MET-EGFR dimerization in lung adenocarcinoma is dependent on EGFR mtations and altered by MET kinase inhibition

Advanced lung cancer has poor survival with few therapies. EGFR tyrosine kinase inhibitors (TKIs) have high response rates in patients with activating EGFR mutations, but acquired resistance is inevitable. Acquisition of the EGFR T790M mutation causes over 50% of resistance; MET amplification is also common. Preclinical data suggest synergy between MET and EGFR inhibitors. We hypothesized that EGFR-MET dimerization determines response to MET inhibition, depending on EGFR mutation status, independently of MET copy number. We tested this hypothesis by generating isogenic cell lines from NCI-H1975 cells, which co-express L858R and T790M EGFR mutations, namely H1975L858R/T790M (EGFR TKI resistant); H1975L858R (sensitized) and H1975WT (wild-type). We assessed cell proliferation in vitro and tumor growth/stroma formation in derived xenograft models in response to a MET TKI (SGX523) and correlated with EGFR-MET dimerization assessed by Förster Resonance Energy Transfer (FRET). SGX523 significantly reduced H1975L858R/T790M cell proliferation, xenograft tumor growth and decreased ERK phosphorylation. The same was not seen in H1975L858R or H1975WT cells. SGX523 only reduced stroma formation in H1975L858R. SGX523 reduced EGFR-MET dimerization in H1975L858R/T790M but induced dimer formation in H1975L858R with no effect in H1975WT. Our data suggests that MET inhibition by SGX523 and EGFR-MET heterodimerisation are determined by EGFR genotype. As tumor behaviour is modulated by this interaction, this could determine treatment efficacy.


Introduction
Epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (EGFR-TKIs) have revolutionised treatment of non-small cell lung cancer (NSCLC) in patients with EGFR mutations.These mutations cause constitutive kinase activity and are oncogenic drivers in 10-20% of Caucasian patients and up to 50% of eastern Asians.[1] Such mutations induce conformational changes in the receptor that alter the dimerization interface, destabilize the inactive state and increase kinase activity to 50 times that of the wild type (WT) EGFR.[2] The EGFR exon 21 L858R and in-frame exon 19 deletions account for 85% of such mutations.[3] Whilst responses are often impressive, resistance is inevitable.The commonest mechanism for resistance is acquisition or clonal expansion of the EGFR exon 20 T790M mutation.
Amplification of the MET receptor represents an important alternative resistance mechanism [4,5,6,7].MET is a high affinity tyrosine kinase receptor for hepatocyte growth factor (HGF). [8] Derailment of normal MET signaling is associated with invasive growth, tumor progression and metastases; [9] aberrant MET signaling can result from MET over-expression, amplification or mutations, all of which are relevant in NSCLC.[4,5,6,7] MET amplification predicts worse survival in NSCLC, [10] it has been implicated in 5-20% of patients with acquired resistance to EGFR TKI [11,12,13,14] and correlates with response to MET inhibitor therapy [13].Blockade of MET is a therapeutic strategy in EGFR TKI resistance.The most advanced agents, METMAb, a MET neutralizing antibody and Tivantinib, a small molecule inhibitor of MET have both failed in phase III clinical trials [15]; despite this, there is considerable interest in the therapeutic potential of MET inhibition in NSCLC.In fact, Crizotinib, a MET proto-oncogene, receptor tyrosine kinase (MET) tyrosine kinase inhibitor (TKI) is currently in clinical trial showing good results for both MET amplification and MET exon 14 skipping [14].
MET may exert its oncogenic effects through crosstalk with other membrane receptors including the EGFR family, as evidenced by MET and EGFR co-expression in lung cancer cell lines, [16] crosstalk between EGFR and MET signaling pathways and direct co-immunoprecipitation.[16,17,18] Moreover, MET amplification in association with EGFR mutations additionally has a worse clinical prognosis than EGFR mutations alone.[10] In light of these observations, we sought to understand the importance of EGFR and MET interaction and we have hypothesized that the efficacy of MET inhibition can be influenced by EGFR mutation status.We explored this hypothesis by evaluating the response of three lung adenocarcinoma cell lines that differ only in their EGFR genotype to the MET inhibitor SGX523 in vitro and in a murine xenograft model derived from the same cells.Our data suggest that EGFR mutations can determine the effect of MET inhibition independently of MET copy number, by changing EGFR-MET dimerisation.As tumor behaviour is modulated by this interaction, this could determine treatment efficacy.

EGFR-MET interaction is modulated by EGFR mutations
To assess if EGFR-MET interaction is modified by EGFR mutations, we first generated two novel cell lines by modification of the NCI-H1975 lung adenocarcinoma cell line that harbours L858R and T790M (L858R/T790M) mutant EGFR (to be referred to from here on as H1975 L858R/T790M ).We used lentiviral shRNA knockdown of EGFR (targeting the 5' UTR of EGFR) in the H1975 L858R/T790M , followed by transfection with a plasmid encoding wild/type (wt) EGFR and EGFR with the L858R mutation, to generate the H1975 WT and the H1975 L858R cell lines respectively.Relative allele frequency (L858R vs. T790M) in these cell lines was quantified by digital droplet PCR.We observed equal L858R and T790M copies in the H1975 L858R/T790M cells and a clear reduction of L858R and T790M EGFR alleles in the H1975 WT cells, confirming their effective knockdown following shEGFR treatment; we also observed decrease of the EGFR-T790M allele frequency in the H1975 L858R cells (Fig 1A).Using Western blot (WB), we showed the total levels of EGFR in the generated cell lines (Fig 1B ) and that the H1975 L858R and H1975 WT cells became sensitive to the EGFR TKI Erlotinib We then explored the interaction between EGFR and MET in the cells lines studied.We observed no differences in basal MET protein levels between the cell lines (Fig 2A).Given that MET expression is not a good marker for MET activation we also treated the three cell lines with HGF for different times to asses for the phospho-MET levels.We could observe no difference in the activity of MET in the different cell lines (Fig B in S1 File).As MET amplification has been implicated as a mechanism by which secondary resistance EGFR TKI can emerge in NSCLC, we used a MET/cep7 fluorescence in situ hybridization (FISH) probe and calculated the ratio of MET to chromosome 7 centromere signals and the mean copy number of MET.The MET/cep7 ratio was <1 in all three cell lines; however, MET copy number was increased in all (Fig 2B ).
Immunofluorescence staining and confocal images showed that both proteins co-localise similarly in the membrane of the three cell types (Fig 2C).Immunoprecipitation result showed there was more interaction between EGFR and MET in H1975 L858R/T790M compared to H1975 L858R and H1975 WT cells (Fig 2D ), which provides evidence of potential differential interaction between MET and different EGFR mutant types.Since a direct protein interaction (nanometer (nm) proximity) between MET and EGFR cannot be inferred by these experiments, we assessed Fo ¨rster Resonance Energy Transfer (FRET) using fluorescence lifetime imaging microscopy (FLIM) which is the gold standard technique for measuring protein proximity within the typically <10nm range [19,20,21,22].Results show that highest FRET efficiency between MET and EGFR occurred in H1975 L858R/T790M cells, in contrast with significantly lower FRET efficiency in the H1975 L858R and H1975 WT cells (Fig 2E).These results demonstrate that EGFR and MET can interact directly at the cell membrane and that the level of interaction is significantly higher in H1975 L858R/T790M compared to H1975 L858R and H1975 WT cells.

Inhibition of MET changes the EGFR-MET interaction both in vitro and in vivo
In view of the observed differences in the interaction between EGFR and MET in our cells, we hypothesised that MET inhibition would also result in different outcomes.To test this, we used SGX523, a selective MET kinase inhibitor.[23] After confirming rapid (15min) and sustained (48hr) SGX523 inhibition of MET phosphorylation in the three cell lines (Fig C in S1 File), we measured change in the interaction between MET and EGFR using FRET before and after MET inhibition by SGX523.We found that the interaction between MET and EGFR in H1975 L858R/T790M cells was significantly reduced in the presence of SGX523.By contrast, SGX523 led to a significant increase in MET-EGFR interaction compared to baseline in H1975 L858R cells (Fig 3A and 3B).There was no significant difference in the FRET between MET and EGFR before and after MET inhibition in the H1975 WT cells.These results provide evidence to support the hypothesis that MET inhibition by SGX523 altered the direct interaction between MET and EGFR in a mutation-specific manner in these cells.
To test whether the differential effects of SGX523 on MET and EGFR interaction could be reproduced in vivo, we generated a xenograft model using the H1975-derived mutant cell lines.H1975 L858R/T790M , H1975 L858R and H1975 WT cells were injected into the flanks of BalbC mice and tumors were allowed to grow for 2 weeks.Mice were then subjected to 12 days of treatment with SGX523 or vehicle administered daily by oral gavage.Tumor volumes were measured every two days until animals were culled and tissue collected (see Fig 4D).WB and We observed the highest FRET between MET and EGFR in H1975 L858R/T790M derived tumors, and this was again significantly reduced by SGX523 (Fig 3C and 3D).In H1975 L858R derived tumors, MET-EGFR interaction at baseline was low and increased significantly by SGX523.There was no significant change in FRET between MET and EGFR in xenograft tissue derived from H1975 WT cells.These results mirror the in-vitro findings with these cells (Fig 3C and 3D).We conclude that MET interacts with EGFR differently in cells that encode WT, L858R and L858R/T790M-EGFR and SGX523 modifies this interaction in opposite directions in L858R and L858R/T790M-EGFR encoding cells in vitro and in vivo.

In vitro and in vivo effects of MET inhibition on cell proliferation
In order to assess functional consequence of MET kinase inhibition, we assessed the proliferation rate of the generated cells before and after MET inhibition.We found that the H1975 L858R cells were more proliferative than the H1975 L858R/T790M or H1975 WT cells (Fig E (i) and (ii) in S1 File).MET inhibition significantly decreased proliferation rate in H1975 L858R/T790M cells only (Fig 4A and 4B).We also assessed proliferation in a 3D environment using an anchorage independent growth assay.As expected, we found that the H1975 L858R cells produced most colonies (Fig E (iii) in S1 File).In the presence of SGX523, there was a significant decrease in the number of large colonies in H1975 L858R/T790M cells.SGX523 had no effect in the 2D or 3D proliferation rate in the case of the H1975 WT cells or in the H1975 L858R cells (Fig 4C ).
We then analysed the in vivo effect of oral administration of SGX523 on tumor growth and proliferation.As was the case in vitro, the H1975 L858R was the most proliferative cell line in vivo (Fig 4D and Fig F (i) in S1 File).The H1975 L858R/T790M -derived tumors in SGX523 treated mice grew more slowly than in vehicle treated mice; this effect was evident after 4 days of SGX523 administration and was sustained up to 14 days.SGX523 had no effect in the proliferation in the H1975 L858R and H1975 WT derived tumors over the 14-day growth period (Fig 4D).This result was confirmed using the proliferation marker phospho-Histone H3 (Fig 4E).

In vitro and in vivo effects of MET inhibition on stroma remodeling
We observed marked morphological differences in the xenograft tumors derived from the different cell lines.H1975 L858R/T790M -derived tumors were dense and cellular with tightly packed cells oriented in similar directions.The H1975 L858R -derived tumors had abundant intratumoral stromal tissue deposition leading to highly segregated cells that appeared to grow by interaction with the stroma (Fig F (ii) in S1 File).Given this different tumours appearance, as well as the important role proposed for MET in tumour-tissue interaction and in cell migration/metastasis, we analysed stromal compartment markers, namely collagen deposition (Masson's trichrome), anti-smooth muscle actin staining (α-SMA), a marker of activated fibroblasts and cd31 staining to assess for neovascularization, before and after SGX523 treatment.The H1975 L858R -derived tumors demonstrated the most substantial collagen deposition

In vitro and in vivo effects of MET inhibition on ERK, AKT and FAK phosphorylation
We finally investigated whether there were also differential effects of MET kinase inhibition on signaling pathways.We treated H1975 L858R/T790M , H1975 L858R and H1975 WT cells with SGX523 for 24 hours.Phosphorylation of MET was inhibited but we saw no change in the phosphorylation of EGFR with the SGX523 treatment.We then assessed phosphorylation levels of ERK, AKT and FAK after SGX523.The only significant finding from these studies was a reduction in ERK phosphorylation in SGX523 treated H1975 L858R/T790M cells (Fig 6A and 6C).We repeated these experiments in the presence of EGF and or HGF in the cells treated or not with SGX523 to better understand the decrease in the stromal markers in the H1975 L858R but we observed no effect in the phosphorylation of AKT or FAK (Fig G in S1 File).Finally, the analysis of xenograft tumors obtained from SGX523 or vehicle treated animals also showed reduction in phosphorylated ERK in H1975 L858R/T790M -derived xenografts (Fig 6B and 6C).These observations provide potential explanation for our observation that SGX523 had antiproliferative effect in H1975 L858R/T790M -derived tumors but not H1975 L858R and H1975 WT cells and tumors.

Discussion
In this study we have demonstrated that the responses to MET inhibition in an in vitro and in vivo lung cancer model differed depending on the EGFR genotype.Using as a starting point the H1975 lung adenocarcinoma cell line that harbours double mutated (L858R/T790M double mutant) EGFR, we have developed a cellular model to minimise the contribution of confounding influences.Using Droplet PCR, WB and EGFR TKI sensitivity, we established that in each line the dominant EGFR genotype was of greatest functional importance.We considered the merits of a CRISPR approach to achieve purer cell populations.However, the validation steps described and the tumour cell heterogeneity inherent to our approach could offer a good reflection of the clinical situation.
Our key novel finding from in vitro and in vivo studies is that MET inhibition by SGX523 reduced proliferation and tumor growth in the H1975 L858R/T790M cells and inhibited indices of stromal deposition in H1975 L858R cells.MET inhibition in H1975 L858R/T790M cells also suppressed ERK signaling consistent with interference with an autocrine signaling loop.Conversely, in H1975 L858R cells the effects of SGX523 behavior was more consistent with a paracrine activity of the MET receptor (see Fig 5D).These variations in SGX523 responses in each cell line led us to hypothesize that these differences could be in part a product of the interaction between the EGFR and MET.
Existing evidence supports EGFR-MET crosstalk.Our own data showing co-localization and co-immunoprecipitation of EGFR and MET at the cell membrane is in agreement with these observations.Importantly, receptor dimerisation is essential for both MET and EGFR signaling and intracellular effect.EGFR can dimerise with itself and other members of the HER family.Previous studies have shown that cell proliferation and tumorigenesis are enhanced in tumor xenografts co-expressing HER family heterodimers compared to those expressing single receptors.[24] For example, the well-characterized HER2-HER3 pair is the most oncogenic in breast cancer.[25] It has also been suggested that MET can interact not only with EGFR but with other proteins that drive receptor activation, such as integrins, [26] plexin B1, [27] and CD44v6.[28] The relevance of any such interactions in the context of different EGFR genotypes and in response to MET inhibition has not been explored.
To the best of our knowledge ours is the first study to utilise FRET FLIM imaging to demonstrate EGFR-MET direct dimerisation.Importantly, we observed that the FRET between EGFR and MET differed according to EGFR genotype, providing a potential novel mechanism by which responses to MET inhibition can differ.In H1975 L858R/T790M cells, SGX523 reduced EGFR-MET FRET.Conversely FRET was increased by SGX523 in H1975 L858R cells but had no significant effect on FRET in H1975 WT cells.SGX523 is known to preferentially bind the less active, unphosphorylated form of MET.[23] Therefore, the differences observed suggest that EGFR L858R/T790M favours the phosphorylated form of MET; by contrast, EGFR L858R only dimerised with MET in the presence of SGX523, which suggests the opposite.No such modulation was seen in H1975 WT suggesting that the altered binding by MET in its active or inactive form was related to the presence of mutated and not WT-EGFR.Altered dimerisation between MET and mutant EGFR in EGFR L858R/T790M and H1975 L858R cells in the presence of SGX523 could be explained by SGX523-induced conformational changes; such conformational changes following treatment with small molecule kinase inhibitors are recognized and can be associated with unexpected responses to treatment.For example HER2-HER3 dimer formation occurs in breast cancer following treatment with Lapatinib.[29] Our studies cannot discriminate between the binding of MET with an EGFR monomer or homodimer.Dimer formation may activate one or other of the receptors or indeed both.In the case of higher order oligomers, MET may facilitate or impede EGFR homodimer formation or modulate the kinase activity of EGFR molecules within the homodimer.EGFR remained phosphorylated after treatment with SGX523 (Fig 5A).Whether MET is additionally capable of phosphorylating EGFR or modulating EGFR activity allosterically is not clear.
We observed clear phenotypic differences between the three cell lines.This may result from different tendency to activation or affinity for substrate.For example L858R and L858R/ T790M have increased ligand-activated signaling activity compared with WT-EGFR, particularly with respect to downstream effectors involved in cell survival and activation of transcription.[30] This is in agreement with published work showing that the EGFR mutants mediate significantly increased ligand-independent activation of the receptor.[31] This could explain that the presence of EGFR L858R mutant increases proliferation in the cells and tumors, and the novel finding that these cells also produce tumors with higher stromal markers.Alternatively, the EGFR-MET dimer offers a novel interpretation of the differences between the lines.In the H1975 L858R/T790M cells, where the EGFR-MET dimer is present at baseline, downstream signals may be driven by this heterodimer.The addition of SGX523 resulted in loss of EGFR-MET heterodimer and reduction in phosphorylated ERK in the H1975 L858R/T790M cells and derived tumours, in agreement with this hypothesis.In the cells expressing EGFR L858R there is no EGFR-MET dimer at baseline.The addition of SGX523 increases the dimerisation between the receptors, which resulted in the decrease of stromal markers such as % of collagen, SMA or cd31 staining, in the absence of inhibition in any of the typical signaling pathways analysed.However, as previously suggested, the interplay of EGFR with MET could provide a secondary set of effectors molecules through MET binding sites and furthermore through other receptors with which MET is reported to interact, such as HER3.[13] It is therefore likely that additional as yet unidentified pathway is active whilst the AKT/FAK pathway may not dominate in MET inhibition in this model.
What is the implication of our findings on the potential role of MET inhibition in lung cancer?Current emphasis is on the role of MET amplification as biomarker to select patients for MET inhibitor therapy in both treated or untreated NSCLC.[15,16,32,33,34].Our observations that the effects of MET inhibition are influenced by EGFR-MET dimerisation and in turn EGFR genotype, suggest novel mechanisms to understand how best to target MET in NSCLC and highlight the importance of understanding the interaction between multiple driver mutations when facing treatment resistance.Designing a FRET-FLIM assay that measures EGFR-MET heterodimerisation could potentially play a role in patient selection for MET targeted therapy.It is also interesting to speculate how our observations on the effects of SGX523, a MET kinase inhibitor, may influence treatment options in NSCLC.Whether it would be preferable in the future to target MET with a monoclonal antibody or through direct targeting of its kinase domain is unknown.A better understanding of crosstalk between therapeutic targets, which could be targeted in parallel, may also be of benefit to treatment efficacy and cost-effective use.
In summary, we have demonstrated using FRET that we can measure the interaction between EGFR and MET and we have made the important observation that response to MET inhibition correlates with EGFR-MET dimerisation, which is influenced by EGFR genotype.This could result from the different EGFR-MET dimers that arise following treatment with SGX523.The next important steps will be to determine how this information can be used to select effective therapies targeting MET.

Reagents and antibodies
Recombinant Human Epidermal Growth Factor (EGF) and Hepatocyte Growth Factor (HGF) were purchased from Peprotech.SGX523 was purchased from GE and Erlotinib from Cayman chemicals; both were solubilised in DMSO.All antibodies (Abs) were purchased from commercial sources as indicated.For tissue and cell staining, Phospho-Histone H3 was from Millipore, α-SMA from AnaSpec Inc., cd31 from Abcam, total cMET (D1C2) from Cell Signaling Technology and EGFR Ab from Novocastra/Leica.For immunoprecipitation, the EGFR Ab was from Santa Cruz Biotechnology (sc-120).For Western Blot, phospho-EGFR (Y1173), total EGFR phospho-MET (Y1234), total MET, phospho-ERK (Thr202/Tyr204), total ERK, phopho-AKT (Ser473), total AKT, phosphor-FAK (Tyr397) and total FAK were from Cell Signaling Technology; tubulin was from Millipore and hsc70 from Santa Cruz Biotechnology.For proliferation assays, anti-Brdu was purchased from Abcam.72˚C for 5 minutes, followed by 2 minutes in 4xSSC/Tween and two rounds of 2 minutes in PBS at room temperature.The coverslips were then mounted onto slides with DAPI counterstain and analysed using a fluorescence microscope.

Immunoprecipitation
Cells were lysed in lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 10% glycerol, 1% Triton X-100, 10 mM NaF, 1 mM Na 3 VO 4 , 10 mM N-ethylmaleimide, 0.01 μM Calyculin A) with Protease inhibitor cocktail set I (Roche).After centrifugation, the supernatants were incubated overnight at 4˚C with anti-EGFR or an irrelevant IgG, and subsequently for an additional hour with protein A/G-Agarose beads (Alpha Diagnostic International Inc.).After centrifugation, the immunoprecipitates were washed and subjected to SDS-PAGE and analysed by WB.

BrdU incorporation
Cell proliferation was determined by 5'-Bromo-Uridine (BrdU) based assay.Cells were plated for sub-confluence on coverslips in a 24-well plate and BrdU was added to the cells for 3 hours.Coverslips were then prepared for immunofluorescence staining as described below with an additional 15 minutes DNA denaturation step using 1.5M Hydrochloric Acid (HCl) and using a BrdU Ab.Hoechst was used to stain all the cell nuclei.Images were obtained LSM510 microscope (Carl Zeiss) and quantified in ImageJ using a macro to automate cell counting for red vs blue nuclei.

Soft-agar growth assay
Anchorage-independent growth was evaluated as described before.[35] Briefly, 1x10 4 cells were plated in complete DMEM containing 0.3% soft agar in 6-cm plates over a solidified DMEM plated in containing 0.7% soft agar layer.Medium was added twice a week to maintain humidity.After 3 weeks, colonies were stained with crystal violet (0.02%) for 1 hour and counted.

Wound healing experiments
Cells were seeded to confluence.A horizontal wound was made through the middle of the wells using a micropipette tip.Cells were allow to migrate (heal the wound) for different times (0, 2, 8, 20 and 28 hours).Images were taken and the wound closure was quantified using Image J software analysis.
In vivo tumorigenicity assay H1975 WT , H1975 L858R and H1975 L858R/T790M cell lines (3x10 6 ) were injected subcutaneously into the two posterior flanks of BALB/c nude mice (Charles River Laboratories).For each cell line, 16 female 5 weeks old mice were used.Mice were followed weekly and tumors allowed growing for 13 days after injection.Tumors were measured with a caliper in long and short axes and volume was determined in based of the equation 0.4xAxB^2 (A, the long axis and B, the short axis of the tumor).At day 14 th after injection mice were divided randomly in two groups (6 animals/group) and vehicle or SGX523 (drug treatment) was started.60 mg/kg of SGX523 (equilibrated suspensions in 0.5% Methocell A4M) was administered for 12 days, daily, by oral gavage.At day 25 th after injection mice were culled using CO 2 and tumours were removed aseptically with dissecting scissors and weighed.All animals were maintained under specific pathogen-free conditions and handled in accordance with the Institutional Committees on Animal Welfare of the UK Home Office (The Home Office Animals Scientific Procedures Act, 1986).All animal experiments were approved by the Ethical Review Process Committee at King's College London and carried out under license from the Home Office, UK.

Tissue samples, immunohistochemistry and histopathological analysis
Tissue samples obtained from xenografts were fixed with formalin and paraffin-embedded.Immunohistochemical analyses were performed using 3 μm sections.Briefly, slides were dewaxed and antigen retrieval was performed using 0.1M citrate pH 6.0 buffer at 120˚C for 10 min, followed by blocking in TBS-Tween 0.1% + 1% BSA + 1% FBS.Primary antibodies were added overnight at 4˚C.As secondary antibodies, peroxidase-conjugated (Envision+) anti-rabbit and anti-mouse Ig reagents from Dako were used for 1h.Non-immune (Santa Cruz Biotechnology) or pre-immune rabbit serum was used as negative controls.Reactions were developed using diaminobenzidine (DAB) as chromogenic substrate.Images from digitalized scans of the glass slide specimens were obtained at magnification ×20 (0.45 μm/pixel resolution) using a Hamamatsu Nanozoomer 2.0 HT.All quantifications were done using Image J.

FRET determination by FLIM measurements
Coverslips or FFPE tissue was prepared as described as above for immunofluorescence except for the added step of quenching of autofluorescence by 15 minutes immersion in NaBH 4 (1 mg/mL).FLIM was performed using time-correlated single-photon counting (TCSPC) with a multiphoton microscope system as described previously.[19] Analysis was performed in Tri2 for time resolved analysis.To analyze FRET in tissue xenografts samples, a dedicated algorithm was used that masks autofluorescent lifetime measurements to reveal true FRET.[20] Enzyme-linked immunosorbent assay (ELISA) Cells were plated in 6 well plates and incubated overnight.Supernatant was collected and EGF and HGF ligands were measured by sandwich enzyme-linked immunosorbent assay manufacturer's instructions.

Statistical analysis
Student t test for parametric data, with ANOVA comparison for more than two groups, determined statistical differences.Probability of p<0.05 was considered significant.Graphs were prepared in Excel or in Prism (GraphPad) software.

Fig 1 .
Fig 1.In vitro validation of the H1975 EGFR mutant cell lines.(A) Relative mutant allele frequency was compared in cDNA from each cell line by Digital droplet PCR.(B) WB of total EGFR levels in the H1975 L858R/T790M cell line before and after lentivirus infection with a shEGFR and in the H1975 cell lines.hsc70 levels are shown as loading control.Values beneath blots are relative levels of T-EGFR compared to the H1975 L858R/T790M cell line from 2 independent experiments (C) WB of phospho and total EGFR in H1975 derivative cell lines untreated or treated with EGF (100ng/mL) for 15 min, Erlotinib (1μM, 5μM or 10 μM) for 1 hour or both.hsc70 levels were used as loading control.Quantification of the WB is shown above the figure from 3 independent experiments.Error bars is SD (*p<0.05 and **p<0.001).doi:10.1371/journal.pone.0170798.g001

Fig 2 .
Fig 2. EGFR-MET interaction is modulated by EGFR mutations.(A) Western blot (WB) of total MET levels in H1975 derived cells.Tubulin levels are shown as loading control.Values beneath blots are relative levels of MET compared to the total levels in the H1975 L858R/T790M cell line from 2 independent experiments +/−SD (B) MET (7q31) (red signal) copy number analysis by FISH in the three H1975 cell lines using the Leica Kreatech C-MET (7q31)/SE7 FISH probe (KBI-10719).The green signal indicates the chromosome 7 centromere control probe.Scale bar 10 mm.Average copy number and ratio between MET and chromosome 7 centromere probe are also indicated (n = 30 cells).(C) Immunofluorescence of total EGFR (Alexa546 -red in the image) and MET (Cyanine 5 -green in the

Fig 3 .
Fig 3. Inhibition of MET changes the EGFR-MET interaction both in vitro and in vivo.(A) Fluorescence lifetime imaging performed in the three H1975 cell lines plated on coverslips with or without treatment with SGX523 (5 μM) for 24 hours.Representative pseudocolour lifetime images showing FRET efficiency accompanied by corresponding grayscale donor (EGFR-Alexa 546) and acceptor (MET-Cyanine 5) intensity images are shown.Scale bar: 50μm.(B) Bar graph showing quantification for average FRET efficiency in the three cell lines with or without SGX523 treatment (*** p < 0.0005, ** p = 0.001).(C) Quantification of average FRET of EGFR-MET interaction performed in xenograft tumors from each H1975 cell line in mice receiving mock or SGX523 treatment (60 mg/kg) for 12 days (*** p<0.001, * p<0.05).(D) Representative lifetime images for EGFR:MET FRET in xenograft tumors accompanied by corresponding grayscale donor (EGFR-Alexa 546) and acceptor (MET-Cyanine 5) intensity images.Scale bar, 50μm.doi:10.1371/journal.pone.0170798.g003

Fig 4 .
Fig 4. In vitro and in vivo effects of MET inhibition on cell proliferation.(A) Representative images of the three cell lines grown at 60% confluence on coverslips and treated or not with SGX523 (5 μM) for 24 hours.Scale bar, 20μm.The BrdU positive nuclei (red) show the cycling cells, and the Hoechst dye was used to stain all the nuclei of the cells (in blue).(B) Quantification of the proliferation rates as described in (A) (*p <0.04, **p<0.005,***p<0.0001).(C) Soft agar colony formation in the H1975 derivate cell lines.The graph shows the number of colonies after 3 weeks of growing in the presence or absence of SGX523 (5 μM) (*p <0.001).Representative images of untreated and SGX523 treated colonies are also shown.(D) Graphs showing the fold change in volumes in 16 xenografts tumors (8 coming from mice subjected to vehicle and 8 from SGX523 treatment) coming from the three H1975 cell lines during the 12 days of treatment.Tumor volumes were measured at the indicated times using a

Fig 5 .Fig 6 .
Fig 5.In vitro and in vivo effects of MET inhibition on stroma remodeling.Quantification and representative images of (A) collagen, (B) α-SMA and (C) cd31 staining of xenografts tumors (FFPE) grown from each H1975 derivate cell line coming from mice treated or not (mock) with SGX523 (60 mg/kg).Bar 250 nm.Quantification of the staining is shown above the images and was performed using Image J. (*p<0.05,**p<0.01,***p<0.001).EGF (D) and HGF (E) quantification assessed by ELISA in each H1975 derivate cell line supernatants.doi:10.1371/journal.pone.0170798.g005 hsc70 is also shown as loading control.(C) Quantification of the P-ERK vs T-ERK WBs shown in (A) and in (B) (*p<0.05).(D) Schematic model demonstrating effect of SGX523 on EGFR-MET dimerization pattern and consequences for tumor characteristics following treatment.doi:10.1371/journal.pone.0170798.g006