Mutation or deletion of Neurofibromin 1 (NF1), an inhibitor of RAS signaling, frequently occurs in epithelial ovarian cancer (EOC), supporting therapies that target downstream RAS effectors, such as the RAF–MEK–ERK pathway. However, no comprehensive studies have been carried out testing the efficacy of MEK inhibition in NF1-deficient EOC. Here, we performed a detailed characterization of MEK inhibition in NF1-deficient EOC cell lines using kinome profiling and RNA sequencing. Our studies showed MEK inhibitors (MEKi) were ineffective at providing durable growth inhibition in NF1-deficient cells due to kinome reprogramming. MEKi-mediated destabilization of FOSL1 resulted in induced expression of receptor tyrosine kinases (RTK) and their downstream RAF and PI3K signaling, thus overcoming MEKi therapy. MEKi synthetic enhancement screens identified BRD2 and BRD4 as integral mediators of the MEKi-induced RTK signatures. Inhibition of bromo and extra terminal (BET) proteins using BET bromodomain inhibitors blocked MEKi-induced RTK reprogramming, indicating that BRD2 and BRD4 represent promising therapeutic targets in combination with MEKi to block resistance due to kinome reprogramming in NF1-deficient EOC.
Our findings suggest MEK inhibitors will likely not be effective as single-agent therapies in NF1-deficient EOC due to kinome reprogramming. Cotargeting BET proteins in combination with MEKis to block reprogramming at the transcriptional level may provide an epigenetic strategy to overcome MEKi resistance in NF1-deficient EOC.
Treatment of epithelial ovarian cancer (EOC) is currently lacking effective targeted therapies (1). Large-scale genomic studies of EOC tumors have identified frequent deletion/mutation (12%) of the tumor suppressor gene NF1 (Neurofibromin 1), which is a Ras-GTPase–activating protein that negatively regulates the RAS family of proteins (1). Loss of NF1 promotes activation of several RAS effector pathways including (i) PI3K-mTOR-AKT, (ii) RAF-MEK-ERK, and (iii) RalGDS signaling resulting in oncogenic growth and survival (2–4). Notably, IHC studies have demonstrated MEK-ERK activation in the majority of NF1-deficient high-grade serous ovarian cancer tumors, and EOC cell lines harboring NF1 alterations exhibit elevated RAS activity (5). Other studies have shown NF1-deficient cells and tumors display sensitivity toward inhibitors targeting downstream RAS effector pathways such as PI3K-AKT or MEK-ERK signaling, suggesting EOC tumors that harbor NF1 defects may be sensitive to targeted PI3K or MEK inhibitors (MEKi; refs. 2, 3, 6).
Several highly specific kinase inhibitors targeting MEK1/2 proteins, including trametinib, have been developed and are currently being evaluated in clinical trials for a variety of cancers including ovarian cancers (7–9). Although promising, single-agent MEKi therapies have shown limited therapeutic benefit in the treatment of cancer due to the rapid development of drug resistance. One mechanism by which tumor cells circumvent the inhibitory action of single-agent kinase inhibitors is through kinome reprogramming, a process characterized by system-wide changes in kinase networks (10, 11). Recent studies have shown that inhibition of key kinase signaling nodes that are critical for tumor growth, such as MEK, triggers the rapid activation of compensatory kinase signaling pathways that facilitates drug resistance. For example, MEK inhibition in triple-negative breast cancer (TNBC) resulted in a rapid reprogramming of the kinome via the induced expression and activation of several receptor tyrosine kinases (RTK), including DDR1, PDGFRB, and AXL, that effectively bypasses the original MEK-ERK inhibition (11). In addition, lung and colon KRAS–mutant cancers have been shown to respond to targeted MEKi by increasing the expression and/or activity of ERBB3 and/or EGFR, whereas KRAS-mutant pancreatic cancers overcome MEK inhibition through activation of the RTKs AXL, PDGFRA, HER2, and EGFR (12, 13). Thus, development of combination therapies that inhibit, block, or prevent MEKi-induced RTK response and subsequent resistance is a highly tractable goal for the treatment of RAS-altered cancers.
In this study, we explored the role of kinome reprogramming as a resistance mechanism to MEKi in NF1-deficient EOC cells. Through the combined use of multiplexed inhibitor beads and mass spectrometry (MIB-MS) and RNA sequencing (RNA-seq), we demonstrated that resistance to trametinib in NF1-deficient EOC cells is mediated by “adaptive kinome reprogramming,” where inhibition of MEK rapidly triggers the expression and activation of a variety of RTKs and their downstream RAF and PI3K signaling overcoming MEK inhibition. We identified the transcriptional coactivators BRD2 and BRD4 as essential for MEKi-mediated kinome reprogramming in NF1-deficient EOC. Inhibition of BRD2 and BRD4 by genetic interference or with BET bromodomain inhibitors blocked MEKi-induced RTK reprogramming and provided more durable therapeutic responses, representing a promising combination therapy for NF1-deficient EOC.
Materials and Methods
Cell lines were verified by IDEXX laboratories and were verified to be Mycoplasma negative (January 7, 2019) using the Hoechst DNA stain method. A1847 cell lines were obtained from the Fox Chase Cancer Center Cell Culture Facility (Philadelphia, PA; deposited by Dr. Thomas Hamilton). The identity of the cells was authenticated by short tandem repeat analysis and comparison to early-passage stocks of the parental cells donated by Dr. T. Hamilton. OVCAR8 cells were maintained in RPMI1640 supplemented with 10% FBS, 100 U/mL Penicillin—Streptomycin, and 2 mmol/L GlutaMAX. A1847 and SKOV3 cell lines were maintained in RPMI1640 supplemented with 10% FBS, 100 U/mL Penicillin–Streptomycin, 2 mmol/L GlutaMAX, and 5 μg/mL insulin. SNU-119 cells were maintained in RPMI1640 supplemented with 10% FBS, 100 U/mL Penicillin–Streptomycin, 2 mmol/L GlutaMAX, and 25 mmol/L HEPES. COV362 cells were maintained in DMEM supplemented with 10% FBS, 100 U/mL Penicillin—Streptomycin, and 2 mmol/L GlutaMAX. CAOV3 cells were maintained in DMEM supplemented with 10% FBS, 100 U/mL Penicillin–Streptomycin, 2 mmol/L GlutaMAX, and 100 mmol/L sodium pyruvate. JHOS-2 and JHOS4 cells were maintained in DMEM/F12 supplemented with 10% FBS, 100 U/mL Penicillin–Streptomycin, 2 mmol/L GlutaMAX, and nonessential amino acids. HIO-80 ovarian surface epithelial cells were a generous gift supplied by Dr. Andrew Godwin (The University of Kansas, Lawrence, Kansas) and were maintained in MCDB 105/Media 199 + 5% FBS, 7.5 μg/mL insulin, 100 U/mL Penicillin–Streptomycin, and 2 mmol/L GlutaMAX. To generate trametinib-resistant cell lines, A1847 cells were chronically exposed to 50 nmol/L trametinib and propagated in the continued presence of 50 nmol/L trametinib. For drug washout experiments, trametinib was removed from resistant cells for approximately 1 month prior to retreatment. A1847 doxycycline–inducible shRNA BRD4 cells were generated by infecting A1847 cells with lentiviral particles made with the trans-lentiviral shRNA packaging system and a TRIPZ lentiviral human BRD4 shRNA plasmid. Cells were selected with and maintained in media containing 2 μg/mL puromycin. For SILAC labeling, cells were grown for seven doublings in arginine- and lysine-depleted media supplemented with either unlabeled l-arginine (84 mg/L) and l-lysine (48 mg/L) or equimolar amounts of heavy isotope-labeled [13C6,15N4]arginine (Arg10) and [13C6]lysine (Lys6) (Sigma) as described previously (14).
BIX02189, bortezomib, GDC-0941, trametinib, LY3009120, and sorafenib were purchased from Selleckchem, and JQ1 was purchased from ApexBio. PP58 (15) and VI16832 (16) were custom synthesized according to previously described methods by The Center for Combinatorial Chemistry and Drug Discovery, Jilin University (Changchun, China). CTx-0294885 (17) was purchased from MedKoo. Conjugation of inhibitors to beads was performed by carbodiimide coupling to ECH Sepharose 4B (CTx-0294885, VI16832, and PP58) or EAH Sepharose 4B (purvalanol B; GE Healthcare).
Cell lysates were subjected to SDS-PAGE using 8% gels. For experiments where multiple antibodies are probed, samples in 6× LSB were prepared in bulk. Twenty micrograms of sample from the same batch was run in multiple gels at the same time to ensure consistency. Samples were then transferred to polyvinylidene difluoride membranes. Following blocking with 5% milk, membranes were cut horizontally to probe multiple proteins of different molecular weights on the same blot. A list of antibodies used can be found in Supplementary Excel S1. Secondary horseradish peroxidase (HRP)-anti-rabbit and HRP-anti-mouse were obtained from Thermo Fisher Scientific. SuperSignal West Pico and Femto Chemiluminescent Substrates (Thermo Fisher Scientific) were used to visualize blots.
Cells were harvested in RTK array lysis buffer containing 20 mmol/L Tris-HCl (pH 8.0), 1% NP-40, 10% glycerol, 137 mmol/L NaCl, 2 mmol/L EDTA, 1× EDTA-free protease inhibitor cocktail (Roche), and 1% each of phosphatase inhibitor cocktails 1 and 2 (Sigma). After incubating on ice for 20 minutes, cell debris was pelleted at 4°C. Lysates (500 μg protein) were applied to R&D Systems Proteome Profiler Human Phospho-RTK antibody arrays. Washing, secondary antibody, and developing steps were performed according to the manufacturer's instructions.
For short-term growth assays, 3,000–5,000 cells were plated per well in 96-well plates and allowed to adhere and equilibrate overnight. Drug was added the following morning and after 120 hours of drug treatment, cell viability was assessed using the CellTiter-Glo luminescent cell viability assay according to the manufacturer's instruction (Promega). Student t tests were performed for statistical analyses and P < 0.05 was considered significant. For colony formation assays, cells were plated in 24-well dishes (3,000–5,000 cells per well) and incubated overnight before continuous drug treatment for 14 or 28 days, with drug and media replaced twice weekly. At the end of treatment, cells were rinsed with PBS and fixed with chilled methanol for 10 minutes at −20°C. Methanol was removed by aspiration, and cells were stained with 0.5% crystal violet in 20% methanol for 20 minutes at room temperature, washed with distilled water, and scanned.
MIBs preparation and chromatography
Experiments using MIB-MS were performed as described previously (11, 18). Briefly, endogenous kinases were isolated by flowing lysates over kinase inhibitor–conjugated sepharose beads (purvalanol B, VI16832, PP58 and CTx-0294885 beads were used) in 10 mL gravity-flow columns. Kinases were eluted from the column by boiling in 2 × 500 μL of 0.5% SDS, 0.1 mol/L Tris-HCl (pH 6.8), and 1% 2-mercaptoethanol. Eluted peptides were reduced by incubation with 5 mmol/L DTT at 60°C for 25 minutes, alkylated with 20 mmol/L iodoacetamide at room temperature for 30 minutes in the dark, and alkylation was quenched with DTT. Kinase peptides were cleaned with PepClean C18 Spin Columns (Thermo Fisher Scientific) and ethyl acetate extraction and subsequent LC/MS analysis was performed.
Mass spectrometry and spectra analysis
Proteolytic peptides were resuspended in 0.1% formic acid and separated with a Thermo RSLC Ultimate 3000 on a Thermo Easy-Spray C18 PepMap 2 μm column with a 235-minute gradient of 4%–25% acetonitrile with 0.1% formic acid at 300 nL/minute. Eluted peptides were analyzed by a Thermo Q Exactive plus mass spectrometer utilizing a top 15 methodology, in which the 15 most intense peptide precursor ions were subjected to fragmentation. The AGC for MS1 was set to 3 × 106 with a maximum injection time of 120 milliseconds and the AGC for MS2 ions was set to 1 × 105 with a maximum injection time of 150 milliseconds and the dynamic exclusion was set to 90 seconds. Protein identification was performed by searching MS-MS data against the Swiss-Prot human protein database downloaded on July 26, 2018, using Andromeda 220.127.116.11 built in MaxQuant 18.104.22.168.
Statistical analysis of MIB-MS
For trametinib-induced MIBs signatures, we performed at least two biological MIB-MS experiments using reciprocal SILAC heavy/light and light/heavy. A one-paired t test (Benjamini–Hochberg P < 0.05) of SILAC MIB-binding ratios from biological replicates of L-drug/H-control or H-drug/L-control was performed to generate MIB-MS response signatures to trametinib using Perseus software v 22.214.171.124. Visualization of MIBs-kinome signatures using heatmaps were generated using Perseus software v 126.96.36.199.
The GeneJET RNA Purification Kit (Thermo Fisher Scientific) was used to isolate RNA from cells according to the manufacturer's instructions. qRT-PCR on diluted cDNA was performed with inventoried TaqMan gene expression assays on the StepOne real-time PCR system. A list of TaqMan gene expression assay probes used can be found in Supplementary Excel S1.
RNAi knockdown studies
siRNA transfections were performed using 25 nmol/L siRNA duplex and the reverse transfection protocol. 3,000–5,000 cells per well were added to 96-well plates with media containing the siRNA and transfection reagent (Lipofectamine RNAiMax) according to the manufacturer's instructions. In experiments where inhibitors were used, the inhibitor was added at the time of transfection. Cells were allowed to grow for 72–120 hours posttransfection prior to CellTiter-Glo (Promega) analysis. Two-to-three independent experiments were performed with each cell line and siRNA. Student t tests were performed for statistical analyses and P < 0.05 was considered significant. For Western blot or MIB MS studies, the same procedure was performed with volumes and cell numbers proportionally scaled to a 60 mm, 10 cm, or 15 cm dish, and cells were collected 48–72 hours posttransfection. A list of siRNAs used can be found in Supplementary Excel S1.
The sequencing libraries were constructed from 500 ng of total RNA using the Illumina's TruSeq RNA Sample Kit V2 (Illumina) following the manufacturer's instruction. The fragment size of RNAseq libraries was verified using the Agilent 2100 Bioanalyzer (Agilent), and the concentrations were determined using Qubit Instrument (Life Technologies). The libraries were loaded onto the Illumina HiSeq 2500 at 8 pmol/L and run the rapid mode for 2 × 40 bp paired end sequencing. The fastq files were generated on the Illumina's BaseSpace, and further data analysis was performed using the TopHat platform for read alignment and the Cufflinks Assembly and DE for gene expression assessment (19).
All P values were two-sided unless otherwise specified.
Single-agent MEK inhibition shows minimal efficacy in NF1-deficient EOC cell lines
Mutation or deletion of the RAS suppressor NF1 has been reported in several EOC cell lines (Fig. 1A; ref. 5, 20). Characterization of NF1 protein levels by Western blot analysis confirmed NF1-altered lines A1847, CAOV3, SNU119, and JHOS-2 lacked NF1 protein, as well as identified NF1–wild-type (wt) lines COV362, JHOS-4, and OVCAR8 to be deficient in NF1 protein (Fig. 1B). SKOV3 cells have a reported mutation in NF1 but expressed NF1 protein, whereas other NF1-wt EOC cell lines KURAMOCHI, OVCAR4, OVSAHO, and OVCAR5 have detectable NF1 protein similar to normal ovarian surface epithelial cells (HIO80; Supplementary Fig. S1A). Differential activation of RAS effector AKT signaling was detected amidst NF1-deficient cells with majority of NF1-deficient cells exhibiting activation of RAF-MEK-ERK activity (Fig. 1B). Treatment of EOC cells with trametinib had minimal impact on cell viability across EOC cell lines, with the exception of JHOS-2 and the KRAS–mutant OVCAR5 cells. Notably, the majority of NF1-deficient cell lines were resistant (9) to trametinib therapy with GI50 values > 100 nmol/L (Fig. 1C; Supplementary Fig. S1B). Moreover, trametinib treatment of NF1-deficient A1847 cells only partially reduced colony formation and failed to induce apoptosis as observed with the KRAS–dependent OVCAR5 cells (Fig. 1D and E). Inhibition of MEK-ERK-RSK1 pathway by trametinib at 4 hours was confirmed by Western blot analysis in A1847 cells, however, activation of ERK phosphorylation returned by 48 hours, consistent with kinome reprograming (Fig. 1F).
MEK inhibition dynamically reprograms the kinome in NF1-mutant EOC cells
To explore adaptive kinase resistance mechanisms to MEK inhibition in NF1-deficient EOC, we applied MIB-MS in conjunction with RNA-seq to measure MEKi-induced transcriptional and proteomic reprogramming (Fig. 2A). Using this proteogenomic approach, we can identify the fraction of the kinome promoting resistance to the MEK inhibitor trametinib in NF1-deficient cells to rationally predict MEKi combination therapies that provide more durable therapeutic responses (11, 21). Kinome profiling of NF1-deficient A1847 cells using MIB-MS and RNA-seq revealed widespread transcriptional and proteomic rewiring of kinase networks following MEK inhibition. Increased MIB binding of the RTKs PDGFRB, DDR1, EPHB3, EPHA4, and MST1R, the tyrosine kinase (TK) FRK and PTK2B, as well as MYLK3, ULK1, MAP2K6, MAP3K3, MAP2K5, and MAPK7 were observed in A1847 cells following 48 hours trametinib treatment (Fig. 2B and C; Supplementary Excel S2A). Reduced MIB binding of EPHA2, AURKA, AURKB, and PIK3R4 was also observed following trametinib treatment. Trametinib treatment of A1847 cells for 48 hours increased RNA levels of several kinases including PIK3C2A, MYLK3, MYLK, PDGFRB, DDR1, RIPK2, JAK1, MST1R, and ATM, and decreased RNA levels of MAP2K7, MAP2K3, STK11, MAP2K2, and MAPK12 (Fig. 2D; Supplementary Excel S2B). Many of the kinases that showed induced MIB binding following trametinib treatment also exhibited increased RNA levels, including PDGFRB, DDR1, MST1R, MAP2K6, MAPK7, and ULK1, suggesting that a large component of the kinome rewiring is transcriptional (Fig. 2E). Notably, the transcriptional induction of RTKs in response to trametinib was observed in several additional NF1-deficient and NF1-wt EOC cells, demonstrating MEKi-induced RTK reprogramming was a common adaptive mechanism in EOC. Trametinib treatment increased expression of PDGFRB in NF1-deficient CAOV3, COV362, OVCAR8, SNU119, and JHOS-2 cells, as well as in NF1-wt SKOV3 and OVSAHO cells. Elevated DDR1 RNA levels were detected in JHOS-2, OVCAR8, OVCAR4, OVSAHO, and KURAMOCHI cells, whereas ERBB2 expression was induced in JHOS-2, OVCAR8, and A1847 cells following MEK treatment (Supplementary Fig. S2).
Consistent with MIB-MS and RNA-seq analysis, increased tyrosine phosphorylation of PDGFRB and DDR1 in response to trametinib treatment was observed by RTK arrays in A1847 cells, as well as the time-dependent upregulation of PDGFRB and downstream AKT, and STAT3 survival pathways (Fig. 2F and G). In addition, an increase in RNA levels of PDGFRB ligands PDGFB and PDGFD were observed in A1847 cells following MEK inhibition, demonstrating the MEKi-induced autocrine/paracrine loop involving upregulation of both the PDGFRB receptor and its activating ligands (Fig. 2H).
MEKi-induced kinase signature is chronically maintained in NF1-deficient cells and is reversible
Chronic exposure of A1847 cells to trametinib abolished the growth inhibition observed following short-term exposure to trametinib (Fig. 3A). Kinome profiling of A1847 cells chronically exposed to 50 nmol/L trametinib (A1847-T) demonstrated induced MIB binding of the majority of the kinases activated by short-term MEKi exposure, including PDGFRB, DDR1, MST1R, BRAF, and MAPK7, signifying trametinib-resistant cells continuously maintain the kinome signature induced at 48 hours (Fig. 3B and C; Supplementary Excel S3A and S3B). Increased MIB binding of JAK3, RIPK2, IKBKB, CHUK, TAOK1, RPS6KA1, and PRKCQ were uniquely observed in A1847-T cells, whereas MIB binding of PDGFRB and JAK1 were further induced relative to short-term trametinib exposure (Fig. 3D). Consistent with activation of several RTKs in A1847-T cells, elevated downstream survival signaling including STAT3, SRC, AKT, and CRAF, as well as the reactivation of ERK, was observed in cells chronically exposed to trametinib (Fig. 3E). Notably, removal of trametinib from A1847-T cells nearly completely reversed the MEKi-induced kinome signature, demonstrating the absolute requirement of trametinib to maintain the RTK-driven resistance signature (Fig. 3F). Furthermore, removal of trametinib resensitized A1847-T cells to trametinib (Fig. 3G).
Collectively, NF1-deficient cell line A1847 dynamically reprogrammed the kinome inducing a number of RTKs, including DDR1 and PDGFRB and their subsequent downstream RAF-MEK-ERK, JAK-STAT, and PI3K-AKT signaling that was maintained indefinitely in the presence of trametinib (Fig. 3H). Following removal of trametinib, A1847-T cells revert to initial kinome state and regain sensitivity to MEK inhibition.
Kinase inhibitor combination therapies overcome adaptive resistance to trametinib in NF1-deficient EOC cells
Knockdown of MIB-MS–nominated kinases in A1847 and A1847-T cells revealed enhanced dependency of cells chronically exposed to trametinib for RTKs (PDGFRB, EGFR, FGFR2, FGFR4, and DDR1), TKs (JAK1 and JAK3), MAPK pathway components (BRAF, MEK1/2, ERK1/2, and ERK5), as well as RIPK2 (Fig. 4A). Genetic depletion or small-molecule inhibition of PDGFRB using the pan-TK inhibitor sorafenib (22), greatly enhanced the efficacy of trametinib in A1847 or A1847-T cells, demonstrating the dependency of these cells for PDGFRB to overcome single-agent MEK inhibition (Fig. 4B–D). In addition, cotreatment of NF1-deficient EOC cell lines JHOS-2, OVCAR8, JHOS-4, SNU119, and CAOV3, as well as NF1-wt KURAMOCHI and OVSAHO cells with trametinib and sorafenib enhanced growth inhibition relative to single-agent therapies (Supplementary Fig. S3A). Targeting the MEK5-ERK5 pathway using MEK5 inhibitor BIX02189 or treatment with RAF inhibitor LY3009120 blocked cell growth to a greater extent in A1847-T cells than parental, demonstrating the dependency of trametinib-resistant cells for MEK5 and CRAF-BRAF signaling (Fig. 4E and F). Activation of RAF following MEK-ERK inhibition has been shown to occur in cancer cells due to alleviation of ERK-dependent negative feedback on RAF (23). Here, the combination of LY3009120 and trametinib provided superior growth inhibition relative to single agents across all NF1-deficient and NF1-wt EOC cells, signifying the MEKi-induced RAF dependency was a global adaptive mechanism to MEK inhibitors in EOC (Supplementary Fig. S3B).
Treatment of RAS-altered cells with MEK inhibitors has been shown to induce PI3K activity through a compensatory feedback mechanism, supporting the combined blockade of PI3K and MEK (2, 3, 24). Consistent with these findings, knockdown of PIK3CA or small-molecule inhibition with GDC0941 blocked cell growth in A1847-T cells to a greater extent than parental cells (Fig. 4G and H). Kinome profiling of A1847-T cells treated with GDC0941 for 48 hours resulted in reduced MIB binding of several cell-cycle protein kinases including PLK1, PLK4, CDK1, BUB1, TTK, AURKA, and AURKB consistent with the growth inhibition observed (Fig. 4I; Supplementary Excel S4). Furthermore, the combination of trametinib and GDC0941 showed significant synergy in A1847 parental cells, as well as across all EOC cell lines with the exception of OVCAR4 and KURAMOCHI cells, demonstrating activation of PI3K-AKT axis was an essential component of the kinome reprogramming response to MEK inhibition in both NF1-deficient and NF1-wt EOC cells (Fig. 4J; Supplementary Fig. S3C).
Taken together, targeting the MEKi-induced kinase signature with a variety of small-molecule inhibitors acting at the level of RTK or their downstream kinase pathways such as RAF, MEK5, or PI3K provides more durable responses relative to single agents alone in NF1-deficient cells. Moreover, these findings demonstrate numerous distinct kinase pathways are triggered in response to MEK inhibition that are integral to the MEKi-induced kinome reprogramming response.
MEKi-mediated degradation of FOSL1 alleviates repression of RTK signaling promoting trametinib resistance
The RAF-MEK-ERK pathway has been shown to regulate the stability of a number of cancer-associated transcription factors including MYC and FOSL1 (25, 26). Phosphorylation of MYC at S62 and/or FOSL1 at S265 by ERK1/2 prevents proteasome-mediated degradation leading to protein stabilization promoting growth and survival. Loss of MYC repressor activity following MEKi therapies has been shown to result in induced expression of RTKs in breast and colorectal cancers promoting drug resistance (11, 13). Herein, reduced phosphorylation of MYC at S62 and total MYC levels were observed at higher concentrations of trametinib (>100 nmol/L) in NF1-deficient A1847 and JHOS-2 EOC cells (Fig. 5A; Supplementary Fig. S4A). Notably, reduced levels of phosphorylated FOSL1 at S265 and FOSL1 total protein were observed at picomolar concentrations of trametinib that coincided with ERK inhibition and the induction of RTKs PDGFRB and DDR1. Reduced FOSL1 and increased PDGFRB RNA and protein levels were detected within 4 hours of trametinib treatment, demonstrating MEK inhibition rapidly mediates transcriptional reprogramming in A1847 cells (Supplementary Fig. S4B and S4C). Furthermore, knockdown of FOSL1 in A1847 cells induced transcription of several RTKs, including DDR1, ERBB2, FGFR3, IGF1R, INSR, MERTK, MET, and PDGFRB, whereas MYC knockdown induced ERBB3, ERBB2, and PDGFRB RNA levels (Fig. 5B). Knockdown of MYC uniquely increased protein levels of ERBB3 in both A1847 and JHOS-2 cell lines, whereas FOSL1 depletion distinctly elevated DDR1 protein levels. Either MYC or FOSL1 knockdown increased PDGFRB protein levels with FOSL1 depletion elevating PDGFRB to a greater extent than MYC ablation (Fig. 5C; Supplementary Fig. S4D). Moreover, preventing MEKi-mediated FOSL1 protein degradation using the proteasome inhibitor bortezomib blocked induction of PDGFRB and DDR1, demonstrating the RTKs are transcriptionally induced by MEKi through destabilization of FOSL1 proteins (Fig. 5D).
Bromo and extra terminal bromodomain proteins BRD2 and BRD4 are essential for MEKi-induced kinome adaptations in NF1-deficient EOC cells
The plasticity of the MEKi-mediated kinome reprogramming response, as well as the significant transcriptional upregulation of RTKs, suggests an epigenetic mechanism may be contributing to the MEKi-resistance in NF1-deficient cells. A key group of epigenetic remodelers referred to as “readers,” contain bromodomains that recognize histone modifications and facilitate the assembly of chromatin-remodeling complexes controlling gene expression (27). To interrogate the role of bromodomain function in transcriptional reprogramming of the kinome in response to MEK inhibition, we performed a MEKi synthetic lethal siRNA screen of 46 bromodomain-containing proteins in A1847 cells chronically exposed to trametinib (A1847-T) or A1847 parental cells (Fig. 6A). Knockdown of ATAD2B, TRIM33, TAF1, SP140L KAT2B, TAF1L, and BRD4 enhanced growth inhibition of MEKi-resistant A1847-T cells relative to parental cells (Supplementary Fig. S5A). Moreover, siRNA-mediated knockdown of BRD4 dramatically improved growth inhibition by trametinib and blocked the formation of trametinib-resistant colonies (Fig. 6B and C; Supplementary Fig. S5B and S5C).
Recently, the BET (bromo and extra terminal) bromodomain proteins were shown to be essential for kinase expression and promoting adaptive kinase transcription responses to inhibitors (18, 28–30), supporting BET proteins as attractive targets for blocking kinase transcription. Here, using qRT-PCR in combination with loss-of-function assays, we determined the role of BET proteins in the regulation of the RTK transcription in NF1-deficient A1847 cells. Differential regulation of kinases at the transcriptional level was observed among BET proteins, where BRD2 knockdown uniquely reduced DDR1, DDR2, IGF1R, and INSR RNA levels, whereas BRD4 knockdown solely inhibited PDGFRB transcription (Fig. 6D). Interestingly, knockdown of BRD4 reduced protein levels of ERBB3 without suppressing RNA levels, suggesting that BRD4 depletion may influence the protein stability of ERBB3 (Supplementary Fig. S5D). Notably, knockdown of both BRD2 and BRD4 resulted in overall greater reduction in transcription of RTKs, reducing RNA levels of AXL, DDR1, EGFR, ERBB2, ERBB3, IGF1R, and PDGFRB in A1847 cells, demonstrating combined blockade of BRD2 and BRD4 provided superior RTK transcriptional repression than knockdown of individual BET proteins (Fig. 6D).
Next, we tested whether downregulation of BRD2, BRD4, or both prevented MEKi-induced transcriptional induction of RTKs in A1847 cells. Knockdown of BRD4 blocked the MEKi-mediated induction of PDGFRB, whereas BRD2 depletion prevented increased expression of DDR1 following MEK inhibition, and depletion of both BRD2 and BRD4 was required to prevent MEKi-induced ERBB3 transcription (Fig. 6E). Furthermore, downregulation of BRD4 in A1847 cells blocked or reduced activation of the majority of the trametinib-mediated resistance signature as shown by MIB-MS, including PDGFRB, BRAF, RAF1, DDR1, MAPK7, JAK1, and RSK1 (Fig. 6F; Supplementary Excel S5).
Taken together, we identified BRD2 and BRD4 as integral to the regulation of kinases induced by MEKi in NF1-deficient cells, as well as demonstrated that blockade of BET protein function prevented the MEKi-mediated upregulation of RTKs RNA and protein levels. These findings highlight the therapeutic implications of targeting BET proteins to block kinome reprogramming to MEKi in NF1-deficient cells.
BET bromodomain inhibition blocks MEKi-induced kinome adaptations sensitizing NF1-deficient cells to trametinib
Small-molecule inhibitors of BET proteins have been developed that interfere with bromodomain binding blocking transcription (31). Combined treatment of the BET bromodomain inhibitors (BETi), JQ1, and trametinib for 48 hours blocked the induction of the majority of the MEKi-mediated reprogramming as determined by MIB-MS kinome profiling (Fig. 7A and B; Supplementary Excel S6). Reduced MIB binding of PDGFRB, RAF1, DDR1, RIPK2, JAK1, MAPK7, BRAF, MAP2K6, RPS6AK1, MET, SRC, and CHUK was observed in response to the combination therapy, as well as the induction a number of kinases involved in growth arrest and apoptosis including TGFBR1 and ACVR2A. Blockade of trametinib-induced activation of PDGFRB and RAF-MEK-ERK signaling by JQ1 was confirmed by Western blot and RTK arrays (Fig. 7C; Supplementary Fig. S6A). The combined treatment of trametinib and JQ1 induced apoptosis in A1847 cells, increasing cleaved PARP levels relative to single-agent trametinib therapy. The trametinib and JQ1 combination repressed MEKi-induced transcription of PDGFRB, DDR1, and ERBB3 in A1847 cells, and blocked trametinib-induced PDGFRB and DDR1 expression in CAOV3, COV362, JHOS-2, and OVCAR8, as well as prevented PDGFRB reprogramming in SKOV3 and SNU119 cell lines (Fig. 7D; Supplementary Fig. S6B and S6C). In addition, BET inhibition blocked MEKi-mediated induction of PDGFRB and DDR1 expression following FOSL1 knockdown, demonstrating BET proteins were required to promote transcription of these RTKs following FOSL1 depletion (Supplementary Fig. S6D). Notably, continued treatment of JQ1 was required to maintain PDGFRB repression in presence of trametinib, where JQ1 removal from media abolished repressive effects on RTK transcription (Supplementary Fig. S6E). The JQ1 and trametinib combination therapy enhanced growth inhibition of A1847 cells relative to single agents alone in 5-day viability assays, as well as blocked the emergence of MEKi-resistant A1847 colonies in long-term colony formation assays (Fig. 7E; Supplementary Fig. S6F). Moreover, cotargeting MEK and BET proteins provided superior growth repression over individual agents in all NF1-deficient cell lines (Fig. 7F; Supplementary Fig. S6G). Notably, NF1-proficient cell lines, KURAMOCHI, OVCAR4, and OVSAHO showed no therapeutic benefit to the trametinib–JQ1 combination, suggesting NF1-deficient EOC cells may display enhanced sensitivity to combined blockade of BET and MEK signaling.
In addition, A1847 cells that acquired resistance to trametinib therapy showed increased sensitivity to JQ1 treatment, where BET inhibition destabilized the MEKi-mediated reprogramming signature in MEKi-resistant cells (Fig. 7G). MIB-MS kinome profiling of trametinib-resistant A1847 cells revealed JQ1 treatment reduced MIB binding of several of the MEKi resistance signature including PDGFRB, BRAF, DDR1, RAF1, and MAPK3, as well as cell-cycle–related kinases AURKA, AURKB, and PLK1 consistent with the growth arrest observed (Fig. 7H). These findings demonstrated the requirement for the presence of both BETi and MEKi to efficiently block reprogramming, as well as show MEKi-resistant cells require BET proteins to maintain kinome adaptive responses.
Loss of RAS suppressor protein NF1 frequently occurs in EOC leading to activation of PI3K-AKT and RAF-MEK-ERK signaling (5), consequently, PI3K and MEK inhibitors have emerged as a promising therapeutic avenue. Moreover, increased frequency of NF1 mutations has also been reported in patients with EOC who have developed resistance to first-line chemotherapy for which there are no current therapies (32). Despite these observations and implications, no comprehensive studies have been carried out testing the efficacy of MEK inhibition in NF1-deficient EOC. Here, we performed a detailed proteomics, genomics, and functional analysis of MEK inhibition in an established NF1-deficient EOC cell line, A1847, and then subsequently validated our findings in several additional NF1-deficient EOC cells.
Our studies showed the MEK inhibitor trametinib was ineffective at providing durable growth inhibition or inducing apoptosis in NF1-deficient cells due to MEKi-mediated kinome reprogramming. Kinome profiling of MEKi-treated or MEKi-resistant NF1-deficient A1847 cells using MIB-MS and RNA-seq revealed cells activated a network of kinases including a variety of RTKs and their downstream PI3K-AKT, and RAF-MEK-ERK signaling overcoming MEK inhibition. Here, we showed targeting MEKi-induced RTKs, such as PDGFRB in combination with MEK inhibitor, improved the efficacy of trametinib in NF1-deficient EOC cells, as well as in some NF1-wt EOC cells. Importantly, recent studies have implicated activation of PDGFRB signaling as a biomarker for poor prognosis in patients with EOC (33). Our studies suggest that combination of trametinib and sorafenib may represent a promising therapy for EOC tumors with elevated PDGFRB and/or NF1 deficiencies. In addition, we showed cotargeting MEK and PI3K or RAF significantly enhanced growth inhibition relative to single agents across all NF1-deficient EOC cell lines tested, demonstrating MEKi-induced PI3K and RAF activation were integral to the kinome reprogramming response promoting trametinib resistance in NF1-deficient EOC cells. Notably, cotargeting MEK and PI3K or RAF enhanced growth inhibition independent of NF1 status, suggesting these combination therapies may have broad activity across EOC. Combination therapies targeting MEK and PI3K have been extensively explored in a variety of RAS-altered cancer models, as well as in clinical trials for RAS-altered cancers (24). However, significant toxicities have been associated with this kinase inhibitor combination with ongoing studies exploring dose regimens and timing of the drug combinations to deal with off-target side effects (34).
Transcriptional induction of RTKs following destabilization of key oncogenic transcription factors, such as MYC, has been observed in response to blockade of RAF-MEK-ERK signaling (11, 13, 30). Here, we demonstrated that treatment of NF1-deficient cells with low nanomolar concentrations of trametinib (<10 nmol/L) resulted in the rapid degradation of the transcription factor FOSL1 that coincided with induced expression of RTKs. FOSL1 has been implicated as a master gatekeeper of the EMT program in TNBC claudin low cells, consistent with a role of FOSL1 in the regulation of expression a diverse array of RTKs, particularly those RTKs involved in epithelial and mesenchymal signaling (35). In addition, we observed differential degradation of MYC or FOSL1 that was dependent on the dose of MEK inhibitor administered, where low doses of trametinib reduced FOSL1 but not MYC protein levels. Moreover, knockdown of MYC or FOSL1 in NF1-deficient cells resulted in induced expression of distinct RTKs, demonstrating the unique repressive functions of these transcription factors on RTK expression. Interestingly, these findings suggest that the composition of MEKi-induced RTK signature may depend on the concentration of trametinib that is administered. Low doses of trametinib may alleviate repression of RTKs regulated by FOSL1 such as DDR1, PDGFRB, and IGF1R, whereas higher doses of trametinib may lead to loss of repression of both FOSL1- and MYC-regulated RTKs such as EGFR, ERBB3, and PDGFRB. The heterogeneity of RTKs induced at higher doses of trametinib, due to destabilization of both MYC and FOSL1, may present a significant challenge in the design of combination therapies involving MEK and RTKs, particularly as ERBB and PDGFR family of receptors are quite distinct in structure and inhibitor sensitivities.
Previous studies have shown that combined blockade of BET bromodomain “reader” proteins and kinase inhibitors such as ERBB2/EGFR, PI3K, or MEK inhibitors prevented kinase adaptive responses providing superior growth inhibition in breast cancer models (28–30). Here, we showed that the bromodomain proteins BRD2 and BRD4 were required for the MEKi-induced RTK transcription and that combination therapies involving BETi and MEKi blocked RTK reprogramming providing more durable therapeutic responses in NF1-deficient EOC cells. Intriguingly, our loss-of-function studies suggest distinct functions for BRD2 and BRD4 in regulation of RTK transcription, where BRD2 and BRD4 were required for transcription of different sets of RTKs following MEK inhibition in NF1-deficient EOC cells. The relevant functions of the distinct BET bromodomain proteins in the epigenetic control of cell growth and survival remains poorly characterized (36), with the majority of studies focusing on defining the role of BRD4 in promoting cancer. Here, we showed BRD2 was required for DDR1 transcription, and that combined blockade of both BRD2 and BRD4 was required to reduce transcription of several RTKs. However, further studies exploring the role of each BET protein in controlling kinase transcription will be required to define BET protein–specific functions in EOC.
In previous reports, NF1-altered cancers were shown to exhibit enhanced sensitivity to epigenetic inhibitors such as JQ1 due to loss of the polycomb repressive complex 2 gene SUZ12, which potentiates RAS-driven transcription (37). Here, we found the majority of NF1-deficienct EOC cells exhibited enhanced sensitivity to the trametinib–JQ1 combination relative to single agents, whereas the majority of NF1-proficient EOC cells tested showed minimal to no therapeutic benefit, suggesting NF1-deficient EOC cells may harbor unique transcriptional plasticity and dependency for BRD4 for MEKi-mediated kinome reprogramming. However, further proteogenomic studies exploring the distinct responses of NF1-wt or NF1-deficient EOC cells to trametinib will be required to decipher the differential impact that RAS activation has on transcriptional reprogramming to MEK inhibition.
Finally, our evaluation of combination therapies using trametinib and JQ1 showed BET inhibition can prevent the induction of MEKi-reprogramming in NF1-deficient EOC cells, as well as disrupt the adaptive response once it is established in MEKi-resistant cells. Moreover, simultaneous administration of the drugs was required to prevent MEKi-mediated transcriptional induction of RTKs, where removal of JQ1 abolished transcriptional repression of MEKi-induced RTKs. Our findings suggest BET inhibitors should be utilized concurrently with trametinib or administered once tumors have adapted to MEKi therapies for optimal inhibition of MEKi-induced kinome reprogramming. However, further exploration of BET and MEK inhibitor combinations in NF1-deficient EOC preclinical models will be required to optimize dosing and timing, as well as define potential toxicities associated with the dual agent therapy.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: A.M. Kurimchak, J. Chernoff, J.S. Duncan
Development of methodology: K.E. Duncan, J.S. Duncan
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.M. Kurimchak, C. Shelton, C. Herrera-Montávez, K.E. Duncan, J.S. Duncan
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.M. Kurimchak, J.S. Duncan
Writing, review, and/or revision of the manuscript: A.M. Kurimchak, J. Chernoff, J.S. Duncan
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.S. Duncan
Study supervision: J. Chernoff, J.S. Duncan
This work was funded by NIH CORE Grant CA06927 (Fox Chase Cancer Center), R01 CA211670 (to J.S. Duncan), R01 CA142928 (to J. Chernoff), and NIH T32 CA009035 (to A.M. Kurimchak). The W.W. Smith Charitable Trust Research Grant (to J.S. Duncan). Kinome profiling studies were funded by the Cancer Kinome Initiative at Fox Chase Cancer Center, which was established by a donation from Don Morel.
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.