Cell-Autonomous and Non–Cell-Autonomous Mechanisms of HGF/MET–Driven Resistance to Targeted Therapies: From Basic Research to a Clinical Perspective

Targeted therapies by means of compounds that inhibit a specific target molecule give a new perspective in the treatment of cancer (1). In contrast to conventional chemotherapy, which acts mainly on dividing cells, targeted drugs act on subpopulations of cells directly involved in tumor progression in a more specific way.

The frequent alteration of receptor tyrosine kinases (RTK) in human malignancies led them to be considered targets for antineoplastic therapies; this resulted in the development of several inhibitors with strong clinical activity. The concept of “oncogene addiction,” that is the dependence of tumor cells on the constitutive activation of a single oncogene for their proliferation and survival, has added a further rationale to the use of targeted therapies (2). In a fraction of patients with cancer, targeted therapies have shown excellent results, leading to dramatic tumor regression. However, a substantial percentage of patients selected to express the target of the drug do not benefit from treatment (primary resistance) and, almost invariably, initially responsive patients develop resistance to treatment and undergo tumor relapse (secondary resistance). Therefore, an important challenge associated with targeted therapies is to predict mechanisms that could cause resistance to treatment and to find ways to solve, or even prevent, this problem.

Recent studies aimed at investigating the molecular mechanisms responsible for primary and secondary resistance to targeted therapies mainly relied on in vitro experiments carried out in tumor cells. These cells are highly heterogeneous and often display genomic instability. The therapeutic treatment can thus select preexisting clones with alterations in signaling pathways that are able to compensate for the drug-inhibited kinase. The use of in vitro models allowed the identification of a number of molecular mechanisms responsible for acquired resistance to tyrosine kinase inhibitors (TKI). Although most of these mechanisms have also been validated in patients, cellular models have clear limitations, such as the artificial growing conditions and the lack of epithelial–stromal interactions that typically occur in vivo. Indeed, very recent data support the concept that the tumor microenvironment plays an important role in sustaining resistance to targeted therapies, for example, by producing ligands that, in a paracrine manner, activate signals able to compensate for the drug-inhibited pathways in tumor cells. Thus, an emerging concept is that resistance of tumor cells to targeted therapies can be due to cell-autonomous and/or non–cell-autonomous mechanisms. Our knowledge of the prevalence of either mechanism in different clinical situations is largely incomplete.

In this review, we will discuss the role of the hepatocyte growth factor (HGF)/MET system in mediating resistance to anticancer kinase inhibitors, both in cell-autonomous and in non–cell-autonomous manners, and the translational implications of these findings.

The MET proto-oncogene encodes the tyrosine kinase receptor for HGF (3, 4). Ligand binding induces MET activation, which drives a complex biologic program defined as “invasive growth,” resulting from the promotion of several biologic activities such as cell proliferation, cell invasion, and protection from apoptosis (Fig. 1). MET-driven invasive growth is a physiologic program that occurs during embryonic development and in adulthood during tissue regeneration. However, it has been shown that the inappropriate activation of this program contributes to several aspects of tumor progression. MET-activated signaling pathways are shared with many other RTKs and include the mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinase (PI3K)/AKT pathways, STAT3, RAC1, and the NF-κB pathway (Fig. 1). Many studies have focused on the specific role played by each of these pathways in different MET-induced biologic activities (for a review, see ref. 5).

MET-mediated signaling results from pathways directly activated by this receptor, but it can also be modulated by the cross-talk between MET and different membrane receptors, acting in complex interacting networks (Fig. 2). In vitro data suggest that this cross-talk is not essential for cell survival but enables a better integration of the signals present in the extracellular environment. Under physiologic conditions, these networks are probably redundant. However, it is likely that under pathologic conditions, these interacting receptors cooperate in promoting tumorigenesis and/or metastasis and in inducing resistance to targeted drugs. The first network of interaction involves MET and adhesive receptors such as CD44 and the α6β4 integrin. CD44, a transmembrane receptor for hyaluronic acid, a major component of the extracellular matrix, has been implicated in tumor progression and metastasis (6). The CD44v6 variant, necessary for the activation of some MET intracellular transducers, functions as an amplifying platform, linking the MET cytoplasmic tail to the actin cytoskeleton and sustaining activation of the MAPK cascade (7, 8). The α6β4 integrin, independent of its adhesive role, acts as a supplementary docking platform for amplification of PI3K, MAPK, and SRC-dependent pathways (9). B family plexins, the receptors for semaphorins, can transactivate MET in the absence of HGF and promote proinvasive signals (10). In addition, the interaction between MET and FAS is important in the modulation of apoptosis (11). MET also interacts with other tyrosine kinase receptors, such as those belonging to the EGF receptor (EGFR) family. In fact, reciprocal transphosphorylation between these receptors has been shown in different systems, and their ability to substitute for each other has been shown in tumor cells (12–14).

In transformed tissues, the gain of an invasive growth program is advantageous for cancer progression and metastasis. In fact, constitutive MET activation can contribute to several aspects of tumor progression, because it induces neoplastic cells to disaggregate from the tumor mass, erode basement membranes, infiltrate stromal matrices, and eventually colonize new tissues to form metastases (for a review, see ref. 15).

Data produced by many laboratories provide compelling evidence that HGF/MET signaling plays an important role in the development and malignant progression of tumors, particularly in tumor invasiveness and metastasis. Preclinical studies show that cells that ectopically overexpress MET or HGF are tumorigenic and metastatic in nude mice, whereas MET inhibition decreases these properties (15). Moreover, cancer cell lines exhibiting MET gene amplification are “addicted” to MET (which means that they are dependent on this receptor for their growth and survival), and MET inhibition results in either a block in proliferation or cell death (16–18).

As shown in Fig. 3, deregulated MET activation in cancer can be due to different molecular alterations. Clinical data show that MET overexpression, in the absence of gene amplification, is the most frequent cause of constitutive MET activation in human tumors and often correlates with poor prognosis. Overexpression can be caused by several factors, such as hypoxia (19), activation of upstream oncogenes (20, 21), inactivation of tumor suppressor genes (22), or loss of miRNAs (23, 24). MET gene amplification, which drives increased expression and constitutive receptor activation, has been described in selected histotypes such as gastroesophageal, colorectal, endometrial, and lung carcinomas, glioblastomas, and medulloblastomas (reviewed in ref. 25). Autocrine MET activation has been described in sarcoma, glioblastoma, breast carcinoma (reviewed in ref. 25), and, very recently, in a high percentage of acute myelogenous leukemia (26). Finally, the unequivocal evidence linking MET and human cancer came from the identification of germline activating mutations in patients suffering from hereditary papillary renal carcinomas (27, 28). Activating mutations in sporadically occurring tumors are relatively rare and have been found mainly in lung and kidney carcinomas and hepatoblastoma (reviewed in ref. 25). These mutations are located in the tyrosine kinase domain, the juxtamembrane portion, or in the extracellular Sema (semaphorin) domain. Although overexpression can render MET activation independent from HGF stimulation, in most cases the ligand is still required for full receptor activation. This is also true for MET receptor variants containing activating mutations that need HGF to fully activate their kinase activity (29).

In light of the functional role played by the HGF/MET axis in different human tumors, in the last decade several strategies have been designed to inhibit the activation of the MET receptor, and multiple agents are currently in clinical trials (for a summary, see Fig. 4).

Clinical practice, as well as experimental evidence, has clearly shown that biologic therapies are effective only in a certain percentage of tumors expressing the appropriate target. Moreover, even responding tumors develop resistance to treatment quite rapidly. These observations prompted researchers to investigate the mechanisms responsible for primary and secondary resistance. The first studies were focused on the changes occurring in tumor cells, considered more likely to be responsible for the phenomenon due to their high genomic instability (Fig. 5). However, many data have unveiled a critical role for the tumor microenvironment in generating and sustaining resistance. The tumor stroma, in fact, produces growth factors and other molecules (such as extracellular matrix components) that can activate signaling pathways in tumor cells that are able to overcome the inhibitory effect of the drug (Fig. 6).

The first clinical evidence of a role of MET in sustaining resistance to EGFR inhibitors came from two studies of patients affected by non–small cell lung cancer (NSCLC) and treated with the reversible EGFR inhibitors gefitinib and erlotinib. Engelman and colleagues (30) started from the in vitro observation that lung cancer cells that were originally sensitive to EGFR inhibitors became resistant to long-term treatment with gefitinib as a consequence of MET gene amplification. They also showed that sensitivity to EGFR inhibitors could be restored through simultaneous treatment with anti-MET and anti-EGFR drugs, thus proving the causative role of MET in sustaining acquired resistance. They observed that MET causes resistance to gefitinib by transphosphorylating HER3 and driving HER3-dependent activation of the PI3K pathway. When they analyzed 18 patients with NSCLC, all displaying acquired resistance to either erlotinib or gefitinib, they found that 22% of these patients displayed MET amplification. Notably, one patient bearing two independent metastases showed MET amplification in only one metastatic lesion, whereas the other displayed the EGFR T790M mutation, known to impair response to these drugs.

The work published by Bean and colleagues (31) confirmed these results. Indeed, using high-resolution genome-wide profiling of NSCLCs before and after treatment with anti-EGFR drugs, they found MET amplification in nine of 43 patients (21%) with acquired resistance, compared with two of 62 (3%) untreated patients. Interestingly, they observed that 40% of the samples with MET amplification also harbored the EGFR T790M mutation. MET amplification was also found in an NSCLC cell line, in which the same cells harbored (i) an EGFR mutation associated with drug sensitivity, (ii) an EGFR mutation associated with drug resistance, and (iii) MET amplification, thus showing that all these genetic lesions can occur within the same cell population. Finally, this work also showed that treatment with a MET inhibitor was able to overcome acquired resistance to EGFR kinase inhibitors, even in cells harboring the T790M mutation. It has to be noted that more recent studies in rebiopsied specimens from larger cohorts revealed a lower frequency of MET amplification in resistant tumors compared with that reported by Engelman and colleagues (30) and Bean and colleagues (31), ranging from 5% to 11% (32, 33). However, given the results of these two studies, combination therapy with MET kinase inhibitors and anti-EGFR drugs should be considered for patients whose tumors have become resistant to gefitinib or erlotinib and harbor MET amplification.

One of the questions stemming from these studies is the origin of the cells displaying MET amplification: did the cell lines and the tumors harbor preexisting MET-amplified clones and did the drug simply create a favorable environment capable of selecting them? Or, alternatively, did the drug induce a stress able to promote DNA breaks in fragile sites such as the one in which the MET gene is located (34)? To answer these questions, Turke and colleagues (35) analyzed the HCC827 NSCLC cell line that, upon gefitinib treatment, became resistant via MET amplification. When they studied the parental HCC827 cells using high-throughput FISH, they identified six of 4,237 cells harboring a significant increase in MET copy number. In contrast, they did not find any MET-amplified subpopulation in the other two NSCLC cell lines, namely H3255 and PC-9, which, upon acquisition of resistance to gefitinib, displayed the EGFR secondary resistance mutation T790M, but not MET amplification. To further validate these observations in the clinical setting, the authors evaluated specimens obtained from 27 patients who developed acquired resistance to either gefitinib or erlotinib and found MET amplification in four of these samples. In all of these patients, the pretreatment specimens revealed rare (<1%) tumor cells with MET amplification. In contrast, only one of eight cases of resistance to anti-EGFR drugs resulting from EGFR secondary resistance mutations displayed rare MET-amplified tumor cells. Altogether, these data suggest that MET-amplified tumor cells preexist, and that these cells are selected during the course of therapy in patients and NSCLC cell lines that are characterized by MET amplification as a mechanism of secondary resistance to EGFR inhibitors. However, it remains to be clarified if this is true only in NSCLC or if it is a common feature of other cancers.

Another interesting observation stemming from this work is that the presence of HGF in the environment, concomitant with EGFR inhibition, strongly favors the emergence of MET-amplified cells. In vitro experiments showed that the exposure of cells to HGF reduced the time required for the onset of resistance to EGFR inhibitors from several months to 14 days. Therefore, it is conceivable that, in the presence of EGFR inhibitors, HGF provides a proliferative advantage to cells with MET amplification and favors their clonal expansion. Thus, the combination of gene amplification and ligand-mediated activation represents a very powerful mechanism that renders lung cancer cells resistant to targeted therapies.

Several studies have shown that in HER2-overexpressing or amplified breast cancers, the anti-HER2 humanized recombinant monoclonal antibody trastuzumab, in combination with chemotherapy, was significantly more effective than chemotherapy alone, both in the metastatic and the adjuvant settings (36, 37). In spite of the effectiveness of this treatment, only one third of patients who are eligible for the treatment respond; moreover, most of the patients who initially respond show disease progression within 1 year of treatment. Among the mechanisms of resistance to trastuzumab treatment (38–40), a role for MET has been reported by Shattuck and colleagues (41), who found that inhibition of MET sensitizes cells to trastuzumab-mediated growth inhibition, whereas MET activation protects cells against trastuzumab. On the base of these results, Minuti and colleagues (42) recently evaluated the relationship between MET and HGF copy number and responsiveness to trastuzumab in 130 patients with HER2-positive metastatic breast cancer using FISH. MET-positive cases (defined as the ratio of mean MET/mean CEP 7 > 2) had a significantly higher trastuzumab failure rate (44.4% vs. 16%; P = 0.001) and a significantly shorter time to progression (5.7 vs. 9.9 months; HR 1.74; P = 0.006) compared with MET-negative cases. Similar results were obtained when HGF FISH status was evaluated. Interestingly, MET and HGF positivity were highly correlated (P < 0.001), suggesting the presence of chromosome aneuploidy, rather than single amplification of either of the two genes. These results, together with the preclinical studies, suggest the possible effectiveness of combined anti-HER2 and anti-MET therapies in a subgroup of patients with HER2-positive metastatic breast cancer.

The EGFR-targeted monoclonal antibodies cetuximab and panitumumab are effective in a subset of metastatic colorectal tumors. The presence of KRAS mutations and deregulation of effectors of the EGFR signaling cascade (PIK3CA, PTEN, NRAS) or of EGFR modulators (HER2, EGFR ligands) are thought to impair the primary response to EGFR blockade (43–45). Moreover, all responding patients develop resistance, which occurs through emergence of KRAS mutations in approximately 50% of the cases (46, 47). A recent report has shown that amplification of the MET proto-oncogene can be responsible for de novo and acquired resistance to anti-EGFR therapy in patients with colorectal cancer (48). Notably, in xenografts derived from MET-amplified tumors, treatment with MET kinase inhibitors overcame resistance to EGFR blockade. These results highlight the role of MET in mediating primary and secondary resistance to anti-EGFR therapies in colorectal cancer and offer novel opportunities to design clinical studies with combined EGFR/MET inhibitors.

Recent work conducted by Jun and colleagues (49) revealed a role of MET in mediating resistance to gefitinib in a preclinical model of glioblastoma multiforme. Inhibition of EGFR resulted in a substantial change in global gene expression levels and, in particular, increased MET expression, which resulted in sustained prosurvival Akt signaling. Pharmacologic inhibition of MET overcame the resistance to EGFR inhibition, further supporting the wide importance of MET/EGFR interaction in multiple tumor types.

As these studies imply a clinically relevant role for MET gene amplification in sustaining resistance to anti-EGFR therapies, a critical point is how to define MET amplification. Even though FISH seems to be the best way to identify MET gene copy number, standardized criteria to define MET FISH–positive tumors have not yet been determined. The two most commonly used criteria are the University of Colorado Cancer Center Criteria (UICC; ref. 50) and the Cappuzzo scoring system (51). Both of these scoring systems, however, evaluate aneuploidy and real MET amplification at the same time. From a biologic point of view, this can potentially confound the results, as chromosome 7, on which MET is located, also contains the genes for HGF and EGFR. Thus, chromosomal duplication increases not only the number of MET copies but also the copy number of its ligand and of EGFR, the therapeutic target. A recent work (52) has compared the Cappuzzo scoring system with the PathVision system, which takes into account only real amplifications. The latter, in fact, considers a ratio between total MET copy number/chromosome 7 copy number ≥2.0 as positive. The analysis conducted on 138 patients with lung adenocarcinoma revealed MET FISH positivity in 15% and 4% of cases according to the Cappuzzo scoring system and PathVision, respectively. This means that in the great majority of cases, the increased MET copy number is not due to a real gene amplification but rather to chromosomal aneuploidy. Whether these two biologic conditions are associated with a different clinical behavior should be assessed in future studies.

What is the clinically relevant number of MET copies to confer resistance to targeted therapies? This question has been addressed by Suda and colleagues (53), who found that the MET gene copy number was augmented in proportion to erlotinib resistance in cells treated with increasing concentrations of the drug. By using FISH analysis, they observed that when the MET gene copy number had increased more than fourfold (that is at least eight gene copies), the cells were able to proliferate in the presence of micromolar concentrations of the drug, a clinically achievable dose. If these data hold true, the threshold suggested by the UICC and the Cappuzzo scoring systems might be too low to identify situations in which the level of MET amplification is high enough to have a strong clinical impact. However, it is essential to consider that as HGF is available in vivo, lower amplification levels can lead to ligand-dependent activation of the MET signaling pathway that is sufficient to drive resistance. At the moment, this question is still open and preclinical and clinical studies are needed to clarify this critical point.

Cell autonomous or “intrinsic” mechanisms of resistance (such as activation of alternative signaling pathways, onset of secondary mutations in drug targets, amplification of the target gene, activation of efflux pumps) have been deeply investigated, and therapeutic strategies and new drugs have been generated to overcome these molecular alterations. In contrast, non–cell-autonomous mechanisms of resistance to therapy have only recently been investigated. In fact, the concept that the tumor behaves as an organ, in which the interaction between cancer cells and the peritumoral environment is critical to sustain its survival, is relatively new. The first focus was on the role of angiogenesis and resulted in many studies aimed at blocking tumor growth by interfering with its vascularization (reviewed in ref. 54). However, it soon became clear that other cells of the microenvironment, such as cancer-associated fibroblasts and inflammatory cells, can also modulate the biologic properties of tumor cells, affecting their growth and invasion potential (55). More recently, it has been shown that the microenvironment can also contribute to both primary and secondary resistance to anticancer therapies. Stromal cells, in fact, secrete factors able to activate redundant signaling programs in cancer cells that render them resistant to targeted treatments. One of the first reports to show this, from Williams and colleagues (56), found that in an animal model of BCR–ABL+ chronic myelogenous leukemia, neoplastic cells were sensitive to the ABL-specific inhibitor imatinib only in vitro but not in vivo, due to the secretion of host cytokines.

Several recent papers have highlighted the notion that HGF secreted by cells of the microenvironment can play a critical role in primary and secondary resistance to several targeted therapies in the context of different tumors. Using a panel of kinase-addicted human cancer cell lines, Wilson and colleagues (57) showed that most of these cells could be rescued from drug sensitivity by exposure to growth factors usually secreted by microenvironmental cells. HGF, fibroblast growth factor (FGF), and neuregulin 1 were the most broadly active ligands and could induce either a “partial rescue” or a “complete rescue” (that is a shift of the IC₅₀ curve of more than tenfold or a complete suppression of drug response) in the tested cell lines. The two downstream pathways commonly engaged by these ligands were the PI3K/AKT and MAPK pathways, but only HGF was able to simultaneously rescue both of these pathways. In the cell lines examined, HGF promoted resistance to several targeted therapies, including not only those against tyrosine kinases (such as ALK and members of the EGFR family), but also therapies against serine/threonine kinases (for example, RAF).

An important question is: how does the stimulation with HGF exert its role in promoting resistance? Does it simply activate downstream signals or, while doing so, does it also select cells able to better respond to HGF? Strong evidence in favor of the latter hypothesis was provided in the work of Turke and colleagues (35), who showed that in EGFR-mutant NSCLCs, HGF treatment promoted the emergence of cell subpopulations displaying MET amplification before exposure to the anti-EGFR drug gefitinib. Indeed, MET-amplified cells do not have any growth advantage in the absence of gefitinib, which targets EGFR-addicted cells, but they are enriched by the selective pressure of the drug during ligand exposure. Similar results were also obtained by Straussman and colleagues (58), who showed that HGF treatment could rapidly promote resistance to lapatinib (a HER2 inhibitor) in HER2-amplified breast cancer cells, likely by driving selection of a subpopulation of MET-amplified cells.

HGF is an angiogenic factor (59) that is able to activate MET in endothelial cells, in part by inducing VEGF production and decreasing thrombospondin-1 expression (60). Although antiangiogenic therapies have shown efficacy and survival benefits in clinical trials, the majority of patients develop resistance during treatment and undergo progression. Several studies have suggested that factors such as FGF2 or platelet-derived growth factor, produced by either cancer or stromal cells, can contribute to the onset of resistance (57, 61). Shojaei and colleagues (62) have shown that HGF production by the tumor stroma can also contribute to resistance to treatment with sunitinib, a multikinase inhibitor targeting the VEGF pathway. Indeed, they found that resistant tumors displayed a greater concentration of HGF and a high MET expression level in endothelial cells. They also showed that combinatorial treatment with sunitinib and an anti-MET TKI was more effective than sunitinib alone. Moreover, systemic injection of HGF conferred resistance to sunitinib treatment by sustaining tumor angiogenesis. Altogether, these data suggest that cotargeting of VEGF/HGF-mediated pathways could be a strategy to circumvent resistance to antiangiogenic therapies.

The role of HGF in mediating resistance to RAF inhibition in melanomas was studied by Straussman and colleagues (58), who showed that HGF secretion by stromal cells resulted in the activation of MET and reactivation of PI3K/AKT and MAPK pathways and in primary resistance to RAF inhibition. Moreover, they found a significant correlation between HGF expression by stromal cells and primary resistance to RAF inhibition. Importantly, dual inhibition of RAF and either HGF or MET reversed drug resistance. Interestingly, none of the other 22 RTK ligands tested were able to rescue BRAF-mutated melanoma cells from the BRAF inhibitor. In accordance with what was described by Wilson and colleagues (52) among all the evaluated ligands, only HGF was able to simultaneously and efficiently activate both the PI3K/AKT and MAPK pathways. These data suggest that a combined anti-RAF and anti-MET/HGF therapy might represent a useful option in patients affected by RAF-mutated tumors, such as melanomas, colon cancer, and glioblastomas.

The first study showing that HGF can induce gefitinib resistance in lung adenocarcinomas with EGFR-activating mutations was that by Yano and colleagues (63), who found that HGF acts by restoring PI3K/AKT activation via phosphorylation of MET (but not of EGFR or HER3); moreover, they showed that strong immunoreactivity for HGF was detected in patients displaying primary or secondary resistance to gefitinib. The same authors subsequently found that high HGF immunohistochemical reactivity was present in 29% of nonresponding patients, and that HGF expression was significantly higher in tumors with acquired resistance than in sensitive ones (64). Plasma levels of HGF were evaluated by Tanaka and colleagues (65), who observed that administration of EGFR kinase inhibitors significantly increased plasma HGF levels 15 days after treatment. Moreover, high levels of plasma HGF before treatment were found in patients with primary resistance to anti-EGFR drugs. Finally, Yamada and colleagues (66) recently showed that, in lung cancer, HGF can induce resistance not only to small TKIs but also to cetuximab, an anti-EGFR monoclonal antibody, and that in preclinical in vivo models this is due to the secretion of HGF by stromal fibroblasts.

Data published by the same authors indicate that a similar HGF-mediated mechanism can induce resistance to ALK inhibitors in lung cancer cells addicted to an activated form of the tyrosine kinase ALK (67). ALK fusion proteins are present in 3% to 7% of unselected NSCLCs; treatment of these patients with ALK inhibitors showed a response rate of 54%, with a 91% disease control rate (partial response + stable disease; ref. 68). However, almost all the responding patients developed resistance to treatment and underwent relapse. Resistance was due to several mechanisms, such as ALK gene amplification, secondary mutations, or coactivation of other RTKs like EGFR and HER2 (69). Yamada and colleagues (66) showed that paracrine activation by either EGFR or MET ligands from the microenvironment can trigger resistance to ALK inhibitors. Similar results were also reported in vitro by Harbinski and colleagues (61). These data suggest that the dual MET/ALK inhibitor crizotinib may be more effective than more specific ALK inhibitors in treating NSCLCs expressing ALK fusion proteins.

Triple-negative breast cancers (i.e., breast cancers negative for the expression of estrogen receptor, progesterone receptor, and HER2) display EGFR overexpression in 54% of cases, but they are not responsive to EGFR-targeting drugs. Mueller and colleagues (70) showed that HGF treatment of breast cancer cells that are sensitive to EGFR TKIs conferred resistance to treatment. Interestingly, EGFR silencing abrogated HGF-induced cell survival, indicating a critical role for EGFR/MET cross-talk. These data support the hypothesis that poor efficacy in the clinical treatment of triple-negative breast cancers overexpressing EGFR may be due to cross-talk between EGFR and MET. Thus, simultaneous targeting of both receptors could be an effective therapeutic strategy.

More than 15% of gastric cancers display amplification of HER2. In vitro studies showed that lapatinib, a dual TKI targeting HER2 and EGFR, strongly decreased the viability of gastric cancer cells with amplified HER2 (71). However, HGF-mediated MET activation rescued cells from lapatinib-induced inhibition and reactivated downstream pathways. Finally, Liska and colleagues (72) showed that in colorectal cancer cells, HGF-induced MET activation could release cells from cetuximab-mediated EGFR inhibition, suggesting that inhibition of the MET/HGF pathway may also improve the response to EGFR inhibitors in colorectal cancer.

The idea of targeting the HGF/MET axis as a strategy to avoid or circumvent resistance to molecular therapies seems very attractive in light of the preclinical data collected in the last few years. Recent works have shown that only a small number of oncogenic tyrosine kinases display a compensatory ability when the “driver” kinase is inhibited. Among the tested kinases, MET seems to be one of the most effective, as it simultaneously activates the RAS/MAPK and the PI3K/AKT pathways with high efficiency. As previously described, preclinical experiments have shown that combinatorial treatments with a drug blocking the driver kinase plus a MET/HGF inhibitor delay the onset of secondary resistance or even overcome resistance to the targeting treatment. Notably, clinical data also support the value of combining a MET/HGF inhibitor with already-approved targeted therapies to delay patient relapse. In particular, a recent phase II clinical trial with MetMAb (onartuzumab, a one-armed antibody designed to compete with HGF by binding the extracellular portion of MET; ref. 73), conducted in patients with advanced NSCLC, has shown the clinical efficacy of the combined therapy. This phase II study showed that patients whose tumors had high MET levels experienced a twofold increase in progression-free survival in response to MetMAb plus erlotinib compared with erlotinib alone. Moreover, the combination of MetMAb plus erlotinib tripled overall survival compared with erlotinib alone in patients harboring tumors with high MET expression (74). Overall, although this trial was not designed to investigate the role of MET in acquired EGFR TKI resistance, these results suggest that combined anti-EGFR/MET treatment can delay the onset of resistance. This is likely the consequence of preventing the outgrowth of clones displaying MET activation, which are favored in the presence of anti-EGFR drugs.

A clinically relevant point is how to select patients who could benefit from the combined treatment. There are currently no available validated biomarkers that can be used to predict the molecular mechanisms that will sustain resistance in different patients. Therefore, at the moment, preventing resistance is more a dream than an imminent reality, as the presence of cell-autonomous and non–cell-autonomous mechanisms leaves many possibilities open. In addition, the clinical perspective is different in the case of patients with primary (innate) or secondary (acquired) resistance.

With regard to primary resistance, preclinical models identified local increases in HGF/MET activation as possible reasons for resistance to targeted therapies. Relying on these data, it would be important to stratify patients on the basis of HGF/MET status in the tumor, to select patients who could benefit from treatment with a single agent and patients who would need combined therapy with MET inhibitors. The main problem related to this stratification is that, at present, no technique to quantify HGF or MET activation in patient-derived samples has been standardized and routinely used. In particular: (i) there are no anti-HGF antibodies of sufficient quality for routine use in a pathologic setting. This issue could be overcome by the use of recently developed in situ hybridization techniques (75). However, because of the high costs and the technical difficulties, this approach will likely be applicable only on a large scale in the future; (ii) the phosphorylation of tyrosine receptors such as MET is rapidly lost in cells and tissues at room temperature. As adequate preservation of the surgical resection/biopsy is not routinely done, immunohistochemical analysis with antiphospho-MET antibodies or phosphoproteomic approaches would rarely be informative. Moreover, the lack of standard in anti-MET antibodies for immunohistochemistry (IHC) and in the scoring system to evaluate MET activation by IHC is a key limitation. These technical problems in evaluating MET activation in patient samples may also suggest that the involvement of the HGF/MET axis in primary and secondary resistance has probably been underestimated; and (iii) intratumoral heterogeneity and multiple layers of mechanisms of pathway activation represent additional problems.

With regard to secondary resistance, molecular analysis of the relapsed tissues is becoming mandatory, as many clinical studies have shown that the expression profile as well as the genomic status of the resistant tumors/metastases can be very different from that of the primary neoplasm. The identification of MET gene amplification in resistant samples will demand the use of a combined anti-EGFR/anti-MET therapy. This kind of analysis (evaluation of MET gene amplification upon relapse, for example by FISH) is, at present, not routinely conducted in all centers, but it will probably become mandatory in the near future. In addition, it would be of great clinical impact to monitor the possible onset of MET gene amplification in the tumor during the treatment targeting the “driver” kinase (e.g., anti-EGFR drugs in lung cancer). It would thus be possible to foresee the development of resistance before the radiologic appearance of the relapsed tumor and to start a combined therapy with MET inhibitors in advance of relapse. At the moment, the only known possibility to repeatedly monitor, to some extent, a tumor or a metastasis in a patient is by analyzing tumor-free DNA in the circulating blood (the recent concept of “liquid biopsy”). This strategy has been successfully used to identify the onset of acquired resistance to anti-EGFR antibodies in colorectal cancer due to KRAS mutations (47). However, as MET alterations implicated in the acquisition of resistance are quantitative and not qualitative (i.e., increased gene copy number and not mutations), their identification in the blood seems more difficult.

To evaluate the efficacy of combined treatments in counteracting both primary and secondary resistance, it is now possible to use mouse models that mimic the complexity of a specific patient's cancer. In practical terms, immunocompromised mice can be directly engrafted with tumor surgical specimens/biopsies to obtain “xenopatients” that, even upon serial passages, retain the histologic and genomic features of their original tumors (44). Once the “signature” of a tumor is identified, different drugs or combinations can be tested in these “avatars,” to select the most efficient combination for the patient. Other than being feasible in relatively few cancer centers, the main problem to evaluate HGF or MET as therapeutic targets through this approach is the observation, often reported in the literature, of the low cross-reactivity between human MET and murine HGF (76). The decreased ability of the mouse ligand to activate human MET could lead to the underestimation of the role played by the stromal murine HGF on tumor cells expressing human MET.

Recently, we have gained a better understanding of the molecular mechanisms underlying primary and secondary resistance to targeted therapies. According to Diaz and colleagues (46), “only a limited number of genes are likely to be able to exert a resistance phenotype,” and recent studies have shown that the HGF/MET axis is one such pathway. Drugs that target this tyrosine kinase are now available and offer the opportunity to treat patients to overcome resistance, but the problem of identifying patients who could benefit most from these treatments still remains. Personalized approaches based on genomic and functional profiles will be critical to select these patients and to evaluate the complex behavior of their tumors over time.

No potential conflicts of interest were disclosed.

Conception and design: S. Corso, S. Giordano

Writing, review, and/or revision of the manuscript: S. Corso, S. Giordano

The authors thank their colleagues for useful discussion and Dr. F. Natale for editing the manuscript.

This work was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC) IG 11819 grant (to S. Giordano) and the AIRC 2010 Special Program Molecular Clinical Oncology 5xMille, Project n. 9970 (S. Giordano).