Abstract
Overcoming cellular mechanisms of de novo and acquired resistance to drug therapy remains a central challenge in the clinical management of many cancers, including non–small cell lung cancer (NSCLC). Although much work has linked the epithelial–mesenchymal transition (EMT) in cancer cells to the emergence of drug resistance, it is less clear where tractable routes may exist to reverse or inhibit EMT as a strategy for drug sensitization. Here, we demonstrate that extracellular signal-regulated kinase (ERK) 1/2 (mitogen-activated protein kinase 3/1, MAPK3/1) signaling plays a key role in directing the mesenchymal character of NSCLC cells and that blocking ERK signaling is sufficient to heighten therapeutic responses to EGF receptor (EGFR) inhibitors. MEK1/2 (MAPKK1/2) inhibition promoted an epithelial phenotype in NSCLC cells, preventing induction of EMT by exogenous TGF-β. Moreover, in cells exhibiting de novo or acquired resistance to the EGFR inhibitor gefitinib, MEK inhibition enhanced the sensitivity to gefitinib and slowed cell migration. These effects only occurred, however, if MEK was inhibited for a period sufficient to trigger changes in EMT marker expression. Consistent with these findings, changes in EMT phenotypes and markers were also induced by the expression of mutant KRAS in a MEK-dependent manner. Our results suggest that prolonged exposure to MEK or ERK inhibitors may not only restrain EMT but also overcome naïve or acquired resistance of NSCLC to EGFR-targeted therapy in the clinic. Cancer Res; 74(1); 309–19. ©2013 AACR.
Introduction
EGF receptor (EGFR) overexpression and activation are hallmarks of many cancers, including non–small cell lung cancer (NSCLC). Consequently, a number of inhibitors and monoclonal antibodies targeting EGFR have been developed and approved for various cancers. Unfortunately, these drugs are generally ineffective. In NSCLC, response to EGFR inhibitors is limited mainly to the rare patients (∼10%) whose tumors harbor somatic, kinase-activated mutants of EGFR (1, 2). Even these patients almost invariably develop resistance to EGFR inhibitors, often through the EGFR “gatekeeper” mutation (T790M; refs. 3, 4) or through upregulation of the HGF receptor c-MET or other receptors (5). Combination therapies present a possible strategy to overcome resistance. In NSCLC, recent investigations suggest promise for combining EGFR inhibitors with chemoradiation (6), the multikinase inhibitor sorafenib (7), or a c-MET inhibitor (8). Scheduling multiple drugs such that initial therapy reprograms cells to respond to another drug is another possible strategy. In one recent example, triple-negative breast cancer cells and NSCLC cells were dramatically sensitized to doxorubicin by pretreatment with the EGFR inhibitor erlotinib (9).
Epithelial–mesenchymal transition (EMT) is another pathway through which cancers of epithelial origin become chemoresistant. EMT is a developmental process whereby epithelial cells lose cell–cell adhesions to become more motile and invasive. Cells undergoing EMT lose expression of epithelial markers (e.g., E-cadherin) and gain expression of mesenchymal markers (e.g., vimentin and fibronectin) through differential expression and activation of transcription factors including Twist, ZEB1, and Snail (10, 11). EMT is frequently hijacked in metastatic progression, and mesenchymal dedifferentiation has been associated with resistance to EGFR inhibitors, chemotherapy, and other targeted drugs in cancers of the lung (12–14), bladder (15), head and neck (16, 17), pancreas (18), and breast (19). In NSCLC, in vitro acquired resistance to the EGFR inhibitor erlotinib can result from the selection of a mesenchymal subpopulation (20), and restoring E-cadherin expression in mesenchymal-like NSCLC cells potentiates sensitivity to EGFR inhibitors (21). In addition, growing evidence for AXL-mediated EGFR inhibitor resistance has been tied to EMT (22). Thus, developing treatments that elicit a mesenchymal-epithelial transition (MET) could be a useful approach for expanding the efficacy of EGFR inhibitors.
Several studies have demonstrated a requirement for extracellular signal-regulated kinase (ERK)1/2 or mitogen-activated protein kinase (MAPK) 3/1 pathway activity in EMT induced by TGF-β in nontransformed cells (23–25). ERK2, but not ERK1, activity also induces EMT in nontransformed mammary epithelial cells (26) and has been implicated as mediating oncogenic KRAS-induced invasion in pancreatic cancer cells (27). Interestingly, ERK2 amplification was recently identified as a mechanism leading to acquired resistance to EGFR inhibitors in NSCLC (28).
Here, we sought to determine ERK's role in governing EMT in NSCLC. In a panel of NSCLC cell lines, inhibition of MEK1/2 (MAPKK1/2) prevented TGF-β–induced EMT and promoted epithelial cellular characteristics when administered alone. Conversely, augmented ERK activation, through KRAS12V expression or ERK2 amplification, promoted mesenchymal characteristics. Furthermore, chronic MEK inhibition for times long enough to observe changes in epithelial and mesenchymal marker expression augmented cellular sensitivity to the EGFR inhibitor gefitinib in cell lines with de novo or acquired resistance to EGFR inhibitors. These changes were reversible and accompanied by shifts in expression of stem cell-like markers CD24 and CD44. These results suggest the potential utility of drug scheduling strategies first targeting ERK to promote epithelial characteristics before targeting EGFR or other oncogenic signaling nodes.
Materials and Methods
Cell culture
H1666 cells were obtained from the American Type Culture Collection. H322, gefitinib-resistant PC9 (clone GR4, referred to as GR henceforth), and WZ4002-resistant PC9 cells (clone WZR12, referred to as WZR henceforth) were provided by Dr. Pasi Jänne (Dana–Farber Cancer Institute, Boston, MA). Parental PC9 cells were provided by Dr. Douglas Lauffenburger (Massachusetts Institute of Technology, Cambridge, MA). Because PC9 cells came from different labs, we confirmed similar expression of important proteins and response to gefitinib for parental stocks from both labs. H358 cells were provided by Dr. Russ Carstens (University of Pennsylvania, Philadelphia, PA). PC9 (all variants), H322, and H358 cells were maintained in RPMI-1640 supplemented with 10% FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, and 1 mmol/L L-glutamine. H1666 cells were maintained in ACL4 (29). Cell culture reagents were from Life Technologies. Cell lines were validated for anticipated responsiveness to gefitinib and were cultured for less than 2 months from low-passage frozen stocks.
Chronic MEK inhibition
H1666, PC9 GR, and PC9 WZR cells were maintained in 10, 5, and 20 μmol/L U0126, respectively, with controls maintained in dimethyl sulfoxide (DMSO). Responses to 1-hour or 2-day treatments with a range of U0126 concentrations were evaluated for each cell line to identify U0126 concentrations yielding significantly inhibited ERK phosphorylation (Supplementary Fig. S1) but minimal background cell death. For culture experiments, media were changed every 2 days or cells were passed as necessary. When passing, cells were subcultured into inhibitor-free media to promote adhesion. U0126 or DMSO was replaced the following day. When appropriate, cells were lysed or gefitinib was added 24 hours later. Time points in figures with chronic MEK inhibition reflect the total time of U0126 exposure before lysing or gefitinib addition. To probe the reversibility of U0126 effects, cells were split from U0126 cultures after 7 days and maintained in DMSO thereafter. All time courses were performed from freshly thawed cells multiple times to demonstrate reproducibility. Naïve cell death response to U0126 and gefitinib was quantified at multiple points during time courses, including the beginning and end, to verify that the baseline response was not changing.
Wound-healing assay
Confluent cell monolayers in 6-well plates were scratched with a pipet tip, and media were immediately changed. Phase contrast images were taken with a Zeiss Axiovert 40 CFL microscope (×10 objective) every 1 to 3 hours for 11 hours or less. Scratch areas were quantified using ImageJ, and closure rates were calculated from linear fits of areas versus time and reported as percentage of total image area closed per hour normalized to the conditions indicated in figures. Where inhibitors were used, cells were plated at subconfluence and treated at appropriate times with media containing inhibitor. Media and inhibitor were changed every 2 days until wells reached confluence.
Transwell migration assay
Untreated PC9 GR and WZR cells or H1666 cells treated with U0126 or DMSO for 4 days were trypsinized and resuspended in media containing 0.1% FBS and U0126 or DMSO for H1666. Of note, 20,000 untreated or DMSO-treated cells or 50,000 U0126-treated cells (adjusted for U0126-mediated changes in adhesion) were added to 8 μm Transwell membranes (Corning), which were placed in 24-well plate wells containing complete media with DMSO or U0126. After 20 hours, cells on the upper surfaces were removed with a cotton swab. Cells on the lower surfaces were fixed in 4% paraformaldehyde for 30 minutes and washed with PBS, and nuclei were stained with Hoescht-33342. Membranes were mounted and imaged (×10 objective), and nuclei were counted using ImageJ. Counts were normalized to the number of adhered cells from parallel wells for each condition.
Flow cytometry
For cell death assays, cells were plated in 6-well dishes from their various culture conditions and treated with inhibitors. Floating and adherent cells were collected 48 hours later, resuspended in PBS containing TO-PRO3 (Life Technologies), and analyzed within 1 hour. For CD24 and CD44 measurements, cells were collected as above, washed with 0.1% bovine serum albumin in PBS (PBSA), blocked for 10 minutes in PBSA, and incubated for 1 hour with 3 μL each of fluorescein isothiocyanate–conjugated antihuman CD44 antibody (BD Pharmingen, #555478) and Alexa-647 conjugated antihuman CD24 antibody (BioLegend, #311109) in 200 μL. Labeled cells were washed again and resuspended in PBS. Cytometry was performed on a BD Biosciences FACSCalibur cytometer, and data were analyzed using FlowJo.
Antibodies and other reagents
pERK T202/Y204 (#4377) and ERK (#4695) antibodies were from Cell Signaling Technology. E-cadherin (sc-8426), vimentin (sc-373717), fibronectin (sc-8422), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; sc-32233) antibodies were from Santa Cruz Biotechnology. Stocks of U0126 and gefitinib (LC Labs) were prepared in DMSO. Recombinant human EGF and TGF-β were from Peprotech. Infrared dye- and Alexa Fluor-conjugated secondary antibodies were from Rockland Immunochemicals and Invitrogen, respectively.
Additional methods
Western blotting, immunofluorescence, KRAS12V expression, experiments involving CI-1040, chronic MEK inhibition in H322, H358, and SKRB3 cells, and the EGFR internalization assay were performed as described in Supplementary Materials and Methods.
Results
NSCLC cell lines undergo MEK-dependent TGF-β–induced EMT
H1666, H322, and H358 cells were cultured in complete media with TGF-β and EGF with or without U0126 for 4 days, with a media change after 2 days (Fig. 1). In response to TGF-β and EGF, E-cadherin expression decreased and vimentin expression increased in all three cell lines, and fibronectin expression increased in H1666 and H358 cells (Fig. 1A). U0126 cotreatment inhibited changes in E-cadherin and vimentin, but did not prevent fibronectin induction. The same trends were found in H1666 cells using the alternative MEK inhibitor CI-1040 (Supplementary Fig. S2). The conditions used in Fig. 1A were explored in H1666 cells by immunofluorescence (Fig. 1B). Treatment with TGF-β and EGF promoted an elongated cellular morphology and the appearance of F-actin fibers, and decreased the E-cadherin intensity. U0126 addition prevented E-cadherin loss and promoted E-cadherin localization at cell–cell junctions.
MEK inhibition promotes epithelial phenotypes in NSCLC cells
We next explored the ability of MEK inhibition alone to promote epithelial characteristics in H1666 cells. As seen when combined with TGF-β and EGF, U0126 promoted E-cadherin localization at cell–cell junctions (Fig. 2A). MEK inhibition also antagonized H1666 wound healing in a treatment time-dependent manner (Fig. 2B and Supplementary Fig. S3). Although U0126 inhibited ERK phosphorylation within 1 hour of addition (Supplementary Fig. S1), U0126 treatment for the duration of the scratch assay (10 hours) did not significantly affect the wound-closure rate (“at scratch” condition vs. DMSO in Fig. 2B). The wound-closure rate decreased significantly with 1 day of U0126 pretreatment and was nearly zero with 4 days of pretreatment. Western blot analyses of lysates created in parallel revealed that E-cadherin expression increased and vimentin expression decreased with increasing the U0126 exposure time, consistent with the time-dependent changes in the wound-closure rate (Fig. 2C and Supplementary Fig. S3). U0126-mediated changes in wound-healing migration were further confirmed by assessing migration across Transwell inserts (Fig. 2D).
Chronic MEK inhibition sensitizes NSCLC cells to EGFR inhibition on a time scale consistent with changes in epithelial and mesenchymal markers
Because epithelial characteristics have been connected to EGFR dependence in NSCLC (14, 21), we explored the ability of MEK inhibition to increase sensitivity to gefitinib (Fig. 3). As described in Materials and Methods, H1666 cells were maintained in 10 μmol/L U0126 or DMSO (control) for up to 3 weeks. Throughout the 3-week period, cells were evaluated for death response to gefitinib. Without U0126 pretreatment, 2 days of EGFR and MEK coinhibition led to only 7% cell death in H1666, but this more than doubled with 3 days of U0126 pretreatment (Fig. 3A and Supplementary Fig. S4A). This enhancement further increased with additional U0126 pretreatment time, reaching a maximum at 11 days. Increased cellular sensitivity to gefitinib was accompanied by increased E-cadherin expression (Fig. 3B and Supplementary Fig. S4B). Beyond 11 days, the synergistic effect began to decrease, which was accompanied by increased fibronectin expression and decreased E-cadherin expression. Similar effects were observed in H1666 cells using CI-1040 (Supplementary Fig. S4C).
H358 and H322 cells and the breast cancer cell line SKBR3 were also evaluated for response to gefitinib following chronic MEK inhibition (Supplementary Fig. S5). H358 cells harbor a KRAS mutation and are ERK addicted (30). Not surprisingly, therefore, H358 response to gefitinib was not augmented by U0126 pretreatment. However, H322 and SKBR3 cells were sensitized to EGFR inhibition with 5 days of U0126 exposure, suggesting fairly broad applicability of this strategy.
We also probed the reversibility of the effects of chronic MEK inhibition in H1666 cells (Fig. 3C–E and Supplementary Fig. S6). After 7 days, a fraction of cells was split from the U0126 culture and maintained in media containing DMSO (control). This “U0126 removal” culture was subsequently plated in parallel with the cells maintained in U0126. After 7 days of U0126 exposure, removal from U0126 for 4 days completely reversed changes in cellular sensitivity to gefitinib and epithelial and mesenchymal marker expression (Fig. 3C and D and Supplementary Fig. S6A).
Stem cell-like subpopulations with intrinsic therapeutic resistance are found within tumors (20, 31) and are typically identified as CD44high/CD24low (20) or E-cadherinlow (32). CD44high/CD24low enrichment from heterogeneous cultures promotes mesenchymal behaviors including resistance to EGFR inhibitors (20). Using flow cytometry, we measured the shift in the CD44high/CD24low subpopulation with U0126 addition and subsequent removal (Fig. 3E and Supplementary Fig. S6B). U0126-treated cells had fewer CD44high/CD24low cells and more CD44low/CD24high cells than controls. As with epithelial/mesenchymal marker expression, U0126 removal for 4 days reversed the effect on the CD44low/CD24high population.
KRAS12V-mediated ERK activation promotes EMT
To further probe the connection between ERK activity and EMT, we expressed KRAS12V in H1666 cells. This produced an anticipated increase in ERK phosphorylation and increased vimentin expression (Fig. 4A). Gefitinib and U0126 promoted an epithelial shift in control cells, as determined by Western blot analysis and wound-closure measurements (Fig. 4A and B). KRAS12V promoted maintenance of ERK phosphorylation in response to gefitinib and largely prevented gefitinib-mediated changes in marker expression and wound closure (Fig. 4A and B). Importantly, the effects of KRAS12V were MEK dependent, as demonstrated by the effects of U0126 (Fig. 4).
MEK inhibition antagonizes mesenchymal phenotypes and acquired resistance to EGFR inhibition in an EGFR-mutant–expressing cell line
We tested the effects of chronic MEK inhibition in a panel of cell lines derived from PC9 cells, which express EGFRdelE746_A750 and are gefitinib sensitive. In parental PC9 cells, MEK inhibition prevented TGF-β–mediated EMT and drove changes in epithelial/mesenchymal marker expression and wound healing consistent with the effects observed in H1666 cells (Fig. 5 and Supplementary Fig. S7).
A previous study derived PC9 cells that acquired gefitinib resistance through EGFRT790M mutation (28). Another cell line was then derived from these cells with secondary resistance to the irreversible EGFR inhibitor WZ4002, which potently inhibits EGFRT790M (28, 33). WZ4002 resistance arose through ERK2 amplification. We used these GR and WZR clones to examine the effects of ERK2 activity, which has been tied to EMT in nontransformed cells (26). WZR cells displayed a more mesenchymal marker expression pattern than parental or GR cells, with increased fibronectin and vimentin expression and decreased E-cadherin expression (Fig. 6A). WZR cells were also more migratory than GR cells as measured by wound closure and Transwell migration (Fig. 6B and C).
The effects of chronic MEK inhibition were tested in GR and WZR cells, using a higher U0126 concentration for WZR because of the ERK2 amplification. Death response to cotreatment was enhanced by approximately two-fold within 3 days of U0126 pretreatment in both cell lines (Fig. 7A and Supplementary Fig. S8A and S8B). As in H1666 cells, sufficiently long U0126 exposure resulted in an eventual decrease in augmented response to EGFR inhibition. Increased sensitivity to gefitinib correlated with reduced vimentin expression and slight increases in E-cadherin expression, and the eventual decrease in augmentation was accompanied by increased fibronectin expression in WZR cells (fibronectin was not detectable in GR cells; Fig. 7B and Supplementary Fig. S8C).
Discussion
We characterized the role of ERK in regulating EMT in NSCLC cells. MEK inhibition prevented TGF-β–induced EMT in NSCLC cell lines expressing wild-type (WT) EGFR and KRAS (H1666 and H322), WT EGFR and mutant KRAS (H358), or mutant EGFR (PC9). This suggests a general role for ERK in promoting mesenchymal characteristics across typical NSCLC cell types. MEK inhibition also antagonized mesenchymal-associated characteristics including migration, resistance to EGFR inhibition, and CD44high/CD24low expression.
The role of ERK in EMT has been explored in other cell types. ERK is an established determinant of EMT in nontransformed cells (23, 34–36), with ERK2 specifically implicated as an EMT driver (26, 27, 37). This is consistent with findings in breast cancer (38), where MEK inhibition reverses miR-21–mediated EMT, and with studies of phosphatases regulating ERK (i.e., MKP3, PTPN14, and SHP2; refs. 39–41), which also suggest a connection between ERK activity and EMT.
A number of factors regulate ERK activity in NSCLC cells. We previously demonstrated a connection of impaired mutant EGFR endocytosis with a diminished functional role of SHP2 and impaired ERK activity (29, 42). More recently, in collaboration with others, we identified a coupling between ERK2 amplification and increased EGFR endocytosis in the WZ4002-resistant PC9 cells used here (28). Interestingly, inhibiting EGFR endocytosis in WZ4002-resistant cells reduced ERK phosphorylation and promoted an epithelial marker expression pattern (Supplementary Fig. S9), drawing a potential connection between EGFR endocytosis and EMT. More recently, Sprouty2 has been shown to promote ERK activity in NSCLC cells with or without EGFR mutations (43).
Of course, KRAS mutations may also promote ERK activity in NSCLC. In H1666 cells, KRAS12V-driven ERK activity promoted mesenchymal marker expression and resistance to gefitinib-mediated MET changes. These findings are generally consistent with studies connecting oncogenic HRAS or KRAS mutants with EMT in mammary epithelia (26), pancreatic cancer (27, 30), and lung cancer (30). The MEK dependence of our observations with KRAS12V may support exploration of MEK and ERK inhibitors in NSCLCs with KRAS mutation.
In addition to promoting epithelial marker expression, MEK inhibition promoted response to EGFR inhibition. Some studies have suggested the efficacy of EGFR and MEK coinhibition in gastric cancer (44) and pancreatic cancer cells (45), and current clinical trials are testing erlotinib combined with MEK inhibitors in NSCLC. In light of this, it is worth noting that the effects we observed with MEK inhibition occurred on a time scale that was much longer than that for ERK inhibition and more consistent with changes in epithelial and mesenchymal marker expression. This may suggest considering a clinical scheduling approach wherein a MEK inhibitor is initially used alone to promote an epithelial phenotype, followed by addition of an EGFR inhibitor. An analogous staggered drug scheduling approach was effective in triple-negative breast cancer cells, which were sensitized to doxorubicin by erlotinib pretreatment in vitro and in mouse tumor xenografts (9). Applying this basic concept to future clinical trial design may allow for lower MEK inhibitor doses to curb toxicities (46) and promote response to EGFR inhibitors. Given that the time windows for augmented response to gefitinib and the useful concentrations of U0126 were variable among cell lines, it will of course be important first to examine this strategy in preclinical animal models to assess general efficacy in an in vivo setting and optimize timing and dosing.
We note as well that some recent data suggest that MET is required for efficient colonization of distant metastases (47, 48). Thus, in evaluating different potential strategies for driving MET, the potential to drive proliferation of metastases will have to be weighed against the potential ability to kill tumor cells. ERK may be an especially attractive target in this regard. In at least one study, ERK inhibition through microRNA-mediated reduction of SHP2 (a positive regulator of ERK signaling) reduced metastases in an in vivo breast cancer model (49), which was due to SHP2′s simultaneous regulation of proliferation, invasion, and transcription factor expression. Thus, promoting MET through ERK inhibition, which should additionally inhibit cellular proliferation, may mitigate risks associated with proliferation of micrometastases.
In all three cell lines we tested for effects of chronic MEK inhibition, enhanced gefitinib response eventually decreased. Though initially unresponsive to U0126, U0126-cultured cells displayed noticeable cell death at mid-range time points (partially reflected by cell death measurements for U0126-only samples in Figs. 3 and 7). This could result in selection for U0126-resistant cells over time. Because the fraction of CD44high/CD24low cells in U0126-treated cultures remained low at time points when resistance was observed (Fig. 3C and E), the potentially inherently resistant population was apparently not CD44high/CD24low. Alternatively, resistance could arise from adaptation of the entire cell population, for example, through increased fibronectin-mediated signaling, which has been linked to AKT activation and docetaxel resistance in ovarian and breast cancer cells (50). Indeed, in all cases, the eventual drop in augmented response to gefitinib was accompanied by increased fibronectin expression. Future work should consider effectors of fibronectin-mediated signaling as signaling nodes whose inhibition may augment EGFR inhibitor response.
Although our data suggest potential relevance of our findings to de novo and acquired EGFR inhibitor resistance, responsiveness to chronic MEK inhibition was not uniform among cell lines, which could result from any of the numerous differences in initial conditions among NSCLC cell lines. For example, although H1666 cells express WT EGFR and KRAS, they also express the low-activity G446V BRAF mutant (51), which could influence their death response to MEK inhibition. In the future, therefore, investigating the cellular initial conditions that may determine responsiveness to this approach and identifying alternative signaling nodes whose inhibition may result in more stable maintenance of an epithelial cell type will also be important.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: J.M. Buonato, M.J. Lazzara
Development of methodology: J.M. Buonato, M.J. Lazzara
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.M. Buonato
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.M. Buonato
Writing, review, and/or revision of the manuscript: J.M. Buonato, M.J. Lazzara
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.J. Lazzara
Study supervision: M.J. Lazzara
Acknowledgments
The authors thank Drs. Pasi Jänne, Douglas Lauffenburger, Deepak Nihalani, and Russ Carstens for generously providing reagents, Dr. Devraj Basu for technical discussions, and Calixte Monast, Chris Furcht, and Alice Walsh for technical assistance and guidance.
Grant Support
This work was supported, in part, by the NIH under Ruth L. Kirschstein National Research Service Award 2T32HL007954 from the NIH–NHLBI. This material is also based upon work supported by the National Science Foundation Graduate Research Fellowship under grant no. DGE-0822.
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.