Induction of MNK Kinase–dependent eIF4E Phosphorylation by Inhibitors Targeting BET Proteins Limits Efficacy of BET Inhibitors

The bromodomain (BRD) and extra-terminal domain-containing (BET) proteins, which belong to the BRD protein family, play an important role in various cellular processes (1). BET proteins, by functioning as transcriptional coactivators, can enhance cell-cycle progression, limit apoptosis, and promote cell motility and metastasis in a broad range of cancers (1). BET proteins mediate their protumorigenic function, in part, by upregulating expression of Myc and other oncogenes (2). Clinically, there is increasing interest in targeting BET proteins, and a number of potent and selective BET inhibitors (BETi) have been evaluated in early-phase clinical trials (3, 4). For example, the BETi OTX-015 (MK-8628, birabresib) has successfully undergone phase I clinical studies in patients with hematologic malignancies and select solid tumors (3, 4).

Although BETis have shown promising results in early-phase clinical trials, resistance to BETis has been reported in preclinical models (4). For example, cancer cells can develop resistance to BETis by activating alternative signaling pathways to upregulate MYC transcription (5–7). Recently, it was shown that increased accumulation of BET proteins also mediates resistance to BETis (8). BET proteins can be degraded through a SPOP-mediated ubiquitination (9, 10); however, cancer cells with SPOP mutation demonstrate increased levels of BET proteins and resistance to BETis (9, 10). Consequently, to overcome resistance due to increased accumulation of BET proteins, there is also increasing interest in developing inhibitors that promote degradation of BET proteins through the use of proteolysis-targeting chimeras (PROTAC; refs. 8, 11).

Cancer cells can also develop resistance by activating the translational machinery (12). For example, breast cancer cells developing resistance to tamoxifen demonstrate evidence of selective reprogramming of mRNA translation mediated by eukaryotic initiation factor 4E (eIF4E; ref. 13). The eIF4E activity can be regulated by MNK kinase–dependent phosphorylation, with eIF4E phosphorylation being critical for malignant transformation and promotion of tumor development in animal models (14, 15). Importantly, the MNK–eIF4E pathway has also been shown to be activated as a feedback survival mechanism in response to acute treatment with chemotherapy and targeted therapies (16, 17). However, the activation of the MNK–eIF4E pathway by BETis and the involvement of the MNK–eIF4E pathway in limiting response to BETis have not been previously evaluated.

In this study, we demonstrate that BETis and eIF4E knockdown decrease growth of cancer cells. Although BETis do not repress eIF4E protein levels, we show that BETis induce MNK kinase–dependent eIF4E phosphorylation. BET PROTACs also induce MNK kinase–dependent eIF4E phosphorylation in cancer cells. Mechanistically, we show that the effect of BETis on MNK-eIF4E phosphorylation is mediated by p38 MAPKs. We also show that BETis suppress RacGAP1 to induce Rac-mediated eIF4E phosphorylation. Significantly, MNK inhibitors and MNK1/2 knockdown enhance the efficacy of BETis in suppressing proliferation of cancer cells in 3D collagen and in a syngeneic mouse model. Together, these results demonstrate increased eIF4E phosphorylation following treatment with BETis and identify combination therapy with BET and MNK inhibitors for the treatment of patients with cancer.

K1, RO82-w-1, and FTC-133 cell lines were purchased from Sigma and cultured according to the manufacturer's instructions. MDA-T22 and MDA-T85 were obtained from MD Anderson Cancer Center and were cultured in RPMI1640 medium containing 10% FBS and antibiotics (100 U/mL penicillin and 100 μg/mL streptomycin) as described previously (18). Human PDAC cell lines, Panc1 and CD18, were obtained from ATCC. Cells were maintained in DMEM containing 10% FBS and antibiotics (100 U/mL penicillin and 100 μg/mL streptomycin; ref. 19). Mouse thyroid cancer cells, TBP-3868, obtained from Dr. Sareh Parangi (Massachusetts General Hospital, Harvard Medical School, Boston, MA; ref. 20), were maintained in DMEM containing 10% FBS and antibiotics (100 U/mL penicillin and 100 μg/mL streptomycin). Cell lines were used within 15 passages of thawing and continuously cultured for less than 6 months. The cells were not tested for Mycoplasma. The appearance and growth characteristics of the mouse and human thyroid cancer cell lines used in this study were compared with published information to ensure their authenticity. Pancreatic cancer cells were authenticated by short tandem repeat profiling in October 2013 and also monitored by their appearance and growth characteristics to ensure their authenticity.

JQ1 (2) and CGP57380 (21) were purchased from Tocris Bioscience; OTX-015 (22) and ARV-825 (11) were obtained from Selleckchem; U0126 (23) was purchased from Cell Signaling Technology; and SB202190 (24) and NSC23766 (25) were purchased from Calbiochem.

Cells were suspended in type I collagen solution (2.2 mg/mL), which was made as described previously (26), and allowed to gel for 15 minutes at 37°C. For morphologic examination of cells, cell colonies in three-dimensional collagen were examined using a Zeiss Axiovert 40 CFL microscope and pictures taken with a Nikon Coolpix 4500 camera.

Approximately 1,000 cells were embedded in 2.2 mg/mL collagen I solution and seeded in 96-well plates. After 3 days of treatment with the different inhibitors, WST-1 reagent (Sigma-Aldrich) was added into the media at a 1:100 dilution factor. Absorbance was measured according to the manufacturer's instructions every hour for 4 hours or until maximal absorbance was reached. Coefficient of drug interaction (CDI) was calculated as CDI = AB/(A × B; refs. 27, 28). AB, relative cell viability of the combination; A or B, relative cell viability of the single-agent groups). CDI > 1 indicates an antagonistic effect; CDI = 1 indicates an additive effect; and CDI < 1 indicates a synergistic effect (with CDI < 0.6 indicates strong synergy; refs. 27, 28).

Relative mRNA expression levels and survival analyses were determined by cBioPortal analysis as described previously (29). To evaluate mRNA expression levels of fibrillary collagens and eIF4E, we used the papillary thyroid carcinoma dataset (30).

The Human Cytoskeleton Regulators RT² Profiler PCR Array, which enables profiling of 84 genes involved in cytoskeleton regulation, was purchased from Qiagen and performed according to the manufacturer's instructions and using quality controls included in the array. Expression of several representative genes from the array was validated by qRT-PCR.

SMARTpool On-Target siRNA targeting MNK1 and MNK2 were purchased from Dharmacon. siRNA targeting eIF4E was obtained from Ambion. All transfections with siRNA were carried out using RNAimax Reagent (Life Technologies) according to the manufacturer's instructions. RacGAP1 plasmid (variant 1) was obtained from Origene. Control plasmid was obtained from Lonza. All transfection with plasmids were carried out using Lipofectamine 3000 reagent (Life Technologies) according to the manufacturer's instructions.

Quantitative gene expression was performed with gene-specific TaqMan probes, TaqMan Universal PCR Master Mix, and the 7500 Fast Real-time PCR System from Applied Biosystems. Data were then quantified with the comparative C_t method for relative gene expression.

Whole-cell extracts of cultured cells were prepared in RIPA lysis buffer supplemented with phosphatase and protease inhibitors and separated on SDS-PAGE gel. The following antibodies and dilution factors were used: phospho-eIF4E (1:1,000, Cell Signaling Technology), total eIF4E (1:1000, Santa Cruz Biotechnology), phospho-MNK1 (1:1,000, Cell Signaling Technology), total MNK1 (1:1,000, Cell Signaling Technology), phospho-ERK1/2 (1:1,000, Cell Signaling Technology), total ERK1/2 (1:2,000, Santa Cruz Biotechnology), BRD4 (1:1,000, Abcam), Rac1 (1:2,000, EMD Millipore), RacGAP1 (1:2,000, Santa Cruz Biotechnology), and HSP90 (1:3,000, Santa Cruz Biotechnology). Blocking agent was 5% BSA. Secondary anti-mouse IgG (A4416) and anti-rabbit IgG (A6667) antibodies were purchased from Sigma and used at a 1:3,000 dilution factor. When necessary, membrane was stripped using Restore Western Blot Stripping Buffer (Thermo Fisher Scientific).

Human thyroid specimens were purchased from US Biomax and were trichrome stained or stained for eIF4E. Antigen retrieval was carried out as described previously (31, 32). Photographs for quantitative comparison were taken using FeinOptic microscope and Jenoptik ProgRes C5 camera. Cell lines (K1, MDA-T85, CD18, and Panc1) growing on glass coverslips were stained for phalloidin (Alexa Fluor 488) at a ratio of 1:500 and pictures were taken using FeinOptic microscope and Jenoptik ProgRes C5 camera.

TBP-3868 cancer cells were injected subcutaneously (1.5 × 10⁶ cells/site) into the flanks of 6- to 8-week-old B6129SF1/J mice (Charles River Laboratory). Mice were divided into four treatment groups: control (DMSO), JQ1 (12.5 mg/kg), CGP57380 (25 mg/kg), and a combination of JQ1 (12.5 mg/kg) and CGP57380 (25 mg/kg). Treatments were administered daily Monday–Friday for 2 weeks in a suspension containing 10% hydroxypropyl-β-cyclodextrin in double-distilled water. The mice were euthanized by CO₂ inhalation and cervical dislocation, and the tumors were excised and photographed.

Error bars represent SD. All statistical analyses were done using GraphPad Instat. P < 0.05 was considered significant.

All animal work and procedures were approved by the Northwestern University Institutional Animal Care and Use Committee. In addition, all animal experiments were performed in accordance with relevant guidelines and regulations.

Papillary thyroid cancer (PTC), the most common subtype of thyroid cancer (33), demonstrates a collagen-rich stromal reaction (Fig. 1A). Thus, we evaluated whether BETis limit growth of thyroid cancer cells in 3D collagen. Initially, we evaluated expression of members of the fibrillar collagen family in PTC samples available in the TCGA database (29). Increased expression of fibrillar collagen was associated with worse outcome (Fig. 1B). We next evaluated the effect of BETis on the growth of PTC cells growing in 3D collagen. Treatment with the BETis JQ1 and OTX-015 significantly decreased proliferation of thyroid cancer cells in 3D collagen (Fig. 1C and D).

Although transcriptional plasticity can reduce the efficacy of BETis (5–7), it is not known whether enhanced translational activity can also result in reduced efficacy of BETis. Thus, we evaluated the role of eIF4E, which regulates cap-dependent translation of specific mRNAs (12), in mediating response to BETis. Initially, we evaluated the role of eIF4E in mediating the growth of cancer cells in 3D collagen. Downregulation of eIF4E decreased growth of cancer cells in 3D collagen (Fig. 2A and B). Significantly, eIF4E knockdown was effective against both BRAF wild-type thyroid cancer cells and thyroid cancer cells expressing the BRAF^V600E mutation, which is present in approximately 60% of thyroid cancers (34). Importantly, eIF4E expression is increased in cancer samples relative to normal thyroid tissue (Supplementary Fig. 1A), with increased eIF4E expression associated with decreased overall survival in these patients (Supplementary Fig. 1B; ref. 29).

Because both BETis and eIF4E knockdown decreased growth of cancer cells in 3D collagen, we evaluated the extent to which BETis repress eIF4E expression in cancer cells. Although the BETis JQ1 and OTX-015 did not decrease eIF4E protein levels, we found that treatment with these inhibitors enhanced eIF4E phosphorylation (Fig. 2C). Importantly, eIF4E phosphorylation, which is regulated by the kinases MNK1 and MNK2, promotes tumor development in animal models (15). BETis also enhanced MNK1 phosphorylation in these cells (Fig. 2C). Significantly, pretreatment with the MNK inhibitor CGP57380 blocked JQ1-induced eIF4E phosphorylation (Fig. 2D). In addition, we evaluated the effect of BET PROTACs on eIF4E phosphorylation in thyroid cancer cells. Treatment with the BET PROTAC ARV-825 decreased BRD4 protein levels (Fig. 2C). As seen with BETis, treatment with ARV-825 also induced eIF4E phosphorylation (Fig. 2C).

Because the induction of eIF4E phosphorylation by BETis and BET PROTACs in thyroid cancer cells was unexpected, we evaluated whether BETis and BET PROTACs also induce eIF4E phosphorylation in other cancer cells. As we had shown previously that BETis decrease growth of pancreatic cancer cells in 3D collagen (5, 19), we evaluated the effect of BETis and BET PROTACs on eIF4E phosphorylation in pancreatic cancer cells. Significantly, treatment with JQ1 and ARV-825 induced MNK1 and eIF4E phosphorylation in pancreatic cancer cells (Fig. 2E). As with thyroid cancer cells (Fig. 2D), treatment with the MNK inhibitor CGP57380 blocked JQ1-mediated eIF4E phosphorylation in pancreatic cancer cells (Fig. 2F).

Because MNK kinases can be phosphorylated by ERK and p38 MAPKs (35), we next evaluated their roles in JQ1-induced eIF4E phosphorylation using specific MEK/ERK (U0126) and p38 MAPK (SB202190) inhibitors. Treatment with U0126 blocked ERK1/2 phosphorylation, but did not affect JQ1-induced MNK1 or eIF4E phosphorylation in cancer cells (Fig. 3A and C). However, treatment with the p38 MAPK inhibitor SB202190 decreased JQ1-induced MNK1 and eIF4E phosphorylation (Fig. 3B and D), establishing a requirement for p38 MAPK in JQ1-mediated MNK and eIF4E phosphorylation.

We noticed that BETis induced spindle-shaped morphology in thyroid and pancreatic cancer cells that was more clearly detected with phalloidin staining (Fig. 4A). As the change in morphology suggested increased Rac signaling, we evaluated whether blocking Rac activity could attenuate the effects of BETis on cytoskeletal changes. Pretreatment of cancer cells with the well-established Rac inhibitor NSC23766 (36) attenuated JQ1-induced morphologic and cytoskeletal changes (Fig. 4A). Importantly, the Rac inhibitor NSC23766 also blocked JQ1-induced eIF4E phosphorylation in both thyroid and pancreatic cancer cells (Fig. 4B).

To understand the mechanism by which BETis induce Rac signaling, we evaluated the effect of JQ1 on genes known to regulate Rac signaling using an RT²PCR array (Supplementary Fig. S2). JQ1 suppressed mRNA expression of ROCK1 and ROCK2 (Supplementary Table S1; Supplementary Fig. S2), which can counteract the effects of Rac signaling (37). While JQ1 not did not affect Rac1 protein levels (Fig. 4C), JQ1 also suppressed expression of RacGAP1 (Supplementary Table S1; Supplementary Fig. S2; Fig. 4C), a negative regulator of Rac activity (37). As these results suggest that BETis induce relative Rac activation by suppressing negative regulators of Rac signaling, we evaluated the extent to which overexpressing RacGAP1 could attenuate the effects of JQ1. Importantly, overexpression of RacGAP1 in cancer cells blocked JQ1-induced eIF4E phosphorylation (Fig. 4D).

We evaluated the effect of cotreatment with BET and MNK inhibitors on the growth of cancer cells in 3D collagen. Combining the MNK inhibitor CGP57380 with the BETi JQ1 significantly suppressed the growth of thyroid and pancreatic cancer cells (Fig. 5A and B). The coefficient of drug interaction between BET and MNK inhibitors was 0.54 in K1 cells and 0.61 in CD18 cells (Supplementary Table S2), indicating synergistic antiproliferative effects of combining BET and MNK inhibitors in cancer cells. Similarly, combining CGP57380 with the BETi OTX-015 or the BET PROTAC ARV-825 also significantly suppressed growth of cancer cells (Supplementary Fig. S3) and demonstrated synergistic antiproliferative effects (Supplementary Table S2).

We also evaluated the extent to which MNK1/2 siRNA enhanced JQ1 suppression of growth of cancer cells in 3D collagen. Initially, we evaluated the effect of MNK1/2 siRNA on JQ1-induced eIF4E phosphorylation. The MNK1/2 siRNA decreased basal and JQ1-induced eIF4E phosphorylation in cancer cells (Fig. 5C and D). The MNK1/2 siRNA also enhanced JQ1 suppression of growth of cancer cells in 3D collagen (Fig. 5E and F) and demonstrated synergistic antiproliferative effects, evidenced by their coefficient of drug interaction of 0.66 for K1 cells and 0.75 for CD18 cells (Supplementary Table S2).

Finally, we evaluated the effect of cotreatment with BET and MNK inhibitors on tumor growth in vivo using a syngeneic model of thyroid cancer. Because human thyroid cancer cells grow poorly in vivo (18), we used the mouse papillary thyroid cancer cell line TBP-3868. This mouse thyroid cancer cell line, derived from thyroid tumors developing in a transgenic mouse model expressing B-raf(V600E) with loss of both copies of p53 in the thyroid tissue (20), has been shown to faithfully recapitulate human papillary thyroid tumors. We initially evaluated the effect of BET and MNK inhibitors on the growth of TBP-3868 cells in 3D collagen. As with human cancer cells, combining the MNK inhibitor CGP57380 with the BETi JQ1 significantly suppressed the growth of mouse TBP-3868 cells (Fig. 6A). We then injected TBP-3868 cells in the syngeneic B6129SF1/J mice, and established tumors were treated with JQ1 and/or CGP57380. Combination treatment with BET and MNK inhibitors significantly reduced tumor growth in vivo (Fig. 6B and C). Significantly, in contrast to JQ1 alone, combination of JQ1 and CGP resulted in tumor shrinkage relative to size of tumors at the start of treatment. Importantly, the combination treatment did not affect the weight of the treated mice (Fig. 6D). Together, these findings demonstrate cross-talk between BET proteins and the MNK–eIF4E pathway and identify combination therapy with BET and MNK inhibitors for the treatment of patients with cancer.

BETis have shown some efficacy in hematologic malignancies, NUT carcinomas, and in select solid tumors in clinical trials (3, 4). However, the efficacy of BETis for most solid tumors is limited by minimal single-agent response or by the development of resistance to BETis (3, 4). For example, Myc-dependent pancreatic cancer and AML cells developing resistance to BETis activate alternative non-BRD transcriptional program to increase Myc protein levels (5, 6). In contrast, ovarian cancer cells activate compensatory prosurvival kinase network to overcome the effects of BETis (38). Also, colorectal cancer cells with intrinsic resistance to BETis have increased activation of the MEK/ERK signaling pathway (39). In this report, we show that cancer cells demonstrate increased MNK and eIF4E phosphorylation following treatment with BETis JQ1 and OTX015 as a compensatory prosurvival feedback mechanism to limit the antiproliferative effects of BETis. We also show that the BET PROTAC ARV-825 induces eIF4E phosphorylation. This is not surprising because ARV-825 consists of OTX-015 attached via a linker to pomalidomide to promote E3 ubiquitin ligase–mediated degradation of BRD proteins (8). Importantly, we show that blocking eIF4E phosphorylation with MNK inhibitors and MNK knockdown enhances the antitumorigenic effects of BETis and BET PROTACs.

We show that the p38 MAPK pathway, but not the MEK/ERK pathway, mediates BETi-induced eIF4E phosphorylation. Using the well-established Rac inhibitor NSC23766 and overexpression of RacGAP1, we also show that the Rac signaling is involved in eIF4E phosphorylation. Importantly, Rac signaling can mediate p38 MAPK activation in a number of model systems. For example, Rac mediates activation of p38 MAPKs following hyperosmotic shock of mammalian cells (40), following treatment of leukemia and breast cancer cells with all-trans-retinoic acid (41), and following treatment of leukemia cells with arsenic trioxide (42). Significantly, similar to our findings with BETis, arsenic trioxide also induces p38 MAPK–mediated eIF4E phosphorylation in leukemia cells (43). While we establish the role of Rac, p38 MAPKs, and MNK kinases in regulating eIF4E phosphorylation following treatment with BETis, Rac signaling can also regulate eIF4E function by modulating the cellular localization of cytoplasmic FMR1–interacting protein 1 (CYFIP1; ref. 44). CYFIP1 is associated with two distinct protein complexes: the wave regulator complex (WRC) to regulate actin remodeling and the eIF4E complex to inhibit translation of specific mRNAs (44). Rac activation causes redistribution of CYFIP1 from the eIF4E complex to the WRC, resulting in increased mRNA translation and actin polymerization (44). Future studies will evaluate the extent to which BETis regulate eIF4E and mRNA translation by modulating the function of CYFIP1.

In contrast to a recent study showing that JQ1 represses eIF4E in lung cancer cells (45), we have found that BETis do not repress eIF4E protein levels in thyroid and pancreatic cancer cells. This discrepancy may be due to differences in tumor types and/or duration of treatment with BETis. However, given the importance of eIF4E in cancer development and progression, the function of eIF4E is tightly regulated in cells under normal physiologic conditions and in response to different signals (14, 35). eIF4E activity can be regulated by two major signaling pathways, the MAPK–MNK and the mTOR signaling pathways (14, 35). Interestingly, both the MAPK and mTOR signaling pathways can mediate resistance to BETis. The mTOR signaling was recently shown to mediate resistance to BETis in SPOP-mutated cancer cells (10), while we demonstrate in this report the importance of the MAPK–MNK pathway in limiting the efficacy of BETis in cancer cells.

Finally, our findings suggest that MNK inhibitors should be combined with BETis and BET PROTACs for treatment of solid tumors. There are a number of BETis and BET PROTACs that are currently in preclinical development or in early-phase clinical trials (4, 46). For example, the BETi OTX-015 was found to have reasonable safety profile in phase I clinical trials in patients with hematologic malignancies and select solid tumors (3, 4, 47, 48). Similarly, given the role of MNK kinases in mediating resistance and demonstrating synergistic activity of MNK inhibitors with other inhibitors (12, 49, 50), there is also interest in moving MNK inhibitors into clinical trials. For example, the MNK1/2 inhibitor eFT508 (Effector Therapeutics) is currently undergoing phase I clinical trial in patients with hematologic malignancies and solid tumors (49). Overall, our results demonstrate activation of the MNK–eIF4E pathway by BETis, and suggest that combination therapy with BET and MNK inhibitors should be pursued for the treatment of solid tumors.

No potential conflicts of interest were disclosed.

Conception and design: T.N.D. Pham, M. Matsangou, H.G. Munshi

Development of methodology: T.N.D. Pham

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.N.D. Pham, K. Kumar, B.T. DeCant, M. Shang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.N.D. Pham, K. Kumar, M. Shang, S.Z. Munshi, H.G. Munshi

Writing, review, and/or revision of the manuscript: T.N.D. Pham, M. Matsangou, H.G. Munshi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Ebine, H.G. Munshi

Study supervision: T.N.D. Pham, H.G. Munshi

Others (clinical collaborator): M. Matsangou

This work was supported by grants from the NCI (R01CA186885, to H.G. Munshi; and R21CA220625, to K. Kumar), a Merit award from the Department of Veteran Affairs (I01BX002922, to H.G. Munshi), and a Translational Bridge Fellowship Award from the Lurie Cancer Center (to T.N.D. Pham).

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.