Increased TGF-α as a Mechanism of Acquired Resistance to the Anti-EGFR Inhibitor Cetuximab through EGFR–MET Interaction and Activation of MET Signaling in Colon Cancer Cells Free

Colorectal cancer is one of the most frequently diagnosed malignant diseases in Europe and one of the leading causes of cancer-related death worldwide (1). Recent therapeutic strategies for metastatic colorectal cancer (mCRC) have been focused on developing molecularly targeted therapies. The EGFR is expressed in 60% to 80% colorectal cancer and plays a key role in the development and progression of human cancers and for this reason it has been proposed as a target for anticancer therapies (2, 3). Cetuximab, an anti-EGF receptor (EGFR) blocking monoclonal antibody (mAb), is an effective treatment as single agent and in combination with standard chemotherapy regimens for patients with KRAS wild-type mCRC (4).

Cetuximab treatment is effective in a subgroup of patients with mCRC because resistance to anti-EGFR therapies could be due to the constitutive activation of signaling pathways acting downstream of the EGFR. Several retrospective clinical studies have demonstrated that patients with mCRC with tumors harboring a KRAS mutation have no clinical benefit following cetuximab treatment, whereas, in patients with wild-type KRAS tumors, a significant clinical benefit in terms of progression-free survival and overall survival has been observed (5, 6). The evidence that KRAS mutations were associated with the lack of response to cetuximab or panitumumab, another anti-EGFR blocking mAb, in patients with mCRC has led the U.S. Food and Drug Administration and the European Medicines Agency to restrict the use of anti-EGFR mAbs to patients with KRAS wild-type tumors. In addition to KRAS gene mutations, several retrospective studies in patients with chemorefractory mCRC have provided evidence that primary resistance to cetuximab could be correlated with mutations in other intracellular downstream effectors of EGFR activation, such as BRAF, NRAS, and PIK3CA exon 20 genes (7, 8). However, even among the molecularly enriched subset of patients with colorectal cancer with KRAS, BRAF, NRAS, and PIK3CA exon 20 wild-type genes, cetuximab is not uniformly clinically effective, suggesting that there are other undefined mechanisms of cetuximab resistance (7–10). Therefore, both intrinsic and acquired resistance mechanisms significantly limit the efficacy of anti-EGFR mAbs in the medical management of patients with mCRC. In this scenario, HER2 gene amplification has recently been identified as a potential mechanism of resistance to cetuximab in a subset of patients with mCRC with wild-type KRAS/NRAS/BRAF/PIK3CA genes (11–13). A proof-of-concept, a multiarm preclinical study in mice bearing human coloreactal cancer tumors with HER2 amplification (“xenopatients”), revealed that the combined inhibition of HER2 and EGFR induced long-lasting tumor regression (12). Other mechanisms of resistance have recently been reported, such as MET and KRAS amplification (14, 15). Therefore, an understanding of these mechanisms is necessary to design effective therapies for patients to prevent or to overcome clinical resistance to cetuximab treatment.

MET is a cell membrane tyrosine kinase receptor for the hepatocyte growth factor (HGF). Deregulation and consequent aberrant MET signaling may occur by different mechanisms and it has been reported that MET is overexpressed in a variety of cancers including colorectal cancer (16, 17). High expression of MET in colorectal is associated with development of distant metastases (18). Increased and deregulated MET signaling has been identified as one of the key pathways to bypass growth inhibition caused by drugs targeting the EGFR in lung adenocarcinoma (19, 20). Several clinical studies have evaluated the combination of MET and EGFR inhibitors in patients with non–small cell lung cancer (NSCLC; refs. 21–23).

Furthermore, a functional link between EGFR and MET has been suggested (24). In this respect, expression of EGFR and MET correlate in multiple malignancies such as chordoma, prostate and ovarian carcinomas, and gastrinoma (25). EGFR has been implicated in HGF-induced hepatocyte proliferation and is required for MET-mediated colon cancer cell invasiveness (26). Despite these studies, the underlying mechanisms of EGFR-induced MET phosphorylation are not yet well understood.

In the present study, we have generated and characterized two human colon cancer cell models of acquired resistance to cetuximab to elucidate the molecular mechanisms of resistance. We have found that in cetuximab-resistant colorectal cancer cells, cell proliferation, and survival pathways are activated by MET. Interestingly, enhanced expression of the selective EGFR ligand TGF-α in cetuximab-resistant colorectal cancer cells is responsible for EGFR–MET interaction and subsequent EGFR-induced MET phosphorylation and activation. Finally, treatment of these cells with a selective MET inhibitor restores cetuximab sensitivity, suggesting that the combined inhibition of both EGFR and MET receptor tyrosine kinases (RTK) could represent a rational therapeutic strategy for preventing and/or overcoming cetuximab resistance in patients with mCRC.

Cetuximab, an anti-EGFR human–mouse chimeric mAb was kindly provided by Merck Serono. PHA665752, a selective MET tyrosine kinase inhibitor (TKI), was purchased from Santa Cruz Biotechnology. PHA665752 was dissolved in sterile dimethyl sulfoxide and a 10 mmol/L working solution was prepared and stored in aliquots at −20°C. Working concentrations were diluted in culture medium just before each experiment.

Four- to six-week-old female balb/c athymic (nu+/nu+) mice were purchased from Charles River Laboratories. The research protocol was approved and mice were maintained in accordance with the institutional guidelines of the Second University of Naples Animal Care and Use Committee (Naples, Italy). Mice were acclimatized at the Second University of Naples Medical School Animal Facility for 1 week before being injected with cancer cells. Mice were injected subcutaneously with 2.5 × 10⁶ GEO cells that had been resuspended in 200 μL of Matrigel (BD Biosciences). When established tumors of approximately 200 to 300 mm³ in diameter were detected, mice were treated continuously by intraperitoneal (i.p.) injection with cetuximab (1 mg twice weekly) for the indicated time periods. Each treatment group consisted of 8 mice. Tumor size was evaluated twice per week by calliper measurements using the following formula: π/6 × larger diameter × (smaller diameter)². The animals were sacrificed when the tumor diameter exceeded 1,500 mm³. Tumor from the treated group, were removed, digested, and suspended as a single cell, which were propagated in in vitro culture. For a period of 6 months SW48 cells were continuously exposed to increasing concentrations of cetuximab. The starting dose was the dose causing the inhibition of 50% of cancer cell growth (IC₅₀). The drug dose was progressively increased to 1 μg/mL in approximately 2 months, to 5 μg/mL after other 2 months, and finally, to 10 μg/mL after additional 2 months. The established cetuximab-resistant SW48 cancer cell line (SW48-CR) was then maintained in continuous culture with this maximally achieved dose of cetuximab that allowed cellular proliferation.

The human GEO colon cancer cell was a gift of Dr. N. Normanno (National Cancer Institute, Naples, Italy). The human SW48 colon cancer cell line was obtained from the American Type Culture Collection. Four GEO cetuximab-resistant clones were established as in vitro cell lines after cancer cell recovery and enzymatic treatment (27) from in vivo GEO tumor xenografts in mice, as previously described (27). GEO and GEO cetuximab-resistant clones were grown in Dulbecco's Modified Eagle Medium (Lonza) supplemented with 20% FBS (Lonza), 1% penicillin/streptomycin (Lonza) and maintained in a humidified atmosphere of 95% air and 5% carbon dioxide (CO₂) at 37°C. SW48 and SW48-CR cells were grown in RPMI-1640 (Lonza) supplemented with 10% FBS, 1% penicillin/streptomycin. All cell lines were routinely screened for the presence of mycoplasma (Mycoplasma Detection Kit, Roche Diagnostics).

Cancer cells were seeded in 24-well plates and were treated with different concentrations of PHA665752 (range, 0.01–10 nmol/L), cetuximab (range, 0.01–20 μg/mL) alone or in combination for 96 hours. Cell proliferation was measured with MTT. The IC₅₀ was determined by interpolation from the dose-response curves. Results represent the median of three separate experiments, each performed in quadruplicate.

GEO, GEO-CR, and SW48-CR cells were seeded in 6-well plates, treated for 72 hours, and stained with Annexin V–fluorescein isothiocynate (FITC). Apoptotic cell death was assessed by counting the numbers of cells that stained positive for Annexin V–FITC and negative for propidium iodide using the Apoptosis Annexin V-FITC Kit (Invitrogen), coupled with fluorescence-activated cell sorting analysis.

GEO, SW48, GEO-CR, and SW48-CR cells were seeded into 100 mm³ dishes and treated for 24 hours. Protein lysates containing comparable amounts of proteins, estimated by a modified Bradford assay (Bio-Rad), were subjected to immunoprecipitation or direct Western blot analysis. Immunocomplexes were detected with the enhanced Chemiluminescence Kit (Pierce Biotechnology Inc.). Anti-EGFR, phospho-EGFR (Tyr1068), HER2, HER3, HER4, p44/42 MAPK, phospho-p44/42MAPK, AKT, phospho-AKT (Ser473), MET, phospho-MET (Tyr 1234/1235), STAT3, phospho-STAT3, FAK, and phospho-FAK were purchased from Cell Signaling Technology. Monoclonal anti-α-tubulin antibody was purchased from Sigma Chemical Co. Secondary antibodies coupled to horseradish peroxidase were from GE Healthcare. Each experiment was done in triplicate. Two milligrams of protein lysates were immunoprecipitated with the required antibodies; immunocomplexes were recovered with protein G Sepharose (Roche Diagnostics) and detected by Western blotting.

Agilent (Agilent Technologies) microarray analyses were performed to assess baseline gene expression profile for GEO and GEO-CR cells using a one color labeling microarray system as previously described (28). The absolute amount and purity (A260/280 nm ratio) of total RNA samples were determined by spectrophotometry (Nanodrop, Thermofisher) and the size distribution was assessed by Agilent Bioanalyzer. Eight hundred nanograms of total RNA were converted into labeled cRNA with nucleotides coupled to a fluorescent dye (either Cy3 or Cy5) following the manufacturer's protocol (Quick Amp Kit, Agilent). Yield and purity (A260/280 nm ratio) of cRNAs were determined by spectrophotometry (Nanodrop, Thermofisher). Of note, 825 ng of cRNA labeled from GEO colon cancer cell lines were hybridized to Agilent Human Whole Genome 4 × 44 k Microarrays. Data were extracted from slide image using Agilent Feature Extraction software (v.10.5). The raw data and associated sample information were loaded and processed by Gene Spring 11.5X (Agilent Technologies). For identification of genes significantly altered in resistant cells, total detected entities were filtered by signal intensity value (upper cutoff 100th and lower cutoff 20th percentile) and flag to remove very low signal entities. Experiments were carried out in triplicate and data were analyzed using Student t test (P < 0.05) with a Benjamani–Hochberg multiple test correction to minimize selection of false positives. Of the significantly differentially expressed RNA, only those with greater than 2-fold increase or 2-fold decrease as compared with the controls were used for further analysis. Functional and network analyses of statistically significant gene expression changes were performed using Ingenuity Pathways Analysis (IPA) 8.0 (Ingenuity Systems; http://www.ingenuity.com). Analysis considered all genes from the data set that met the 2-fold (P < 0.05) change cutoff and that were associated with biologic functions in the Ingenuity Pathways Knowledge Base. The significance of the association between the data set and the canonical pathway was measured in two ways: (i) ratio of the number of genes from the data set that map to the pathway divided by the total number of genes that map to the canonical pathway is displayed; (ii) Fisher exact test was used to calculate a P value determining the probability that the association between the genes in the data set and the canonical pathway is explained by chance alone.

The small inhibitor duplex RNAs (siRNA) (ON-target plus SMARTpool) siMET (human: #L-003156-00), siTGF-α (human: #L019737), and siHBEGF (human: #L019624) were from Dharmacon. The siCONTROL Non-targeting Pool (#D-001206-13-05) was used as a negative (scrambled) control. Cells were transfected with 100 nmol/L siRNAs using Dharmafect reagent following the manufacturer's instructions. The day before transfection, the cells were plated in 35 mm dishes at 40% of confluence in medium supplemented with 5% FBS without antibiotics. Cells were harvested 48 hours after transfection. Western blot analysis for MET, and PCR for TGF-α and HBEGF expression was done. RNA extraction was performed by the RNeasy Kit (Qiagen) following the manufacturer's instructions. The RNA was quantified by Nanodrop (Thermo Scientific) and RNA integrity was analyzed by the 2100 Bioanalyzer (Agilent Technologies). For semiquantitative PCR reactions, random-primed first-strand cDNA was synthesized in a 50 μL reaction volume starting from 2 μg RNA by using the GeneAmp RNA PCR Core Kit (Applied Biosystems). PCR amplification was performed using the GeneAmp RNA PCR Core Kit system following the manufacturer's instructions. To exclude DNA contamination, each PCR reaction was also performed on untranscribed RNA. We used the amplimers for TGF-α, MET, and HBEGF and levels of β-actin transcripts were used for normalization. Sequence of amplimers used is available on request. The siRNA effects on cell signaling were evaluated by Western blot analysis. siRNA effects on cell proliferation and apoptosis were evaluated by MTT and apoptotic assay as previously described. Briefly, cells were seeded into 24-multiwell cluster dishes and transfected with MET, TGF-α, and HBEGF siRNA. Of note, 96 hours after transfection, cells treated with MET, TGF-α, and HBEGF siRNA received 5 μg/mL of cetuximab and cell proliferation and apoptosis were determined 24 hours later.

The concentrations of HGF, TGF-α, HBEGF in the conditioned medium were analyzed by using Luminex technology (29) analysis with multiplex beads suspension array plates according to the manufacturer's instructions (Invitrogen). Assays were done in triplicate. Results were normalized for the number of producing cells and reported as pg of ligands per 10⁶ cells per 72 hours.

Cell migration was assessed using a commercially available chemotaxis assay. Briefly, cells were incubated in serum-free medium for 24 hours and were left untreated or treated with the indicated doses of cetuximab, PHA665752, or their combination following which they were detached from flasks, suspended in quenching medium (serum-free medium containing 5% bovine serum albumin) and EDTA, and seeded into Boyden migration chamber inserts placed in a 24-well plate. The inserts contain a microporous membrane with an 8 mm pore size. Inserts were placed over wells containing serum-free media plus chemo-attractant (10% FBS). After a 48-hour treatment period, cells/media were discarded from the top side of the migration chamber insert and the chamber was placed in the wells of a new 24-well plate containing cell detachment solution. Following incubation for 30 minutes at 37°C, the insert was discarded, and a solution of lysis buffer and CyQuant GR dye was added to each well. CyQuant is a green fluorescent dye that exhibits strong enhancement of fluorescence when bound to cellular nucleic acids released by the lysis buffer, enabling assessment of the relative number of migrated cells. Fluorescence was determined with a fluorimeter at 480/520 nm. Assays were performed in triplicate.

GEO, SW48, GEO-CR, and SW48-CR cells were seeded into 100 mm³ dishes for 72 hours. DNA extraction was performed by the QIAamp DNA Mini Kit (Qiagen) following the manufacturer's instructions. The DNA was quantified by Nanodrop (Thermo Scientific). Ten nanograms of GEO, SW48, GEO-CR, and SW48-CR genomic DNA was analyzed by next generation sequencing using the Ion AmpliSeq colon and lung panel as previously described (30).

The statistical analyses of in vitro and in vivo data were carried out using Prism version 4.02 (GraphPad Software, Inc). The Student t test was used to evaluate the statistical significance of the results. All P values represent two-sided tests of statistical significance with P value <0.05.

To study the mechanisms of cetuximab-induced resistance, we generated cetuximab-resistant cells starting from cetuximab-sensitive GEO colon cancer cells, which were grown as tumor xenografts in vivo in immunodeficient mice that were exposed to continuous treatment with an optimal therapeutic dose of the drug (Fig. 1A). Continuous treatment caused tumor growth suppression within the first weeks of treatment. However, after approximately 12 weeks, in most mice GEO tumors resumed growth despite cetuximab treatment, reaching a growth rate comparable to untreated control GEO tumors within 18 to 20 weeks (Fig. 1A). Four cetuximab-resistant tumors were surgically removed and homogenized into single-cell suspensions used to generate in vitro GEO-CR cell lines, which displayed resistance to the growth inhibitory effects of cetuximab treatment (Fig. 1B). Only one of these cell lines (named GEO-CR cells) was further characterized and used for the subsequent experiments. As illustrated in Fig. 1B and C, GEO-CR cells were not sensitive cetuximab treatment at doses up to 20 μg/mL, as compared with parental GEO cells.

The investigation of EGFR-dependent intracellular pathways revealed that treatment of GEO cells with cetuximab completely blocked EGFR phosphorylation and consequently the activation of downstream mitogen-activated protein kinase (MAPK) and AKT signals (Fig. 1D). Although cetuximab treatment markedly reduced EGFR phosphorylation in GEO-CR cells, the downstream MAPK and AKT pathways were not completely blocked (Fig. 1D).

To investigate the potential molecular pathways involved in cetuximab resistance, mRNAs from GEO and GEO-CR cells were extracted and assessed for global gene expression changes by microarray analysis. In GEO-CR cells, 39 and 54 genes were identified whose expression was upregulated or downregulated, respectively, as compared with GEO cells (P < 0.05). Among the 39 upregulated genes in GEO-CR cells, we identified 17 genes involved in the HGF–MET-dependent pathways (Supplementary Table S1 and Fig. 2A) including MET (31–34). Moreover, the mRNAs of two EGFR selective ligands were upregulated in GEO-CR cells as compared were GEO cells: TGF-α and heparin-binding EGF–like growth factor (HB-EGF). Among the 54 downregulated genes in GEO-CR cells, 7 genes were involved in DNA repair and 4 genes were involved in cell-cycle regulation (Supplementary Table S2A and S2B). By Western blot analysis, we found a moderate increase in total MET protein expression, which was not accompanied by a significant increase in MET mRNA in GEO-CR cells as compared with parental GEO cells. However, higher levels of activated phospho-MET protein were depicted (Fig. 2B).

To determine whether MET activation could be involved in the acquisition of cetuximab resistance in GEO-CR cells, we investigated whether reduction of MET expression could restore cetuximab sensitivity. Transfection with a specific MET siRNA for 48 hours significantly reduced MET protein expression in GEO-CR cells, as shown in Fig. 3A. As illustrated in Fig. 3B and C, MET siRNA treatment slightly reduced cell growth with a little increase in apoptosis in GEO-CR cells. Although single-agent cetuximab treatment did not affect GEO-CR proliferation, cetuximab treatment in combination with MET silencing determined a statistical significant cell growth inhibition and proapoptotic effects in GEO-CR cells. MET silencing also restored the ability of cetuximab to inhibit MAPK and AKT activation in GEO-CR cells, as shown by downregulation of both phospho-MAPK and phospho-AKT levels (about 50% of inhibition, respectively; Fig. 3D).

To further evaluate whether MET could confer resistance in a different model of cetuximab-sensitive colorectal cancer cells, we used the SW48 colorectal cancer cell line, harboring KRAS, NRAS, BRAF, and PIK3CA wild-type genes, known to be cetuximab sensitive. Following continuous exposure to increasing doses of cetuximab for up to 6 months of SW48 cells, SW48-CR cells were obtained. As depicted in Supplementary Fig. S1A, SW48-CR cells were not sensitive to the growth inhibitory effect of cetuximab treatment at doses up to 20 μg/mL. The EGFR-dependent intracellular signaling pathways were not inhibited by cetuximab treatment in SW48-CR cells as compared with parental SW48 cells (Supplementary Fig. S1B). It has recently suggested that a mechanism of acquired resistance to cetuximab continuous treatment in human colorectal cancer cell lines that are initially sensitive, is the appearance of cancer clones with a mutated RAS gene (9, 10). To test this hypothesis, we have evaluated the presence of mutations in both the KRAS and the NRAS genes in GEO-CR and SW48-CR cells as compared with parental GEO and SW48 cells by next generation sequencing using the Ion AmpliSeq colon and lung cancer panel that evaluates the presence of mutations in 22 genes including exon 2, 3, and 4 KRAS and exon 2 and 3 NRAS (30). No additional mutations were observed in KRAS and NRAS genes in GEO-CR and SW48-CR cells (data not shown). Moreover, increased MET protein phosphorylation was observed in SW48-CR but not in parental SW48 cells, (Supplementary Fig. S1B). Similarly to GEO-CR cells, the inhibition of MET expression by siRNA treatment in combination with cetuximab determined a statistical significant cell growth inhibition and proapoptotic effects in SW48-CR (Supplementary Fig. S2A–S2C). Moreover, MET protein downregulation by siRNA treatment also restored the ability of cetuximab to inhibit MAPK and AKT activation in SW48-CR cells (Supplementary Fig. S2D).

We next evaluated whether MET activation could confer resistance in two cetuximab-sensitive colorectal cancer cell lines (GEO and SW48). These colorectal cancer cell lines were extremely sensitive to cetuximab cell growth inhibition with IC₅₀ of approximately 0.1 and 0.2 μg/mL, respectively (Fig. 3E). Treatment with the MET ligand HGF, significantly reduced cetuximab-induced cell growth inhibition in both cancer cell lines (Fig. 3F). In addition, despite treatment with cetuximab, HGF treatment caused MET phosphorylation and partial reactivation of MAPK and AKT in both GEO (Fig. 3G) and SW48 cells (Fig. 3H) compared with cetuximab treatment alone.

To further evaluate the role of MET activation in cetuximab resistance, treatment of GEO-CR and SW48-CR cells with PHA665752, a selective MET TKI, was performed. PHA665752 treatment of GEO-CR and SW48-CR cells induced a dose-dependent inhibition of cell growth with an IC₅₀ of approximately 0.5 and 5 nmol/L, respectively (Fig. 4A and C). Moreover, altough cetuximab treatment had no effect on cell growth in GEO-CR and SW48-CR cells, the combined treatment with PHA665752 restored the sensitivity of GEO-CR and SW48-CR cells to cetuximab in a dose-dependent manner (Fig. 4B and D). In particular, in GEO-CR cells the IC₅₀s for treatment with cetuximab in combination with PHA665752 (0.01 or 0.05 nmol/L) were 0.1 and 0.01 μg/mL, respectively (Fig. 4B). We next assessed by Western blot analysis the effects of treatment of GEO-CR and SW48-CR cells with cetuximab and/or the selective MET inhibitor on the expression and activation of key proteins, which act downstream to EGFR and MET. As shown in Fig. 4E and F, single-agent cetuximab or single-agent PHA665752 treatment were able to completely inhibit EGFR and MET phosphorylation, respectively, while inducing a modest decrease of MAPK and AKT phosphorylation. However, the combined treatment with both drugs was able to fully abrogate phosphorylation of both MAPK and AKT in GEO-CR and SW48-CR cells. (Fig. 4E and F).

GEO and SW48 cells are differentiated colon carcinoma cells with little or no ability to migrate or to invade in vitro (35). GEO-CR and SW48-CR cells acquired the ability to migrate in an in vitro migration assay (Fig. 4G and H). To investigate whether the migration properties of GEO-CR and SW48-CR cells could be due to MET activation, we performed a migration assay on GEO-CR and SW48-CR cells in the presence of cetuximab, PHA665752 or their combination. Although single-agent cetuximab treatment had no effect on GEO-CR and SW48-CR cancer cell migration, PHA665752 significantly reduced the number of cancer cells that were able to invade through the migration chambers. Moreover, this effect was markedly potentiated by the combined inhibition of EGFR and MET (Fig. 4G and H).

Once demonstrated the MET role in the acquired cetuximab resistance of GEO-CR and SW48-CR cells, we further evaluated the potential mechanism(s) responsible for MET activation. One major mechanism of cell membrane growth factor receptor activation in cancer cells is the specific ligand stimulation. In this respect, HGF-driven autocrine or paracrine activation of MET has been observed in several cancers (16). However, little or no HGF secretion was detected in GEO, SW48, GEO-CR, and SW48-CR cells, suggesting that autocrine MET activation was not occurring in these cancer cell models (Fig. 5A and E). Moreover, it has been shown that several members of the EGFR family can transactivate MET by inducing the formation of heterodimers even in absence of HGF (36). Therefore, a Western blot analysis was performed to evaluate the expression of the four members of the EGFR family. Although EGFR was expressed in GEO, SW48, GEO-CR, and SW48-CR cells, no expression of ERBB2, ERBB3, and ERBB4 was found in these cancer cell lines (data not shown). According to these results, we hypothesized that MET activation could be caused by interaction with the EGFR in GEO-CR and SW48-CR cells. By microarray gene expression analysis, the two specific EGFR ligands TGF-α and HB-EGF, were upregulated in GEO-CR cells as compared with GEO cells (Fig. 2A and Supplementary Table S1). ELISA assays confirmed that both growth factors were significantly secreted in the conditioned medium derived from GEO-CR cells as compared with parental GEO cells (Fig. 5A). Furthermore TGF-α levels were also significantly higher in the conditioned medium derived from SW48-CR cells (Fig. 5E). To examine whether TGF-α and/or HB-EGF enhanced expression could be causally involved in cetuximab resistance, GEO-CR cells were transfected with specific siRNAs for TGF-α or HB-EGF (Fig. 5B and C). TGF-α silencing, but not HB-EGF, was functionally relevant because it sensitized GEO-CR cells to the antiproliferative effects of cetuximab (Fig. 5B). Similarly, TGF-α downregulation by siRNA sensitized SW48-CR cells to cetuximab (Fig. 5F and G). Furthermore, inhibiting TGF-α expression by siRNA together with cetuximab treatment of GEO-CR and SW48-CR cells caused a significant suppression of MET phosphorylation as well as of downstream effectors MAPK, AKT, STAT3, and FAK (Fig. 5D and H).

Taken together, these results suggest a role for TGF-α overexpression in acquired cetuximab resistance in GEO-CR and SW48-CR cells. To test this hypothesis, further experiments were carried out in GEO and SW48 parental cells to evaluate whether exogenous TGF-α treatment by itself could lead to the acquisition of resistance to cetuximab. TGF-α stimulation of both cell lines resulted in decreased cetuximab sensitivity, because TGF-α could partially overcome the cetuximab induced growth inhibition (Fig. 6A). Moreover, TGF-α treatment of GEO and SW48 cells led to both EGFR and MET phosphorylation with an increase in MAPK activation and did not show evident effects on AKT (Fig. 6B). However, combined treatment with cetuximab partially rescues the cetuximab-mediated effects on MAPK and AKT phosphorylation. (Fig. 6B).

We next analyzed whether a functional cross talk between EGFR and MET in GEO-CR cells could be due to the interaction of the two growth factor receptors. For this purpose, GEO and GEO-CR protein extracts were immunoprecipitated with a specific anti-MET antibody and assayed by Western blotting with a specific anti-EGFR antibody. As shown in Fig. 6C, EGFR immunoprecipitated together with MET in GEO-CR cells, but not in GEO cells. Moreover, to elucidate the potential role of TGF-α in inducing EGFR–MET interaction, GEO, SW48, GEO-CR, and SW48-CR cells were treated with TGF-α in the presence or in the absence of cetuximab and lysates were immune-precipitated with the anti-MET antibody and then assayed by Western blotting with the anti-EGFR antibody. As reported in Fig. 6D, TGF-α treatment induced EGFR–MET interaction in GEO and SW48 cells and increased the interaction GEO-CR and SW48-CR cells. The EGFR–MET interaction was also observed following combined treatment of cells with both TGF-α and cetuximab, suggesting that, even in the presence of cetuximab, TGF-α could induce a cross-interaction between EGFR and MET.

Elucidating the mechanisms of cancer cell resistance to anticancer drugs is critical for the development of effective therapies. In this respect, mechanistic insights gained from preclinical models can be used to design novel treatment strategies. In the past few years, an extensive effort has been made to understand the resistance mechanism(s) that cancer cells develop to overcome the anticancer efficacy of EGFR inhibitors (3, 4, 7).

The aim of this study was to examine the signaling mechanisms operating in human colorectal cancer cells that are sensitive to the antitumor activity of the anti-EGFR mAb cetuximab and that become resistant following continuous exposure to the drug. In fact, although cetuximab is an effective anticancer agent in the treatment of patients with mCRC, the occurrence of acquired resistance is a major clinical limitation. It has been shown that intrinsic resistance to cetuximab as well as of the other anti-EGFR mAb panitumumab can be the result of constitutive activation of KRAS signaling, due to the presence of KRAS gene mutations (4). For this reason, cetuximab therapy is limited to patients with mCRC with a tumor harboring wild-type KRAS gene. Moreover, the constitutive activation of other proteins by specific gene mutations are responsible of intrinsic resistance to anti-EGFR therapies in patients with colorectal cancer, including BRAF, NRAS, PI3KCA (exon 20) genes, or inactivation of the PTEN phosphatase (7, 37). It has recently been shown that in initially responding patients with colorectal cancer with a KRAS wild-type tumor, the resistance to anti-EGFR antibodies may occur by selection of cancer cell clones harboring a KRAS gene mutation (9, 10). However, approximately 25% of patients with colorectal cancer not responding to EGFR inhibitors are wild-type for KRAS, BRAF, NRAS, PIK3CA, and PTEN genes and the mechanism of resistance in these patients is still unknown (38). Therefore, both primary and acquired resistance mechanisms significantly limit the efficacy of anti-EGFR mAbs in the medical management of patients with colorectal cancer.

In the present study, we have generated and characterized two models of cetuximab resistance in human colon cancer cells. In the parental, cetuximab-sensitive, GEO, and SW48 colorectal cancer cells, cetuximab treatment was able to induce cell growth inhibition and apoptosis with MAPK and AKT phosphorylation reduction. On the contrary, this was not observed in the derived GEO-CR and SW48-CR cells, in which activation of MAPK and AKT, was not blocked, despite EGFR inhibition by cetuximab treatment. MAPK activation that could bypass EGFR inhibition is one of the potential mechanisms of acquired resistance to anti-EGFR therapies (39). In this respect, it has recently been shown that one mechanism of acquired resistance to cetuximab treatment could be the HER2 (ERBB2) activation (11). Similarly, Bertotti and colleagues identified HER2 gene amplification as a mechanism of resistance to cetuximab in mCRC that harbor wild-type KRAS, NRAS, BRAF, and PIK3CA genes (12). In this respect, although HER2 gene amplification was found to occur in only 2% to 3% of unselected mCRC specimens, a significantly higher frequency was found in metastatic KRAS wild-type patients with colorectal cancer that did not benefit from treatment with anti-EGFR mAbs (12). For this reason, we decided to examine the expression and the activation of HER2 and of other EGFR family members in GEO-CR and SW48-CR cells. However, no gene amplification and/or enhanced expression of HERB2, HERB3, or ERBB4 was found.

Our results suggest that MET activation may play a relevant role in determining acquired resistance to cetuximab. In particular, gene expression profiling of GEO and GEO-CR cells identified a series of genes involved in HGF–MET-dependent pathway, that were upregulated in GEO-CR cells as compared with GEO cells. Intriguingly, inhibition of MET correlated with a partially restored sensitivity to cetuximab in GEO-CR and SW48-CR cancer cells. In fact, MET silencing restores cetuximab ability to inhibit MAPK and AKT and cell proliferation. In contrast, HGF-mediated activation of MET in parental GEO and SW48 cells reduced the sensitivity cetuximab. The role of MET in EGFR resistance was also confirmed by using the selective MET inhibitor PHA665752 in combination with cetuximab that overcome the resistance in GEO-CR and SW48-CR cells.

Several studies have demonstrated that in lung adenocarcinoma–derived cells the EGFR inhibition can be overcome by MET (19, 22). Moreover, HGF-dependent MET activation also proved to be a mechanism of intrinsic resistance to gefitinib in NSCLC cells with EGFR-activating mutations and no MET gene amplification (22). MET amplification is associated with acquired resistance to anti-EGFR treatment in patients with mCRC who have not selected a KRAS mutation during the therapy (14). In agreement with these results, it has been recently shown that HGF stimulation rescues cetuximab-sensitive colorectal cancer cells from EGFR inhibition by preventing cell-cycle arrest and inducing cell proliferations (40). Similarly, in HER2-overexpressing breast cancer cells, MET contributes to trastuzumab resistance (12). Conversely, MET-amplified gastric cancer cells were shown to be resistant to a MET-specific TKI when stimulated with EGF or heregulin-β (11). In these studies treatment of cancer cells with both MET and EGFR inhibitors could overcome resistance to a single inhibitor.

We have investigated the potential mechanisms by which MET can be activated. Although HGF-driven autocrine loop as a mechanism of MET activation was not demonstrated in GEO-CR and SW48-CR cells, here we provide evidence of a cross talk between EGFR and MET. In fact, we found that EGFR immunoprecipitated together with MET in GEO-CR and SW48-CR cells, but not in GEO and SW48 cells. This interaction was observed in parental cells following TGF-α treatment. In agreement with these results, several studies have shown an interaction between EGFR and MET that allows activation of MET following stimulation of different cell lines (A431, AKN-1, HepG2, AKN-1, HuH 6, and MRC5) with the EGFR selective ligands EGF or TGF-α (24). Stimulation of cells expressing both MET and EGFR with EGF resulted in phosphorylation of MET, and stimulation with ligands for both receptors resulted in synergistic activation of downstream modulators, indicating mutual activation of these two pathways (41).

EGFR ligands have been studied in several cancers as potential biomarkers for EGFR-targeted therapy. However, the results have been controversial depending on different tumor types and different clinical specimens used for testing (42, 43). In a study by Cohen and colleagues, changes in serum TGF-α levels in patients with head and neck cancer treated with gefitinib, were not associated with clinical response to gefitinib (43). In a study by Mutsaers and colleagues, TGF-α levels were increased in the plasma of patients with colorectal cancer during cetuximab treatment (44). However, increased TGF-α levels did not associate with cetuximab response (45). In contrast, more consistent results support the view that increased mRNA levels of amphiregulin (AREG) and epiregulin (EREG) in the tumors of patients with colorectal cancer are associated with cetuximab sensitivity. Khambata-Ford and colleagues were the first to demonstrate that AREG and EREG were significantly upregulated in chemotherapy refractory patients with colorectal cancer that obtained a clinical benefit from cetuximab treatment (42). The predictive value of EREG and AREG expression for cetuximab sensitivity was confirmed by Jacobs and colleagues, which analyzed primary tumors from refractory patients with mCRC treated with cetuximab-based therapies (46). Furthermore, in the 045 phase I study of cetuximab monotherapy as first-line treatment of patients with mCRC, Tabernero and colleagues found that AREG and EREG mRNA levels were significantly higher in tumor samples from patients with clinical response to cetuximab as compared with patients with disease progression (47). Although high level of expression of AREG and EREG ligands may predict a better response to treatment with cetuximab in patients with mCRC, HB-EGF and TGF-α overexpression could confer lack of sensitivity to EGFR inhibitors. In fact, in 045 clinical trial, TGF-α mRNA levels were significantly elevated in the tumors from patients with mCRC with disease progression after cetuximab therapy as compared with patient with clinical response (47). Furthermore, increased HB-EGF may be correlated with cetuximab resistance in head and neck squamous cell carcinoma (48).

Taken together, these results suggest that enhanced expression of different EGFR-specific ligands could have a different effect on the clinical activity of cetuximab. In this respect, the observation of TGF-α overexpression with activation of MET signaling in human colorectal cancer cells with acquired resistance to cetuximab treatment, as described in the present study, is in agreement with the above clinical findings.

The results of the present study suggest that overexpression of TGF-α, which induces EGFR–MET interaction, plays a relevant role in determining the development of acquired resistance in cetuximab-treated colorectal cancer cells. Therefore, the combined inhibition of both EGFR and MET RTKs could represent a rational therapeutic strategy for preventing and/or overcoming cetuximab resistance in patients with colorectal cancer.

No potential conflicts of interest were disclosed.

Conception and design: T. Troiani, E. Martinelli, A. Capasso, A. Nappi, L. Berrino, R. Bianco, F. Ciardiello

Development of methodology: T. Troiani, E. Martinelli, S. Napolitano, D. Vitagliano, A. Nappi, R. Bianco, F. Ciardiello

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Troiani, F. Morgillo, V. Sforza, A. Nappi, L. Berrino, F. Ciardiello

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Troiani, E. Martinelli, L.P. Ciuffreda, S. Costantino, F. Morgillo, V. Sforza, A. Nappi, R. De Palma, E. D'Aiuto, R. Bianco, F. Ciardiello

Writing, review, and/or revision of the manuscript: T. Troiani, E. Martinelli, F. Morgillo, A. Nappi, R. De Palma, L. Berrino, F. Ciardiello

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Nappi, F. Ciardiello

Study supervision: R. De Palma, R. Bianco, F. Ciardiello

This research has been supported by a grant from the Associazione Italiana per la Ricerca sul Cancro (AIRC).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.