It has been established that development of a carcinoma is due to the accumulation of somatic mutations that occur in oncogenes and tumor suppressor genes in epithelial cells. A better understanding of genetic changes that occur in cancer cells would also facilitate progress in new therapeutic approaches such as “molecular target therapy.” On the other hand, the neoplastic transformation of epithelial cells and the malignant behavior of carcinoma cells are influenced by their interactions with neighboring stromal components, including fibroblasts, blood vessels, inflammatory cells and the extracellular matrix.1, 2 Recombinant grafts in which nontumorigenic epithelial cells are combined with particular cancer-associated fibroblasts induce tumorigenic progression.3 Recent studies have demonstrated that experimentally induced genetic alterations in stromal fibroblasts induce epithelial neoplasia and invasive carcinoma.4, 5 Meanwhile, cancer cells themselves alter their adjacent stroma to form a permissive and supportive microenvironment (known as the “reactive” tumor stroma) by producing growth factors and cytokines (e.g., vascular endothelial cell growth factor (VEGF), basic fibroblast growth factor (bFGF), platelet-derived growth factor (PDGF), transforming growth factor-β (TGF-β), interleukins and colony stimulating factors).1 These factors induce stromal reactions such as angiogenesis and inflammatory reaction, and also activate stromal cells in a paracrine manner, leading to the secretion of additional growth factors and proteases. An understanding of substances that mediate mutual interactions between epithelial cells of normal or neoplastic and surrounding stromal cells would provide new insights into tumor biology and therapeutics.
Hepatocyte growth factor (HGF) has attracted considerable attention as a stromal-derived mediator in tumor–stromal interactions, particularly based on its close involvement in cancer invasion and metastasis. Since the fatal nature of carcinoma is the result of invasion and subsequent metastasis, processes that are characteristic to malignant cells, new therapeutic approaches targeting HGF and its receptor, Met tyrosine kinase, are ongoing. This review focuses on the significance of the HGF–Met system in tumor–stromal crosstalk.
HGF and Met: brief background
HGF was first identified as a mitogenic protein for hepatocytes6 and its molecular cloning indicated that HGF had no structural similarity to other known growth factors but had a structural similarity to plasminogen.7, 8 HGF is a heterodimeric protein composed of an α-chain containing the N-terminal domain and 4 kringle domains, and a β-chain containing a serine protease-like motif. The receptor for HGF was identified as Met, a transmembrane tyrosine kinase.9 Met is the only known functional receptor for HGF, and HGF is the only natural ligand for Met.
Under normal conditions, HGF and Met play roles in embryonic development, epithelial–mesenchymal transition, angiogenesis and tissue regeneration, including the liver.10, 11 In 1990, the purification and determination of the partial amino acid sequence of scatter factor indicated that it is structurally similar to HGF,12 and it was subsequently shown to be identical to HGF. Scatter factor was originally identified as a fibroblast-derived cell motility factor for epithelial cells.13 Likewise, a fibroblast-derived morphogenic factor, which induces epithelial branching tubulogenesis, was also determined to be HGF.14 These early studies suggest the possibility that HGF might be a mesenchymal-derived mediator in epithelial–mesenchymal interactions, a fundamental tissue interaction involved in the development and morphogenesis of distinct types of organs.
During development, HGF and Met provide essential signals for development of the liver, placenta and skeletal muscle.11 In HGF−/− or Met−/− murine embryos, the liver is reduced in size and the impaired placental development is responsible for their embryonic death. Likewise, impaired delamination and migration of skeletal-muscle-progenitor cells from the dermomyotome to the limbs and diaphragm results in the absence of muscle groups in these mice. In adults, HGF plays roles in the regeneration and protection of various organs, including the liver.10 The hepatocyte-specific disruption of the Met gene results in a markedly reduced ability of the liver to regenerate.15, 16
Fibroblast-dependent tumor invasion
The involvement of stromal fibroblast-derived factor in the invasion of carcinoma cells was first demonstrated using oral squamous cell carcinoma cells cultured on a collagen gel matrix.17 When the cancer cells were cultured alone, no invasion was observed, whereas they invaded collagen gels in which fibroblasts were incorporated or when fibroblast-derived conditioned medium was added to the culture. These findings indicate that fibroblast-derived soluble factor is responsible for the induction of invasive behavior in oral squamous cell carcinoma. The factor was subsequently shown to be HGF.18 On the other hand, during the purification and characterization of scatter factor (prior to the confirmation that it was HGF) Weidner et al. reported that it induced the invasion of pancreatic cancer cells into collagen gels, indicating that HGF plays the role of a fibroblast-derived factor in the invasion of cancer cells.12
Following these early studies, the enhancement of migration and invasion of cancer cells in the presence of stromal fibroblasts was demonstrated in a variety of cancer cells, including squamous cell carcinoma cells, breast carcinoma, gallbladder carcinoma, esophageal cancer and prostate cancer.19, 20, 21, 22, 23 An example of the remarkable invasive behavior in human gallbladder cancer cells in the presence of fibroblasts is shown in Figure 1a. In these studies, the enhancement in cancer cell invasion by cocultivation with fibroblasts was inhibited by NK4, a specific competitive antagonist against HGF–Met (see later for details) or a neutralizing antibody against HGF. The induction and enhancement of cancer cell invasion by HGF has been demonstrated in a wide variety of cancer cells.23 These results indicate that the invasion of various cancer cells is enhanced by fibroblast-derived factors and that HGF is a significant factor responsible for cancer cell invasion mediated by tumor–stromal interactions.
The invasion of tumor cells is regulated by distinct cellular functions, including cell–cell adhesion, cell–matrix association, the proteolytic breakdown of the extracellular matrix and cellular locomotion. HGF stimulates (1) the dissociation of cancer cells at the primary site, (2) invasion through the basement membrane and host stroma by enhancing cell–matrix interactions, protease networks for the breakdown of extracellular matrix proteins, and motogenic responses (Fig. 1b).11, 23, 24 Experimentally induced autocrine activation of Met tyrosine kinase by the transfection and expression of HGF or the Met gene in nontransformed cells resulted in tumorigenic transformation, and the progression to invasive and metastatic tumors in nude mice.25, 26, 27 Aberrant regulation of Met receptor expression or mutational Met activation is associated with malignant progression and the behavior of a variety of cancers in humans.11, 23, 28 In human gastric adenocarcinoma tissues, the in situ tyrosine phosphorylation of Met was detected in poorly differentiated adenocarcinoma, including in the invasion front, whereas Met was only weakly phosphorylated in a normal gastric mucosa even though Met was expressed.29 Likewise, the Met receptor was in situ tyrosine-phosphorylated in invasive carcinomas in mice, whereas the inhibition of Met tyrosine phosphorylation was associated with the suppression of the invasive behavior.30 These results indicate that the in situ activation of Met is associated with the invasive and metastatic progression of cancers.
Crosstalk mediated by HGF-inducers and HGF
Stromal components associated with carcinoma cells are composed of extracellular matrix and distinct types of cells, including those comprising the vasculature (endothelial cells, pericytes and smooth muscle cells), inflammatory cells (lymphocytes, neutrophils and macrophages), fibroblasts and tumor (or carcinoma)-associated fibroblasts with myofibroblast-like characteristic. Among these cell types, fibroblasts derived from a variety of tissues express HGF. In addition, it has been shown that vascular endothelial cells, vascular smooth muscle cells, neutrophils and macrophages are all cellular sources of HGF.24 The coexpression of HGF and Met in several different types of carcinoma cells have been noted, however, the collective expression of HGF is restricted not exclusively but predominantly to stromal cells in a variety of carcinoma tissues. For example, in human lung adenocarcinoma tissues, a detailed analysis of the expression of HGF and Met using the laser-beam microdissection of cancer cells and tumor-associated stromal cells indicated that HGF mRNA is expressed exclusively in the stromal cells in different tumor samples.31 Although the mechanisms by which carcinoma cells acquire the ability to express HGF are unknown, the autocrine activation of Met may possibly be associated with the epithelial–mesenchymal transition of cancer cells, a phenomenon that is occasionally observed in highly aggressive cancer cells.
Consistent with a lack of autonomous potential to aggressively invade into the scaffold of extracellular matrix in many carcinoma cells, most carcinoma cells do not secrete HGF. However, carcinoma cells not only receive influence of stromal-derived HGF but also facilitate HGF production in stromal cells. Several reports noted that many types of carcinoma cells secrete a variety of HGF-inducers, through which HGF production in stromal fibroblasts is up-regulated.19, 20, 22, 32, 33 These carcinoma-cell-derived HGF-inducers were identified as interleukin-1β (IL-1β), bFGF, PDGF, transforming growth factor-α (TGF-α) and prostaglandin E2 (PGE2). The expression of HGF in fibroblasts in culture is consistently enhanced by the presence of cocultured carcinoma cells, whereas inhibitors or antibodies against cancer cell-derived HGF-inducers suppressed the expression of HGF in fibroblasts, which are associated with the decrease in invasiveness of carcinoma cells (Fig. 2a). These observations indicate that the presence of crosstalk between carcinoma cells and stromal fibroblasts, mediated by HGF and the HGF-inducers loop: carcinoma cells secrete HGF-inducers for stromal fibroblasts, while stromal fibroblasts secrete HGF, which in turn stimulates cancer cell invasion and metastasis (Fig. 2b). In addition to the role of these mediators that are capable of stimulating HGF expression, mediators that are expressed in cancer tissues are involved in modulating a microenvironment such as angiogenesis, the inflammatory response, and the growth of stromal cells.1, 34, 35, 36, 37, 38
In this context, it is noteworthy that prostaglandins (e.g., PGE1, PGE2 and PGI2), synthesized by the cyclooxygenase-2 (COX-2)-mediated metabolic pathways, are among the most potent inducers of the gene expression of HGF in fibroblasts.39 Interestingly, while COX-2 is often expressed in cancer cells in the human colon and other organs, its expression in precancerous benign polyps is found in stromal fibroblasts and endothelial cells but not in mucosal epithelial cells in both human familial adenomatous polyposis and the mouse model Apc mutants.40 Likewise, IL-1 stimulates HGF expression through the induction of COX-2 in fibroblasts (our unpublished data). COX-2 plays a key role in carcinogenesis such as colon cancer.41 Collectively, these results suggest the potential involvement of HGF in COX-2-related carcinogenesis and cancer progression through tumor–stromal interactions.
Crosstalk mediated by TGF-β and HGF
Recent studies have revealed the presence of a unique crosstalk, mediated by TGF-β and HGF, between epithelial and stromal cells.4 When the TGF-β type II receptor gene was inactivated specifically in stromal fibroblasts in mice, the mice showed prostatic intraepithelial neoplasia and invasive squamous cell carcinoma of the forestomach, both of which are associated with an increased number of stromal cells. Importantly, fibroblasts derived from prostatic and forestomach stroma overexpressed HGF, and an increase in the tyrosine phosphorylation of Met was seen in forestomach carcinoma cells. The results suggest that the enhanced activation of paracrine HGF signaling caused by impaired TGF-β-signaling in fibroblasts is one mechanism, leading to epithelial neoplasia and tumor progression to invasive carcinoma (Fig. 3). The TGF-β signaling pathway is known to suppress tumor formation when acting on epithelial cells, while it can also indirectly inhibit epithelial proliferation when acting on adjacent stromal fibroblasts. Because TGF-β is a potent suppressor of the gene expression of HGF in several types of cells including fibroblasts,42 the impaired TGF-β receptor-mediated signal transduction in the fibroblast-specific TGF-β type II receptor knockout mice seems to result in the overexpression of HGF in fibroblasts, thereby leading to paracrine activation of the Met receptor (Fig. 3).
It is also noteworthy that human mammary fibroblasts that are engineered to ectopically overexpress HGF or TGF-β alone or a mixture of both types of fibroblasts induce human mammary epithelia to develop ductal carcinoma in situ, adenocarcinoma, and invasive cancer, whereas transplantation of the same epithelial cell population with wild-type fibroblasts did not.5 In this study, unique “humanized mammary fat pads” were prepared as modified humanized stroma by transplanting human mammary fibroblasts. The development of human breast cancer in the fat pads were seen following the implantation of mammary epithelial cells prepared from 1 out of 10 patients, although cancer formation was not seen in 9 out of 10 cases. The development of invasive carcinoma caused by genetic modification of fibroblasts to overexpress HGF appear to be mechanistically similar to the impaired TGF-β signaling in fibroblasts, at least, from the aspect of the paracrine activation of Met receptor in epithelial cells. On the other hand, although TGF-β down-regulates the expression of HGF in fibroblasts, continuous stimulation by TGF-β may possibly induce the angiogenesis and transdifferentiation of fibroblasts into myofibroblasts, thereby changing characteristic of stroma susceptible to the development of cancer and its subsequent progression.
HGF-Met in activated stromal fibroblasts, myofibroblasts
Stromal fibroblasts in cancer tissues become “activated” into myofibroblasts, and frequently express myofibroblastic markers such as α-smooth muscle actin and fibroblast activation protein,43, 44 although the characteristics and cellular source of tumor-associated myofibroblasts are not completely understood. In addition to transdifferentiation of resident fibroblasts into myofibroblasts, recent studies have provided evidence that bone-marrow contributes to the myofibroblast and fibroblast populations in tumor-associated stroma, at least in part.45 The presence of myofibroblasts in tumor-associated stroma has been demonstrated in various types of carcinomas, including the uterine cervix, colon, ovary, breast, skin, prostate and lung. Myofibroblasts are also produced in other physiological and pathological conditions such as wound healing and fibrotic disorder in a variety of tissues.
In human lung adenocarcinoma, the Met receptor is expressed in tumor-associated myofibroblasts in the invasive area.31 Myofibroblasts secrete HGF and HGF stimulates the proliferation of tumor-associated myofibroblasts through an autocrine loop. Moreover, a significant relationship exists between Met expression in tumor-associated myofibroblasts and the shortened survival of patients with lung adenocarcinoma. These results suggest that the HGF–Met system may be involved in the expansion of tumor-associated myofibroblasts through autocrine activation loop and that HGF derived from myofibroblasts plays a role in the invasion and metastasis of lung adenocarcinoma. On the other hand, HGF was reported to suppress growth and stimulate apoptosis in myofibroblast-like cells of the liver and lung with fibrotic change and this may be associated with improved tissue fibrosis by HGF.46, 47 Perhaps heterogeneity and the unique characteristics of myofibroblasts that are associated with differences in their cellular origin and pathological conditions may be involved in the distinct biological actions of HGF on myofibroblasts or myofibroblast-like cells.
Involvement of HGF–Met in malignant behaviors caused by radiation and hypoxia
Radiation inhibits cell proliferation or induces apoptotic cell death, and is one of the major adjuvant treatments for malignant tumors. However, irradiation promotes the invasion and metastasis of cancer cells, as well as neoplastic progression.48, 49 Experiments using fibroblasts that had been irradiated to cause sublethal damage provide further evidence for a role of fibroblasts in epithelial neoplasia and progression to malignant cancer. When mammary epithelial cells were transplanted into the fat pads of mice containing irradiated fibroblasts, the incidence of breast tumors was markedly increased compared to the incidence seen in the nonirradiated condition.50 Tissue recombination of pancreatic cancer cells with irradiated fibroblasts resulted in a more aggressive and invasive cancer than when nonirradiated fibroblasts were used.51 Importantly, the increased invasiveness of pancreatic cancer cells in combination with irradiated fibroblasts was associated with an increased tyrosine phosphorylation of the Met receptor in cancer cells. This result indicates a role of HGF–Met signaling in invasive and metastatic behaviors of cancer cells, those being promoted by tumor–stromal interaction and irradiation.
Solid tumors acquire the ability to induce angiogenesis for growth and tumor growth depends on the formation of vasculatures.52 However, since the architecture of the tumor vasculature is disorganized compared to normal tissues, tumor tissues are susceptible to hypoxia.53, 54 Pathological and clinical studies have indicated that the presence of hypoxic regions within neoplastic lesions is correlated with a poor prognosis and an increased risk for developing distant metastases.55 Likewise, hypoxia can directly increase the invasiveness of tumor cells.56 Pennacchietti et al. demonstrated that hypoxia induced the transcriptional activation of the Met receptor gene and the subsequent amplification of HGF–Met signaling, thereby promoting the invasiveness of cancer cells.55 Consistently, the Met receptor was overexpressed in hypoxic areas of tumors and specific inhibition of Met expression inhibited the hypoxia-induced invasive growth of cancer cells. Therefore, a connection between hypoxia and the Met receptor appears to explain why hypoxia is often correlated with the acquisition of invasive and metastatic behavior and a poor prognosis. On the other hand, the inhibition of angiogenesis in tumor tissue has been new therapeutic strategy for the treatment of malignant tumors.57, 58 Hypoxia might starve the main mass of the tumor, but hypoxia, caused by the inhibition of angiogenesis, could also stimulate the invasive and metastatic characteristics of cancer cells at the tumor–host interface. Perhaps, simultaneous inhibition of tumor angiogenesis and the HGF–Met pathway may be a strategy for augmenting the effectiveness of antiangiogenesis therapy of cancers, if the inhibition of tumor angiogenesis is associated with an increase in hypoxic regions, thereby causing cancer cells to be more aggressive.
Therapeutic approaches targeting HGF–Met system
The close involvement of the HGF–Met pathway in tumor–stromal interactions and cancer invasion and metastasis suggests that the HGF–Met pathway represents a potentially important molecular target in cancer treatment. Several lines of distinct approaches for inhibiting the HGF–Met pathway have been demonstrated in experimental models, including small molecules that inhibit the tyrosine kinase activity of the Met, ribozyme, small-interfering RNA (siRNA), neutralizing monoclonal antibodies, soluble Met receptor, and variant molecules of HGF (Fig. 4).
Two structurally related inhibitors, OHA-665752 and SU11274, were reported to be small molecules that inhibit tyrosine kinase of Met.59, 60 These molecules are ATP-competititors for the catalytic activity of Met tyrosine kinase, and exhibited >50-fold selectivity for Met, compared to several other receptor tyrosine kinases, in inhibiting kinase activities. In a xenograft model of human gastric cancer in mice, administration of PHA-665752 inhibited Met tyrosine phosphorylation and was associated with the inhibition of tumor growth.59
Knock-down of the expression of HGF or Met by means of a ribozyme or siRNA resulted the in vitro inhibition of migration and invasion of cancer cells and the in vivo inhibition of tumor growth in distinct types of cancers, including breast cancer, glioblastoma, gastric cancer, prostate cancer, and colon cancer.61, 62, 63, 64, 65 The knock-down of Met receptor expression by recombinant adenovirus-mediated expression of siRNA against Met resulted in the inhibition of cell growth, scattering and invasion, and an increase in apoptotic cell death in different types of tumor cells, including prostate cancer, sarcoma, glioblastoma, and gastric cancer.61 Likewise, intratumoral infection by the adenovirus for met siRNA suppressed the growth of prostate cancer in mice.66
To target HGF, 2 different methods are available for it neutralization: utilization of monoclonal antibodies against HGF and a soluble recombinant Met receptor composed of the extracellular domain of Met. A combination of monoclonal antibodies with different epitopes could block the biological activities of HGF and the administration of these monoclonal antibodies were found to potently suppress the growth of glioblastoma xenografts in mice.67 On the other hand, a soluble Met receptor corresponding to its entire extracellular domain showed a high affinity binding to both HGF and Met, since the extracellular domain of Met is conceivably involved in receptor homodimerization.68 Thus, the soluble Met receptor interfered with both the binding of HGF to Met and the homodimerization of Met, thereby being capable of inhibiting HGF-dependent and HGF-independent activation of Met. In a xenograft model of human breast cancer in mice, infection by the lentiviral vector for the expression of soluble Met inhibited growth, tumor angiogenesis, and metastasis.
HGF is biosynthesized and secreted as a biologically inactive single-chain proform and the proteolytic processing into the two-chain mature form by specific serine proteases is associated with the conversion of the proform into the biologically active form. Based on the finding that a single-chain HGF can bind to Met but does not activate it,69 uncleavable HGF prepared by the substitution of an amino acid in the processing site was examined, to determine whether the uncleavable single chain HGF could behave as a competitive inhibitor for HGF.70 The uncleavable HGF inhibited both the protease-mediated conversion of single chain HGF into mature 2-chain form and activation of the Met receptor induced by HGF. The lentivirus-mediated expression of uncleavable HGF potently suppressed tumor growth, angiogenesis, invasion, and metastasis of breast cancer in mice.
Among the inhibitory molecules targeting HGF–Met pathway, NK4 has been the most extensively examined for developing a therapeutic approach. NK4 was prepared as a competitive inhibitor for HGF–Met.21, 71 NK4 is composed of the N-terminal 447 amino acids of the α-chain of HGF and contains the N-terminal domain and 4 kringle domains (designated NK4). NK4 can bind to Met but does not activate it, thereby competitively inhibiting the HGF–Met pathway. Importantly, NK4 acts as an angiogenesis inhibitor and this activity appears to be independent of its original activity as an HGF-antagonist.71, 72 NK4 inhibits proliferation, migration, and tube formation in endothelial cells stimulated by bFGF and VEGF, as well as HGF.72 It is noteworthy that the kringle domains in some kringle-containing proteins have angioinhibitory activity, including angiostatin, kringle-2 of prothrombin, and the kringle domains of tissue plasminogen activator, urokinase type plasminogen activator, and apolipoprotein(a).71 Therapeutic effects of NK4 has been demonstrated in a variety of cancers using animal models, including lung, breast, colon, gallbladder, gastric, ovarian, pancreatic, and prostate carcinomas, glioblastoma, and lymphoma.71 In these experiments, NK4 gene expression or the administration of recombinant NK4 inhibited tumor growth, angiogenesis, invasion, and metastasis, and prolonged life span in several models. Given that the inhibition of tumor angiogenesis is associated with the more aggressive behavior of cancer cells due to hypoxia, the simultaneous inhibition of tumor angiogenesis and the HGF–Met pathway by NK4 seems to be particularly an effective strategy for blocking the highly aggressive behavior of cancers.
Under physiological conditions, the HGF–Met system plays a role in tissue regeneration and wound healing, including liver regeneration.10, 15, 16 Thus, the retardation of tissue regeneration and wound healing is a conceivable side effect associated with the inhibition of HGF–Met pathway when patients have wounded tissues. Details of the potential side effects associated with the inhibition of the HGF–Met pathway remain to be evaluated in preclinical studies, though appreciable side effects were not noted in these experimental cancer treatments, which involved the inhibition of the HGF–Met pathway.
Tumor–stromal interactions are reminiscent of epithelial–mesenchymal (or -stromal) interactions, fundamental tissue interactions involved in dynamic growth, morphogenesis, and regeneration of several types of organs. However, a definitive difference between epithelial–mesenchymal and tumor–stromal interactions is that the expression and functions of mediators and their receptors participating in the former interactions are tightly regulated, both temporally and spatially; once normal tissue architectures are constructed or reconstructed, the expression and functions of mediators and their receptors become quiescent in epithelial–mesenchymal interactions.
Studies completed in the past decade indicate a more important role for stromal cells (particularly fibroblasts/myofibroblasts and endothelial cells) in carcinogenesis and progression to metastatic cancer than had previously been appreciated. Genetical alterations involved in carcinogenesis occur not only in epithelial cells but also in stromal cells such as fibroblasts. Thus, the fundamental question remains as to how signal exchange and/or direct cell–cell contact in epithelial–stromal crosstalk induce genetical alteration and instability, which facilitate the malignant transformation of normal cells. The mechanism by which normal cells acquire genomic changes may be elucidated by further investigations on tumor–stromal interactions, which will likely result in new therapeutic approaches.
HGF and its partner Met play a definitive role in tumor–stromal interactions, particularly leading to invasive and metastatic cancers. The invasion and subsequent establishment of metastasis are devastating events in patients with cancer, but many past approaches have not addressed what are perhaps the most important issues in cancer treatment, i.e., invasion and metastasis. A therapeutic strategy targeting the HGF–Met axis is based on the suppression of the intrinsic characteristics of malignant tumors, i.e., invasion and metastasis. Undoubtedly, therapeutic approaches that target the HGF–Met axis warrant further investigation and attention as potential cancer therapy for humans.