was administered to a rat model of myocardial infarction, cardiac damage was exacerbated [24]. Such a key role of endogenous HGF has been seen in acute and chronic organ diseases [8,25]. Thus, compensation for the loss in intrinsic HGF by HGF administration is a logical strategy to improve organ failures [8,25].

Since constitutive activation of the MET signal is one of the key oncogenic events, it is important to discuss its molecular basis, focusing on downstream MET. Using some mutations of MET identified in patients with familial PRC, Giordano et al. found that some mutated MET enhance the Ras signaling pathway [32]. Other mutations are devoid of transforming potential but are effective in inducing protection from apoptosis, associated with the efficient interaction of PI3Kinase. Thus, different mutations in the MET gene may elicit tumorigenesis via Ras-based mitogenesis and PI3Kinase-based protection pathways [32].

β-catenin is an oncogenic protein involved in the regulation of cell–cell adhesion and gene transcription. Once β-catenin is activated, it is translocated from sub-membrane to nucleus, and then β-catenin functions as an activator of Tcf-transcriptional factor, leading to upregulation of mitogenic proteins, such as c-Myc and cyclin-D1. HGF activates β-catenin via a direct assembly of phospho-MET and β-catenin. Likewise, active MET mutants (such as M1268T) also activate the nuclear β-catenin function [33]. Thus, aberrant activation of nuclear β-catenin by active MET leads to tumor formation via the upregulation of c-Myc and cyclin-D1.

Src kinase activity is important for mutant MET to acquire tumorigenic potentials. In the NIH-3T3 clone (F4) that overexpresses the mutant MET (M1268T), malignant formation was evident in both soft agar and implanted nude mice [34]. These malignant phenotypes were abolished by dominant negative c-Src in the NIH-3T3-F4 clone. A recent report described how the direct assembly of MET by a membrane Fas-ligand is also required for tumorigenic potentials, possibly via sequential phosphorylations from MET to the downstream Stat3, a key regulator for enhancing anchorage independent growth [35].

MET overexpression and its crosstalk with other receptors, such as EGFR, Her2, Her3 are also critical for tumor development. EGFR tyrosine kinase inhibitors (EGF-TKI) intercept EGFR-mediated downstream signaling pathways (such as PI3K-AKT) via de-phosphorylation of Her3, while some lung carcinoma cell lines acquire resistance via amplification of MET [36]. In this process, amplified MET trans-phosphorylates Her3, and then impaired AKT pathways are restored, resulting in resistance to EGFR-TKIs, including Gefitinib (Iressa^®) [36,37]. Such a redundant signaling via amplified MET may participate in breast cancer malignant behaviors during the treatment with anti-Her-2 antibody, Trastuzumab (Herceptin^®) [38], possibly due to association (and re-activation) of Her2 with amplified MET [39,40]. Thus, MET inhibition can be a target to block the redundant pathways, required for drug resistance (see Section 6.4).

In summary, several lines of human and animal studies have demonstrated that not only MET mutations, but also HGF over-productions, lead to oncogenenic outcomes via over-activation of MET-associated molecular pathways, such as Ras, β-catenin etc. (Table 1).

Tumor cell scattering in the primary tumors is the first step for metastasis. In tumor tissues, stroma-secreted HGF is required for cancer cells to infiltrate neighboring tissues, such as vascular beds, across the basement membrane. The paracrine loop between HGF-producing stroma and MET-expressing tumor cells acts as a local switch for the invasive phenotypes, as follows.

In addition to adjacent invasion, the HGF–MET axis is important for distant metastasis, through extravasation, anti-anoikis and homing. HGF promotes cancer metastasis via inducing vascular bed formation and enhancing chemokine-induced homing. HGF also protects cancer cells from anoikis or immunological challenge. These sequential events lead to successful metastasis.

In addition to oncogenic mutants, the stromal environment plays a key role in cancer prognosis. Carcinoma-associated fibroblasts (CAF), but not normal fibroblasts, promote tumor progression of initiated non-tumorigenic epithelial cells. Stroma-derived HGF is responsible for tumor invasive growth [11–13]. Thus, the molecular basis of HGF production should be discussed.

Numerous types of carcinoma cells secrete soluble factors that induce HGF production in stromal cells (i.e., HGF-inducers). For example, conditioned medium of breast cancer cells enhances HGF production in fibroblasts, along with a raise in prostaglandin-E2 [64]. Inhibition of prostaglandin-E synthesis by indomethacin leads to downregulation of HGF production and suppression of tumor migration, indicating that cancer-derived prostaglandins are important for stromal HGF production [64]. Other cancer-derived HGF-inducers are IL-1β, b-FGF, PDGF, and TGF-α [12,65]. Such a paracrine loop, mediated by the cancer-derived HGF-inducers and stroma-secreted HGF, confers a key mechanism of tumor metastasis [66].

In addition to CAF, tumor-associated macrophages (TAM) are a major source of HGF in tumor tissues. TAM isolated from 98 primary lung cancer tissues shows the higher ability to produce HGF. This is also the case for HGF produced by neutrophils infiltrating bronchiolo-alveolar subtype pulmonary adenocarcinoma [67]. Serum levels of HGF are elevated in patients with recurrent malignant tumors [68], thus suggesting an endocrine HGF delivery system. In this regard, peripheral blood monocytes are known to produce HGF, contributing to the increase in blood HGF levels via an endocrine mechanism [69].

In summary, stroma-derived HGF is required for cancer invasion (i.e., paracrine system). Moreover, blood leukocytes inhibit anoikis and enhance SDF1 signaling in tumor cells during blood flow via a secretion of HGF into blood (i.e., endocrine system). The roles of HGF during tumor growth (see Section 3) and metastasis are summarized in Table 1.

NK4 was initially purified as a fragment from elastase-digested HGF samples [15,70]. The amino acid sequence revealed that NK4 is cleaved between Val⁴⁷⁸ and Asn⁴⁷⁹ by elastase. The N-terminal structure of NK4 is the same as undigested HGF (i.e., 32nd pyroglutamate), indicating that NK4 is composed of the N-terminal hairpin domain and 4-kringle domains (thus designated NK4) (Figure 2A). Precursor HGF is cleaved between Arg⁴⁹⁴ and Val⁴⁹⁵, while NK4 is cleaved between Val⁴⁷⁸ and Asn⁴⁷⁹. In other words, NK4 is identical to HGF α-chain that lacks C-terminal 16 amino acids (i.e., Asn⁴⁷⁹ to Arg⁴⁹⁴) (Figure 2A). The domains that are responsible for high-affinity binding to MET are the N-terminal hairpin and the K1 domains in NK4.

Although NK4 is a MET-binder, NK4 alone does not phosphorylate MET tyrosine residues, indicating that NK4 acts as a complete HGF-antagonist (Figure 2B). Actually, MET tyrosine phosphorylation occurs in A549 lung carcinoma cells within 10 min after HGF addition, while NK4 inhibits the HGF-mediated MET activation. HGF induces invasion and migration of the gallbladder cancer cells in Matri-gels, while NK4 inhibits HGF-induced invasion [15]. These anti-invasive effects of NK4 are also seen in distinct types of cancer cells, strengthening the common role of NK4 in cancer migration [66].

Other HGF fragments, NK1 (composed of hairpin-loop and K1 domain) and NK2 (hairpin-loop and K1–K2 domains), are natural variant forms of HGF. NK1 is often agonistic to MET, particularly in the presence of heparin, but it sometimes elicits antagonistic effects. NK2 is antagonistic but sometimes agonistic to HGF–MET signaling. Indeed, forced expression of NK2 transgene leads to the acceleration of metastasis in mice bearing melanoma [71]. In contrast to these variants, NK4 is a complete antagonist (i.e., without agonistic features). Therefore, this fragment is reasonable as a practical tool for anti-metastatic strategies (see, Section 6).

Vascular EC expresses MET, and HGF stimulates mitogenic and morphogenic activities in EC [4,61]. Thus, it is possible that NK4 inhibits HGF-induced angiogenesis. Expectedly, NK4 potently inhibited the HGF-mediated proliferation of EC in vitro[72]. Strikingly, NK4 also inhibited EC proliferation, induced by other angiogenic factors, such as b-FGF and VEGF [54]. In this process, HGF and VEGF phosphorylate MET and KDR/VEGF-receptor, respectively, whereas NK4 inhibits HGF-induced MET tyrosine phosphorylation, but not KDR activation [54]. Nevertheless, NK4 inhibits VEGF-induced proliferation, without modification of VEGF-mediated ERK1/2 activation. The same effect of NK4 was also seen in vivo: b-FGF-induced angiogenesis in the rabbit cornea was inhibited by NK4. Thus, another mechanism (i.e., MET-independent pathway) may be involved in NK4-mediated arrest of angiogenesis. This alternative pathway is demonstrated in 2009, as follows.

The fibronectin-integrin signal in EC is required for angiogenesis. The NK4-mediated growth arrest of EC is due to a loss of the fibronectin-integrin signal [73]. The affinity purification with NK4-immobilized beads revealed that perlecan is a counterpart to trap NK4. Perlecan is a heparan sulfate proteoglycan that interacts with basement membrane ECM, such as fibronectin. Deletion of perlecan expression by siRNA diminished the fibronectin assembly and EC spreading, indicating a key role of fibronectin-perlecan interaction during EC migration. Notably, NK4-perlecan interaction reduced the assembly of fibronectin by perlecan. As a result, FAK activation became faint due to the inhibition of perlecan-fibronectin complex with NK4. Under such a loss of integrin signals, EC growth was impaired [73]. This MET-independent pathway is involved in the inhibitory effect of NK4 on VEGF-mediated angiogenesis (Figure 3).

In addition to pancreatic protease, endogenous proteases can generate an NK4-like fragment. Leucocyte-derived proteases, such as mast cell chymase, are able to convert HGF to fragments, including NK4 and remnant HGF-β [74,75]. Endogenous cathepsin-G and MMP also produce NK4-like fragment via its enzymatic reaction, suggesting that de novo synthesized NK4 may have a counteractive role in inhibiting HGF-mediated functions under pathological conditions.

For example, NK4-like fragments have been detected in the effusive fluid from skins with intractable ulcers in patients [75]. In this process, kallikrein, an endogenous serine-protease, is critical for HGF-to-NK4 conversion at the local sites. Endogenous HGF is required for skin regeneration [8]. Thus, it is likely that the synthesis of HGF-antagonistic NK4, in addition to the decrease in HGF (i.e., intrinsic repair factor), underlies the key mechanism whereby epidermal regeneration is impaired in the patients with intractable leg ulcers [75].

During tumor progression, cancer cells may produce NK4 for tumor growth retardation. Indeed, NK4-like fragments are detectable in the culture supernatants of certain tumor cell lines that produce HGF [76,77]. Although it is still unclear whether NK4-like fragment is generated via an enzymatic cleavage or variant transcription, NK4-like fragment(s) could act as a natural HGF-antagonist to minimize tumor metastasis (as a negative feedback). Thus, exogenous NK4 supplementation may be a reasonable approach to arresting tumor malignancy, as follows.

HGF, or co-cultured fibroblasts, induces invasion of gallbladder carcinoma cells (GB-d1) across Matrigel, an experimental mimic of the basement membrane [46]. NK4 competitively inhibits the binding of HGF to MET on GB-d1 cells. As a result, NK4 diminishes HGF-induced, or fibroblast-induced, motogenic activities [16]. Such a major role of HGF was also seen in vivo. Subcutaneous inoculations of GB-d1 cells in nude mice allow for primary tumor growth and invasion to adjacent muscular tissues. Using NK4 in this model, we provided the first evidence that HGF- antagonist can arrest tumor invasion in vivo[16]: recombinant NK4 inhibited the growth and muscular invasion in mice bearing GB-d1 carcinoma. Consistent with growth arrest, apoptotic change was evident in NK4-treated mice. This is the first proof of concept to show that HGF-antagonist is useful for anti-tumor therapy.

In culture of EC, NK4 produces anti-angiogenetic effects via a MET-independent pathway, as reported [54,73]. These effects are also observed in animal models of malignant tumors: recombinant NK4 suppressed the primary tumor growth, metastasis of Lewis lung carcinoma, and Jyg-MC(A) mammary carcinoma in mice [54], although neither HGF nor NK4 affected proliferation and survival of these tumor cells in vitro. NK4 treatment resulted in a remarkable decrease in vessel density and an increase in apoptotic cells in the tumor tissues [54]. NK4-induced anti-angiogenic effects are reproduced in various types of cancers [79–81]. Since inhibition of angiogenesis by NK4 results in tumor hypoxia, hypoxia-primed apoptosis may be involved in cancer growth arrest by NK4.

Anti-tumor effects of NK4 is also observed in advanced carcinoma with metastasis [79]. When NK4 treatment was initiated on day 10 (a time when cancer cells were already invading surrounding tissues), NK4 potently inhibited the tumor growth, peritoneal dissemination, and ascites accumulation at four weeks after the tumor inoculation. As a result, NK4 prolonged the survival time of mice at an end-stage of cancer (Figure 4). Because effective systemic therapy for pancreatic cancer is currently not available, and diagnosing pancreatic cancer in its early stages is difficult, the highly invasive and metastatic behaviors lead to difficulty in attaining a recurrence-free status. Blockade of HGF-mediated invasion may prove to be potential therapy for advanced stages of pancreatic carcinoma.

Chemotherapy, radiation therapy and anti-angiogenic treatment are applicable in patients with inoperable tumors. Recent reports demonstrated the synergic anti-tumor effect of NK4 on these conventional therapies in rodents, as follows.

We provided evidence in 2001 that not only NK4 but also anti-HGF antibody is useful for improving the prognosis of malignant tumors, using pancreatic cancer-bearing mice [79]. In the same year, another group also demonstrated the anti-tumor effect of anti-HGF antibody in a xeno-graft model [105]. These studies stimulated the pharmaceutical companies to develop several types of antibodies specific to HGF or its receptor, MET, as follows.

Small molecule MET tyrosine kinase (TK) inhibitors may be practical in a clinical setting, due to an oral administration. Binding of HGF to MET induces a dimer–dimer complex (2:2) and phosphorylates kinase domain active loop (i.e., Y1230/34/35) via recruiting ATP. These kinase domain activations rapidly elicit phosphorylation of multi-docking sites (i.e., Y1349/56), and then various activities are producible via recruiting downstream adapter, Grab2 [114]. Thus, blockade of this trans-phosphorylation can be an anti-cancer strategy, as follows.

PHA-665752 was identified as a small-molecule, ATP-competitive inhibitor of the catalytic activity of the MET kinase domain loop (i.e., Y1230/34/35). PHA665752 reduced NCI-H69 (small-cell lung cancer) and NCI-H441 (non-small-cell lung cancer) tumorigenicity in mouse xenografts by 99% and 75%, respectively [115]. PHA665752 inhibited MET phosphorylation and c-Cbl binding sites in mouse xenografts derived from non-small cell lung cancer cells (NCI-H441 and A549) and small cell lung cancer cells (NCI-H69) [115].

ATP-competitive inhibitors may suppress MET mutation-induced (i.e., HGF-independent) auto-phosphorylation by counteracting the kinase domain (Y1230/34/35) activation. However, some MET mutants are not sensitive to ATP-competitive inhibitors. Under such a resistant state, the MET multi-docking region may be a downstream target. For example, MK-2461, a novel multi-targeted TK inhibitor, suppresses phosphorylation of Y1349, but not Y1234/35 [116]. As a result, MK-2461 reduced the malignant phenotypes in mice bearing MET-mutated tumors.

Collectively, not only targeting the kinase domain with ATP-competitive inhibitors but also inhibiting multi-docking sites may provide a therapeutic approach. Although small-sized TK inhibitors facilitate an oral administration (i.e., systemic exposure), an effective dose might impair non-tumor normal organs, since endogenous HGF is essential for organ homeostasis [8]. Clinical trials of MET TK inhibitors (such as PHA-665752 or SU-11274) are now being developed [117]. The tumor-specific delivery of these drugs warrants ongoing attention.

Decorin is a member of the small leucine-rich proteoglycan gene family and it can impede tumor growth, but its molecular basis has not been elucidated. A recent study identified decorin as a MET antagonist. Decorin binds to MET and phosphorylates tyrosine sites in the kinase domain (i.e., Y1234/35) in Hela-cells within 15 min. post-addition [118]. Nevertheless, decorin does not phosphorylate the MET multi-docking site, Y1349. As a result, decorin inhibited tumor growth in vivo, probably via the inhibition of the β-catenin pathway.

Norleual, an angiotensin-IV analog, exhibits a structural homology with the hinge region of HGF. Norleual competitively inhibited the binding of HGF to MET in mouse liver membranes, with an IC₅₀ value of 3 pM [119]. Predictably, norleual inhibited HGF-dependent signaling, proliferation, migration and invasion in the cultures of multiple cell types at concentrations within the picomolar range. In vivo, norleual suppressed the pulmonary colonization by B16-F10 melanoma in mice [119]. Thus, angiotensin-IV analogs have utility as agents in metastatic disorders via the interference with HGF–MET systems.

There are now numerous candidates for HGF-antagonistic and MET-targeted strategies (Table 3). Soluble MET-Sema and phage display-based peptides against MET may be useful as a decoy or masking agent [120,121]. Aspirin is a classic type COX2 inhibitor that reduces tumor malignancy via an inhibition of prostaglandin-dependent HGF production [122]. In addition, natural products included in foods or plants, such as diet-derived flavonoids [123] and green-tea catechins [124] exert unique activity that blocks HGF-mediated malignant behaviors. These drugs or natural products will produce or support anti-tumor outcomes.