EGF Receptor activates MET through MAP kinases to enhance non-small cell lung carcinoma invasion and brain metastasis

Abstract

MET amplification as a mechanism of acquired resistance to EGFR targeted therapies in non-small cell lung carcinoma (NSCLC) led to investigation of novel combinations of EGFR and MET kinase inhibitors. However, promiscuous interactions between MET and ERBB family members have made it difficult to evaluate the effects of MET on EGFR signaling, both independent of drug treatment and in the context of drug resistance. We addressed this issue by establishing a 32D model cell system wherein ERBBs or MET are expressed alone and in combination. Using this model, we determined that EGFR signaling is sufficient to induce MET phosphorylation, although MET activation is enhanced by co-expression of ERBB3. EGFR-MET crosstalk was not direct but occurred by a combined regulation of MET levels and intermediary signaling through MAP kinases. In NSCLCs harboring either wild-type or mutant EGFR, inhibiting EGFR or MAP kinases reduced MET activation and protein levels. Furthermore, MET signaling promoted EGFR-driven migration and invasion. Lastly, EGFR-MET signaling was enhanced in a highly metastatic EGFR mutant cell subpopulation, compared to the indolent parental line, and MET attenuation decreased the incidence of brain metastasis. Overall, our results establish that EGFR-MET signaling is critical for aggressive behavior of NSCLCs and rationalize its continued investigation as a therapeutic target for tumors harboring both wild-type and mutant EGFR at early stages of progression.

Introduction

Activating mutations in Epidermal Growth Factor Receptor (EGFR) are commonly found in NSCLC, and cells expressing mutant alleles depend on EGFR signaling for survival. Drugs targeting the EGFR kinase domain are effective for NSCLCs; unfortunately, responding tumors eventually progress owing to acquired resistance. Mechanisms of resistance to EGFR tyrosine kinase inhibitors (TKI) include secondary mutations in EGFR and activation of compensatory receptor tyrosine kinases, such as MET. In 5-20% of recurrent patients, MET activation sustains tumor cell survival and is associated with relapse (1-4). Accordingly, EGFR and MET combination therapies are in clinical trials for patients harboring MET amplification and resistance to EGFR TKIs (5).

Bi-directional signaling between EGFR and MET occurs in both EGFR TKI resistant and drug-naïve NSCLC cells. Notably, in NSCLC cells addicted to EGFR mutations, EGFR inhibition reduces basal MET phosphorylation (6-8). Conversely, MET inhibition in cells with high basal MET activity or MET amplification reduces basal phosphorylation of EGFR and ERBB family members, ERBB2 and ERBB3 (1, 2, 9, 10). In MET amplified cells, MET signaling through ERBB3 maintains PI3K/Akt cell survival signaling despite EGFR inhibition (1). Additionally, in NSCLC with wild-type EGFR or EGFR TKI resistance mutations, ligand stimulation of EGFR induces MET activation, and vice versa (8, 11, 12).

Mechanisms and outcomes of these biochemical events are difficult to interpret owing to the complexity of inter-ERBB interactions, which may involve promiscuous ERBB heterodimerization or indirect signaling cascades. Proposed mechanisms of ERBB-MET crosstalk include direct interaction, activation through autocrine regulation, and signaling through intermediary proteins (12, 13). Moreover, receptor crosstalk in some cells may not be bi-directional, but rather unilateral from EGFR to MET (14). In cancer cells encoding wild-type EGFR, MET regulates invasion and cell motility in an EGF-dependent manner (8, 12). Significantly, it is unknown if EGFR-MET crosstalk in lung cancers with EGFR mutations modulates similar phenotypes, particularly in patients with advanced disease and metastatic relapse. Defining the specific biological contexts of ERBB-MET crosstalk, identifying their mechanisms of action, and characterizing their function may help expand clinical settings where combinatorial EGFR/MET therapies would benefit NSCLC patients.

To systematically examine ERBB-MET crosstalk, we employed a cellular system enabling analysis of functional interactions between individual ERBBs and MET in isolation or combination. We found that EGFR is sufficient to stabilize and cross-activate MET in the absence of other ERBBs. Activation is indirect, involving MEK/p38MAPK signaling, and is enhanced by ERBB3. This EGFR-MET axis is also active in NSCLC cells, where MET facilitates EGF-induced migration and invasion irrespective of EGFR mutational status. Finally, EGFR-MET activation is enhanced in an experimental model of metastatic NSCLC cells and potentiates brain metastasis.

Methods

Cell Culture

32D mouse myeloid cells [American Type Culture Collection (ATCC)] were grown in RPMI with 10% heat-inactivated (HI) fetal bovine serum (FBS), 10% WEHI-3B conditioned medium (CM). WEHI-3B (ATCC) cells were grown in RPMI with 10% HI-FBS. For WEHI-CM, WEHI-3B cells were grown in RPMI with 1% FBS for 4 days, medium was collected, and passed through a 0.22μm filter. A549, H441, H2030, H1650, HCC827, and H1975 cells (ATCC) and PC9, PC9-BrM3 cells (15) were grown in RPMI with 10% FBS. SYF−/− mouse embryonic fibroblasts (MEFs) (16), were grown in DMEM with 10% FBS. All media contained 1% penicillin/streptomycin, L-glutamine, and sodium pyruvate.

Cloning

Site-directed mutagenesis was performed with QuikChange Multi Site-Directed Mutagenesis Kit (Agilent Technologies) following manufacturer's protocol. Primers were designed using manufacturer's guidelines and obtained from Integrated DNA Technologies. The pRK5TKneo-MET-V5/His (Genentech) plasmid was used to make pRK5TKneo-MET-K1110A-V5/His, pRK5TKneo-MET-Y1234/1235F-V5/His, and pRK5TKneo-MET-Y1349/1354F-V5/His.

Plasmids, Electroporation, and Virus Production

32D cells were electroporated at 240V, one pulse, 35ms, with pBABE-EGFR (Addgene), pBABE-ERBB3 (Nancy Hynes, Friedrich-Miescher Institute), pBABE-EGFR-L858R (Don Nguyen), pCDNA3(-)-EGFR-L858R/T790M (Katerina Politi, Yale University), pRTK-V5/His-Met, pRTK-V5/His-Met-K1110A, pRTK-V5/His-Met-Y1234F/Y1235F, or pRTK-V5/His-Met-Y1349F/Y1354F. Cells were selected then maintained at 2μg/ml and 1μg/ml puromycin (R&D Systems) or 100μg/ml and 50μg/ml G418, respectively (GIBCO).

Control and MET lentivirus was produced by cotransfecting HEK 293T cells with pLKO-shRNA constructs (Sigma-Aldrich), pMD2.G, and psPAX2 (Addgene) using lipofectamine (Invitrogen). Supernatants were collected daily for 3 days, combined, and concentrated with Centricon plus-20 filters (Millipore Corporation). Cells were infected overnight in 4μg/ml polybrene, selected and maintained at 2μg/ml and 1μg/ml puromycin, respectively.

Growth Factor Stimulation and Drug Treatment

Cells were stimulated with 10ng/ml EGF, 50ng/ml NRG, or 50ng/ml HGF (R&D Systems) for the indicated times. Kinase inhibitors Sunitinib, Imatinib, Nilotinib, Tozasertib, Gefitinib, U0126, SB203580, Dasatinib (LC Laboratories); BMS-754807, AZD-4547, BI2536 (ChemieTek); BMS-536924, MK2206, PLX4032, PD0332991 (Sellek Chemicals); PP2, AR-A014118 (Sigma Aldrich); Jnk Inhibitor II, Syk Inhibitor II (EMD Chemicals); PF-573208, CMPD1, BRD7389, SB41542, PHA665752 (Tocris); AZD-7762, NSC625982 (Axon Medchem); CIP-1374 (AlloStem Therapeutics); and PP2 (Calbiochem) were used at 1μM for the indicated times. Actinomycin D (Sigma) was used at 10μg/ml, Cycloheximide (Sigma) at 10ng/ml, and bortezomib (LC Laboratories) at 100nM.

Immunoprecipitation

Cells were lysed in TX100 lysis buffer (20mM Tris-HCl pH 7.5, 150mM NaCl, 1mM EGTA, 1mM EDTA, 1% Triton X-100). Immunoprecipitations were performed overnight at 4°C with 2mg protein lysate, 4μg anti-Met or anti-Cbl antibody (Santa Cruz), and Protein A/G Ultralink Resin (Invitrogen), washed thrice in TX100 buffer, and re-suspended in 2X Laemmli sample buffer (SB).

Immunoblotting

Cell lysates were prepared in 2X SB or NETN lysis buffer (150mM NaCl, 1mM EDTA, 50mM Tris HCl pH 7.8, 1% NP40 substitute [Fluka]). Immunoblots on PVDF were blocked in 5% nonfat milk/PBST (Dulbecco's phosphate-buffered saline, 0.1% Tween-20). Antibodies against phospho-EGFR-Y1068, phospho-ERBB3-Y1197, phospho-MET-Y1003, phospho-MET-Y1234/1235, phospho-Akt-S473, Akt, phosphop44/42MAPK(Erk1/2)-T202/Y204, p44/42MAPK(Erk1/2), phospho-p38MAPK-T180/Y182, p38MAPK (Cell Signaling Technology); EGFR, ERBB3, MET, Ubiquitin, cCBL, GAPDH (Santa Cruz Biotechnology); and V5 (Invitrogen) were incubated overnight at 4°C in 5% milk/PBST. Anti-phospho-tyrosine (Cell Signaling Technology) antibody was incubated 3 hours at room temperature. Membranes were washed in PBST and incubated 1 hour in horseradish peroxidase-conjugated secondary antibodies in 5% milk/PBST.

RNA Isolation and real-time PCR

RNA was isolated using the RNeasy Mini Plus kit with QIAshredder columns (Qiagen) and cDNA was synthesized using the iScript kit (Bio-Rad). Real-time PCR was performed on a Bio-Rad iCycler real-time PCR machine by combining cDNA. TaqMan universal master mix, and premixed FAM-labeled TaqMan probes (Applied Biosystems). mRNA quantity was calculated relative to GAPDH using the 2-ΔΔCt method.

Migration, Invasion, and Growth Assays

For migration assays, cells were pre-incubated 3 hours with 10ng/ml EGF, 1μM PHA665752, or 1μM PF-04217903 then plated at 5×10^4 cells/well in 24 well plates with 8μm filter inserts (BD Biosciences). Conditions were maintained and cells migrated overnight from 0.1% FBS toward 10% FBS. Cells were stained and cell number was averaged from three fields of view per chamber. Invasion assays were performed similarly but with 1.4×10⁵ cells/well in Matrigel invasion chambers (BD Biosciences). For growth assays, cells were plated at 1×10⁴ cells/well in a 12 well dish and counted daily for 5 days. Results are the average of at least three replicates and p-values were calculated by T-test.

Colony Formation Assay

Cells were plated at 2000 cells/well in a 6 well dish and grown for 7 days then washed in PBS, fixed 10 minutes on ice with cold methanol, and stained 2 hours with 0.1% crystal violet/PBS. Cells were rinsed 30 minutes in PBS and air-dried. Colonies were counted using scanned images with ImageJ software.

In vivo Metastasis Assay

Animal procedures were performed in accordance with the Yale Institutional Animal Care and Use Committee. 2.5×10^4 cells were suspended in 0.1ml PBS and injected into the right ventricle of six week old nude athymic mice. Metastasis was detected by bioluminescence with an IVIS Spectrum as previously described (17). Tumor incidence was quantified by luminescent signal in limbs, spine, or brain in whole body images at given time points and presented as Kaplan-Meier curves. P-values were calculated by Log rank test. When possible, animals were sacrificed for harvest of tissues to confirm metastasis by quantifying organ-specific luminescence.

Results

Signaling through EGFR leads to delayed activation of MET

Discrepancies in ERBB-MET interactions among NSCLC cell lines arise from underlying differences in ERBB mutational status and relative expression levels. Since ERBBs cross-activate, it was uncertain which individual ERBB(s) interact with MET or if intermediary signaling pathways indirectly integrate MET with ERBB signaling. To distinguish these possibilities, we employed 32D cells that lack ERBBs and MET. Since epithelial cell lines express one or more ERBBs, use of murine hematopoietic 32D cells has been invaluable for elucidating ERBB interactions in reconstruction experiments (18, 19). RT-PCR verified that 32D cells are devoid of EGFR, ERBB2, ERBB3, ERBB4, MET, and agonists EGF, TGFα, AREG, BTC, NRG1, and HGF, consistent with previous studies (data not shown) (18, 19). We then ectopically expressed EGFR and MET alone or in combination (Figure 1A).

Figure 1. EGFR activation increases MET protein and phosphorylation in 32D cells.

A, 32D cells were transfected with constructs for expression of human EGFR or MET, and protein expression was tested by immunoblotting. B, 32D cells expressing EGFR and MET were incubated without serum for 4 hours then stimulated with 10ng/ml EGF for the indicated times. Following stimulation, cells were lysed and analyzed by immunoblotting with the indicated antibodies. C, mRNA was collected from 32D cells after a 4 hour incubation without serum followed by stimulation with or without 10ng/ml EGF for 4 or 12 hours. Relative expression of human MET mRNA normalized to mouse GAPDH was determined by quantitative RT-PCR. D, The ratio of phospho-MET to total MET against duration of EGF stimulation was determined by densitometry of panel B. E, EGFR/MET 32D cells were incubated without serum for 4 hours, then incubated for 4 hours with 1μM PHA665752, 1μM gefitinib, and/or 10 ng/ml EGF as indicated and lysed for immunoblotting with the indicated antibodies. F, EGFR L858R/MET 32D cells were incubated for 4 hours without serum, then incubated with 1uM gefitinib and/or 10ng/ml EGF for 4 hours and lysed for immunoblotting with the indicated antibodies.

In 32D cells engineered to express human EGFR or MET singly, EGF stimulated EGFR phosphorylation and HGF stimulated MET phosphorylation, as expected (Supplementary Figure S1A). EGF did not induce MET phosphorylation and HGF did not induce EGFR phosphorylation, verifying specificity of ligands for their receptors (Supplementary Figure S1A). In cells expressing both EGFR and MET (EGFR/MET), HGF failed to induce phosphorylation of EGFR at Y1068 (Supplementary Figure S1B). However, EGF increased MET phosphorylation and protein levels (Figure 1B). Although EGF induces rapid phosphorylation of EGFR, EGF-induced phosphorylation of MET was delayed, occurring after two hours and increasing through ten hours (Figure 1B). EGF stimulation induced an increase in total MET levels without affecting the abundance of MET mRNA (Figure 1C). Also, quantification of phospho-MET to total MET revealed increased stoichiometry of MET phosphorylation induced by EGF (Figure 1D). Overall, EGFR upregulates MET abundance and relative Tyr phosphorylation over an extended time frame, and does not require other ERBB family receptors.

EGFR and MET kinases are required for EGF-dependent activation of MET

EGF-induced MET phosphorylation could occur through direct phosphorylation of MET by EGFR, EGF-dependent activation of MET autophosphorylation, or another intermediary kinase. To investigate the importance of kinase activity in this receptor interaction, small molecule tyrosine kinase inhibitors gefitinib and PHA665752 were used to inhibit EGFR and MET, respectively. Specificity of inhibitors for the cognate receptors was verified (Supplementary Figure S2A and S2B). Although slight changes in phospho-EGFR were seen with single inhibitors, these changes reflected similar differences in total protein level and were not seen consistently (Figure 1E). Treatment of EGFR/MET 32D cells with gefitinib prevents EGF-induced MET phosphorylation, and PHA665752 completely abolishes both EGF-induced and basal phosphorylation of MET (Figure 1E). Similarly, EGF was unable to induce robust phosphorylation of kinase-inactive MET mutants K1110A and Y1234F/Y1235F, while still activating wild-type MET receptor and docking site mutant Y1349F/Y1354F (Supplementary Figure S2C). Hence, EGF-induced MET phosphorylation in 32D cells requires activity of both EGFR and MET kinases.

Enhanced MET signaling is one mechanism by which NSCLC patients with EGFR activating mutations, such as EGFR L858R, become resistant to therapy with the EGFR inhibitor erlotinib (2-4). Recent evidence suggests that MET and activated EGFR also interact in drug-naïve lung cancer cells independent of drug resistance mechanisms (6-9, 12). To determine if activated EGFR alleles commonly found in NSCLC activate MET, MET was expressed in 32D cells in combination with EGFR L858R. Similar to wild-type EGFR, EGFR L858R induced MET phosphorylation after 4 hours of EGF stimulation, and activation was prevented by gefitinib (Figure 1F). These results show for the first time that EGFR activates MET independent of other ERBB family members, and that both wild-type and activated EGFR induce MET phosphorylation.

ERBB3 enhances EGFR induced MET phosphorylation

ERBB3 is nearly devoid of intrinsic catalytic activity, and functions through heteromers with other ERBBs (20, 21). Although ERBB3 phosphorylation can be dependent on MET in gefitinib-resistant NSCLC cells, it is unclear whether ERBB3 and MET interact in drug-naïve cells or if another ERBB family member is required (1). To investigate signaling interactions between ERBB3 and MET, ERBB3 was co-expressed with MET alone or with EGFR/MET in 32D cells. Co-expression of ERBB3 enhanced EGF-induced MET phosphorylation compared to EGFR/MET alone (Figure 2A). Similar to EGFR/MET cells, stimulation of EGFR/ERBB3/MET cells with ERBB ligands did not affect MET mRNA (Figure 2B). Increased EGF-dependent phospho-MET relative to total MET occurs within four hours in EGFR/ERBB3/MET cells and is maximal by seven hours (Figure 2C). Hence, ERBB3 enhances EGF-induced MET phosphorylation, likely through signal amplification and broadening by EGFR/ERBB3 heteromers. In fact, co-expression of EGFR and ERBB3 alters EGF-induced downstream signaling, leading to decreased phospho-Akt, increased phospho-Erk, and possibly increased total Erk (Figure 2D).

Figure 2. ERBB3 enhances EGFR activation of MET and activates MET when co-expressed with EGFR.

A, 32D cells expressing EGFR/MET or EGFR/ERBB3/MET were incubated without serum for 4 hours, stimulated with 10ng/ml EGF for the indicated times, then lysed for immunoblotting with the indicated antibodies. B, mRNA was harvested from 32D cells expressing EGFR, MET, and ERBB3 after a 4 hour incubation without serum followed by stimulation with or without 10ng/ml EGF or 50ng/ml NRG for 4 or 12 hours. Relative expression of human MET mRNA normalized to mouse GAPDH was determined by quantitative RT-PCR. C, the ratio of phospho-MET to total MET over a time course of EGF stimulation was determined by densitometry of panel A. D, 32D cells expressing EGFR and MET or EGFR, ERBB3, and MET were incubated without serum for 4 hours then stimulated with 10ng/ml EGF for the indicated times and lysed for immunoblotting with the indicated antibodies. E, 32D cells expressing ERBB3 and MET or ERBB3, EGFR, and MET were incubated without serum for 4 hours, stimulated with 50ng/ml NRG for the indicated times, then lysed for immunoblotting with the indicated antibodies.

Having determined that ERBB3 alters signaling to MET by EGFR, we investigated whether MET can be activated by ERBB3 alone. Stimulation of ERBB3/MET cells with the ERBB3 ligand neuregulin 1 (NRG) for up to 10 hours failed to activate MET or Akt (Figure 2E). Lack of ERBB3/MET signaling activity suggests these receptors cannot act as direct dimerization partners. However, when EGFR is co-expressed in ERBB3/MET cells, NRG induces MET phosphorylation (Figure 2E), indicating that ERBB3 can induce MET phosphorylation when it has a catalytically active signaling partner. Therefore, MET cannot act as a functional dimerization partner for ERBB3, but ligand-activated ERBB3 is able to induce MET phosphorylation when co-expressed with EGFR.

EGFR activates MET through protein stability and MAP Kinase signaling in 32D cells

EGFR may activate MET through direct phosphorylation, increased autocrine MET signaling, or both. EGF-induced MET activation is Actinomycin D-sensitive and thus requires transcription (Supplementary Figure S3A). However, MET activation was not associated with changes in MET mRNA levels (Figure 1C) and HGF mRNA was not expressed (data not shown). Therefore, transcriptional dependency in 32D cells involves another component of the pathway.

The changes observed in MET levels without an increase in MET mRNA suggest instead that EGFR stabilizes MET protein. 32D cells expressing MET or EGFR/MET were treated with the protein synthesis inhibitor, cycloheximide (CHX), and the half-life of MET protein was evaluated. EGF failed to stabilize MET in cells expressing MET alone, as expected (Figure 3A). However, MET was stabilized by co-expression of EGFR and further stabilized when EGFR was activated by EGF (Figure 3A). Overexpression of EGFR increased the MET half-life from 3.5 to 5.5 hours, while EGF-stimulation of EGFR further increased the half-life to 6.3 hours (Figure 3B). It is likely that the increase in MET half-life by EGFR expression is caused by basal EGFR activity from high expression levels. Overall, MET protein is stabilized by EGFR signaling in 32D cells.

Figure 3. EGFR signaling stabilizes MET protein and induces MET phosphorylation through MAP kinase signaling.

A, 32D cells expressing MET or EGFR and MET were incubated without serum for 4 hours, pretreated with or without 10ng/ml EGF for 15 minutes, then incubated with 10ng/ml cycloheximide for the indicated times and lysed. Total MET (V5) and GAPDH levels were analyzed by immunoblotting. B, The half-life of MET was calculated based on densitometry of panel A. Results from three experiments were averaged and p-values were calculated by t-test. C, EGFR/MET 32D cells were incubated without serum for 4 hours, incubated with 1uM gefitinib (Gef), dasatinib (Das), or PP2, incubated with or without 10ng/ml EGF, and lysed for immunoblotting with the indicated antibodies. D, 32D cells expressing EGFR and MET or E, EGFR L858R/T790M and MET were incubated without serum for 4 hours, incubated with 1μM of the indicated agents, stimulated with 10ng/ml EGF for 4 hours, then lysed for immunoblotting with the indicated antibodies.

In addition to regulating MET protein stability, it is possible that EGFR activates MET through signaling proteins such as Src, which acts as an intermediary in EGFR-induced MET phosphorylation of the EGFR wild-type NSCLC cell line, 201T (12). To determine if this is a general mechanism of EGFR-induced MET phosphorylation, Src inhibitors were evaluated for the ability to prevent EGFR-MET cross-talk in 32D cells. Interestingly, neither Dasatinib nor PP2 prevented EGF-induced MET phosphorylation (Figure 3C). To confirm this result, MET was expressed in SYF−/− MEFs with constitutional knockout of Src family members Src, Yes, and Fyn (16). EGF stimulation led to turnover of EGFR, causing decreased EGFR protein, however, activated EGFR was still detectable, indicating that EGFR signaling does occur (Supplementary Figure S3B). Despite the lack of Src proteins, EGF activated MET accumulation and phosphorylation, indicating Src family proteins are not absolutely required for EGF-induced MET phosphorylation (Supplementary Figure S3B). Thus, although Src is involved in EGFR-induced MET activation in some contexts, it is not a unique mediator of and is not required for EGFR-MET crosstalk.

To identify intermediaries in EGF-induced MET phosphorylation in 32D cells, pools of protein kinase inhibitors were tested for interference with this process (Supplementary Figure S3C). Of these, inhibitor pools A, C, and D prevented EGF-induced MET phosphorylation. Pool C yielded inconsistent results among experiments, and only partially inhibited Met phosphorylation. Pool A includes at least one agent, BMS754807, which is now reported to have some inhibitory activity on MET. For these reasons, we focused on Pool D, comprised of MAPK signaling inhibitors.

To evaluate the role of MAP kinases in EGFR-MET signaling, inhibitors of MEK1/2 and p38MAPK, U0126 and SB203580, were tested for ability to prevent EGFR-MET crosstalk. These inhibitors had moderate or no effect on EGF-induced MET phosphorylation singly, but nearly completely prevent it when combined (Figure 3D). This suggests that EGF-induced MET phosphorylation works through the RAS-ERK-p38MAPK pathway, and compensation occurs through parallel MAP kinase cascades when only one is inhibited.

While gefitinib inhibits EGF-induced MET phosphorylation in cells with wild-type EGFR or EGFR L858R (Figure 1E and 1F), it is ineffective against the resistance mutation, T790M. To determine if MEK and p38MAPK inhibitors prevent EGFR activation of MET in cells with T790M mutations, EGFR L858R/T790M (LT) receptors were co-expressed with MET in 32D cells. Gefitinib did not affect EGFR or MET phosphorylation in these cells, but the combination of U0126/SB203580 substantially decreased MET phosphorylation (Figure 3E). Hence, MEK/p38MAPK inhibitors can inhibit EGFR-MET signaling in cells where gefitinib is not effective. Overall, EGFR activation of MET in 32D cells occurs through increased stability of MET protein and intermediary signaling through MAP kinases.

EGFR regulates MET at multiple levels in NSCLC cell lines

To confirm the presence of EGFR-MET crosstalk in NSCLC cell lines, the EGFR mutant cell line, HCC827, was treated with gefitinib for up to 24 hours. Gefitinib treatment induced rapid decreases in MET phosphorylation and delayed decreases in MET protein levels (Figure 4A). These different rates of change suggest EGFR can affect MET phosphorylation independent of MET protein levels. Unlike 32D cells, EGFR signaling in HCC827 cells regulates MET protein and MET mRNA with similar kinetics following gefitinib treatment (Figures 4B and 4C), emphasizing the complex dynamics of EGFR-MET signaling in NSCLC cell lines. Similar changes in MET phosphorylation, protein, and mRNA were observed following gefitinib treatment of another EGFR mutant cell line, PC9 (Supplementary Figure S4).

Figure 4. EGFR regulates MET phosphorylation, protein, and mRNA levels in NSCLC cells.

A, HCC827 cells were incubated with 1μM gefitinib for the indicated times then lysed for immunoblotting with the indicated antibodies. B, MET protein levels were determined by densitometry of panel A and normalized to GAPDH. Results were averaged among three experiments and p-values were calculated by t-test. C, HCC827 cells were incubated with 1μM gefitinib for the indicated times. mRNA was isolated and the relative expression of human MET mRNA normalized to human GAPDH was determined by quantitative RT-PCR. D, NSCLC cell lines with the indicated EGFR mutations were incubated overnight without serum then with 50ng/ml HGF for 10 minutes as marked. Cells were lysed and MET or Cbl immunoprecipitations (“IP”) were performed followed by immunoblotting with the indicated antibodies. E, HCC827 cells were incubated overnight with 1μM of the indicated agents and lysed for immunoblotting with the indicated antibodies. * p<0.05.

Although changes in MET protein levels following gefitinib treatment may be caused by decreased mRNA abundance, there is evidence for EGFR-dependent modulation of MET at the protein level through MET ubiquitination. Following stimulation by HGF, MET is ubiquitinated by the E3 ligase, c-CBL, and internalized for intracellular trafficking or degradation (22-25). Although three cell lines with wild-type EGFR underwent robust MET ubiquitination following ligand stimulation, three out of four cell lines with EGFR activating mutations did not (Figure 4D). Importantly, MET ubiquitination was associated with phosphorylation of c-CBL (Figure 4D). This corroborates our finding that, in 32D cells, EGFR stabilizes MET protein levels independent of transcriptional changes and indicates that EGFR modulates MET at multiple levels in NSCLC cell lines.

To determine if intermediary signaling crosstalk occurs through MAP kinases in NSCLC cell lines, HCC827 cells were treated with U0126/SB203580. Interestingly, treatment with either gefitinib or U0126/SB203580 caused a reduction in total MET levels, accompanied by elimination of phospho-MET (Figure 4E). This confirms that signaling observed in 32D cells resembles NSCLC cell lines and that MEK/p38MAPK pathways act as signaling intermediaries between EGFR and MET in multiple cell types. Although NSCLC cells display a more complicated relationship between EGFR and MET than 32D cells, NSCLC cells similarly show independent regulation of MET protein and phosphorylation by EGFR signaling. Importantly, the existence of EGFR-MET crosstalk in multiple NSCLC cell lines indicates that EGFR induced MET activation may have a biological significance.

MET regulates EGFR-induced migration and invasion in NSCLC cells

MET regulates many cellular phenotypes, including cell growth and motility. We compared EGFR-only to EGFR/MET 32D cells and found that EGFR-MET crosstalk had no effect on cell viability (data not shown). Since MET signaling is also important for cell motility, it is possible that EGFR activation of MET modulates this phenotype. MET is required for EGF-induced cell invasion and motility in EGFR wild-type NSCLC cells but it was unclear if this is true in cells with EGFR activating mutations (12). Hence, we confirmed the previously reported MET-dependence for EGF-induced migration and invasion in the EGFR wild-type NSCLC cell line A549 (Supplementary Figure S5A) and tested additional cell lines with various EGFR mutations. Indeed, PHA665752 prevented EGF-induced migration of NSCLC cell lines with both wild-type and activated EGFR (Figure 5A). Similarly, MET inhibition prevented EGF-induced invasion through Matrigel of EGFR wild-type and mutant NSCLC cells (Figure 5B). This was confirmed with an additional highly specific MET inhibitor, PF-04217903, that has strong activity in NSCLC cells (Supplementary Figures S5B and S5C). Therefore, EGFR-MET signaling is active in NSCLC cell lines with multiple EGFR activating mutations and mediates aggressive phenotypes including migration and invasion.

Figure 5. MET promotes EGF-induced cell migration and invasion of NSCLC cell lines.

NSCLC cell lines were pretreated for 3 hours with 10ng/ml EGF (EGF), 1uM PHA665752 (PHA), or both (EGF+PHA) then plated in migration or invasion chambers and allowed to migrate overnight. Results are represented relative to no treatment controls (none) and are the average of at least three experiments. p-values were calculated by t-test.

EGFR-MET signaling is important for metastatic behavior

MET promotes EGF-induced NSCLC cell invasion, so it may have a role in the biology of metastatic lung cancer cells addicted to EGFR. To investigate this, we compared EGFR-MET signaling in an EGFR mutant NSCLC cell line, PC9, and its metastatic subpopulation, PC9-BrM3 (15, 26, 27). Both parental PC9 and PC9-BrM3 cell lines harbor constitutively active EGFR Δ746-750 and are dependent on EGFR signaling for survival. However, PC9-BrM3 cells have a marked increase in the capacity to invade and colonize distant organs, most notably the brain (15).

EGFR inhibition in both PC9 and PC9-BrM3 cells decreases MET phosphorylation and protein levels, confirming the presence of EGFR-MET signaling in these cells (Figure 6A). However, the mechanism of EGFR-MET signaling differs as MEK/p38MAPK inhibition reduces MET phosphorylation and protein levels in PC9-BrM3 cells, but has little effect on PC9 cells (Figure 6A). Different requirements for MAP kinases in EGFR-MET signaling may correlate with lower Erk phosphorylation observed in PC9 compared with PC9-BrM3 cells (Figure 6B). Additionally, mRNA levels of both EGFR and MET trended higher in PC9-BrM3 than PC9 cells (Figure 6C). Although possibly coincidental, these differences in EGFR-MET signaling raised the possibility that this pathway contributes to the highly metastatic behavior of PC9-BrM3 cells.

Figure 6. EGFR-MET signaling is enhanced in metastatic cells.

A, PC9 or PC9-BrM3 cells were incubated for 24 hours with 1μM of the indicated agents then lysed for immunoblotting with the indicated antibodies. B, PC9 or PC9-BrM3 cells were lysed and immunoblotting was performed with the indicated antibodies. C, RNA was isolated from PC9 and PC9-BrM3 cells and expression levels of EGFR and MET mRNA normalized to GAPDH were determined by quantitative RT-PCR. D, Growth of PC9 and PC9-BrM3 cells with control (shScr) or MET knockdown was counted over 5 days and is represented as cell number averaged among three experiments. E, PC9 and PC9-BrM3 cells with control (shScr) or MET knockdown were pretreated with or without 10ng/ml EGF for 3 hours then plated in migration or invasion chambers and allowed to migrate overnight. Results are represented as EGF-induced migration or invasion relative to shScr and are the average of at least three experiments. F, Clonogenic colony formation of PC9 and PC9-BrM3 cells with control (shScr) or MET knockdown was tested by plating cells at low density and staining for colonies after 6 days of growth. The number of colonies was quantified using Image J software and is averaged among at least 6 wells. Representative stained plates are shown. A-F, p-values were calculated by t-test, * p<0.05, ** p<0.01.

To directly evaluate the role of MET in EGFR-driven phenotypes of PC9 and PC9-BrM3 cells, a lentiviral system was used to induce MET knockdown (Supplementary Figure S6A). Similar to growth studies in 32D cells, MET knockdown did not affect growth of PC9 or PC9-BrM3 cells (Figure 6D). In contrast with other NCSLC cells, MET was not required for EGF-induced cell invasion of parental PC9 cells and knockdown inconsistently affected cell migration (Figure 6E). Interestingly, MET knockdown prevented both EGF-induced cell migration and invasion in PC9-BrM3 cells (Figure 6E). This was confirmed with PHA665752 treatment, demonstrating consistent cellular responses to MET inhibition and knockdown (Supplementary Figure S6B). The role of MET in clonogenic potential also differed between the cell lines. MET knockdown with two different shRNAs yielded inconsistent changes in clonogenic colony formation of PC9 cells, suggesting these effects were not MET specific (Figure 6F). However, knockdown of MET in PC9-BrM3 cells consistently reduced clonogenic colony formation (Figure 6F). Although EGFR-MET crosstalk exists in parental PC9 cells, the functional consequence of EGFR-MET crosstalk is different in metastatic cells where it enhances clonogenicity and invasion, independently of cell survival. Thus, although NSCLC cells generally require mutant EGFR signaling for tumorigenesis, crosstalk with MET may further enhance metastatic progression.

MET signaling modulates metastasis in many epithelial cancers (28-31). In NSCLC patients, MET expression and phosphorylation correlate with the incidence of brain metastasis, but the functions of MET in lung cancer metastasis remain uncharacterized (32). A rate-limiting step in metastasis is invasion and colonization of distant organs by cancer cells following dissemination into the bloodstream (33). To assess effects of EGFR-MET signaling on metastatic colonization, we injected luciferase-marked PC9-BrM3 cells into the arterial circulation of immunocompromised mice. Incidence and burden of metastasis were compared following intracardiac injection of PC9-BrM3 cells stably expressing control or MET knockdown vectors. Surprisingly, no significant difference was seen in overall tumor burden between control and MET knockdown (Figure 7A). While MET knockdown had no effect on bone metastasis, incidence of brain metastasis was significantly delayed and/or attenuated (Figures 7B,C). When possible, animals were sacrificed to confirm metastasis by brain imaging. Although there was no significant difference in the brain tumor burden in animals that harbor metastasis at endpoint, MET knockdown decreased the frequency of brain metastasis from 100% to 30% (Figure 7B). This corroborates in vitro results of MET knockdown decreasing cell invasion but not growth (Figure 6). In summary, MET signaling enhances in vivo brain metastatic invasion by EGFR-addicted NSCLC cells. This supports the clinical correlation observed between MET expression and metastatic relapse in NSCLC patients, and emphasizes the significance of EGFR-MET signaling in aggressive stages of lung cancer progression (32).

Figure 7. MET knockdown decreases the incidence of brain metastasis of EGFR mutant NSCLC cells.

A, Metastatic tumor burden was measured by normalized whole body total photon flux at 35 days after injection of 2.5×10⁴ PC9-BrM3 cells expressing control (shScr) or MET knockdown shRNA. B, incidence of bone and brain metastasis was determined by organ-specific luminescent signal in the whole body image. Data are represented as metastasis-free survival in a Kaplan-Meier curve. p-values were determined by a log rank test. C, where possible, metastasis was confirmed by isolating and imaging the brain from animals expressing control (shScr) (N = 7) or MET knockdown (N = 10) shRNA. Representative images are shown for brain metastasis-positive and -negative animals.

Discussion

Existence of signaling interactions between MET and the ERBBs is well established, but mechanism(s) are incompletely understood. For reconstruction of defined sets of receptors, it was essential to use a receptor-null cell background, which was not possible for epithelial lines since they express one or more ERBBs. We used murine 32D cells, which do not express endogenous MET or any ERBBs, to investigate interactions between MET and individual ERBBs. EGFR signaling was sufficient to induce MET phosphorylation and increase MET protein levels, but MET activation was augmented by co-expression of ERBB3. Rather than direct cross-phosphorylation, EGFR activation of MET occurs through combined stabilization of MET protein and intermediary signaling through MAP kinases. EGFR-MET crosstalk in lung cancer cells is more complex, and involves multiple levels of regulation including RNA, protein, and phosphorylation. Significantly, in lung carcinoma, EGFR-MET signaling is exaggerated in metastatic cells where MET enhances EGFR-mediated aggressive phenotypes including migration, invasion, and metastasis, independently of cell growth.

EGFR, ERBB2, and ERBB3 have all been implicated in ERBB-MET signaling crosstalk. However, promiscuous ERBB dimerization and co-expression of multiple ERBBs in NSCLC cells makes it difficult to determine which ERBB(s) activate MET. Using a model system, we determined that EGFR activation by ligand or mutation is sufficient to induce MET phosphorylation (Figure 1). Although we did not observe activation of EGFR by MET, only one EGFR phosphorylation site was tested and we cannot rule out such crosstalk entirely. Additionally, ERBB3 enhances EGFR-driven phosphorylation of MET and activates MET itself when part of an active dimer (Figure 2). Consequently, MET activation may occur in NSCLC tumors with active EGFR but could be enhanced in tumors co-expressing other ERBB receptors. In addition to enhancing EGFR-induced MET activation, ERBB3 alters downstream signaling of EGFR through MAPK and PI3K pathways (Figure 2). Therefore, co-expression of ERBB3 in NSCLC may affect not only ERBB-MET crosstalk but intracellular signaling as well.

Multiple different mechanisms may regulate EGFR activation of MET including direct receptor interactions, regulation of MET autocrine signaling, or indirect signaling through intermediary proteins. We have excluded direct receptor cross-phosphorylation based on delayed EGF-induced MET phosphorylation, requirement for both EGFR and MET kinase activities, and the inability of MET to act as a dimerization partner for ERBB3 (Figures 1 and 2). Instead, we see evidence for ERBB regulation of MET through both autocrine and intermediary signaling.

In some cell lines, EGFR activates MET by inducing MET transcription, thereby increasing the concentration of MET at the membrane, leading to more receptor collisions, homodimer formation, and MET activation (12, 34, 35). In our model system, EGFR regulates MET at the protein level by extending the MET half-life (Figure 3), which would also increase MET availability and enhance autocrine activation. In papillary renal carcinoma, MET mutations lead to increased recycling and decreased degradation, causing increased activation of MET that drives tumorigenesis (36). We see decreased MET ubiquitination in EGFR mutant NSCLC cells, which would similarly alter MET recycling and degradation (Figure 4). Therefore, active EGFR in NSCLC tumors may regulate MET through altered receptor trafficking to enhance already robust tumorigenic signaling.

Dulak et al., reported that Src is required for EGFR-induced MET activation independent of MET transcription, indicating MET can also be regulated through intermediary signaling pathways (12). We found that MAP kinase signaling correlated with higher ERBB-driven MET phosphorylation and was required for EGFR-dependent MET activation in cells with both wild-type and mutant EGFR (Figures 3 and 4). Different requirements for Src and MAP kinase signaling may depend on the cell line, and it is possible that both pathways act as intermediates in different contexts. Additionally, Src can enhance activation of MAPK in cancer cells directly or by binding to EGFR, so these pathways may have additive functions (37). Investigation of additional experimental models and tumor samples could reveal the relative roles of these proteins in EGFR-MET crosstalk in different cell types and contexts.

Existence of EGFR-MET signaling in multiple normal and cancer cell lines suggests MET modulates biological outcomes of EGFR signaling. Although both receptors activate PI3K/Akt signaling, MET was not required for cell survival in 32D or NSCLC cells. EGFR and MET also commonly regulate cell motility and MET promotes EGF-induced cell motility and invasion in EGFR wild-type NSCLC cells (8, 12). NSCLC cells with EGFR mutations have constitutively hyperactive EGFR signaling and it was unclear if MET could regulate EGFR driven phenotypes in these cells. We found, however, that MET facilitates EGF-induced migration and invasion in EGFR mutant NSCLC cells (Figure 5), demonstrating that MET signaling enhances other aggressive phenotypes of EGFR mutant lung cancers.

Moreover, we discovered differences in EGFR-MET crosstalk in NSCLC cell lines of varying metastatic potential. In metastatic PC9-BrM3 derivatives, but not parental PC9 cells, EGFR-MET signaling enhanced migration, invasion, and colony formation, and was dependent on MAP kinase signaling (Figure 6). Differences between PC9 and PC9-BrM3 cells reveal context dependent requirements for EGFR-MET signaling in metastatic NSCLC cells. Since PC9-BrM3 cells were selected to be highly metastatic in vivo, enhanced signaling in these cells suggests EGFR-MET crosstalk could be a mediator of metastasis and that the mechanism of crosstalk can change throughout progression.

MET regulates metastasis of gastric cancer, renal papillary carcinomas, and breast cancer and MET expression and activity in NSCLC correlate with occurrence of brain metastasis (29, 32, 38, 39). Additionally, MET copy number, expression, and phosphorylation are enriched in NSCLC brain metastases compared to primary tumors (32). Despite association of MET with metastasis, a direct role for MET in metastasis of NSCLC has not been previously identified. We observed that MET knockdown in highly metastatic NSCLC PC9-BrM3 cells caused significant reduction in metastasis to the brain (Figure 7). Differences between brain and other organ sites may depend on the brain being a more stringent organ to invade, which is supported by delayed detection of brain metastasis versus other sites in our model. Additionally, differences in brain tumor incidence but not burden suggest that an important role of MET in lung cancer metastasis may be to promote invasion of the brain environment. The significance of this result is underscored by the fact that the brain is the major metastatic site and source of morbidity in lung cancer patients (40). The inability of mouse HGF to activate human MET makes it difficult to address the relative contributions of EGFR-MET crosstalk versus HGF paracrine MET activation towards metastasis with this model, but does not detract from the importance of MET in promoting NSCLC progression. These results support further investigation of the role of MET in EGFR-mutant NSCLC brain metastasis and the use of MET inhibitors to prevent metastatic progression of NSCLC in patients.

Previous clinical trials for MET inhibitors in NSCLC were focused on patients who were resistant to EGFR TKIs due to MET amplification. The ability of both wild-type and mutant EGFR to activate MET, as well as MET regulation of EGFR-driven migration, invasion, and metastasis, suggests that MET inhibitors may be beneficial to patients with varying mutational status. In fact, a recent clinical trial of combination EGFR/MET inhibitors that did not pre-screen patients for mutation status found increased progression free and overall survival in wild-type EGFR NSCLC patients (41). Additionally, EGFR-MET signaling becomes more important at later stages of NSCLC progression, indicating that use of MET inhibitors in early stage cancer patients may prevent occurrence of invasion and metastasis and should be evaluated as a therapeutic option. Further investigation of EGFR-MET crosstalk will be useful for determining which patients may benefit most from combination therapies and for identifying potential new targets to prevent EGFR induced MET activation.