C3G downregulation induces the acquisition of a mesenchymal phenotype that enhances aggressiveness of glioblastoma cells

Glioblastoma (GBM) is the most aggressive tumor from the central nervous system (CNS). The current lack of efficient therapies makes essential to find new treatment strategies. C3G, a guanine nucleotide exchange factor for some Ras proteins, plays a dual role in cancer, but its function in GBM remains unknown. Database analyses revealed a reduced C3G mRNA expression in GBM patient samples. C3G protein levels were also decreased in a panel of human GBM cell lines as compared to astrocytes. Based on this, we characterized C3G function in GBM using in vitro and in vivo human GBM models. We report here that C3G downregulation promoted the acquisition of a more mesenchymal phenotype that enhanced the migratory and invasive capacity of GBM cells. This facilitates foci formation in anchorage-dependent and -independent growth assays and the generation of larger tumors in xenografts and chick chorioallantoic membrane (CAM) assays, but with a lower cell density, as proliferation was reduced. Mechanistically, C3G knock-down impairs EGFR signaling by reducing cell surface EGFR through recycling inhibition, while upregulating the activation of several other receptor tyrosine kinases (RTKs) that might promote invasion. In particular, FGF2, likely acting through FGFR1, promoted invasion of C3G-silenced GBM cells. Moreover, ERKs mediate this invasiveness, both in response to FGF2- and serum-induced chemoattraction. In conclusion, our data show the distinct dependency of GBM tumors on C3G for EGF/EGFR signaling versus other RTKs, suggesting that assessing C3G levels may discriminate GBM patient responders to different RTK inhibition protocols. Hence, patients with a low C3G expression might not respond to EGFR inhibitors.

C3G function in human cancer varies with tumor type and stage 4,16 . C3G prevents malignant transformation induced by several oncogenes 5,6,17 . Accordingly, C3G expression is reduced in cervical squamous cell carcinoma 18 . In contrast, C3G is upregulated in non-small-cell lung cancer 19 and, hepatocarcinoma (HCC), inducing tumor growth 20,21 . High expression of p87C3G isoform is also associated with chronic myeloid leukemia development 22 and C3G mutations correlate with lymphomas development 23 . In colorectal cancer (CRC), C3G plays a dual role. It promotes tumor growth, while inhibits migration/invasion 13 . C3G also reduces migration in highly invasive breast carcinoma cells 24 . Additionally, C3G is important in the tumor stroma. In particular, C3G promotes the release of pro-angiogenic and prometastatic factors from platelets, enhancing tumor growth and metastasis 25 .
Even though C3G expression is ubiquitous in humans, C3G levels are higher in brain than in other tissues 1,4,32,33 . C3G regulates important functions in the CNS, such as neural differentiation, neurite outgrowth, and survival 28,30,31,34 . C3G also controls migration and the size of different neuron populations during brain development 11,[35][36][37][38] . However, little is known about its role in brain tumors. In particular, its function in glioblastoma (GBM) remains unknown.
GBM is considered the most aggressive CNS tumor due to its malignant nature and the lack of effective treatments. CNS tumors are classified by World Health Organization according to the histological and molecular criteria 39,40 . GBMs grade IV astrocytomas, characterized by a high proliferation, low differentiation, high cell heterogeneity, and invasive capacity 39 , are subdivided into IDH-wt (90%) or mutated (10%) 40 . The most frequent genetic alterations in GBMs affect RTK, Rb, and p53 signaling pathways 41,42 . GBMs can also be subdivided into classical, neural, pro-neural, and mesenchymal subtypes 40,43,44 . However, GBM classification remains controversial due to the heterogeneity of tumors.
GBM aggressiveness hinders its treatment. Moreover, tumor relapse after treatment is very common in GBM patients 44 and there is no good second-line treatment available. Therefore, more studies are necessary to understand the mechanisms controlling this aggressive pathology in order to find more effective therapies. Given the relevance of C3G in brain development, its key function in adhesion and migration and its role in other cancer types, we hypothesized that C3G could play a relevant role in the tumorigenic and invasive properties of GBM cells. To explore this, we used different human GBM cell models for in vitro and in vivo studies and patient data.

Results
C3G is downregulated in glioblastoma, favoring the acquisition of a more invasive phenotype The function of C3G in GBM remains unknown. Hence, we first analyzed its expression using databases. A detailed analysis from TCGA-RNA-seq database showed that RAPGEF1 levels normalized with GUSB (Fig. 1A), ACTB or UBC (Supplementary Fig. 1A) were significantly downregulated in patient GBM tumors. Moreover, RAP-GEF1 downregulation was sex and age independent ( Supplementary Fig. 1B), suggesting a putative role for C3G in GBM onset and/or development. In addition, C3G protein levels were decreased in a panel of human GBM cell lines as compared to human astrocytes (Fig. 1B). We also found an inverse correlation between C3G levels and the mesenchymal marker, Vimentin (Fig. 1B), associated with higher aggressiveness in GBM cells [55][56][57] .
To characterize the role of C3G in GBM, a permanent C3G silencing was performed in a standard GBM U87 cell line using specific shRNAs. C3G protein expression was efficiently down-regulated (60-70%), while non-targeting control (NTC) shRNAs had no effect (Fig. 1C). U87shC3G cells displayed a phenotype with few cell-cell contacts (Figs. 1D and E) and more F-actin migratory structures (mainly, filopodia and blebs) (Fig. 1F). Such morphological changes were not observed in control shRNA cells (Fig. 1D). This was confirmed in 12Ф12D cells, a cell line derived from a patient 58,59 , where C3G knock-down ( Fig. 1G) induced similar morphological changes (Supplementary Fig. 2A) and an increase in stress fibers Fig. 1H.
Next, we evaluated the effect of C3G downregulation on GBM migration/invasion and adhesion. We found an enhanced invasiveness in U87shC3G cells as compared to parental or NTC U87cells ( Fig. 2A) using serum as chemoattractant. Similarly, C3G silencing increased invasion (Fig. 2B) and migration ( Supplementary Fig. 2B) of 12Ф12D cells. Moreover, MMP-2 activity was higher in U87shC3G than in non-silenced cells ( Supplementary Fig.  2C). Accordingly, adhesion of U87 and 12Ф12D cells decreased following C3G downregulation (Fig. 2C-E), which would facilitate migration.
To get further insights into the mechanisms used by C3G to regulate invasion in GBM cells, we analyzed EMT markers. C3G silencing increased Vimentin (mesenchymal marker) protein levels in both U87 and 12Ф12D cells and reduced E-cadherin (epithelial marker) levels in 12Ф12D cells (Fig. 2F). In addition, mRNA levels of EMTassociated transcription factors, TWIST1 and ZEB2, were up-regulated in U87shC3G cells (Fig. 2G).  1 C3G is downregulated in GBM promoting changes in cell morphology. A C3G (RAPGEF1) mRNA levels in GBM patients and normal brain (171 patients) obtained from TCGA database normalized with GUSB. B Western-blot analysis of C3G and Vimentin protein levels in the indicated GBM cell lines and human astrocytes (HAs) normalized to β-actin. Densitometric quantification of C3G/β-actin and Vimentin/β-actin ratios are shown. C, G Western-blot analysis of C3G normalized to β-actin to confirm silencing in U87 and 12Ф12D cells as compared to parental cells and/or cells with control shRNA (NTC). Densitometric quantification of C3G/β-actin ratio is shown. D Phase-contrast microscopy images of parental U87, U87shC3G and NTC cells maintained either in the presence (10% FBS) or absence (0% FBS) of serum for 24 h. Scale bars: 50 µm. E Representative images of β-catenin (red) and DAPI (blue) staining analyzed by confocal microscopy. Scale bars: 25 µm. F and H Left panels, immuno-fluorescence microscopy images of phalloidin staining (red) in parental and C3G-silenced U87 and 12Ф12D cells, maintained as indicated. Cell nuclei were stained with DAPI (blue). Scale bars: 50 µm. An amplification of cells inside the square is also shown. Right panels, histograms showing the quantification of the number of cells presenting migratory structures (filopodia, blebs, stress fibers and/or lamellipodia) expressed as fold increase.

C3G downregulation alters tumorigenic properties of glioblastoma cells
To determine the role played by C3G in GBM tumorigenic properties, we first performed in vitro functional analyses. Anchorage-dependent and -independent growth assays revealed an increased number of foci upon C3G silencing in U87 cells as compared to parental (Figs. 3A and 3B, respectively) or NTC-U87cells (Supplementary Fig.  3A and 3B). Similar results were obtained using 12Ф12D cells ( Supplementary Fig. 3C and Fig. 3C). Interestingly, foci formed by C3G-silenced cells contained fewer and more disseminated cells and cell-cell interactions were reduced, leading to a scattered phenotype ( Fig. 3A-C). This suggested that C3G-silenced cells might generate more foci by cell scattering, while being unable to repopulate them. In support of the latest, cell proliferation was decreased in U87shC3G cells, under adherent and nonadherent conditions, according to cell cycle and Ki67 staining analyses (Figs. 3D and 3E). However, no differences were found in apoptosis (data not shown).
To evaluate the in vivo function of C3G in GBM tumor growth, xenograft assays were performed. Tumors generated by U87shC3G cells showed a significantly increased size at the end point (Fig. 4A).

C3G facilitates activation of EGFR signaling favoring its membrane localization
Signaling by RTKs is broadly implicated in driving GBM onset and development. EGFR alterations and/or overexpression are frequent in GBM patient tumors 62 . Therefore, we analyzed if C3G regulated EGFR signaling in GBM cells. C3G silencing reduced ligand-induced EGFR phosphorylation in U87 cells, but not total EGFR levels ( Fig. 5A). In addition, p38MAPK, ERKs, and Akt phosphorylation levels decreased in response to EGF in C3G-silenced cells (Fig. 5A). Next, we studied the functional consequences of the defective EGF/EGFR signaling in GBM cells with C3G downregulation. We observed a significant decrease in EGF-induced invasion upon C3G silencing in the absence of chemoattractant (Fig. 5B).
C3G controls the exocytosis of angiogenic factors by platelets through interaction with VAMP7 25 , a protein involved in vesicle trafficking 63 . Therefore, we studied if the alterations in EGFR signaling and functionality could be a consequence of changes in receptor endocytosis and/ or recycling. We detected a reduction in activationinduced EGFR internalization in C3G knock-down cells (Fig. 5C). This suggested that C3G is involved in EGFmediated EGFR endocytosis and/or recycling. Using flow cytometry as a complementary approach, we found that EGFR levels on cell surface were higher in parental U87 than in U87shC3G cells, both in the presence and absence of serum (Fig. 5D). As a positive control of EGFR endocytosis, serum-deprived cells were treated with EGF for 2 h. In parental U87 cells, the presence of EGFR in the membrane was highly reduced upon EGF treatment ( Fig. 5E), according to the rapid internalization observed by microscopy (Fig. 5C). However, in U87shC3G cells the low levels of EGFR at cell surface remained unchanged in response to EGF (Fig. 5E). Moreover, inhibition of recycling by monensin 64 treatment significantly reduced EGFR levels at cell surface in non-silenced cells, while it had no effect on C3G-silenced cells (Fig. 5F). These findings indicate that C3G would be required for EGFR membrane localization in GBM cells, mainly by favoring receptor recycling.

C3G downregulation enhances the activation of several tyrosine kinase receptors in glioblastoma cells
Although invasion was enhanced in U87shC3G cells using serum as chemoattractant, EGF driven invasion was decreased. Therefore, we evaluated the phosphorylation of RTKs using a proteome profiler human phospho-RTK array, searching for potential mediators of the increased invasion induced by C3G downregulation in response to serum. This analysis revealed multiple changes induced by C3G knock-down in U87 cells ( Fig. 6A and Supplementary Fig. 5). Among them, we found a significant upregulation in the phosphorylation of FGFR1, erbB2, Eph family members (e.g., EphA6, EphB1, EphB2, and EphB4) and Tie1/2. Several of these RTKs have been associated with cell migration and invasiveness.
Taking into account the relevance of FGFR1 in GBM 65 , we evaluated the effect of its ligand, FGF2, on invasion. U87shC3G cells stimulated with FGF2 showed higher invasiveness than parental cells (Fig. 6B). Moreover, when serum was used as chemoattractant, invasion was prevented by infigratinib, a FGFR1/2/3 inhibitor (Fig. 6C). Altogether, these results indicate that C3G differentially regulates the localization and functionality of selected RTKs in GBM cells, decreasing EGFR activation, while increasing FGFR1 and other RTKs activation. This results in an overall enhanced invasiveness, in part mediated by FGFR1.
ERKs are involved in the pro-invasive effect of C3G downregulation in glioblastoma cells C3G regulates several intracellular signaling pathways involved in migration and invasion 12,13 . In particular, C3G downregulation in MEFs and HCT116 colon carcinoma cells enhanced ERKs activation 7,13 . In GBM cells, EGFinduced ERKs phosphorylation decreased upon C3G downregulation. However, in response to serum, C3G To assess if C3G effects were mediated by ERKs, we inhibited this pathway with PD98059 66 . PD98059 treatment, which decreased P-ERKs levels in parental and U87shC3G cells (Fig. 7C), significantly reduced seruminduced invasion of U87shC3G cells to the levels of parental cells (Fig. 7D) and migration of 12Ф12DshC3G cells ( Supplementary Fig. 6). The effect of C3G downregulation on anchorage-dependent (Fig. 7E) and -independent (Fig. 7F) growth was also prevented by ERKs inhibition. Moreover, cell proliferation in non-silenced U87 cells was reduced by PD98059 ( Fig. 7G and 7H), having no effect on C3G-silenced cells.

Discussion
C3G regulates relevant processes for tumor progression including proliferation, apoptosis, adhesion, invasion and angiogenesis 7,12,13 . However, C3G function in cancer appears to depend on the context 4,16 . In this study, we uncover a novel key function for C3G in GBM. Public databases revealed that C3G (RAPGEF1) mRNA levels are downregulated in GBM patient samples. C3G protein expression is also reduced in human GBM cell lines, suggesting that C3G levels could be downregulated during GBM onset and/or progression.
We demonstrate that C3G downregulation enhanced migration/invasion of GBM cells (U87 and 12Ф12D). These data are in agreement with the increased migratory properties of C3G deficient MEFs, C3G knock-down CRC and HCC cells 13,21 and with the inhibitory effect of C3G overexpression on the migration of breast carcinoma cells 24 . In CRC HCT116 cell line, C3G downregulation induces actin cytoskeleton reorganization and increases MMP2/9 activities through upregulating p38αMAPK activity 13 . In GBM cells with C3G downregulation, the enhanced migration/invasion might be a consequence of an EMT-like process that promotes the acquisition of a more mesenchymal phenotype, supported by increased Vimentin levels, MMP2 activity, and TWIST1 and ZEB2 mRNA levels 67 . Although the role of EMT in GBM is controversial due to brain plasticity, its non-epithelial characteristics, similarity with astrogliosis and the mesenchymal profile of most GBMs 46,68 , it is accepted that GBM cells can acquire a more mesenchymal phenotype [55][56][57] , associated with high Vimentin levels, invasiveness, and poor prognosis.
Previously, we reported that C3G downregulation in CRC and HCC cells reduced tumor growth and size 13,21 . However, C3G-silenced GBM cells form bigger tumors in xenograft and CAM assays, even though proliferation is decreased both in vitro and in vivo. This correlates with the higher number of foci with less cells/focus formed by C3G knock-down GBM cells in anchorage-dependent and -independent growth assays and the lower number of tumor cells in CAM-derived tumors. Therefore, this lower cell density, loss of cell-cell interactions, and higher mobility of cells with C3G downregulation lead to larger tumors. Moreover, tumors originated by C3G-silenced cells in CAM assays present higher levels of α-SMA and MECA32, pointing to enrichment in stroma and blood vessels, most likely by infiltration of host-niche cells. Such a scenario could contribute to increase the size of tumors with C3G downregulation. Although cellularity usually correlates with a higher aggressiveness and poor prognosis, the invasive capacity of GBM cells is also associated with treatment resistance, recurrence and poor overall survival 69 . Therefore, C3G downregulation in GBM cells would induce a higher aggressiveness.
Our data also unveil a novel function of C3G controlling the signaling elicited by RTKs, which represent the most commonly altered molecules in glioma 70 . C3G downregulation in GBM cells reduces the amount of EGFR at the cell surface. This reflects defects in receptor recycling and would explain the defective EGFR-mediated signaling and pro-invasive effect. Although the precise mechanisms involved need to be characterized in detail, the regulation of actin cytoskeleton organization by C3G might play a role in the proper recycling and membrane localization of EGFR 71 . C3G controls exocytosis of angiogenic factors in platelets through interaction with VAMP7 25 , which is involved in the secretory pathway that allows EGFR localization in membrane microdomains, also regulating EGFR endocytosis and signaling 72 . Moreover, other SNAREs collaborate with VAMP7 and other VAMPs in cellular vesicle trafficking, including that of EGFR to the membrane 73 . Therefore, C3G might favor EGFR membrane localization and recycling, acting through VAMP proteins. However, other mechanisms might contribute to the defective EGF/EGFR signaling, such as the reduced formation of EGFR-Crk-C3G complexes 28 .
Based on the increased phosphorylation of several, but not all, RTKs in C3G-silenced GBM cells, C3G seems to differentially regulate signals mediated by distinct RTKs, leading to an overall pro-invasive phenotype when C3G levels are low. For example, we observed a strong upregulation of FGFR1 phosphorylation and FGF2-mediated (see figure on previous page) Fig. 5 Effect of C3G knock-down on EGFR activity and membrane localization. U87 and U87shC3G cells were used. A Representative westernblot analysis of phosphorylated and total levels of EGFR, ERKs, p38MAPK, and Akt proteins normalized to β-actin. Densitometric quantification of these proteins versus β-actin expressed as the fold increase of the value of untreated cells from the corresponding genotype (non-silenced and C3G silenced) (n = 3). Untreated shC3G versus non-silenced cells ratios are: P-EGFR/β-actin = 1, P-Akt/β-actin = 2, P-ERKs/β-actin = 1.6, P-p38MAPK/ β-actin = 2.6. Serum-starved cells (for 16 h) were stimulated with EGF for 5-60 min or maintained untreated.  ERKs activation in GBM cells with C3G downregulation, leading to increased invasiveness. This is in agreement with FGFR1-induced expression of EMT-associated genes in GBM 65 . Moreover, the effect of C3G silencing resembles the pro-invasive action of Abl downregulation or inhibition, but not its anti-tumorigenic effect 74 . This could reflect the differential impact of Abl versus C3G on the activity or expression of individual RTKs. Our data also revealed a remarkable increase in the levels of phosphorylated EphB2 in GBM cells with C3G downregulation, which promotes migration and invasion of GBM and GBM cancer stem cells 75,76 . Therefore, EphB2 might contribute to favor invasion of C3G knockdown GBM cells. Similarly, the up-regulation of ErbB2 phosphorylation found in C3G-silenced cells might also facilitate migration 77 and could contribute to induce resistance to EGFR inhibition therapy 78 .
Tie2 and Tie1 phosphorylation is also upregulated in C3G-silenced GBM cells, which suggests that C3G downregulation might facilitate an endothelial transdifferentiation, a process occurring in GBM, associated to therapy resistance 79,80 .
Finally, the upregulation of ERKs activation in response to serum and FGF2 in C3G silenced cells might be responsible of enhancing invasiveness of GBM cells, as ERKs inhibition prevents this effect. Moreover, the upregulation of the activity of several RTKs, prominently FGFR1, detected in the array might be responsible for this high ERKs activity.
In summary, we uncover C3G as a novel key player in GBM biology and tumor progression (Fig. 8). C3G is down-regulated in GBM, promoting the acquisition of a more mesenchymal and invasive phenotype, giving rise to larger tumors with less proliferation, but more stromal cells and vessels. Furthermore, the defective EGFR signaling might contribute to resistance to anti-EGFR therapy in patients with low levels of C3G. In contrast, other RTKs and ERKs might represent alternative therapeutic targets in these patients based on the upregulation of their activities. Future studies would allow to further characterize C3G function in RTKs regulation, which may be also of value for designing novel and personalized therapeutic approaches.

Materials and methods
Analysis of RAPGEF1 expression in glioblastoma patients using public genomic databases C3G (RAPGEF1) mRNA-seq normalized expression in GBM patients was analyzed using 171 samples from TCGA (The Cancer Genome Atlas), grouped into control and GBM tumor. mRNA-seq and clinical data of TCGA GBM patients was downloaded from FIREBROWSE (http://firebrowse.org/).

Cell culture and treatments
The human glioblastoma U87MG (from ATCC) and the non-commercial GBM 12Ф12 cell line with a stem-like phenotype (mycoplasma free), derived from a patient GBM tumor 58,59 , were used. Cells were grown in DMEM supplemented with 10% fetal bovine serum (FBS). This induces the differentiation of 12Ф12 cells (12Φ12D cells).

Western blot analysis
Protein extracts and western blot analysis were carried out as described 13 . Membranes were probed with primary

RNA extraction and RT-qPCR analysis
Total RNA isolation and RT-qPCR analysis was performed as described 21 . Total RNA was isolated using NucleoSpin RNA kit (Macherey-Nagel# 740955.50) and reverse transcribed using SuperScript III RT kit (Invitrogen). cDNA was amplified using specific primers for TWIST1, ZEB2 and GUSB to normalize and detected by SYBR Green (Roche 04913850001) using 7900 Fast Real Time System (Life Technologies 4329001). Ct (threshold cycle) for a gene minus Ct for GUSB = ΔCt and then, referred to non-silenced control values (sample ΔCt-nonsilenced ΔCt = ΔΔCt) to calculate RQ (2 −ΔΔCt ).

Analysis of F-actin organization
Cells seeded on 2% gelatin-coated glass coverslips were fixed with 4% paraformaldehyde (PFA) for 20 min. F-actin was stained with rhodamine-conjugated phalloidin (Sigma Aldrich-P1951) as described 22 . Samples were visualized in a Leica TCS-SL confocal microscope.

EGFR endocytosis analysis
Cells seeded on 2% gelatin-coated glass-bottomed cell culture dishes were incubated for 30 min in phenol-redfree DMEM medium supplemented with 0.1% BSA. Cell dishes were placed into a LSM510 confocal microscope equipped with a thermostated chamber. Cells were stimulated on-stage with Alexa-Fluor488-labeled EGF (50 ng/ml; Invitrogen-E13345) and imaged alive at different time points.

Adhesion assay
Cells were seeded on either Matrigel (5μg/cm 2 )-coated or uncoated plates in a medium containing 10% FBS. Adhered cells at 15 min were fixed, stained with crystal violet and counted using Eclipse TE300 Nikon microscope.

Invasion assay
Invasion was assayed in Matrigel (333 µg/cm 2 ; Corn-ing#356234) coated transwells (BD#353097). 25000 cells were seeded in the upper chamber in serum-free medium. 10% FBS-medium, placed in the lower chamber, acts as chemoattractant. After 24 h, cells from the lower chamber were fixed with 4% PFA, stained with crystal violet and counted using an Eclipse TE300 Nikon microscope. To evaluate the effect of EGF, FGF-2 or PD98059 on invasion, cells in the upper chamber were treated with these compounds and no serum was added into the lower chamber.

Cell cycle analysis
Cells in the medium and trypsinized cells were centrifuged at 2500 rpm 5 min at 4°C, fixed with cold ethanol (70%) and washed with PBS. Cells resuspended in PBS were incubated 30 min with RNase (0.1 µg/ml) at 37°C. Propidium iodide (0.05 µg/µl) was added and cell cycle analyzed by flow cytometry. Cells in suspension (6 h) were directly pelleted and processed with the same protocol.

Analysis of cell surface EGFR and its recycling
Cells were detached using Ca 2+ /Mg 2+ -free PBS, resuspended in PBS and incubated with EGFR affibody conjugated with FITC (Abcam#ab81872; 1:200) for 30 min. FITC intensity was measured in an Accuri BD FACS Flow Cytometer. To evaluate recycling, a monensin 64 (10 μM) 1 h pretreatment was performed. Fluorescence intensity was measured using ImageJ Software and was expressed as integrated density (ID): intensity mean value of the image multiplied by the positive area percentage.

Wound healing and anchorage-dependent and -independent growth assays
Wound healing and anchorage-dependent and -independent growth assays were performed as previously described 21,22 . For anchorage-dependent growth, cells (300/60 mm plates) were seeded in a medium supplemented with 10%FBS. To measure anchorageindependent growth, cells (3000/24 multiwell plates) resuspended in 0.35% agar (Sigma Aldrich #A9414) were poured onto a layer of 0.5% agar (in complete medium). Medium was renewed every 3 days. After 14 days, colonies were stained with 0.005% crystal violet.

RTK phosphorylation analysis
To evaluate phosphorylation of several RTKs, serumstarved cells (60-70% confluence) treated with 10% FBS for 4 h were washed with PBS and lysed using lysis buffer 17 from kit, supplemented with aprotinin (10 µg/ml) and leupeptin (10 µg/ml) and analyzed with human phospho-RTK array Kit (R&D#ARY001B) following the manufacturer's protocol. Once membranes were blocked with buffer 2, they were incubated overnight at 4°C with lysates (300 µg proteins) diluted in array buffer 1, washed and incubated for 2 h at RT with anti-phospho-tyrosine-HRP antibody. Then, after washing, membranes were incubated with the kit Chemi Reagent and visualized with 'VWR Imager Chemi Premium' documentation system.

Statistical analysis
Data are represented as the mean values ± S.E.M (n ≥ 3) of independent experiments. Unpaired Student's t-test was used for comparison of two experimental groups and one-way or two-way ANOVA analyses to compare more than two groups with one or two variables using Graph-Pad Prism 7.0 software. Differences were considered significant when p value was p ≤ 0.05.

Ethics statement
In this study, no patient samples have been used. Therefore, it does not apply. All animal experiments were carried out in compliance with the European Community Council Directive (2010/63/EU), following guidelines for animal research approved by Comunidad Madrid (Spain) with reference PROEX028/17.